Materials for energy conversion devices

Materials for energy conversion devices 1 Materials for solar cells M A G R E E N, University of New South Wales, Aus...

Author: Charles C. Sorrell | Sunao Sugihara | Janusz Nowotny

305 downloads 2468 Views 4MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Materials for energy conversion devices

1 Materials for solar cells M A G R E E N, University of New South Wales, Australia

1.1

Introduction

Solar cells are one of the most benign options yet suggested for generating the world’s future energy needs. Present photovoltaic technology, based on silicon wafers similar to those used in microelectronics, works extremely well and underpins a rapidly expanding industry. However, to be used on the very large scale that is technically possible, less material-intensive approaches need to be developed to reduce costs. This chapter reviews the current status and emerging trends with both the silicon wafer-based approaches as well as with the ‘second generation’ thinfilm technologies likely to play an increasing role in the future. These thin films include thin films of silicon in amorphous and polycrystalline phases, as well as in an intermediate ‘microcrystalline’ mixed phase. Other interesting thin film materials include chalcogenide-based polycrystalline compound semiconductors such as copper indium diselenide and cadmium telluride, which have attractive features such as the relative electronic inactivity of grain boundaries in these materials. Finally, progress with dye-sensitised and organic materials is described.

1.2

Present market status

Solar cells based on silicon wafers have been the workhorse of the photovoltaic industry over the past decades. Recent major investments in new manufacturing facilities for such cells ensure this role will continue well into the future. A recent market survey reports that, of all solar photovoltaic module sales in 2003, 32% were based on monocrystalline silicon wafers, essentially the same wafers as those used in microelectronics (Schmela, 2004). A further 57% were based on lower-quality multicrystalline silicon wafers. These are large-grained polycrystalline wafers produced by slicing from large ingots of directionally solidified silicon, an approach developed specifically for photovoltaics. A further 4% of sales were based on multicrystalline silicon 3

4


ribbons and self-supporting silicon sheet, technologies again developed specifically for photovoltaics, but with the advantage of not requiring slicing into wafers. Combined, these ‘bulk’ silicon approaches accounted for a total of 94% of annual production in 2003 (Schmela, 2004). Most of the remaining production was made up of thin-film amorphous silicon solar cells, including multijunction stacked ‘tandem’ cells. Only 1–2% of total production was accounted for by thin films not involving silicon, specifically by cells based on the chalcogenides Te and Se in the form of polycrystalline thin films of CdTe and CuInSe2 (CIS).

1.3

Bulk silicon

1.3.1

Market overview

Figure 1.1 shows a sampling of the nominal performance of commercial modules from different manufacturers, for modules based on bulk silicon wafers, ribbon and sheet. As noted, such modules accounted for 94% of 2003 production. These modules have a well-established reputation for reliability 20

Ribbon/sheet

Multicrystalline

Monocrystalline

Efficiency (%)

15

10

5

HIP 190/G751

BP 7180/70s

NT-185/175U1

I-165/159

Shell SM110/100

APi-165

Blue Pwr 180/165

KC167/158G

ND-165UE

ASE 200/205

BP 3160/3150

Shell S115/105

ASE-300-DG

EC-115

AP × 140

0

1.1 Survey of bulk crystalline silicon solar module performance. The chart shows the nominal energy conversion efficiency under standard test conditions for several module types based on the manufacturer’s rating and the module’s total framed area. The bar at the top shows the likely range for delivered product falling within specifications.

Materials for solar cells

5

and durability, with many systems installed over 20 years ago still performing creditably (Realini et al., 2001; De Lia et al., 2003). Most commercial bulk silicon modules have energy conversion efficiency, the ratio of electrical power output to solar power on the total module area, in the 10–15% range. Those at the high end of this range are based on monocrystalline silicon wafers, those in the mid-range generally are based on multicrystalline wafers, while modules based on silicon ribbon and sheet occupy the low end of the range. Trends apparent with each of these different bulk substrate types are discussed in the following sections.

1.3.2

Monocrystalline silicon wafers

Czochralski ingot growth Silicon is the second most abundant element in the earth’s crust, behind only oxygen. Although its main use is as an alloying agent in the steel industry, a smaller, (but higher-value) use is in microelectronics. After conversion of ‘metallurgical grade’ material to a purer ‘semiconductor grade’ polycrystalline form, cylindrical crystalline ingots are grown predominantly by the Czochralski technique, although some ingots are grown by the float-zone method. In the Czochralski method, silicon is melted in a crucible and growth is onto a small rotating seed crystal introduced at the top of this melt (Fig. 1.2). Although ingots of 30 cm diameter and weighing over 100 kg are produced

Seed

Crystal

1.2 Czochralski growth of cylindrical silicon ingot.

6


for microelectronics, smaller diameter ingots in the 15–20 cm diameter range are of more interest to photovoltaics, since these are more robust when sliced into thin wafers. Over the last decade, the preferred technique for slicing has switched from the use of inner-diameter saws to wire saws. One manufacturer now offers monocrystalline wafers for photovoltaics prepared using the float-zone technique (Vedde et al., 2004). The float-zone technique requires a high-quality polycrystalline feed-rod, which is crystallised by passing a thin molten zone along its length (Fig. 1.3). The disadvantage is the more demanding feedstock requirements while the advantage for photovoltaics is the higher cell performance possible from ingots prepared this way.

Polysilicon feed rod

RF coal Molten zone

Seed

1.3 Float-zone growth of silicon ingot.

Screen-printed cells Most monocrystalline cells fabricated during 2003 used the screen-printed cell structure shown schematically in Fig. 1.4. By the early 1980s, this approach had displaced alternatives to emerge as the commercial standard (Green, 1995). One advantage was that, as well as being based on silicon wafers developed for microelectronics, it was able to use the same screen


7

150–200 µm 3 mm

Patterned metal contact

Phosphorus Bulk of wafer

n++ Rear metal contact p-type p+

Metal

1.4 Screen-printed solar cell (not to scale) (Green, 1995).

printers, drying and firing furnaces for applying cell contacts as developed for thick-film, hybrid microelectronics. The main cell processing steps (Jester, 2002) consist of wafer cleaning and chemical etching, usually anisotropically to form the micron-sized, crystallographically defined pyramids covering the wafer surface in Fig. 1.4. This is followed by p-n junction formation, either in the same dopant diffusion furnaces as used in microelectronics or using simpler approaches. Examples are the spraying or spinning-on of dopant sources and their subsequent diffusion in the same type of belt furnaces used for contact firing. Contacts are applied as metal pastes, with their pattern on the cell surface defined by printing through an appropriately patterned emulsion mask or ‘screen’. A quarter wavelength dielectric antireflection coating is usually applied either before or after printing and subsequent firing of the top contact pattern. The dielectric has often been TiO2, although silicon nitride is rapidly becoming more popular. The strengths of this technology are the simplicity of cell processing and the ready availability of the required processing equipment. However, compromises in cell design are required to accommodate the less than ideal features arising from the use of screen-printing as a method for applying the cell’s top contact. These constrain cell performance to levels below those that are fundamentally possible (Green, 1995) giving rise, over recent years, to the higher performance sequences described on pages 7–10. The p-type wafers used in these cells are doped with boron during the preparation of the cylindrical ingots from which they are sliced. This is partly for historical reasons, since regions with such doping are more resistant

8


to damage by the high energy particles found in space, where the early commercial applications of silicon cells were found. The complementary ntype surface diffusion using phosphorus is also less demanding than the ptype boron diffusion required if an n-type wafer were used. Recent practice has been to push to progressively thinner wafers to reduce silicon material costs (Jester, 2002). Over recent years, a disadvantage arising from the use of p-type wafers has been recognised and is the subject of much recent research (Rein et al., 2000). Oxygen is also unintentionally incorporated into the wafer during the growth of the original ingot, seeping from the quartz crucibles holding the molten silicon. This oxygen is mostly inert and, in fact, improves wafer strength. However, under illumination, some interacts with the boron dopants to form an electrically active boron–oxygen complex that detracts from device performance. Module output drops about 3% relative under the first few hours of light exposure as a consequence (Eikelboom and Jansen, 2000; De Wolf et al., 2000). This is accommodated within the manufacturer’s warranty, which is generally specified as involving less than 10% module output loss over the first 10 years, and less than 20% over the first 20 years. Elimination of this effect would have obvious advantages, as well as less obvious ones, since it would allow an increase in the stabilised performance potential of higher efficiency cell processing sequences (Rein et al., 2000), such as the buried contact sequence of the following section. Buried contact solar cells The buried-contact cell design of Fig. 1.5 was developed by the author’s group in the early 1980s as a low-cost approach to incorporating some of the gains in laboratory performance of this era into production (Green, 1995). The key feature of this approach is the use of a laser to form grooves into the top surface of the cell, through a previously lightly diffused layer and dielectric coating. These grooves expose fresh silicon that can be heavily doped during a second diffusion, confined by the dielectric to the grooved region. Similarly, the dielectric confines a subsequently electrolessly plated metal layer to this region. Finally, the dielectric serves as an antireflection coating for the final cell. The advantage of this approach is that the quality of the silicon in the surface region of the cell need not be sacrificed, as required for good contact using screen-printing, allowing full response to blue wavelengths. There are also advantages in reduced shadowing of the top surface of the cell by the narrower fingers resulting from this approach and inherently lower series resistance (Green, 1995). The resulting 10–20% performance advantage compared to the screenprinting approach translates to a nearly proportionate cost advantage for similar production volumes. This is due primarily to the shared high material


Oxide or nitride

9

n+ n++

p+

p-type

Plated metal (buried contact)

Metal

1.5 Buried contact solar cell (not to scale) (Green, 1995).

costs when combined with similar processing costs. Since the sequence can extract the full performance benefits from improved wafer quality, control of the boron–oxygen defects mentioned on pages 5–7 would give even greater advantages for this technology (Rein et al., 2000). BP Solar, one of the largest solar cell manufacturers in 2003, has invested in commercialising this technology as its premium ‘Saturn’ product line and has recently expanded production capacity, targeting 80 MW/year by 2006 (Mason et al., 2002). HIT cell An alternative approach to a higher efficiency commercial solar cell is the HIT (heterojunction with thin intrinsic layers) cell of Fig. 1.6. This cell combines both crystalline and amorphous silicon cell design features in the one structure. Hydrogenated amorphous silicon, prepared by plasma-enhanced chemical vapour deposition (PECVD), has a higher band gap than crystalline material (Section 1.4.2). Consequently, this material forms a high band gap hetero-interface with the underlying silicon wafer, providing a very effective, low recombination cap on this wafer. The uppermost thin heavily doped ptype amorphous silicon layer forms a junction with the underlying n-type crystalline wafer. An intervening, very thin intrinsic amorphous silicon layer plays an important role in obtaining high performance levels (Sakata et al., 2000). A reversed polarity structure on the rear of the wafer provides the equivalent of a ‘back surface field’ (Green, 1995). Since the conductivity of even heavily doped amorphous silicon is quite low, due to poor carrier mobility, transparent conducting oxides are required on both front and rear

10

Materials for energy conversion devices Metal

TCO p+ (a-Si) i(a-Si) n(c-Si) i(a-Si) n+(a-Si) Bottom electrode

Metal

1.6 HIT solar cell (not to scale).

surfaces to allow lateral carrier transport to metal contacts screen-printed on both surfaces. There are several other interesting technical features. The quality of surface passivation offered by the amorphous silicon layers is so high that nearrecord output voltages have been confirmed for this approach (Sakata et al., 2000), forming the basis for its high energy conversion efficiency. Also, the approach uses n-type, phosphorus doped wafers, almost uniquely within the industry at present. This overcomes the issues with boron–oxygen defects mentioned on pages 5–7, since there appears to be no corresponding problems with phosphorus–oxygen defects. The main technical weakness of the approach is that the required transparent conducting oxide layers are neither perfectly transparent nor perfectly conducting. This forces a trade-off between light absorption in these layers and lateral resistance losses. Light absorbed in the heavily-doped amorphous layers in these devices is also wasted. These absorption losses result in 10– 15% current loss, tending to offset the high voltage outputs previously noted, but still producing the highest performing modules in Fig. 1.1. Sanyo is reported to be expanding production targets from 32 MW in 2003 to 120 MW in 2005 (Schmela, 2004) which, along with BP Solar’s expansion plans, suggests an increasing market share for high efficiency, monocrystalline cells.

1.3.3

Multicrystalline silicon wafers

Multicrystalline ingot growth Multicrystalline silicon wafers are produced by crystallising molten silicon by directional solidification in large crucibles as shown in Fig. 1.7(a). These


11

Silicon

Mould (a)

(b)

1.7 Multicrystalline silicon casting and sawing.

large ingots are then sawn into smaller units as in Fig. 1.7(b), and these are then sliced into wafers. Wafers are generally square with sides in the 10–20 cm range. The advantage of this technique compared to the Czochralski monocrystalline technique is its much higher throughput and its ability to tolerate poorer feedstock material. The disadvantage is the lower cell performance and the greater variability in the quality of the final wafer, resulting in about 20% variability in the final cell performance. Multicrystalline cell processing Most solar modules produced during 2004 used multicrystalline silicon wafers rather than monocrystalline ones. Grains are generally much larger than the wafer thickness (0.3 mm) and hence extend through the wafer as shown in Fig. 1.8. All commercially processed multicrystalline wafers are presently processed with a screen-printing sequence similar to that outlined for monocrystalline wafers on pages 6–8 although with some differences. The lack of control over grain orientation means that the crystallographic surface texturing used for monocrystalline silicon cells as in Fig. 1.4 is not effective. Reflection control using quarter-wavelength antireflection coatings is therefore essential for good performance. There is also a higher density of crystallographic defects in multicrystalline wafers, not only represented by grain boundaries but also by intragrain dislocations and point defects. Often there will be higher levels of impurities as well, since less pure source material and more rugged manufacturing equipment can be tolerated. There will also be a wide variation in crystallographic quality and impurity levels from wafers cut from the same ingot. Wafers from different suppliers will also often differ in terms of the processing conditions that give best results. Introducing hydrogen in atomic form into the wafer during processing is a particularly effective way of reducing the scatter in final cell performance resulting from the above variances. A method for incorporating hydrogen demonstrated in the early 1980s was by the plasma enhanced chemical vapour

12

Materials for energy conversion devices 150–200 µm 3 mm

Patterned metal contact

Phosphorus Bulk of wafer

n++ p+

p-type

Rear metal contact

Metal

1.8 Screen-printed multicrystalline silicon solar cell.

deposition (PECVD) of a silicon nitride antireflection coating during cell processing, using silane (SiH4) and usually ammonia (NH3) as the source gases. These sources ensure the abundance of atomic hydrogen during deposition, together with its incorporation into the deposited coating and its subsequent diffusion into the underlying wafer. High throughput equipment for such deposition has recently become available (von Aichberger, 2003), allowing these advantages to be captured by an increasing number of manufacturers. This development helps bridge the gap in performance between monocrystalline and multicrystalline cell performance. At the module level, the gap is further reduced by the higher packing density possible for the generally square multicrystalline wafers, as opposed to the circular or trimmed ‘quasi-square’ monocrystalline wafers that are the most economical option, when cut from originally cylindrical ingots. As in the monocrystalline case, the industry is moving towards progressively thinner multicrystalline wafers. There is also, in principle, more flexibility in controlling oxygen levels in these wafers, which may allow the boron-oxygen defect issue to be addressed satisfactorily. The use of higher efficiency processing sequences, such as the buried contact, is also anticipated with excellent laboratory results with this approach demonstrated recently (Jooss et al., 2002).

1.3.4

Silicon ribbon and sheet

Directly preparing silicon in the form of a ribbon or sheet saves the cost of wafering as well as wasting less of the silicon source material. In comparable


13

production volumes, this is expected to give a cost advantage compared to the wafer approach, provided similar yields and energy conversion efficiencies can be maintained. The approach accounting for most ribbon cell production during 2003 (Schmela, 2004; Kalejs, 2003) is the EFG (edge-defined film-fed growth) method shown schematically in Fig. 1.9(a). In the present commercial

Polycrystalline ribbon

Molten silicon moves by capillary action

Carbon die

Molten silicon

(a)

Silicon dendrites or carbon string

Molten silicon

(b)

1.9 (a) Edge-defined film-fed growth (molten silicon moves up the graphite die by capillary action with the ribbon shape defined by the shape of the top of the die); (b) web or string approach (as dendrites or strings are drawn out of the melt, a thin layer of initially molten silicon is trapped between them).

14


implementation, an octagonal tube of multicrystalline material is grown using an appropriately shaped carbon die with wafers then cut out of this tube (Schmidt et al., 2002). This approach is used by the manufacturer, RWE Schott Solar, and results in module performance only a few percent relatively lower than by the same manufacturer using standard multicrystalline wafers. Smaller quantities of ribbon cells were also produced by Evergreen Solar during 2003 by a variant of the web or string approach shown in Fig. 1.9(b). A film of molten silicon is trapped between two ‘wires’ drawn through a molten bath, with the trapped film then solidifying as ribbon (Wallace et al., 1997). Module efficiency is lower than with the EFG approach (von Aichberger, 2004) although the gap may close with time. Given the generally lower performance of ribbon substrates, the defect passivation approaches mentioned for multicrystalline wafers are even more critical for these materials. Another issue can be surface morphology which ranges from rough for the EFG approach to very smooth. Variants of the standard screen-printing approach have been developed to accommodate rough morphology (Schmidt et al., 2002). The ‘Apex’ silicon sheet approach developed by Astropower started as an approach where silicon was deposited onto stainless steel, evolved to silicon deposited onto ceramic-coated stainless steel, then to silicon deposited onto ceramic substrates and finally to silicon deposited onto a preformed silicon template, bearing some relationship to an earlier silicon sheet from powder approach (Eyer et al., 1987). When wafers are cut from the continuously formed sheet, the appearance and the required processing conditions are not too different from a uniformly grained multicrystalline wafer (Culik et al., 2002). At the present stage of development, however, performance is more modest than with such wafers, with this material giving the lowest performance of the bulk silicon approaches surveyed in Fig. 1.1.

1.4

Thin-film silicon

1.4.1

Thin-film status

Over recent years, cautious quantifies of thin-film photovoltaic modules intended for outdoor applications have become available commercially (Green, 2003). A survey of the performance of thin-film modules on the market during 2003 is summarised in Fig. 1.10. Although excellent progress has been made in the laboratory with chalcogenide-based CdTe and CuInSe2 (CIS) polycrystalline thin-film cells (Green, 2003), their market introduction has been quite subdued accounting for only 1–2% of total production in 2003. The difficulty with CdTe has been the lack of market acceptance of a photovoltaic product based on toxic material in what is fundamentally a


15

20

Efficiency (%)

15

a-Si Double/ triple junction

a-Si Single junction

a-Si/µc-Si Double

CdTe

CIS

10

5

ST-36-40

WS 11007

FS 50

ATF 36/50

Apollo 980

Hybrid

US-64

Millenia 43/50

RNE30-DG-UT

EPV-40

FEE20-12

LSU

DS40

B108D

FEE14-12

AMP-1815

0

Module

1.10 Thin-film module survey (see Fig. 1.1 for detailed explanation).

‘green’ market. The apparent lack of stability of unencapsulated CdTe and CIS devices under damp-heat accelerated testing suggests more stringent encapsulation requirements to match the reliability of bulk silicon (Oszan and Dunlop, 2002; Erler et al., 2003). With CIS, reports of low manufacturing yields due to the complexity of the material system, the well-known volatility and reactivity of elements such as Cu, and metastabilities that do not appear to be completely understood seem to have dampened enthusiasm for large investments in increased manufacturing capacity. Fundamentally, there also appears to be a problem with resource availability. All the world’s known resources in In, if used to make CIS cells, would allow these to match the installed capacity of wind generators worldwide at the end of 2004 (Andersson, 2000), but not contribute significantly to the major issues driving the development of photovoltaics. The above difficulties with the non-silicon approaches has improved prospects for three silicon-based thin-film approaches, specifically those based on single-junction and multiple-junction tandem amorphous silicon, hybrid tandem cells based on amorphous and microcrystalline silicon and, most recently, single-junction cells based on polycrystalline silicon thinfilms deposited onto glass.

1.4.2

Amorphous silicon cells

Silicon deposited at low temperature by PECVD using the gas silane (SiH4) as the silicon source forms amorphous material with a high level of hydrogen

16


incorporated (~ 10%). This hydrogenated material behaves vastly differently from pure amorphous silicon, with band gap increasing from 1.1 eV to about 1.7 eV and with greatly improved electronic properties (Green, 2003). These electronic properties, however, remain quite modest compared to crystalline silicon. Unfortunately, the beneficial effects of hydrogen upon material quality are partly undone by exposure to sunlight. Commercial cells have to be designed around the ‘stabilised’ quality of the material after light exposure, rather than the initial ‘as-deposited’ quality. Laboratory devices can be designed around the latter, however, giving a large difference in performance between the best ‘unstabilised’ laboratory devices and what can be produced commercially. This explains the relatively low efficiencies of the commercial single-junction amorphous silicon cells in Fig. 1.10, and also why amorphous silicon cells have not attained the market dominance once thought likely. To accommodate the relatively poor material quality, particularly the low carrier mobilities, device design differs from standard crystalline devices in two ways (Green, 2003). To aid the collection of the photo-generated carriers, these carriers need to be generated in a region where an electric field is present. This is achieved by using the p-i-n structure of Fig. 1.11. By inserting a lightly doped ‘intrinsic’ layer between p- and n-type layers, the high electric field region arising from the work function difference between these regions can be stretched over a large volume. Stretching too far can be self-defeating as this reduces the field, making it less effective in aiding carrier collection.

Transparent conductor

p-layer

i-layer

n-layer

Rear metal contact

1.11 Amorphous silicon p-i-n solar cell structure.


17

The second difference is that the low mobilities in the doped regions combined with their small thicknesses means they have insufficient conductivity to allow lateral flows of carriers within them. A transparent conducting oxide, normally SnO2, is required to provide this lateral conductance on the cell surface exposed to light. Unlike the schematic shown in Fig. 1.11, this is often deposited with surface texture or ‘haze’ to improve the optical performance of the completed cell (Lechner and Schade, 2002). Apart from the greatly reduced material costs, an advantage of the thinfilm approach is that cells deposited over a large glass sheet can be interconnected automatically during deposition by appropriate patterning between deposition steps, as shown in Fig. 1.12. Light

glass SnO2 a-Si:H Al EVA Glass

1.12 Interconnection scheme for amorphous silicon cells.

A large fraction of the thin-film cells produced commercially during 2003 had the single-junction structure so far discussed. An improvement is to go to the stacked ‘tandem’ cell structure of Fig. 1.13. One advantage of stacking two cells is that this allows an approximate doubling of the thickness of the intrinsic i-layer. Some manufacturers such as RWE Schott Solar are content to capture this advantage alone, and use a stack of two cells from the same material (Lechner and Schade, 2002). However, a performance advantage is possible if the underlying cell is made from different material with a smaller band gap. An alloy of amorphous silicon and germanium (Ayra and Carlson, 2002) has been the standard way of implementing this lower cell. If two stacked cells are good, three is even better if cost is no barrier. One company, United Solar, uses a triple-junction stack with different Ge content in the two underlying cells to produce modules with a stabilised efficiency of 6.3% (less than half that of the best crystalline module of 15.2%). This company is in the process of commissioning a new 25 MW manufacturing facility based on this approach, where the cells are deposited continuously

18


Glass SiO2 SnO2 p-layer i-layer

a-Si:H

n-layer p-layer i-layer

a-Si:H, a-Si:Ge:H, or µc-Si:H

n-layer ZnO Rear metal contact

1.13 Tandem a-Si:H solar cell.

on a sheet of stainless steel several kilometres long in a roll-to-roll process. Unfortunately, depositing onto a stainless steel sheet does not allow one of the advantages of a thin-film approach (automatic interconnection during deposition) to be fully exploited. Although the three cells in the tandem stack are so inter-connected, these need to be cut from the stainless steel roll and interconnected within the module, as in a bulk crystalline cell module.

1.4.3

Amorphous/microcrystalline silicon tandem cells

Dilution of the source gases for amorphous Si deposition in hydrogen can have a large impact on the structure of the deposited film (Green, 2003). At high dilutions, the silicon is deposited as a mixed-phase, microcrystalline form as shown in Fig. 1.14(a). As opposed to single-phase polycrystalline material shown in Fig. 1.14(b), the mixed-phase microcrystalline material consists of small crystalline regions, as indicated, linked by regions with a high amorphous content. Despite the similarity to amorphous silicon deposition conditions and the high levels of hydrogen incorporated, the material exhibits properties more similar to rudimentary crystalline than to amorphous silicon. Importantly, the material seems immune from the degradation effects that affect the latter. The relatively poor carrier mobilities again make p-i-n structures essential


19

Column Crystallite Crystallite

Disorder Substrate µc-Si:H (a)

Substrate poly-Si (b)

1.14 (a) Schematic showing key features of the structure of mixedphase, microcrystalline silicon layers; (b) structure of higher temperature polycrystalline silicon layers as used on the ‘crystalline silicon on glass’ approach (Fuhs, 2002).

for good carrier collection, although good collection is maintained for thicker layers than for amorphous silicon. For interconnected tandem cells as in Fig. 1.13, the current from the top amorphous silicon cell has to match that from the bottom microcrystalline device. It turns out that it is difficult for the top amorphous cell to do this (Green, 2003). Although high efficiency is possible if this top cell is made thick, this will result in a marked drop in performance as the quality of this material ‘stabilises’ under sunlight exposure. As for standard amorphous silicon cells, this can lead to large differences between the excellent unstabilised laboratory cell performance demonstrated by this approach and that of stabilised commercial devices. The first commercial devices using this approach were produced by Kaneka in 2002 with nominal energy conversion efficiency in the 8–9% range (Fig. 1.10). Although there is not the same amount of independent information about stabilisation properties available as for the other amorphous silicon technologies, one recent data set provides an indication of present relativities. In side-by-side testing (LEEE, 2002), the efficiency of a triple-junction amorphous silicon module degraded 22% after three months in the field, from 6.4% to 5.0%, consistent with the results of other field studies (LEEE, 2000; Carr and Pryor, 2001). The hybrid amorphous/microcrystalline device degraded 19%, almost exactly the same, with its efficiency decreasing from 8.7% to 7.0%. The hybrid therefore demonstrates a clear efficiency advantage for a similar level of stability, justifying the increasing interest in this approach.

20

1.4.4


Thin-film polycrystalline silicon on glass

This new ‘crystalline silicon on glass’ (CSG) approach, which the author has helped develop, attempts to combine the well-established strengths of the bulk silicon approach with the advantages of a thin-film technology. This results in the stability, durability, abundance and non-toxicity of bulk silicon being retained, while capturing the key advantages associated with thin films, namely greatly reduced material costs and large area monolithic construction. A schematic of a module manufactured by this approach is shown in Fig. 1.15. An anti-reflection coating and the doped silicon layers are deposited in the same deposition chamber onto textured glass by low temperature PECVD to a total thickness of 1–2 microns. The silicon is then crystallised by a high temperature step, to produce single-phase polycrystalline material as in Fig. 1.14(b). The quality of this material is sufficient for the normal diffusive collection of carriers to be possible, as in a standard crystalline wafer cell. Also, carrier mobilities are sufficiently high for good lateral conductance, removing the need for transparent conducting oxide layers that cause performance loss, add cost, and can give rise to durability problems (Oszan and Dunlop, 2002). A unique ‘crater’ and ‘dimple’ approach is used to contact the opposite polarity regions of the cell and to provide monolithic interconnection. This approach is a recent development and has been evaluated only at the pilot-production level. The reported rate of pilot-line module efficiency improvement has been quite remarkable (Fig. 1.16), with 8% energy conversion efficiency confirmed during 2002. Very high manufacturing yield well above 90% has also been reported, even at this early stage of development (Basore, 2002a, 2002b).

Metal

Resin p+ p n+

Silicon ‘Crater’

‘Groove’

‘Dimple’

Textured glass Light in

1.15 ‘Crystalline silicon on glass’; cell diagram and interconnection schematic.

Testing to date has indicated excellent stability and durability for these modules. As well as passing the standard IEC 61646 qualification test, eight


21

Best module efficiency (%)

9 8 7 6 5 4 3 2 1 0 1998

1999

2000

2001

2002

1.16 Aperture-area efficiency of pilot-line ‘crystalline silicon on glass’ module measured at Sandia National Laboratories. The efficiency reported is the aperture-area efficiency under standard reporting conditions for laminated modules having an area between 480 and 900 cm2 as measured outdoors on a tracking structure at Sandia National Laboratories (Basore, 2002b).

modules were subjected to a much harsher version of this test whereby the same module is exposed to the standard thermal cycling, humidity freeze and damp heat. All eight modules survived two rounds of these combined cycles, a feat not able to be matched by all the very reliable wafer-based modules tested at the same time (Basore, 2002b). This already exceptional performance is not totally unexpected, since many of the normal degradation modes associated with standard wafer-based modules are eliminated by this approach. There is no organic material (EVA) on the sun-facing side of the silicon sheet to degrade under ultraviolet exposure, no wafers to crack and no cell interconnect wires to fatigue. Estimated production costs per unit area are comparable to the simplest single-junction amorphous silicon thinfilm sequences, but with the advantages of higher output power, high yields, stable output and good module durability (Basore, 2002b).

1.5

Chalcogenide-based cells

1.5.1

CdTe cells

A key strength of CdTe cells are that they can be prepared by a range of simple techniques and still give good properties (Bube, 1998). This is attributed to the ability of post-deposition treatments to increase the grain size and to reduce the activity of grain boundaries in this material. Best results have been obtained with the CdTe/CdS/SnO2 heterojunction structure of Fig. 1.17. The deposition of the transparent conducting oxide (TCO) layer of SnO2 onto the glass substrate is followed by the deposition of a thin layer of CdS,

22


Glass TCO Window Alloy layer

Absorber

SnO2 CdS CdSxTe1–x

CdTe

Metal contact

1.17 CdTe cells.

often by chemical bath deposition. This layer is usually heat treated in a reducing atmosphere or in CdCl2 to increase grain size and reduce defect density (Chu and Chu, 1995; Meyers and Birkmire, 1995). Next, the CdTe layer is deposited by one of a variety of techniques, followed by a heat treatment in CdCl2 or another chlorine-containing compound. The heat treatment not only increases grain size and reduces defect density, as for the earlier CdS treatment, but results in the interdiffusion of the CdS and CdTe layers. The junction thereby moves into the CdTe, rather than remaining at the original metallurgical interface, an effect that is thought to improve the junction quality. The CdTe layer is usually thicker than required for optimal light absorption, to reduce shunting during back contact formation (Bonnet, 2001). The rear contact to the CdTe presents special challenges (Bonnet, 2001) and a variety of approaches to making this contact has been investigated. These generally are based on a two-layer approach, with the first layer being a heavily doped layer making good electrical contact vertically with the CdTe, while the second layer is metallic and provides good lateral conductivity. Most fabrication procedures include (Chu and Chu, 1995; Meyers and Birkmire, 1995): an etch or surface preparation step, which may produce a Te-rich surface layer; creation of the primary layer, either by deposition of a p+ layer of SnTe-Cu, HgTe or PbTe or by modification of the CdTe surface by supplying a p-type dopant such as Cu, Hg, Pb or Au; a subsequent heat treatment above 150°C; and application of the secondary contact by sputtering, vacuum evaporation or screen-printing. One CdTe deposition approach that has been the focus of a commercial sequence for fabricating large area modules is close-spaced-sublimation (CSS).


23

In the CSS process (Chu and Chu, 1995; Bonnet, 2001), a heated CdTe source dissociates into its Cd and Te constituents in gaseous form. These recombine on the cooler substrate surface to reform CdTe. In the commercial sequence, both CdS and CdTe are sequentially deposited onto a SnO2 coated glass substrate by a modified CSS technique. After a post-deposition heat treatment, electrical contact is made to the CdTe by deposition of a Ni/Al bilayer contact. Laser scribing is used at various stages during processing to pattern the SnO2 layer, the CdS/CdTe active layers and the rear contact layer to provide automatic series interconnection of cells within a module, identical to the approach previously described for a-Si (Fig. 1.12). Efficiencies up to 8% were demonstrated in the early 1990s for large area modules processed in this way. Manufactured product tends to be less than half the efficiency of the best laboratory devices. One contributing feature is that high-temperature borosilicate glass is used to produce the latter, which is considered too expensive for commercial use. Another major factor is the criticality of the CdS layer thickness. This can be much thinner in laboratory devices than possible in production, where high yields are required. Also, such metals as Cu must be avoided in commercial devices, due to their deleterious effect upon device durability (Bonnet, 2001). Environmental issues stemming from the toxicity of Cd and its compounds have slowed the introduction of this CdTe based technology (Schmela and Kruitmann, 2002; Meyers and Birkmire, 1995). Issues arise during manufacture, during deployment in the field and after disposal at ‘end-of-life’. Manufacturing hazards undoubtedly can be controlled. Hazards during deployment stem mainly from such incidents as fire which could cause the release of toxic vapours. Due to the potential for leaching of Cd into groundwater, special attention may have to be given to the final disposal of these modules. Some manufacturers believe the Cd-based materials can be recycled, although collection of product dispersed into widely different cultural and geographical regions would pose significant challenges. The lack in continuity of manufacturing efforts with this technology raises additional issues in relation to recycling.

1.5.2

Copper indium diselenide and its alloys

Copper indium diselenide (CIS) is a direct band gap semiconductor with a band gap of 1.04 eV at room temperature. A small cell of a reported efficiency of 12% was made by the evaporation of CdS onto a CuInSe2 single crystal in 1974. Soon after, the first thin-film cells were reported (Kazmerski et al., 1975). In the early 1980s, efficient thin-film cells were made with this material using co-evaporation of the Cu, In and Se elemental constituents (Mickelsen and Chen, 1982). By the late 1990s, thin-film cell efficiency approaching

24


19% (Contreras et al., 1999) had been demonstrated by incorporating CuGaSe2 into CuInSe2 to increase the band gap of the material. (CuInS2 is another wide band gap candidate that has also given good results.) The generic structure of such a thin-film alloy cell is shown in Fig. 1.18. A molybdenum back contact is deposited onto a glass substrate by sputtering or electron beam evaporation. After deposition of the main Cu(In, Ga)(S, Se)2 absorber layer by techniques to be described, a thin CdS or Cd1–xZnxS window layer is deposited by evaporation or, for best results, by solution growth. This is followed by the deposition of ZnO, by RF sputtering or by chemical vapour deposition. In the best devices, a two-step process is used whereby about 50 nm of lightly doped ZnO is deposited followed by 300 nm of Al-doped material, to reduce lateral resistance. Ni/Al contacts are applied to contact the ZnO. In the best laboratory devices, a MgF2 anti-reflection coating is added to the top of the ZnO (Contreras et al., 1999), although this would be detrimental in an encapsulated device.

TCO Window

ZnO CdS

Absorber

Cu (Ga, In) (Se, S)2

Contact Substrate

Mo Glass

1.18 Solar cell based on copper indium diselenide and related alloys.

Three techniques have been used to deposit the Cu(In, Ga)(S, Ge)2 absorber layer for cells which display over about 16% efficiency (Bloss et al., 1995): 1. co-evaporation of the elements onto a heated substrate; 2. selenisation of sputtered or evaporated Cu/In precursors in a H2Se or Se atmosphere; 3. diffusion of Cu and Se into (In,Ga)2Se3 precursors. The first process is used in pilot production by Würth Solar (Powalla and Dimmler, 2003), while the second is used by Shell Solar (Tarrant and Gay, 2003). Each has its strengths and weaknesses in a production setting. With the first, it would appear to be difficult to control stoichiometry over the


25

large areas required for commercial modules while, for the second, the massive expansion of the volume of the precursors during selenisation is said to be the source of adhesion problems that can become evident under accelerated environmental testing. For the highest performance devices, attempts are made to control composition and hence the alloy band gap across the thickness of the absorber layer. Not all aspects are well understood (Rau and Schock, 2001). With excess Cu concentration, large crystallites are grown, an effect believed to be promoted by the segregation of a copper chalcogenide phase to the surface of the films. This phase can be removed by subsequent chemical treatment (with KCN) or converted to a more desirable compound. The surface composition of grains also depends on the material composition. In indiumrich material, a stable CuIn3Se5 phase has been observed on film surfaces, having a larger band gap than bulk regions (Bloss et al., 1995). It has been proposed that layers of this type are located at the interface with the CdS window layer. Furthermore, Na out-diffusion from soda-lime glass substrates is found to have beneficial effects upon cell performance, by increasing grain size in the absorber layer. Such material complexities may be responsible for reported difficulties in attempts to commercialise this material in the late 1980s (Kazmerski, 1997; Zweibel, 1995). Module efficiencies above 12% have been reported in associated pilot production activities (Powalla and Dimmler, 2003; Tarrant and Gay, 2003). The completed devices use the glass layer as a substrate rather than a superstrate as for the a-Si and CdTe technologies described earlier. The same laser patterning steps can be used, however, although conducted in the reverse order from that shown in Fig. 1.12. The very high efficiencies already obtained with this material in small thin-film polycrystalline cells would seem to make it a strong candidate for a future, low-cost photovoltaic product. Small commercial modules (5 W and 10 W rating) were first reported to be in manufacture in mid-1998 (Siemens Solar, 1998), with 40–60 W rating modules now available. An issue of increasing importance for this technology is that of its moisture sensitivity (Oszan and Dunlop, 2002). Unencapsulated devices are reported to degrade under damp-heat exposure for all alloy compositions of the base (Malmström et al., 2002). Unlike CdTe, where the moisture sensitivity is due to problems at the rear contact, the moisture sensitivity of the CIS devices is usually attributed to metastabilities in the junction region (Oszan and Dunlop, 2002), although a variety of different degradation processes have been identified (Malmström et al., 2002). Hermetic sealing of the devices by the module encapsulation may be required to produce similar durability to the silicon wafer-based standard (Oszan and Dunlop, 2002; Erler et al., 2003). Field experience to date suggests that, although CIS modules appear to be showing good performance in some installations, unacceptably large module

26


degradation (above 10%) has been reported in the first six to twelve months in the field in a number of other installations (Elkelboom and Jansen, 2000; Carr and Pryor, 2001; LEEE, 2000; Lam et al., 2003). It is unclear whether the reported degradation arises from the moisture sensitivity of the technology or other causes, such as the adhesion issue for the selenisation process, or simply is due to a change in the dynamic response of the module (LEEE, 2000). Recent outdoor results suggest it may be more than the latter (Lam et al., 2003). Another issue likely to become increasingly important is the CdS junction layer in this device, deposited by chemical bath deposition in the most efficient CIS devices due to the substantial benefits described elsewhere (Rau and Schock, 2001). However, replacing this wet chemical process by a ‘dry’ process would be more consistent with streamlined module manufacturing (Rau and Schock, 2001). Also, although much less Cd is involved than with CdTe cells, concerns about material toxicity may increasingly shift to CIS if CdTe disappears as a commercial option. Promising materials to replace CdS are In(OH,S), Zn(OH,S) and ZnSe, although performance to date generally is lower and additional processing steps are required compared to CdS (Rau and Schock, 2001). Stability may also be even poorer (Bär et al., 2003). A final issue concerns material availability. Indium is a scarce element. All presently known resources would limit the generating capacity of CISbased photovoltaics to about the worldwide installed capacity of windgenerators in 2004 (Andersson, 2000). Although representing a useful amount of generation, about 1% of present electricity demand, it is not enough to significantly impact large issues such as greenhouse gas abatement. An indium replacement strategy for this technology is required if it is to have a longterm impact.

1.6

Dye-sensitised cells

Photoelectrochemical cells are based on junctions formed between liquids and semiconductors (Chapter 2). In such a cell, the liquid induces a barrier in the semiconductor much in the same way as does a metal. The liquid contains a species known as a redox couple with two charge states. The species changes from an oxidised to a reduced state if it accepts an electron, or undergoes the opposite process of oxidisation if it gives up an electron. Light is absorbed in the semiconductor, creating an electron-hole pair, as in a standard cell. In 1991, a new solar cell was reported that bears some similarity to a photoelectrochemical cell but that more closely mimics photosynthesis in its operation (O’Reagan and Grätzel, 1991). Rather than the sunlight being absorbed in a semiconductor, the cell absorbs light in dye molecules containing ruthenium ions. These dye molecules are coated onto nanocrystals of the


27

wide band gap semiconductor, TiO2 as indicated in Fig. 1.19. Light causes excitation of an electron in the dye (not the semiconductor in this case, since its band gap is quite wide) to an energy where it is injected into the conduction band of the adjacent n-type TiO2. The electron is transported through the TiO2 to the transparent conducting oxide at the front side of the cell, through the load, to the counter-electrode. Here, it reduces tri-iodide to iodide, which then diffuses to the photo-oxidised dye molecules to reduce these molecules back to their original state. Glass

Transparent conductor

Platinum Electrolyte

Die molecule on TiO2 nanocrystal Transparent conductor

Glass

Light

1.19 Nanocrystalline TiO2 dye sensitised solar cell.

The cells are very simple to prepare. The starting point is a sheet of glass coated by transparent conductive oxide (TCO). TiO 2 is prepared in nanocrystalline form, applied to the TCO coated glass and sintered. It is then soaked in a solution containing the dye, which results in the impregnation of the dye into the porous TiO2 resulting in a large interfacial area, maximising prospects for photoabsorption in the dye and electron injection into the TiO2. The counter electrode is again a sheet of TCO coated glass, coated with a thin layer of Pt to catalyse the reduction reaction of this electrode. After

28


sealing, the electrolyte is introduced through holes that are then resealed. The electrolyte consists of a solution of methyl-hexiglimidazolium iodide and t-tert-butylpyridine in acetonitrile. Quantitative modelling of the performance of these devices suggests that one of the key loss processes is the loss of photoexcited electrons in reducing the electrolyte in areas of TiO2 not covered by the dye. With the standard ruthenium-based dyes, which have peak absorption at 550 nm, energy conversion efficiency up to 8.5% has been confirmed for small area (0.25 cm2) cells. With a new ‘black’ dye, this efficiency has been increased to 11.0%. This is quite a remarkable result given both the relatively short history of development and the simplicity of the approach. However, there are some disadvantages which may be overcome with further work. One is the use of a liquid electrolyte and the difficulty of reliably sealing such an electrolyte. Solid electrolytes are being investigated but efficiencies with such solid electrolytes are presently very low. Another disadvantage is that such organic materials as the ruthenium dye or electrolyte components can be subject to degradation, particularly under the harsh and hot outdoor conditions to which cells are exposed. Another disadvantage is that some materials used in present cells such as acetonitrile are flammable and toxic. A further is that, although these cells are probably the simplest to fabricate in the laboratory, this does not necessarily transfer to the lowest possible costs in high volume production. For example, present nanocrystalline dye cells require processing of two glass sheets to form the module (coating of glass by TCO plus one other layer), rather than the single sheet used in other thin-film technologies, as well as the use of more complex cell interconnection approaches. Regardless of these challenges, dye sensitised cell technology is already finding commercial applications in such consumer products as digital watches and bathroom scales. Unique features such as the ability to produce ‘transparent’ modules based on infrared absorbing dyes also make the technology of considerable longer-term interest.

1.7

Organic and plastic cells

Although definitions vary, an organic molecule is one involving carbon, although some would qualify this by including hydrogen as a necessary constituent, while others would merely exclude carbonates, cyanides and cyanates (Internet Google search: ‘define: organic material’). Polymer is a high molecular weight substance usually organic, composed of long chains of repeating units, each relatively light and simple. Plastic is a generic term for high-molecular-weight polymers, capable of flowing under heat and pressure and hence capable of being moulded or extruded into various shapes, including films or filaments (Internet Google searches: ‘define: polymer’, and ‘define:


29

plastic’). The low processing temperatures and low costs associated with processing large areas of organic and plastic materials make them particularly attractive for photovoltaics. The impending ubiquitousness of organic and plastic light-emitting diodes (LEDs) has also encouraged a massive developmental effort, of direct relevance to photovoltaics. However, even though there are many ‘throw-away’ applications where a short operational life is not a problem, improving the durability of these devices is where the challenge lies. Without appropriate encapsulation, durability is in some cases measured in minutes rather than days, with devices sometimes measured ‘in-situ’ in the vacuum chamber in which they are prepared. Unlike other photovoltaic devices, where performance can be verified independently, this instability does not always make independent measurement possible for these devices (historically, performance has been overestimated by as much as a factor of 2 when measured ‘in-house’). Despite these possibly sobering caveats, it seems likely that low-cost, flexible organic or plastic solar cells can be developed that will be sufficiently efficient and rugged at least for short-term applications. The interesting electronic properties of organic molecules arise when they have a conjugated chemical structure. This means they are represented as having alternating double and single carbon-to-carbon bonds within their structure. The actual structure is more symmetrical with each carbon atom having three of its valence electrons in sp2 hybrid orbitals forming covalent bonds with its two carbon neighbours and with a hydrogen atom or other group. The fourth electron occupies a pz orbital, which can interact with one of its neighbours to form the second bond of the double bond representation of a conjugated molecule (Fig. 1.20). However, collectively, the pz orbitals can overlap to form delocalised π bonds which can extend the full length of the molecule.

1.20 Bonding in conjugated polymers.

Two recent reviews (Brabec et al., 2001) use device structure as a method of organising the nearly bewildering range of present activities in this field. The three types of device of present interest are shown in Fig. 1.21. The first, a single-layer device, also known as a metal-insulator-metal (MIM) device,

30

Materials for energy conversion devices Glass

ITO Device layer(s) Metal (a)

(b)

Interpenetrating polymers

(c)

1.21 Three different device structures used with organic/plastic and hybrid cells: (a) single-layer cells; (b) double-layer cell; (c) dispersed heterojunction cell.

uses an organic or plastic layer sandwiched between a transparent electrode (usually indium-tin-oxide, ITO, or a thin layer of metal) and a non-transparent metal. This is the same structure as used in organic/plastic LEDs but generally gives the lowest photovoltaic efficiencies. The second, a heterojunction device, as shown in Fig. 1.21(b), uses a junction between two different organic layers with different electronic properties, specifically different ionisation energies and electron affinities. One layer is an electron acceptor, while the other is a hole acceptor. The basic underlying operational concept is that light is absorbed in either layer of the device by creating excitons that diffuse to their interface. Here, the excitons dissociate by the transfer of the electron to the electron-accepting layer and the hole to the hole-accepting layer. The third type of device, known as a ‘bulk heterojunction’, is extremely interesting in that the electron and hole-accepting materials are mixed together with phase separation occurring in the final stages of film preparation. A relatively fine scale to the final two-phase material means that no exciton is generated too far from an interface. However, once separated, carriers may have to tunnel between appropriate domains to reach their respective contacts. A variation upon this theme is to prepare materials with a diffuse interface, retaining some of the previous advantages, while avoiding the latter problem. Incorporation of the football-shaped fullerene molecule, C60, into both heterojunction and bulk heterojunction devices has given recent exciting results. Incorporation of inorganics such as CIS in the form of quantum dots is also an interesting research path. A stable 5% organic solar cell and integration into disposal consumer electronics are challenging, but attainable, mid-term goals.

1.8

Conclusion

With ongoing improvements, photovoltaics is expected to provide an attractive approach to large-scale electricity generation for the twenty-first century.


31

The technology’s key strengths are its environmental friendliness, deployability, modularity and potential for low cost. Historically, the technology proved its merits in terms of reliability and durability in the space programmes of the 1960s. Since then, the costs for terrestrial use have reduced quite markedly, even though the mainstream product remains based on the same crystalline silicon wafer product of previous decades. Crystalline silicon wafer technology continues to withstand the challenge mounted by chalcogenide-based polycrystalline thin films, judging from recent manufacturing investments (Schmela, 2004). Notable recent trends with the wafer-based approach have been an increasing market share for high energy conversion efficiency monocrystalline silicon wafer sequences, increased market share for multicrystalline silicon wafers, the increased use of nitride-based anti-reflection coatings for the hydrogenation of the latter, and the emergence of reasonable production volumes of cells based on silicon ribbon and sheet. To reach its full potential as a non-polluting energy source, it seems likely that mainstream photovoltaic technology has to shift from the above waferbased approaches to less material-intensive, thin-film approaches. The reason this has not already happened is due not only to the well-known strengths of the established approach but also, in the author’s opinion, to fundamental difficulties with the thin-film options that have been the focus of most effort until recently. Some of the most promising thin-film technologies at present are also based on silicon, taking advantage of silicon in its amorphous, microcrystalline and polycrystalline phases. A recent development has been the emergence of amorphous/microcrystalline silicon hybrid tandem cells that are showing clear advantages over amorphous silicon/germanium hybrids (Green, 2003). Another recent development has been the emergence of a single-phase, singlejunction, thin-film polycrystalline silicon on glass technology that appears capable of challenging the silicon wafer-based incumbents not only in manufacturing cost, but even in the latter’s strengths of high manufacturing yields and product durability (Basore, 2002b). The chalcogenide-based compound semiconductor materials, cadmium telluride (CdTe) and copper indium diselenide (CIS), have produced excellent solar cells in the laboratory but appear problematic as a solution to the pressing need for a low-cost photovoltaic device. The toxicity of CdTe immediately raises issues about its acceptability in a market driven by environmental concerns. The difficulty in differentiating between the relative hazards posed by this material and other electricity generation options, such as nuclear, raises additional issues. CIS has the problem of limited resources of indium which severely limits its potential impact on the larger energy scene. The commercialisation of photovoltaic devices based on both of the above materials, however, has been hampered by more basic problems arising

32


from the complexity and reactivity of the associated materials system. This has made it difficult in the past to manufacture solar modules using these materials with high yield, even if efficiency expectations are severely relaxed from the best seen in the laboratory. The moisture sensitivity of unencapsulated devices compared to silicon is also likely to place more stringent requirements upon encapsulation for comparable field life. Finally, possibly targeting a different consumer-product orientated market, are the organic and plastic solar cell approaches. Excellent progress is being made on several fronts, with the main challenges being improved efficiency and durability.

1.9

Acknowledgements

The author acknowledges the award of an Australian Government Federation Fellowship and the support of the Centre of Excellence for Advanced Silicon Photovoltaics and Photonics by the Australian Research Council.

1.10

References

Andersson, B.A. (2000), ‘Materials availability for large-scale thin-film photovoltaics’, Progress in Photovoltaics, 8, 61–76. Ayra, R.R. and Carlson, D.E. (2002), ‘Amorphous silicon PV module manufacturing at BP Solar’, Progress in Photovoltaics, 10, 67–8. Bär, M., Rusu, M., Riess, J., Glatzel, Th., Sadewasser, S., Bohne, W., Strub, E., Muffler, H-J., Lindner, S., Röhrich, J., Niesen, T.P., Karg, F., Lux-Steiner, M.Ch., and Fischer, Ch.-H. (2003)‚ ‘Insights into the degradation mechanisms of CIGSSe devices based on different heterojunctions’, in Conf. Record, 3rd World Conference on Photovoltaic Energy Conversion, May, 335–9. Basore, P.A. (2002a), ‘Pilot production of thin-film crystalline silicon on glass modules’. In Conference Record, 29th IEEE Photovoltaic Specialists Conference, New Orleans, USA, 49–52. Basore, P.A. (2002b), ‘Pilot production of thin-film crystalline silicon on glass modules’, PV in Europe Conference, Rome, Italy, 236–9. Bloss, W.H., Pfisterer, F., Schubert, M. and Walter, T. (1995), ‘Thin-film solar cells’, Progress in Photovoltaics, 3, 3–24. Bonnet, D. (2001), ‘Cadmium telluride solar cells’, in Clean Electricity from Photovoltaics, M.D. Archer (ed.), Imperial College Press, London, 245–75. Brabec, C.J., Sariciftci, N.S. and Hummelen, J.C. (2001), ‘Plastic solar cells’, Adv. Funct. Mater., 11, 15–25. Bube, R.H., (1998), Photovoltaic Materials, Imperial College Press, London. Carr, A.J. and Pryor, T.L. (2001), ‘A comparison of the performance of different PV module types in high ambient temperatures’, Murdoch University, paper presented at ISES Conference, Adelaide, November. Chu, T.L. and Chu, S.S. (1995), ‘Thin film II-VI photovoltaics’, Solid-State Electronics, 38, 533–49. Contreras, M.A., Egaas, B., Ramanathan, K., Hiltner, J., Swartzlander, A., Hasoon, F. and


33

Noufi, R. (1999), ‘Progress towards 20% efficiency in Cu(InGa)Se2 polycrystalline thin-film solar cells’, Progress in Photovoltaics, 7, 311–16. Culik, J.S., Goncharovsky, I.S., Rand, J.A. and Barnett, A.M. (2002), ‘Progress in 15MW single-thread silicon-filmTM solar cell manufacturing systems’, Progress in Photovoltaics, 10, 119–28. De Lia, F., Castello, S. and Abernante, L. (2003), ‘Efficiency degradation of c-silicon photovoltaic modules after 22 year continuous field exposure’ in Conf. Record, 3rd World Conference on Photovoltaic Energy Conversion, Osaka, 2105–8. De Wolf., S., Choulat, P., Szlufcik, J., Périchaud, I., Martinuzzi, S., Haßler, C. and Krumbe, W. (2000), ‘Light-induced degradation of very low resistivity multi-crystalIine silicon solar cells’, in Conference Record, 28th IEEE Photovoltaic Specialists Conference, Anchorage, Anchorage, Alaska, 53–6. Eikelboom, J.A. and Jansen, M.J. (2000), ‘Characteristion of PV modules of new generations; results of tests and simulations’, Report ECN-C-00-067, 18. Erler, B., Degiampietro, S., Pertl, P., Plessing, A.K., Skringer, A. and Kessler, F. (2003), ‘Multilayer materials for the encapsulation of thin film modules’, in Conf. Record, 3rd World Conference on Photovoltaic Energy Conversion, Osaka, 1896–1898. Eyer, A., Schindler, R., Reis, I. and Schllinger, N. (1987), ‘Crystal structure and electrical properties of silicon sheets grown from powder (SSP-method)’, in Conference Record, 19th IEEE Photovoltaic Specialists Conference, Louisiana, May, 951–954. Fuhs, W. (2002), ‘Crystalline silicon thin-film technology for photovoltaics’, in PVNET Workshop Proceedings, RTD Strategies for PV, pp. 172–5, Ispra, Italy. Green, M.A. (1995), Silicon Solar Cells: Advanced Principles and Practice, (Bridge Printery, Sydney, 1995) (available from www.pv.unsw.edu.au). Green, M.A. (2003), ‘Thin-film photovoltaics’, Advances in Solar Energy. Jester, T.L. (2002), ‘Crystalline silicon manufacturing progress’, Progress in Photovoltaics, 10, 99–106. Jooss, W., Fisher, B., Fath, P., Bucher, E., Roberts, S. and Bruton, T.M. (2002), ‘Processing and characterisation of large area buried contact solar cells on multicrystalline silicon with a record efficiency of 17.5%’, in Conference Record, PV in Europe, Rome, 230– 5. Kalejs, J.P. (2003), ‘An overview of new developments in crystalline silicon ribbon material technology for solar cells’, in Conf. Record, 3rd World Conference on Photovoltaic Energy Conversion, Osaka, 903–8. Kazmerski, L. (1997), ‘Photovoltaics: a review of cell and module technologies’, Renew. and Sustain. Energy Reviews, 1, 71–170. Kazmerski, L.L., Ayyagari, M.S. and Sanborn, G.A., (1975), ‘CuInS2 thin films: preparation and properties’, J. Appl. Phys., 46, 4865. Lam, K-H., Close, J. and Durisch, W. (2003), ‘Degradation study on a commercial copper indium diselenide module’, in Conf. Record, 3rd World Conference on Photovoltaic Energy Conversion, Osaka, 479–82. Lechner, P. and Schade, H. (2002), ‘Photovoltaic thin-film technology based on hydrogenated amorphous silicon’, Progress in Photovoltiacs, 10, 85–98. LEEE (2000), LEEE News – Newsletter of the Laboratory of Energy, Ecology and Economy, 1, 1 (see also Cereghetti, N., Realini, A., Chianese, D. and Rezzonico, S. 2001, ‘Power and energy production of PV modules’, in Conference Record, 17th European Photovoltaic Solar Energy Conference, Munich, 710–13). LEEE, (2002), LEEE News – Newsletter of the Laboratory of Energy, Ecology and Economy, 3, 1.

34


Malmström, J., Wennerberg, J. and Stolt, L. (2002), ‘A study of the influence of the Ga content on the long-term stability of Cu(In, Ga)Se2 thin-film solar cells’, Symposium B, Thin-Film Chalcogenide Photovoltaic Materials, European Materials Research Society, Strasbourg, June. Mason, N.B., Bruton, T.M. and Balbuena, M.A. (2002), ‘Laser grooved buried grid silicon solar cells – from pilot line to 50 MWp in ten years’, in Conference Record, PV in Europe, Rome, 227–9. Meyers, P.V. and Birkmire, R.W. (1995), ‘The future of CdTe photovoltaics’, Progress in Photovoltaics, 3, 393–402. Mickelsen, R.A. and Chen, W.S. (1982), ‘Polycrystalline thin-film CuInSe2 solar cells’, in Conf. Record, 16th IEEE Photovoltaic Specialists Conference, San Diego, 781–5. O’Reagen, B. and Grätzel, M. (1991), Nature, 353, 373. Oszan, E. and Dunlop, E. (2002), ‘Workshop on stability and yield issues in module production’, Symposium B, Thin-Film Chalcogenide Photovoltaic Materials, European Materials Research Society, Strasbourg, June. Powalla, M. and Dimmler, B. (2003), ‘New developments in CIGS thin-film solar cell technology’, in Conf. Record, 3rd World Conference on Photovoltaic Energy Conversion, Osaka, May, 313–18. Rau, U. and Schock, H.W. (2001), ‘Cu (In, Ga)Se2 solar cells’, in Clean Electricity from Photovoltaics, M.D. Archer (ed.), Imperial College Press, London, 277–345. Realini, A., Bura, E., Cereghetti, N., Chianese, D. and Rezzonico, S. (2001), ‘Study of a 20 year old PV plant (MTBF Project)’, in Conference Record, 17th European Photovoltaic Solar Energy Conference, Munich, 447–50. Rein, S., Warta, W. and Glunz, S.W. (2000), ‘Investigation of carrier lifetime in P-type CZ-silicon: specific limitations and realistic prediction of cell performance’, in Conference Proceedings, 28th IEEE Photovoltaic Specialists Conference, Anchorage, 57–60. Sakata, H., Nakai, T., Baba, T., Taguchi, M., Tsuge, S., Uchihashi, K. and Kiyama, S. (2000), ‘20.7% highest efficiency large area (100.5 cm2) HITTM cell’, in Conference Proceedings, 28th IEEE Photovoltaic Specialists Conference, Anchorage, 7–12. Schmela, M. (2004), ‘This is a Sharp world Market survey on world cell production in 2003’, Photon International, March, 46–53. Schmela, M. and Kruitmann, A. (2002), ‘The Cheap and Clean?’, Photon International, April, 38. Schmidt, W., Woesten, B. and Kalejs, J.P. (2002), ‘Manufacturing technology for ribbon silicon (EFG) wafers and solar cells’, Progress in Photovoltaics, 10, 129–40. Siemens Solar (1998), Announcement at IEEE 2nd World Conference on Photovoltaic Energy Conversion, Vienna. Tarrant, D.E. and Gay, R.R. (2003), ‘CIS thin-film development and production status at Shell Solar, May 2003’, in Conf. Record, 3rd World Conference on Photovoltaic Energy Conversion, Osaka, 325–8. von Aichberger, S. (2003), ‘Deeper than blue: market survey on deposition systems for silicon nitride’, Photon International, March, 50–55. von Aichberger, S. (2004), ‘Market survey on solar modules’, Photon International, February, 46–55. Wallace, R.L., Hanoka, J.I., Rohatgi, A. and Crotty, G. (1997), ‘Thin silicon string ribbon’, Sol. En. Matls. Sol. Cells, 48, 179–86. Zweibel, K. (1995), ‘Thin films: past, present, future’, Progress in Photovoltaics, 3, 279– 94.

2 Materials for photoelectrochemical devices S U M K H A N, Duquesne University, USA

2.1

Introduction

There are mainly two photoelectrochemical devices for the generation of energy using sunlight for which searches for new materials are in progress. One of the most important energy producing devices involves the photoelectrochemical splitting of water into hydrogen and oxygen where the former is an abundant, renewable source of clean energy. This is because combustion of hydrogen in air regenerates water. This device uses at least one of the electrodes as a light absorbing semiconductor material which can absorb most of the photons in the solar spectrum. Another device is the photoelectrochemical solar cell in which redox reactions are carried out under illumination of sunlight to generate energy in the form of electricity. This device also uses at least one of the electrodes as a light absorbing semiconductor material. After the discovery by Fujishima and Honda1 in 1972, numerous studies2–23 were carried out for the generation of hydrogen by photoelectrochemical splitting of water using various semiconductor materials and their combinations. After the discovery of the photoelectric effect by Becquerel24 in 1839 many studies were devoted to the development of both efficient solid-state solar cells and more recently the photoelectrochemical wet solar cells.25–34 In this chapter, we focus on both the hydrogen-producing photoelectrochemical cell (HPPEC) and the electricity-producing photoelectrochemical cells (EPPEC). It is important to note that EPPEC must be self-driven such that no external potential is needed to drive the cell. In this case the photon generated voltage must be sufficient to drive redox reactions involving low standard redox potential (< 1.0 volt). The HPPEC are generally of two types, one is non-self-driven and the other is selfdriven. Hydrogen is generally produced by splitting water in a PEC involving one semiconductor photoelectrode and another dark metal electrode. In this case the HPPEC is not generally self-driven due to high voltage (≥ 1.23 volt) needed for the water splitting reaction. Semiconducting materials which 35

36


have band gap energies > 3.0 eV and appropriate conduction and valence band positions can split water only under UV light illumination as a selfdriven system. However, such self-driven HPPECs have negligible solar to hydrogen production efficiency (photoconversion efficiency) for the splitting of water. This is because sunlight has only 5% UV photons. Hence, the HPPEC involving one semiconductor and one metal electrode needs an external bias potential to split water into hydrogen and oxygen. The lower the band gap of the semiconductor electrode used in the HPPEC system, the higher will be the need for an external bias potential and vice versa. Note that the amount of external bias potentials also depends on the band positions of the semiconductor electrode relative to H2/H2O and H2O/O2 energy levels in solution. The HPPECs can be made self-driven for efficient water splitting if an appropriate combination of a p-type semiconductor (photocathodes) and an n-type semiconductor (photoandes) are used. Alternatively, it can be made self-driven for efficient water splitting if a semiconductor material having appropriate band gap energy between 1.7 and 2.2 eV is used in tandem with a solar cell as a photoelectrode and metal as a counter electrode. The photovoltage of the solar cell in tandem supplies the needed bias potential for the water splitting. The semiconductor materials used as photoelectrodes in a PEC for the efficient production of hydrogen, must (i) be highly stable, (ii) be abundant and inexpensive, (iii) have a conduction band minimum that is higher than the H2/H2O level and a valence band maximum lower than the H2O/O2 level, and (iv) can absorb most of the photons of solar spectrum to become efficient (≥10 %) for the photoelectrochemical splitting of water. Materials available at the present time rarely meet all four criteria for the efficient splitting of water to hydrogen in a PEC. Hence, in this chapter we shall concentrate on the present state of the development of materials, both theoretical and experimental methods used to identify the appropriate materials for efficient hydrogen and electricity-producing PECs with more emphasis on the former.

2.2

Photoelectrochemical kinetics

2.2.1

Theory of photocurrent at photoelectrode-solution interface for water splitting

Several theories 35–40 have been developed to determine the rate of photogenerated electron or hole transfer reactions at the semiconductor solution interface and these rates were expressed as photocurrent density and then compared with the experimentally observed results. In these theoretical expressions, various important properties of semiconductor electrode material, properties of the semiconductor-solution interface and the properties of the

Materials for photoelectrochemical devices

37

reacting ions in solution were included. In the following we provide the main aspect of one theory40 which included most of these properties and that will help to provide the guideline in choosing the appropriate materials for efficient hydrogen producing photoelectrochemical devices. In this theory, the rate of photon generated electron transfer reaction to produce hydrogen at a p-type semiconductor solution interface was expressed in the form of photocurrent density, jp as, jp =

∫

vm

e 0 k ct n s (ν ) d ν

vc

=

∫

vm

e 0 k ct (1 – Rv ) I v /( k ct + k sr + k br )

vc

–1 × [1– e – α ν W /(1 + α ν L–1 D )(1 + g1 L D ) – g 2 /( L D + g1 )] d ν 2.1

where e0 is the electronic charge, kct, ksr and kbr are the rate constant of the charge transfer at the interface, surface recombination and bulk recombination rate constants, respectively, ns(ν) is the number of photoelectrons generated by light of frequency, ν, that reached to the surface to undergo transition at the interface, Rν is the reflection coefficient of the light of frequency, ν, αν is the absorption coefficient of light at frequency, ν, Iν is the frequency dependent intensity of light, LD is the diffusion coefficient of photogenerated electrons (minority carriers) in the semiconductor material, W is the width of the space charge region, g1 = (πkBT/4e0Vbb)1/2W and g1e–ανW where kB is the Boltzmann constant and Vbb is the band bending in space charge region (SCR) in the semiconductor electrode (see Fig. 2.1). It is important to note that Eq. (2.1) takes into account the properties of semiconductor materials such as frequency dependent absorption and reflection coefficeient of light, width of the space charge region, W, diffusion coefficient of photogenerated minority carriers, LD, band bending potential in the semiconductor–solution interface, Vbb, and the surface and bulk recombination rate constants. The band gap energy, Eg of semiconductor materials is taken into account in the lower limit of the integral, νc = Eg/h where h is Plank’s constant. The properties of light are taken into account by the frequency dependent intensity of light and the maximum frequency, νm, available in the light source such as sunlight AM 1.5. The properties of the ions in solution are included in the charge transfer rate constant at the interface. Thus, equation (2.1) can be utilized to identify the semiconductor materials of appropiate properties for photoelectrochemial devices for the efficient splitting of water into hydrogen and oxygen gases. This is because high photocurrent density, jp at semiconductor materials of optimum properties at a fixed total intensity of light, I, will correspond to a high photoconversion efficiency for the water splitting to H2 and O2. Thus, this theory40 which takes into account the

38

Materials for energy conversion devices Vacuum level Interfacial barrier

1 SCR

2 FFR

Ece

∆Eo(Vac scale)

Xsc E0

Vo Eo VH

Eoee

Ey Eve

Ground state of acceptor, H5O+

Distributed states

Adsorbed water dipoles OHP p-type semiconductor

Specifically adsorbed ions

Solution

2.1 Schematic representation of a model of a p-type semiconductor 0 solution interface. In the semiconductor side, ECB, EVB, EF, EG, E SS represent the energy at the bottom of the conduction band, the energy at the top of the valence band, the Fermi level energy, the band gap energy, and the energy of the ground state of the surface state inside the band gap, respectively. SCR and FFR represent respectively the ‘space-charge region’ and the ‘field free region’. VS represents the potential drop inside the semiconductor. XSC is the electron affinity of the semiconductor. In the solution side of the interface there are absorbed ion water dipoles, the specifically absorbed ions and the acceptor ions (e.g., H3O) in the outer Helmholz plane (OHP). VH represents potential drop in the Helmholtz layer. E0 and ∆E0 represent the ground-state energy of the acceptor ion with respect to the bottom of the conduction band level at the surface and in terms of vacuum scale, respectively. The diagram contains the interfacial barrier at the double layer and the distribution of higher acceptor (e.g., H3O+) energy states and also the distribution of the surface-state energy states (from ref. 40).

details of light absorption in the semiconductor photoelectrodes, transport of photogenerated carriers inside the semiconductor and their recombination in the bulk and in the surface, and the properties of double layer at the semiconductor-solution interface, the classical and quantum mechanical charge transfer at the interface and the properties of reacting species in solution, and can be used to develop criteria for the choice of materials for the photoelectrochemical devices. The results of model calculations of photocurrent density versus potential at a photocathode are given in Fig. 2.2.

2.2.2

Photoelectrocatalysis

The effect of electrocatalyst on the photoelectrode surface is very dominant for the water splitting where either hydrogen evolution or oxygen evolution


39

3.5

3.0

Experimental

Photocurrent, ip (solar) mAcm–2

Theoretical 2.5

2.0

1.5

1.0

0.5

0.0 0.6

0.4

0.2

0.0 –0.2 –0.4 Potential, V(volt)/NHE

–0.6

–0.8

–1.0

2.2 Plot of photocurrent, ip (solar), as a function of electrode potential, V (volts)/NHE. The theoretical plot indicates the computed result using Eq. (2.1) and the experimental plot indicates the results obtained under xenon lamp illumination (from ref. 40).

or both occur at the semiconductor-solution interface. This is because in most situations the charge transfer at the semiconductor–solution interface becomes the rate determining step and the overall current is not always dominated by the transport of photogenerated minority carriers inside the semiconductor. Only at relatively high potentials at the interface does the transport inside the semiconductor become rate limiting when the charge transfer rate constant, kct becomes >> (ksr + kbr) and at relatively lower potential charge transfer at the semiconductor–solution interface become rate limiting when (ksr + kbr) >> kct (see Eq. 2.1). The charge transfer rate constant was expressed at the photocathode as:40

k ct = Se

∫

∞

T ( E ) f ( E , hν ) Da ( E 0 , E ) d E

2.2

Ec

where Se is the drift velocity of outgoing electrons in the surface region of the semiconductor, T(E) is the tunneling probability of photoelectrons across the interfacial barrier in the double layer, f (E, hν) is the Fermi distribution

40


of photoelectrons in the conduction band of the semiconductor which is influenced by the energy of the photoelectrons, hν, Ec the energy of the photoelectrons at the bottom of the conduction band of the semiconductor photocathode and influence of the electrocatalyst deposited on the semiconductor surface is included in the expression of density of electron acceptor states in the species in the semiconductor–solution interface, Da(E0, E). The ground state energy of hydrogen ion, E0 at the interface depends on the adsorption energy of hydrogen, ∆Hads and this energy depends on the electrocatalyst and highly influences the photocurrent density. It is thus essential to choose the proper electrocatalyst that has a suitable adsorption energy of hydrogen, ∆Hads. Several studies41–46 were carried out to verify the influence of electrocatalytic metal on semiconductor photoelectrode surface. For example, photocurrents due to hydrogen evolution in acidic solution on p-Si and p-InP (photocathodes) were found to be significantly affected by electrodeposition of metal islets. The onset potential shifts systematically in an increasingly positive direction for electrodeposited metal islets of enhanced electrocatalytic effect (see Fig. 2.3). However, the oxygen evolution reaction in the alkaline media on nTiO2 (photoanode) is also influenced to some extent by electrodeposition of metal islets46. Electrocatalytic RuO2 was also found to enhance the rate of oxygen evolution reaction when deposited on photoanode. This is because electrocatalysts can help the faster transfer of carriers from semiconductor to ions in solution and vice versa. Most importantly, these electrocatalysts can help to lower the hydrogen and oxygen evolution onset potential by + 0.1 V to 0.4 V and also enhance the stability of the photoelectrodes. According to an alternative model called the Schottky barrier model,44 the onset potential shift, ∆E for hydrogen evolution reaction should increase with decrease of work function of electrodeposited metal islets on semiconductor photoelectrodes. However, the experimental observation shows the opposite behavior (see Fig. 2.4). On the other hand, the electrocatalytic model was confirmed from the experimental observation of the increased shift in the onset potential, ∆E, for the hydrogen evolution reaction at the photocathode with increased exchange current density on the electrocatalyst metals (see Fig. 2.5). Also it increased with the increase of adsorption energy of hydrogen atoms on electrocatalytic metal surface (see Fig. 2.6).45,46 Thus, it is impotant to choose a suitable metal or metal oxide electrocatalyst material to enhance the performance of semiconducting photoelectrode materials. The metal islets should be chosen using the electrocatalytic model, i.e. at which the adsorption energy of H or OH– is neither too high or too low and at the top of volcano plots.47


41

–10

P-C p-In

P-Co p-InP

d

P-Ni

p-In

–4

p-In

p-In

P-Au

p-In

i/mA cm–2

P-Pt

–6

p-In

P-Pb

–8

–2

0 0.8

0.6

0.4

0.2 E/V NHE

0.0

–0.2

–0.4

2.3 Potentiodynamic experiments with p-InP-Me (Me = metal) photocathodes (1M H2SO4, 50-mW cm–2 Xe light) (from ref. 41).

2.2.3

Theory of matching photoanodes and photocathodes for an efficient self-driven HPPEC

It would be ideal to have two low band gap semiconductors to be used in an efficient photoelectrochemical cell because such materials could absorb most of the photons of solar spectrum. However, a low band gap and the absorption of more photons are not sufficient conditions for an efficient PEC. The combination of two photoelectrodes must supply the required photovoltage to overcome the necessary thermodynamic potential for the reaction (e.g., standard potential and overpotential of the reaction). Generally, low band gap materials such as Si, Ge and InP can produce low photovoltages compared to high band gap materials such as TiO2 and SrTiO3. Furthermore, in low band gap materials the recombination of photogenerated minority carriers are much higher compared to high band gap ones. Hence, the matching of

Materials for energy conversion devices Au 100

80

Cathodic shift, ∆E/mV

Pd

60

Ir

Ni Ru Rh

Pt

40 Co Ag 20 Cd Pb

Au +

0

+ Ru

4

Pd Ir–+++ +Pt Rh

4.25 4.5 4.75 5 Electron work function, Φe /eV

5.25

2.4 Dependence of the cathodic displacement ∆E on the work function Φe (from ref. 42).

0.4 Pt

Ni

∆E /V

42

0.2 Au 0.0

Si

Co

Pb –0.2

Cd –12

–10

–8 –6 log(i0/Acm–2)

–4

–2

2.5 The dependence of the E shift caused by the presence of metal islets on log i0 for hydrogen evolution in the dark on the corresponding massive metals (from ref. 43).


43

Au

100

80

Cathodic shift, ∆E /mV

Pd Ir 60 Ru Pt

Ni Rh

40 Co Ag

20

Cd Au +

0

125

167

+ Ir

Pd ++ Pt + Ru

+ Ru

209 251 D (M-OH)/kJ mol–1

Pb

293

335

2.6 Dependence of the cathodic displacement ∆E on the bond energy D(M-OH) (from ref. 42).

photocathodes and photoanodes is not straightforward. We outline here the main aspect of the theory48 which provides the guideline in choosing appropriate photocathodes and photoanodes for an efficient self-driven PEC. This theory utilizes the expression of photocurrent densities given earlier.40 According to this theory of semiconductor matching, the cell photocurrent density can be expressed as:48,49 j p (cell) = 2 j dp j dn {( j dp + j dn )

+ [( j dp – j dn ) 2 + 4 j dp j dn A exp ( e 0 Vcell /2 k B T ]1/2 } –1 2.3 where j dp and j dn are the diffusion-limited photocurrent density at p-type and n-type semiconductor photoelectrodes, respectively, which can be obtained from Eq. (2.1) when kct >> (ksr + kbr) and the cell potential is given by: Vcell = (Vp – Vn)

2.4

44


where Vp and Vn are the potential drop at the p-type and n-type semiconductorsolution interface respectively, and the constant: A = N ssn ∂ nsr N ssp ∂ srp ( Sth2 / Se Sh )exp{[( E 0a – E 0d – e 0 ( Vsp – Vsn )]/2 k B T ) ]}

2.5

where the N ssn and ∂ nsr are, respectively, the density of surface states and the surface recombination cross section on n-type semiconductor electrode (photoanode) surface, N ssp and ∂ srp are, respectively, the density of surface states and surface recombination cross section on p-type semiconductor electrode (photocathode) surface, S th is the thermal velocity of the photogenerated carriers, Se and Sh are the drift velocity of electrons and holes in the semiconductor electrode.48 E 0a and E 0d are, respectively, the ground state of acceptor and donor ions in solution; Vsp and Vsn are, respectively, the band bending potential at the p- and n-type semiconductor–solution interface and kB is the Boltzmann constant. One can use Eq. (2.3) to compute in iterative method to determine the matching of p-type and n-type semiconductors having properties for the efficient conversion of sunlight to hydrogen by photoelectrochemical splitting of water. Using Eq. (2.3) one can write: log [jp(cell)] = Constant – e0Vcell/4kBT

2.6

rev = 1.23 V and according to Eq. (2.6) the For the case of water splitting, Vcell max maximum cell current, j p (cell) is obtained when Vcell → 0. The maximum efficiency for water splitting when two photoelectrodes are used can be written as: max %ε eff = [ j pmax × 1.23]/[( A n + A p ) Wsolar ]

n

p

2.7

where A and A are the area of n-type and p-type semiconductor photoelectrodes, respectively, and Wsolar is the input solar power corresponding to AM 1.5 illumination (= 100 mW cm–2). Equation (2.6) was used by Kainthla et al.48 to compute the photocurrent potential characteristic HPPEC using a combination of p-type and n-type semiconductors as photoelectrodes. The values of various quantities in Eq. (2.5) used were Se = Sh = Sth = 107 cm s–1, the recombination cross-section from atomic dimensions as ∂ nsr = ∂ srp = 10 –6 cm2 and N ssn = N ssp = 10 15 cm–2 for the monolayer coverage of the surface. The other parameters used were given by Kainthla et al.48 Excellent agreements between the experimental and the theoretical computational results obtained using Eqs (2.3–2.6) was found.48 The computed maximum photoconversion efficiencies for the selfdriven PEC for various combination of photocathodes and photoanodes were also given.48


2.3

45

Photoelectrochemical wet solar cells for electricity generation

In conventional solid state solar cells, electron-hole pairs are generated by light absorption in a semiconductor, with charge separation and collection accomplished under the influence of electric fields within the semiconductor. In fact both n-type and p-type semiconductors of mainly same material, for example p/n-Si solar cell, are used to enhance the electric field inside the semiconductor. This enhancement of electric field reduces the recombination of photon generated electrons and holes, thus making the solar cell efficient. However, in the case of wet solar cells generally one semiconductor photoelectrode (e.g., n-type TiO2) in combination with a metal electrode is used in an electrolyte solution having equal concentrations of both oxidized and reduced forms of ions in a redox couple. The semiconductor generates electron-hole pairs under the illumination of light. In an n-type semiconductor, the photogenerated holes (the minority carriers) oxidize the reduced form of ions in solution and the photogenerated electrons go through the circuit via the load to the counter metal electrode and reduced oxidized form of the redox couple. Alternatively, the semiconductor electrode acts as an electron acceptor from a photoexcited sensitizer dye adsorbed on an electrode surface. The reduced form of the sensitizer dye molecules are oxidized by photoexcitation and transfer the electrons to the semiconductor. Under the influence of the electric field inside the semiconductor these electrons go through the circuit via the load to the counter metal electrode where the oxidized ions of the redox couples are reduced and these reduced ions are oxidized by reducing the photooxidized sensitizer dye (see Fig. 2.7).

2.3.1

Redox reactions

For the operation of wet electricity producing photoelectrochemical solar cells (EPPEC), it is critical to choose the proper redox couples and supporting electrolytes. Wet EPPEC are run in a buffer, acidic, basic solution or in room temperature molten salts. Different redox couples were used by different investigators in this field. The most common redox couples used were 25 I2/I–, Fe3+/2+, Ce4+/3+, V2+/3+, Fe(CN) 64–/3– , Se2–/Se3– and S 2– /S 3– n . Gratzel’s EPPEC became well known because it showed for the first time high stability as well as efficiency as high as 7.9%, which was later improved up to 10.4%. 26,27 The success of the Gratzel cell was due to use of tetrapropylammonium iodide and iodine-iodide redox couple in a non-aqueous solvent, acetonitrile. The stability of this system was enhanced tremendously to at least 108 cycles (which is equivalent to 20 years’ lifetime) when solvents like valaronitrile or γ-butyrolactone was used. The redox reactions involved in Gratzel cells are as follows:25

46

Materials for energy conversion devices –2.4

l2 (S+/S+) l1

V (vs. SCE)

–1.5 cb –0.7 hν1 +0.2

TiO2 (R/R–) S

+0.8 +2.5

Usable voltage

hν2

(S+/S) vb

Load

2.7 Schematic diagram of Ru(H2L’)2(NCS)2/TiO2 solar cell showing the TiO2 valence band (vb) and conduction band (cb), Ru-(H2L’)2(NCS)2 sensitizer (S), redox agent (R), and load. The sensitizer may be excited from its ground slate (S+/S) to an excited state (S+/S*); two are shown here, namely those represented by the first intense peak in the absorption spectrum (hv1, 11 state and the second (hv2, 12 state). Other arrows show the path of a current-producing electron around the cell (from ref. 34).

I – + I 2 → I 3–

2.8

The product I 3– is then reduced at the cathode as:

I 3– + 2e – → 3I –

2.9

where electrons transferred from photoexcited and reduced form of the dye sensitizer (S) moves through the field drop in the n-type semiconductor via the load to metal cathode (see Fig. 2.7) such that the sensitizer is oxidized as 2S + light → 2S* and 2S* → S+ + 2e–. The product I– is then oxidized in solution to I 3– as:

3I – → I 3– + 2e –

2.10

The conversion of oxidized form of sensitizer dye, S+ to its reduced form, S as: 2S+ + 2e– → 2S

2.11

and the iodide ions, 3I– to its original forms, I2 and I–, and complete the cycle. These reactions show the non-destructive nature involved in dye containing wet solar cells.


47

The semiconductor material used in the Gratzel cell was the extremely high surface area nanoparticles of titanium dioxide which absorb only 5% of solar photons between 250 to 414 nm. Thus, for this EPPEC to be efficient the monolayers of sensitizer ruthenium containing dyes such as Ru 2,2bipyridyle-4,4’-dicarboxylate and cis-dithiocyanotobis(2,2’-bipyridyl-4,4’dicarboxylate)-Ru(II) were used and these can absorb solar light up to 650 nm (1.9 eV) and 775 nm (1.6 eV), respectively. However, the main limitation in incorporating a dye in a solar cell assembly is its extreme pH sensitivity and a small shift in pH can drastically effect the charge transfer properties of the dye.50 Furthermore, photoageing of the wet solar cell surface can cause cracking, thus permitting air and water to contact the sensitive sensitizer on the semiconductor surface, modifying its properties to an order of magnitude less efficient.32,33 To overcome this difficulty in the liquid electrolyte, studies are in progress to use electrolyte in the form of a paste or solid.50–52

2.3.2

Materials for EPPEC

It should be noted that much of the work on wet EPPEC has focused on ntype II/IV or III/V semiconductors using the redox systems mentioned above. The most widely used semiconductor was the n-TiO2 which was sensitized by Ru dye due to its large band gap energy (3.0–3.2 eV) which hinders the absorption of most solar light. However, for the fabrication of EPPEC without the use of sensitizer low band gap (Eg) semiconductors such as, CdSe (Eg = 1.7 eV), CdTe (Eg = 1.3 eV), GaAs (Eg = 1.4 eV, GaP (Eg = 2.25 eV), CdS (Eg = 2.25), and Fe2O3 (Eg = 2.1 eV), etc., were used as photoanode and metal counter electrode as the cathode. A redox system should be chosen such that the energy position of the redox couple in the electrolyte solution falls within the band gap energy of the semiconductor photoelectrode (see Fig. 2.8). For the application of sensitizer on nanocrystalline or mesoscopic wide band gap semiconductors, such as n-TiO2, ZnO, SnO2 and Nb2O5 are used. These materials have extremely large surface areas (e.g., 1000 times) and consequently the adsorption of a monolayer of sensitizer molecules can absorb far more of the incident light. Efficient EPPECs can be developed without the use of sensitizer dye if multiple semiconductor layers are used in cascade such as TiO2/Fe2O3, InP/GaAs and InP/GaInAs, etc. The last two systems suffer instability, whereas the first one suffers inefficiency. To overcome such difficulties the multi-layer systems, such as a two-layer carbon modified (CM)-n-TiO2/n-InP/Pt or three-layer CM-n-TiO2/p-Fe2O3/ n-InP/ITO can be utilized. Instead of p-Fe2O3 other p-type semiconductors such as p-CuO can be used. Figure 2.9 shows the energy condition, photogeneration of electron hole pairs and their direction of movement for a three-layer cascade EPPEC. Note that in these systems only the stable CM-

48

Materials for energy conversion devices Vacuum 0

E NHE

–3.0

–1.5

–3.5

–1.0

–4.0

–0.5

SiC GaP

–4.5

–0.0

–5.0

0.5

–5.5

1.0

–6.0

1.5

–6.5

2.0

–7.0

2.5

–7.5

3.0

–8.0

3.5

GaAs CdSe ∆E = 1.4eV 2.25 eV

CdS ZnO

TiO2

Eu2+/3+ H2/H2O

WO3 Fe2O3

SnO2

3.2 1.7 eV eV 2.25 2.1 eV eV

3.2 eV 2.6 eV

3.0 eV

3.8 eV

[Fe(CN)6]3+/4Fe2+/Fe3+ H2O/O2 Ce4+/3+

2.8 Band positions of several semiconductors in contact with aqueous electrolyte at pH 1. The lower edge of the conduction band and upper edge of the valence band are presented along with the band gap in electron volts. The energy scale is indicated in electron volts using either the normal hydrogen electrode (NHE) or the vacuum level as a reference (from ref. 67).

n-TiO2 and the Pt metal or ITO (indium-doped tin oxide) will be exposed to electrolyte solution. The performance of the materials of EPPEC can be best determined by calculating the overall conversion efficiency using the expression:28,53 % Efficiency = ηglobal = [Isc × Voc × ff/Is] × 100

2.12

where Isc is the short circuit current density, Voc is the open circuit potential, Is intensity of solar light and the fill factor, ff can be expressed as: ff = ImVm/IscVoc

2.13

where Im and Vm are the measured current and voltage at a given point, respectively. The current, I – voltage, V (forward bias) can be obtained from the relation, I = IL[1 – exp(eV/kBT)] Where IL is the limiting current, when V = 0.

2.14


49

e– Load

n-TiO2

Indium Tin Oxide

e–

p-Fe2O3

e–

n-InP

R/R–

Indium Tin Oxide

Energy

e–

Light (hν) h+

e–

R–/R

h+

2.9 Schematic diagram of three-layer wet EPPEC system.

2.3.3

Advantages and limitations of EPPEC

The wet EPPECs have the advantage of low cost, since such systems do not require extremely purified single crystal semiconductor materials like presently available solid state solar cells. This type of wet solar cell will be less subject to solar roasting during outdoor use unlike their solid-state counterparts. However, such wet solar cells have the limitation of low efficiency compared to their solid-state ones. Also, such wet EPPECs will be heavier due to presence of liquid, paste or solid electrolytes. There is also the danger of electrolyte leaking out during long-term use.

2.4

Photoelectrochemical cell (PEC) for hydrogen production

It is critical at the begining of the twenty-first century to have a low-cost, renewable and clean source of energy to reduce the dependence on fastdepleting fossil fuels and to minimize environmental pollution and global warming. An (hydrogen) economy supported by the photoelectrochemical splitting of water to produce hydrogen would address this if methods were developed to achieve efficient, low-cost and stable sunlight-driven production of hydrogen. Hydrogen is an environmentally clean fuel because it produces pure water after combustion or when used in a fuel cell. Furthermore, a US Department of Energy, EIA report (March, 2001), noted that global oil reserves and production (with 3% growth) will start to decline sharply as early as

50


2030. This suggests an alarming urgency in the search for a low-cost, renewable and clean energy source in the form of hydrogen from water to sustain the world’s fast growing economy.

2.4.1

Principles

Here we provide the main operating principles of the photoelectrochemical cell for water splitting to hydrogen and oxygen in both non-self-driven and self-driven HPPEC. A non-self-driven HPPEC consists of a light-absorbing semiconductor (photoelectrode) and a metal counter electrode. Absorption of light having energy greater than or equal to the band gap energy of the semiconductor can generate an electron-hole pair per photon. These electrons and holes move in opposite directions and generate the photopotential. If this photopotential is higher than the potential needed to split water (1.23 volt) then HPPEC can run as a self-driven system. In general, a PEC composed of a single photoelectrode and a metal counter electrode cannot generate enough photopotential to split water as a self-driven system. Such non-self-driven HPPEC needs some external potential for water splitting. If the photoelectrode is a n-type semiconductor it acts as a photoanode where photogenerated holes react with water to oxidize it to oxygen. The photogenerated electrons move to metal counter electrode (cathode) and react with water to reduce it to hydrogen. Alternatively, if the photoelectrode is a p-type semiconductor it acts as a photocathode where water is reduced to hydrogen and in the metal counter electrode (anode) water is oxidized to oxygen.

2.4.2

HPPEC with a semiconductor and a metal electrode combination

The reactions for the photoelectrochemical production of hydrogen from water at the photocatalyst electrodes are the following: Hydrogen from water (4 electron-hole transfer reaction): 4H2O → 4H+ + 4OH–

in solution 2.15

n-type semiconductor (photoanode) + sunlight → 4h+ + 4e– at photoanode 2.16 4OH + 4h → O2 + 2H2O –

+

4H + 4e → 2H2 +

–

at photoanode 2.17 at metal cathode 2.18

Eqs (2.15)–(2.18) give the overall reaction as: 2H2O + (semiconductor photocatalyst) + sunlight → 2H2 + O2 2.19


51

In the overall reaction (2.19) the photocatalysts (p or n-type semiconductor) are not consumed even though they are not shown on the right-hand side. Note that most difficult reaction in the water-splitting process is the oxygen evolution reaction which occurs on the n-type semiconductor (photoanode) as shown in Eq. (2.17). Hence, studies on oxide n-type semiconductors are important because of their low-cost, high oxidative power, stability and nontoxicity. Alternatively, when a p-type semiconductor (photocathode) electrode is used instead of n-type semiconductor (photoanode) the above reactions (2.16)– (2.18) will be replaced by: p-type semiconductor (photocathode) + sunlight → 4h+ + 4e– at photocathode (2.20) 4H+ + 4e– → 2H2 4OH– → O2 + 2H2O + 4e–

at photocathode 2.21 at metal anode 2.22

Addition of Eq. (2.15) and (2.20)–(2.22) gives the overall reaction as shown in Eq (2.19). Thus a non-self-driven HPPEC involves one semiconductor photocathode or photoanode and a metal counter electrode. The p-type semiconductors act as photocathodes and the n-type ones act as photoanodes and consequently hydrogen is generated at p-type semiconductor photoelectrodes and oxygen is generated at the counter metal electrode. Alternatively, oxygen evolution occurs at the n-type semiconductor photoelectrodes and the hydrogen is generated at the counter metal electrode.

2.4.3

HPPEC with two semiconductors

The photoelectrochemical cell (PEC) for the splitting of water to hydrogen can be made self-driven (i.e., operates without the use of any external bias potential) if appropriate combinations of photoanodes and photocathodes are identified.48,49 A number of approaches were attempted to overcome the need for an externally applied potential for the photoelectrochemical splitting of water. Photoelectrochemical cells (PEC) with two semiconducting photoelectrodes were used.54 For example, an 8.2% efficient self-driven water-splitting system involving two semiconductors, such as single crystals p-InP and n-GaAs, was reported.54 In such a system, both p-type semiconductor (photocathode) and a n-type semiconductor (photoanode) are used and both electrodes are illuminated by light. For the hydrogen production by photoelectrochemical splitting of water, the reactions given in Eqs (2.16) and (2.17) occur at photoanode and Eqs (2.20) and (2.21) occur at photocathode. The overall reaction is given as in Eq. (2.19). The advantage of such a system is that the appropriate combination of two

52


semiconductor photoelectrodes can generate enough photopotential for the splitting of water without the use of any external bias potential and thus the system becomes fully solar energy driven. However, the disadvantage is that the efficiency becomes half if semiconductor photoelectrodes of equal area are used. This is because both semiconductor electrodes are exposed to light in such a PEC. To overcome this difficulty self-driven tandem HPPECs were developed4 as explained in the next section.

2.4.4

HPPEC in tandem with a solar cell

Khaselev and Turner4 demonstrated an important advance in photosplitting of water in a self-driven photoelectrochemical cell (PEC) with > 10% efficiency using a solar cell in tandem with a semiconductor photoelectrode and a metal wire as the counter electrode (see Fig. 2.10). Recently, Licht et al.6 also reported a further improvement in efficiency (≥ 15%) for solar splitting of water using a solar cell as a photoelectrode. In these devices the required bias potentials for the splitting of water were provided internally by its builtin photovoltaic (p/n) component. In this case one photoelectrode in tandem with a solar cell can split water into hydrogen if the solar cell in the back of the photoelectrode can supply the photopotential ≥ 1.23 volt required for the water splitting. Hence, the efficiency does not become half since a single photoelectrode in tandem with a solar cell is used in combination with a metal counter electrode in this design.

I

Ohmic contact

p -GaInP2

n -GaAs

Pt

p -GaAs

A

⇐hν

Interconnect

2.10 Schematic of the monolithic PEC/PV device (from ref. 4).


2.4.5

53

Materials for HPPEC

The material quality for a practical hydrogen producing photoelectrochemical cell (HPPEC) must be such that it is highly stable in a harsh atmosphere of either acidic or alkaline electrolyte, the band gap energy must be greater than 1.7 eV and less than 2.4 eV to be able to absorb most of the photons of solar spectrum, the mobility of photogenerated carriers must be high enough to minimize the recombination of carriers prior to their reactions with species (e.g., H+ and OH–) in solution and the absorption coefficient of light must be high enough to absorb most photons closer to surface. The surface of the semiconductor photoelectrodes must be less reflective, highly porous and nanocrystalline to have a high effective surface area. The materials must be stable, inexpensive and abundantly available. The conduction and valence band edges must be close to H2/H2O and O2/H2O standard state redox potentials, respectively, so that need for the external bias potential for water splitting is minimal. At present the semiconductor photoelectrode materials which satisfy these criteria are rare. Recently, a p-GaInP2/p/n-GaAs tandem photoelectrode in combination with a counter platinum metal electrode gave rise to a selfdriven photoconversion efficiency of 12.4% for water splitting to hydrogen.4 Note that the p/n-GaAs solar cell supplied the necessary photopotential and p-GaInP2 acted as a photocathode and Pt metal as counter electrode (dark anode). Licht et al.6 also reported further improvement in efficiency (16.3%) for solar splitting of water using a AlGaAs/Si solar cell. In these devices the required potential for the splitting of water is provided internally by its builtin photovoltaic (p/n) component. An externally biased 12% efficient photoelectrochemical water-splitting system involving a single crystal p-InP photoelectrode and a Pt counter electrode was also reported earlier.55 These photoelectrodes are not stable enough to develop a practical device. For this several attempts were made earlier to lower the band gap of n-TiO2 by transition metal doping.56–58 No noticeable changes in band gap energy of TiO2 to absorb in the visible region were observed by transition metal doping. Consequently, the photoconversion efficiency of n-TiO2 for water splitting was found to be less than optimal (≤ 1–2%). However, in a recent discovery, a chemically modified (CM)-n-TiO2 photocatalyst was reported2 to photosplit water to hydrogen and oxygen with a maximum photoconversion efficiency of 8.35% with minimal bias potential of 0.3 volt (Fig. 2.11). This novel photocatalyst boosted hopes of bringing the long sought goal of efficient (10–15%) solar production of hydrogen within reach, where 10% is the US Department of Energy’s benchmark for a commercially viable photocatalyst. The synthesis of this CM-n-TiO 2 photocatalyst was achieved by incorporation of carbon during oxidation of a Ti-metal sheet in a natural gas

54

Materials for energy conversion devices 9 CM-n-TiO2 (Flame)

Photoconversion Efficiency (%)

8 7 8.35%

6 5 4

1.08%

3 2

n-TiO2 (Oven)

1 0 0

0.2

0.4

0.6 0.8 Eapp (Volts)

1

1.2

1.4

2.11 Photoconversion efficiency as a function of applied Eapp at a chemically modified (CM)-n-TiO2 and the reference n-TiO2 (electric tube furnace- or oven-made) photoelectrodes (from ref. 2).

flame under a controlled amount of oxygen at an elevated temperature of 850oC.2 Carbon dioxide and steam (H2O), the combustion products of natural gas, helped to incorporate carbon, as well as enhancing the thickness of the titanium oxide film, respectively. The chemical modification of n-TiO2 by carbon to an average composition of n-TiO1.85C0.15 helped to lower its band gap from 3.0 eV to 2.32 eV (Fig. 2.12). This helped the CM-n-TiO2 to absorb in the visible region of sunlight up to wavelength 535 nm. This lowering of the band gap energy of CM-n-TiO2 photocatalyst helped to enhance the efficiency of solar production of hydrogen from water. At a molecular level, the interaction of the atomic orbital of carbon with the molecular orbital of TiO2 possibly helped to lower its band gap energy. Two band gaps, as observed in Fig. 2.12, indicate that if appropriate amounts of carbon could be incorporated uniformly in the photocatalyst, CM-n-TiO2, a single low band gap (≤ 2.32 eV) material could be synthesized. In other words, the portion of photocatalyst that has a band gap of 2.82 eV could be lowered to at least 2.32 eV by incorporation of an optimum amount of carbon. Consequently, it will be possible to achieve a higher photoconversion efficiency of 10–12%. The PEC involving CM-n-TiO2 and Pt electrodes can be easily coupled to fuel cell systems to supply its H2 fuel, and utilize a minimal amount of electrical power from it to overcome the fraction (e.g., 0.3 V) of the total thermodynamic barrier (1.23 V) to split water under


55

Absorbance (Arbitrary Units)

CM-n-TiO2 (Flame)

n-TiO2 (Oven)

320

370

420 470 Wavelength (nm)

520

570

2.12 The UV-visible spectra of CM-n-TiO2 (flame-made) and reference n-TiO2 (electric tube furnace- or oven-made). The flame-made sample shows threshold wavelengths of 535 nm (band gap of 2.32 eV) and 440 nm (band gap of 2.82 eV); the electric tube furnace- or ovenmade sample shows a threshold wave of 414 nm (band gap of 3.0 eV) (from ref. 2).

sunlight illumination. Alternatively, it is important to develop a self-driven system involving this CM-n-TiO2 in combination with a suitable p-type semiconductor (p-Si, p-GaInP2 and p-InP, etc.) that will be able to supply > 0.3 V to overcome the need of an externally applied potential. Recently, methods of chemical synthesis of visible light absorbing carbon modified (CM) n-TiO2 were also reported.59 Also, the synthesis of visible light absorbing carbon modified n-TiO2 was reported by oxidation of TiC.60 The visible light absorbing nitrogen doped n-TiO2 was also reported.61 These materials are important and viable candidates for efficient and stable photocatalyst for water splitting. Other low-cost stable materials for photoelectrochemical devices are the n- and p-type iron oxide semiconductors. These materials have the appropriate band gap energy of 2.0 to 2.2 eV but have high resistivity. However, spray pyrolytically synthesized nanocrystalline thin films of these are suitable materials to be used as the front layer in a tandem cell to protect the unstable but efficient back layer such as amorphous Si solar cell, p-Si, p-GaInP2, pInP and p-CdTe, etc. Some of these materials could be synthesized either by electrodeposition or by spray pyrolytic methods5 instead of expensive single crystal fabrication. A 4.5% efficient water splitting tandem cell was reported28

56


in which the tandem cell involved in the front, WO362 and on the back, dye sensitized TiO2. The spray pyrolytically synthesized thin films of carbon modified n-TiO22,59,60,63 and nitrogen doped n-TiO261 on a transparent conducting glass substrate could also be used as a stable front layer to protect unstable p-type semiconductors and develop a self-driven hydrogen producing water-splitting PEC.

2.5

Photoconversion efficiency of HPPEC

The efficiency for the EPPEC can be easily calculated using eq. (2.12) given above. The computation of photoconversion efficiency for the self-driven HPPEC is straightforward and can be calculated using the following expression:64,65 0 % efficiency = % E eff = [( j p × E rev × 100]/ I

2.23

where photocurrent density in mA/cm2 and the intensity of solar light, I, can 0 be used as mW/cm2 which is 100 mW/cm2 for AM 1.5 radiation and E rev is standard state potential which is 1.23 V for the water-splitting reaction. However, the expression of photoconversion efficiency for non-self-driven HPPEC is not straightforward. Various expressions were used in the literature. For example, a photoconversion efficiency expression in terms of saving of electrical applied potential due to photopotential generated at the semiconductor photoelectrode solution interface can be expressed as:55 %Eeff(saving) = [jp × ∆E × 100]/I

2.24

∆E = Eonset (dark) – Eonset (light)

2.25

where

where Eonset (dark) is the onset potential for water-splitting reaction, when the electrocatalytic metal electrode such as Pt is used in the dark and Eonset (light) is the onset potential for the water-splitting reaction under illumination of light. The limitation of Eq. (2.24) is that it is dependent on the type of electrocatalyst metal used.66 0 Considering j p E rev as the maximum power output, Bockris and Murphy66 proposed an equation to calculate practical photoconversion efficiency as, 0 % E eff (practical) = [ j p E rev – j p E app )100]/( I + j p E app )

2.26

where is Eapp the applied (bias) potential at working photoelectrode with respect to a reference electrode in a three-electrode PEC or in a two-electrode PEC (where the reference electrode is grounded to the counter electrode), Eapp is the bias cell potential to counter electrode. This equation is erroneous in the sense that the photocurrent is not observed under bias potential but only under illumination. The bias potential is needed to overcome the


57

thermodynamic barrier and this contribution is subtracted in the numerator 0 and it should not be used in the denominator. from maximum output, j p E rev 66 The authors did not mention the way the bias electrode potential should be determined. If it is determined with respect to a reference electrode in a three-electrode PEC, after adjusting for pH = 0 (using ~ 60 mV/pH), it is not possible to determine the pH adjusted applied potential, Eapp to use in Eq. (2.26) when a p-type semiconductor photocathode is used as the working electrode and a metal is used as the counter electrode. This means that Eq. (2.26) is not applicable for photocathode/metal (e.g., Pt) PECs. Taking into account the above limitations, so that the efficiency equation is applicable for PEC involving either photocathode or photoanode, the following photoconversion efficiency equation for non-self-driven HPPEC was reported2,5,64 as: 0 % E eff = j p [ E rev – | E app |]100]/ I

2.27

where | Eapp | is the absolute value of the applied potential and can be expressed as: Eapp = Emeas – Eaoc

2.28

where Emeas is the potential of working photoelectrode (with respect to a reference electrode, e.g., saturated calomel electrode, SCE) at which the photocurrent density, jp was measured and Eaoc is the potential of the working photoelectrode (with respect to same reference electrode, e.g., SCE) at open circuit conditions under same illumination intensity that was used for the measurement of jp. Note that Eaoc is not the open circuit cell voltage (cell voltage) as is commonly understood. It is rather the electrode potential of the photoelectrode under open circuit condition under illumination of light. However, for two-electrode systems (when the reference electrode probe in the potentiostat is connected to the counter electrode), Eaoc will represent the commonly understood open circuit cell voltage.

2.6

Some criteria of suitable semiconductor photoelectrodes for efficient water-splitting

For efficient photoelectrochemical water splitting the ideal semiconductor photoelectrodes should satisfy the following conditions. 1. For water-splitting reaction 1.23 V is needed and only 70% of the band gap energy can contribute to photopotential needed to split water. Hence the band gap energy of the semiconductor photoelectrode should ideally be at least 1.76 eV. However, if overpotential (the extra potential over the 1.23 V) that is needed to split water is between 0.3 V to 0.6 V, then the band gap would be in the range of 2.06 to 2.36 eV. On the other hand,

58


for reasonable solar efficiencies, most of the visible light spectrum must be absorbed, which puts an upper limit on the band gap of less than about 2.3 eV where photons in sunlight are plentiful. This is because the band gap must be less than or equal to the energy of sunlight in the visible region. There are more light photons in the visible region of sunlight. Hence, if the band gap energy of a semiconductor is about 2.0 eV most of the visible light photons in the sunlight will be absorbed by it. 2. Suitable band edge position of semiconductors7,67 so that the conduction band (CB) minimum is above H2/H2O level (– 4.5 eV from the vacuum (zero energy) level) and the valence band (VB) maximum is below H2O/ O2 level (– 5.73 eV from the vacuum level) (Fig. 2.8). 3. Absorption coefficient of light (both in the UV and visible regions) and the mobility of electrons and holes must be very high in the semiconductor materials.40,48,49 4. Semiconductor materials must be highly efficient (>10%), stable, inexpensive and plentiful. However, such a single junction semiconductor has not been discovered yet to photosplit water efficiently without the use of an external bias potential or an internal bias potential from a tandem solar cell. For example, the most stable semiconductors are oxides, but their band gap energies are either too large to absorb the visible part of the solar spectrum or their band edges do not match well with H2/H2O and H2O/O2 levels.

2.7

Conclusions

Although major research in the field of photoelectrochemical splitting of water started in 1972 following Fujishima and Honda,1 we are only recently observing some major breakthoroughs2,4,6 after more than 30 years of research. It is highly conceivable that with vigorous future research activity involving nanocrystalline stable materials such as CM-n-TiO2 and n-Fe2O3 or p-Fe2O3 thin films as protective front layer, coupled with some tandem amorphous silicon solar cell systems in the back, will soon be proved to be inexpensive and highly stable at levels much above 10% efficient sunlight driven watersplitting system for the generation of clean hydrogen fuel.

2.8

References

1. Fujishima, A. and Honda, K., ‘Electrochemical Photolysis of Water at n-TiO2 Semiconductor Electrode’, Nature, 238 (1972) 37–40. 2. Khan, S.U.M., Al-Shahry, M. and Ingler W.B. Jr, ‘Efficient Photochemical Water Splitting by a Chemically Modified n-TiO2’, Science, 297 (2002) 2243–5. 3. Bolton, R., ‘Solar Production of Hydrogen: A Review’, Solar Energy, 57 (1996) 37– 50.


59

4. Khaselev, O. and Turner, J.R., ‘A Monolithic Photovoltaic-Photoelectrochemical Device for Hydrogen Production’, Science, 280 (1998) 425–7. 5. Khan, S.U.M. and Akikusa, J., ‘Photoelectrochemical Splitting of Water at n-Fe2O3 Thin Film Electrodes’, J. Phys. Chem. B., 103 (1999) 7184 –9. 6. Licht, S., Wang, B., Mukerji, S., Soga, T., Umeno, M. and Tributsch, H., ‘Efficient Solar Water Splitting, Exemplified by RuO2 Catalyzed AlGaAs/Si Photocatalysis’, J. Phys. Chem., 104 (2000) 8920–5. 7. Bak, T., Nowotny, J., Rekas, M. and Sorrel, C.C., ‘Photo-electrochemical Hydrogen Generation from Water using Solar Energy. Material-Related Aspects’, Intl. J. Hydrogen energy, 27 (2002) 991–1022. 8. Getoff, N., ‘Photoelectrochemical and Photocatalytic Methods of Hydrogen Production: A Short Review,’ Int. J. Hydrogen Energy, 15 (1990) 407–17. 9. Hironori, A., ‘Hydrogen Production from Water and Sunlight (Visible Light). Development of Visible Light Responding Photocatalysts for Water-Splitting’, Optronics, 245 (2002) 169–73. 10. Ye, J., Zou, Z., Arakawa, H., Oshikiri, M., Shimoda, M., Matsushita, A. and Shishido, T., ‘Correlation of Crystal and Electronic Structures With Phortophysical Properties of Water-Splitting Photocatalysts INMO4(M = V5+, Nb5+, Ta5+)’, J. Photochem. Photobiol., A:Chemistry, 148 (2002) 79–83. 11. Akihiko, K., ‘Development of Photocatalyst Materials for Water Splitting with the Aim at Photon Energy Conversion’, J. Cer. Soc., Japan, 109 (2001) S81–88. 12. Zou, Z., Ye, J., Sayama, K. and Arakawa, H., ‘Direct Splitting of Water Under Visible Light Irradiation with an Oxide Semiconductor Photocatalyst’, Nature, 414 (2001) 625–7. 13. Gao, X., Kocha, S., Frank, A.J. and Turner, J.A., ‘Photoelectrochemical decomposition of water using modified monolithic tandem cell’, Int. J. Hydrogen Energy, 24 (1999) 319–25. 14. Srivastava, O.N., Karn, R.K. and Misra, M., ‘Semiconductor-septum photoelectrochemical solar cell for hydrogen production’, Int. J. of Hydrogen Energy, 25 (2000) 495–503. 15. Babu, K.S.C., Pandey, R.N. and Srivastava, O.N., ‘Photoelectrochemical Semiconductor Septum (CdSe/Ti and TiO2/Ti) Solar Cells in Relation to Hydrogen Production’, Int. J. Hydrogen Energy, 20 (1995) 771–5. 16. Majumder, S.A. and Khan, S.U.M., ‘Photoelectrolysis of Water at Bare and Electrocatalyst Covered Iron Oxide Electrodes’, Intl. J. Hydrogen Energy, 19 (1994) 881–7. 17. Misra, M., Pandey, R.N. and Srivastava, O.N., ‘Solar Hydrogen Production Employing n-TiO2/TiSC-SEP, Photoelectrochemical Solar Cell’, Int. J. Hydrogen Energy, 22 (1997) 501–10. 18. Babenko, S.D., Balakai, A.A., Lavrushko, A.G., Ponomarev, E.A. and Simbirtseva, G.V., ‘Kinetics of the photoelectrochemical evolution of hydrogen at p-type Si’, J. Electroanal. Chem., 382 (1995) 175–7. 19. Arriaga, L.G., Fernandez, A.M. and Solorza, O., ‘Preparation and Characterization of (Zn, Cd)S Photoelectrodes for Hydrogen Production’, Int. J. Hydrogen Energy, 23 (1998) 995–8. 20. Bard, A.J. and Fox, M.A., ‘Artificial Photosynthesis: Solar Splitting of Water to Hydrogen and Oxygen’, Acc. Chem. Res., 28 (1995) 141–5. 21. Gondal, M.A., Hameed, A., Yamani, Z.H. and Suwaiyan, A., ‘Laser Induced Photocatalytic Oxidation/Splitting of Water over α-Fe2O3, WO3, TiO2 and NiO Catalysts: Activity Comparison’, Chem. Phys. Lett., 385 (2004) 111–15.

60


22. Abe, R., Sayama, K. and Arakawa, H., ‘Significant Effect of Iodide Addition on Water Splitting into H2 and O2 over Pt-loaded TiO2 Photocatalyst: Suppression of Backward Reaction’, Chem. Phys. Lett., 371 (2003) 360–4. 23. Yamakata, A., Ishibashi, T. and Onishi, H., ‘Kinetics of the photocatalytic water splitting reaction on TiO2 and Pt/TiO2 studied by time resolved infrared absorption spectroscopy’, J. Mol. Catal. A: Chemical, 199 (2003) 85–94. 24. Bequerel, E., ‘Recherches sur les effects de la radiation chemieque de la lumiereslaire, au moyen des courants electriques’, C.R. Acad. Sci., 9 (1839) 145–9. 25. O’Regan, B. and Gratzel, M., ‘A Low-Cost High Efficiency Solar Cells’, Nature 355 (1991) 737–40. 26. Hagfeldt A. and Gratzel, M., ‘Molecular Photovoltaics’, Acc. Chem. Res. 33 (2003) 269–77. 27. Gratzel, M., ‘Sol-Gel Processed TiO2 Films for Photovoltaic Applications’, J. SolGel Sci Technol., 22 (2001) 7–13. 28. Gratzel, M., ‘Photoelectrochemical Cells’, Nature, 414 (2001) 338–44. 29. Ishida, T. Kouno, H. Kbayashi, H. and Nakato, Y., ‘Dependences of Photovoltages of Spray-Deposited Indium Tin Oxide/Silicon Oxide/Silicon Junction Solar Cells on Spray Solvents’, J. Electrochem. Soc., 141 (1994) 1357–61. 30. Licht, S. Clark., ‘Electrolyte Modified Photoelectrochemical Cells’, Sol. Ener. Mat. Sol. Cells, 38 (1995) 305–19. 31. Goetzberger, A., Luther, J. and Willeke, G., ‘Solar Cells: Past, Present and Future’, Sol. Ener. Mat. Sol. Cells, 74 (2002) 1–11. 32. Kruger, J., Back, U. and Gratzel, M. ‘High Efficiency Solid State Photovoltaic Device due to Inhibition of Interface Charge Recombination’, Appl. Phys. Lett. 79 (2001) 2085–7. 33. Ragay, F.W., Marti, A., Araujo, G.L. and Wolter, J.H., ‘Experimental Anslysis of Efficiency of Hetero Structures GaAs-AlGaAs Solar cells’, Sol. Ener. Mater. and Sol. Cells, 40 (1996) 5–21. 34. Monat, J.E., Rodriguez, J.H. and McCusker, J.K., ‘Ground and Excited-State Electronic Structures of the Solar Cell Sensitizer Bis(4,4’-dicarboxylato-2, 2’ – bipyridine) bis(isocyanato) Ru(II)’, J. Phys. Chem. A, 106 (2002) 7399–406. 35. Graetner, W.W., ‘Depletion-Layer Photoeffects in Semiconductors’, Phys. Rev., 116 (1959) 84–7. 36. Ginley, S. and Butler, M.A., ‘The Photoelectrolysis of Water using Iron Titanate Anodes’, J. Appl. Phys., 48 (1977) 2019–21. 37. Reiss, H., ‘Photocharacteristics for Electrolyte-semiconductor Junctions’, J. Electrochem. Soc., 125 (1978) 937–49. 38. Wilson, R.H., ‘A Model for the Current-voltage Curve of Photoexcited Semiconductor Electrodes’, J. Appl. Phys., 48 (1977) 4292–7. 39. Alberty, W.J. and Bartlet, P.N., ‘The Recombinaion of Photogenerated Minority Carriers in the Depletion Layer of Semiconductors’, J. Electrochem. Soc., 130 (1983) 1699–705. 40. Khan, S.U.M. and Bockris, J.O’M., ‘A Model for Electron Transfer at the Illuminated p-Type Semiconductor-Solution Interface’, J. Phys. Chem., 88 (1984) 2504–615. 41. Szklarczyk, M. and Bockris, J.O’M., ‘Photoelectrochemical evolution of hydrogen on p-indium phosphide’, J. Phys. Chem., 88 (1984) 5241. 42. Bockris, J.O’M., Szklerczyk, M., Contractor, A.Q. and Khan, S.U.M., ‘On Photoelectrocatalysis of Hydrogen and Oxygen Evolution’, Intl. J. Hydrogen Energy, 9 (1984) 741–6.


61

43. Szklarczyk, M., Contractor, A.Q., Bockris, J.O’M., Young, V.Y., Barnard, L.A. and Sparrow, G., ‘Spectroscopic and Microscopic Studies of Photoelectrode Surfaces’, Sol. Energy Mater., 11 (1984) 105–21. 44. Tsubomura, H., Tonomura, S. and Nakato, Y., ‘The Catalytic Effect of Electrodeposited Metals on the Photo-reduction of Water at p-Type Semiconductors’, Ber. Bunsenges. Phys. Chem., 80 (1976) 1289–93. 45. Kautek, W., Grobrecht, J. and Gerischer, H., ‘The Applicability of Semiconducting Layered Materials for Electrochemical Solar Energy Conversion’, Ber. Busenges. Phys. Chem., 84 (1980) 1034–40. 46. Szklarczyk, M. and Bockris, J.O’M., ‘Photoelectrocatalysis and Electrocatalysis on p-Silicon’, J. Phys. Chem., 88 (1984) 1808–15. 47. Appleby, A.J., in: Comprehensive Treatise of Electrochemistry, Vol. 7, Bockris, J.O’M., Yeager, E. Khan, S.U.M. and White, R.E. (eds), Plenum Press, New York, 1983, pp. 173–239. 48. Kainthla, R.C., Khan, S.U.M. and Bockris, J.O’M., ‘A Theory of Electrode Matching in Photoelectrochemical Cells’, Intl. J. Hydrogen Energy, 12 (1987) 381–92. 49. Bockris, J.O’M. and Khan, S.U.M., Surface Electrochemistry: A Molecular Level Approach, Plenum Press, New York, 1993. Chapter 5. 50. Gratzel, M., ‘Highly Efficient Nanocrystalline Photovoltaic Devices – Charge Transfer Sensitizers Based on Ruthenium and Osmium Achieve Outstanding Performance’, Platinum Met. Rev., 38 (1994) 151–9. 51. Ruile, S., Kohle, O., Pettersson H. and Gratzel, M., ‘Unusual Properties of Ruthenium(II) Diphenylcyanamide Complexes: Chemistry and Application as Sensitizers of Nanocrystalline TiO2’, New J. Chem., 22 (1998) 25–31. 52. Papageorgiou, N., Liska, P., Kay, A. and Gratzel, M., ‘Mediator Transport in Multilayer Nanocrystalline Photoelectrochemical Cell Configurations’, J. Electrochem. Soc., 146 (1999) 898–907. 53. Sze, S.M., Physics of Semiconductor Devices, John Wiley & Sons, New York, 1981 p. 807. 54. Kainthla, R.C., Zelenay, B. and Bockris, J.O’M., ‘Significant Efficiency Increase in Self-Driven Photoelectrochemical Cell for Water Photoelectrolysis’, J. Electrochemical Society, 134 (1987) 841–5. 55. Heller, A. and Miller, B., ‘Some Recent Progress in Semiconductor-liquid Junction Solar Cells’, Electrochim. Acta, 25 (1980) 29. 56. Gosh, A.K. and Maruska, H.P., ‘Photoelectrolysis of Water in Sunlight with Sensitized Semiconductor Electrodes’, J. Electrochem. Soc., 124 (1977) 1516–22. 57. Choi, W., Termin, A. and Hoffmann, M.R., ‘The Role of Metal Ion Dopants in Quantum-Sized TiO2: Correlation between Photoreactivity and Charge Carrier Recombination Dynamics’, J. Phys. Chem., 98 (1994) 13669–79. 58. Anpo, M., ‘Photocatalysis on Titanium Oxide Catalysts. Approaches in Achieving Highly Efficient Reactions and Realizing the Use of Visible Light’, Catal. Surv. Japan, 1 (1997) 169–79. 59. Sahthivel, S. and Kisch, H., ‘Daylight Photocatalysis by Carbon Modified Titanium dioxide’, Angew Chem. Int. Ed., 42 (2003) 4908–11. 60. Irie, H., Watanabi, Y. and Hashimoto, K., ‘Carbon-Doped Anatase TiO2 as VisibleLight Sensitive Photocatalyst’, Chem. Letters, 32 (2003) 772–3. 61. Asahi, R., Morikawa, T., Ohwaki, T., Aoki, K. and Taga, Y., ‘Visible-Light Photocatalysis in Nitrogen-Doped Titanium Oxides’, Science, 293 (2001) 269–71.

62


62. Santato, C., Ulman, M. and Augustanski, J., ‘Photoelectrochemical Properties of Nanostructured Tunsten Trioxide Films’, J. Phys. Chem. B, 105 (2001) 936–40. 63. Akikusa, J. and Khan, S.U.M., ‘Stability and Photoresponse of Nanocrystalline nTiO2 and n-TiO2/Mn2O3 Thin Film Electrodes During Water-Splitting Reactions’, J. Electrochem. Soc., 145 (1998) 89–93. 64. Khan, S.U.M. and Akikusa, J., ‘Photoresponse of p-SiC Towards Water-splitting Reaction’, Int. J. Hydrogen Energy, 27 (2002) 863–70. 65. Parkinson, B., ‘On the Efficiency and Stability of Photoelectrochemical Devices’, Acc. Chem. Res., 17 (1984) 431–7. 66. Bockris, J.O’M. and Murphy, O.J., ‘Photoconversion Efficiencies for Photo-assisted Electrolysis of Water’, Appl. Phys. Commun., 2 (1983) 203. 67. Memming, R., ‘Processes at Semiconductor Electrodes’, in: Comprehensive Treatise of Electrochemistry, Vol. 7, Bockris, J.O’M., Yeager, E., Khan, S.U.M. and White, R.E. (eds), Plenum Press, New York, 1983, pp. 529–92.

3 Photosensitive materials H T R I B U T S C H, Hahn-Meitner-Institut, Germany

3.1

Introduction

This chapter overviews photosensitive materials that absorb light and, in so doing, attain properties that are distinctively different from those of nonexcited materials. By absorbing energy from light, these materials temporarily change their solid-state, molecular and/or interfacial properties. In this way, they may become active in terms of photoconductivity, photoluminescence, photon energy conversion, or photocatalysis. By temporarily storing and converting solar radiation, they may act as solar cells, photodiodes, photodetectors, or photocatalysts. If the interface is sufficiently photoactive, these materials may also react with their molecular environments and generate chemical energy in the form of energy-rich compounds. Photosensitive materials are expected not only to absorb light in the desired or required energy spectrum but they often are also expected to possess interfacial properties that allow the separation of electronic charge carriers. This occurs through either inbuilt electrical fields or kinetically determined mechanisms. Finally, photosensitive materials are sometimes expected to provide electronic or even ionic charge carriers suitable for interaction with chemical reactants. The applications of photoactive materials range from single-crystal electronically tailored devices, such as silicon solar cells, to photographic emulsions and photocatalytically self-cleaning surface layers, which presently are available in the form of TiO2-covered architectural facades and technical interfaces. In all of these applications, the photon energy is utilised in order to provide some well-defined properties indicative of photosensitive materials. While the scientific understanding of photosensitive materials gradually has grown along with the progress in well-defined macro-scale crystalline materials, viz., silicon solar cells, technology is pushing toward less defined nano-structured materials, as in photography and, more recently, new types of solar cells, such as nano-structured dye-sensitised solar cells and composite polymer-fullerene solar cells. While the level of knowledge of these complicated systems is not very advanced at present, they are becoming increasingly 63

64


important because they consist of abundant and easy-to-handle materials – a pre-condition of cost efficiency. Further, the properties of other nano-structured photosensitive materials are based on those that are exploited by natural biological systems. For example, with photosynthetic membranes, charge separation and electronor ion-conducting molecular particles are dispersed in a lipid membrane, which provides efficient functioning of the system. Similarly, in composite polymer-fullerene solar cells, the photoactive polymer and the fullerene are simply mixed together, which yields a reasonably effective performance in a solar cell. The present work discusses new scientific frontiers and industrial developments. Therefore, the discussion of photosensitive materials is structured into classical and new systems under development. The classical photoactive materials are those in which light absorption and charge separation are accomplished within a single crystalline environment. However, with the new nano-crystalline systems, the particles are too small to provide a suitable environment for the build-up of electrical fields. In these cases, other forms of charge separation, typically by kinetic mechanisms, must be achieved in order to obtain favourable macroscopic photosensitive behaviour. Finally, some molecular dynamics mechanisms of relevance to photosensitive materials are considered since they may foreshadow the evolution of present photosensitive materials. It is well known that the applications of materials and devices are led by economic considerations. This is evident particularly for the current range of commercial solar cell materials, which generally are not considered cost-effective relative to the conventional alternatives. This has delayed and, in some cases, prevented the former’s implementation on a large scale, despite the clear environmental benefits. The present work aims to point toward the emerging scientific challenges that must be faced before other economically more feasible photoactive materials can be developed and commercialised.

3.2

Absorption and transport by same materials

3.2.1

Macrocrystalline and microcrystalline materials

Conventional semiconductors The reason why semiconductors, isolators, and dye molecules but not metals are useful as photosensitive materials is straightforward: excited states must survive a reasonably long time (10–10–10–7 s) before the excitation energy is converted into thermal energy. Therefore, the electronic structure must have a reasonably large energy gap that separates: (i) the ground states from the excited states or (ii) a valence band from a higher conduction band. Excited electrons that recombine across such a large energy gap would be required to

Photosensitive materials

65

Energy

activate many phonons – vibrational quanta – more or less simultaneously in order to dissipate the energy. However, this is hindered kinetically in the material, which allows the excited electrons to survive for a reasonably long period of time, during which the material can develop its photosensitive properties. Figure 3.1 shows a typical semiconductor, such as silicon, rendered in the form of an electronic energy scheme. This allows the visualisation of how electrical fields can build up in the presence of interfaces between faces of different free energies – Fermi levels – of electrons. The presence of the electric field is evident from the bending of the energy bands, which is toward the interface. These energy bands describe the work necessary to transfer electrons from a vacuum to the corresponding location in the electronic scheme. Work has to be performed or is generated when an electron is moved against or with an electric field, respectively. Interface

Light Transport

Valence band

Energy gap

Conduction band

Electric field Holes break chemical bonds

Crystallised material (e.g. Si)

3.1 Energy scheme of a crystallised photoconductor (silicon) showing energy bands and band bending in regions where the electrical field is present. Photoexcitation and conduction occurs in the same material. Photoexcitation breaks essential chemical bonds.

Photosensitive materials of this type provide light absorption and charge transport within the same material. Photons are absorbed within the region

66


covered by the electrical field or in the neighbourhood where charge carriers can still diffuse into the field-determined region. Materials belonging to this class must be well crystallised and of dimensions at least in the µm range. The smaller the grains of these materials become, the greater the challenge with the materials technology because grain boundaries assume an increasingly greater effect on the properties of the material. A typical semiconductor of this type is silicon, which has been studied as a photoactive material for solar cells for more than 50 years. During this long period, a significant cost reduction has been achieved but silicon solar cells still generate electricity at a price approximately ten times that of conventional electricity. At present, the biggest challenge remains the economic production of microcrystalline materials of adequate performance. An overview of the current range of inorganic, photoactive, semiconductor materials, some of which are either used or under development for photovoltaics or optoelectronics, is given in Fig. 3.2. It can be seen that the range of energy gaps is dependent on the chemical nature of the compounds. The forbidden energy gap typically decreases within the chalcogenides (Group VIA of the Periodic Table) from the oxides downwards to the sulphides, selenides, and tellurides. Also, this gap is relatively small for the pnictides (Group VA), including phosphides and arsenides. Curiously, although many of these compounds are toxic, they have been subject to development while less toxic materials have been bypassed or studied only superficially. Interfaces of photosensitive materials play a significant role since they are important to both energy conversion and catalytic reactivity. Many of the semiconductors listed in Fig. 3.2 have the disadvantage that positive charges generated in the valence band through excitation of electrons disrupt existing chemical bonds. This means that, when holes accumulate at the interface, they lead to deterioration of chemical stability. It was for this reason that attention been given to a special class of semiconductors, sulfides and selenides where such bond breakage can be avoided (Tributsch and Bennett, 1977; Ennaoui and Tributsch, 1984; Jaegermann and Tributsch, 1988). These semiconducting materials are characterised by valence and conduction bands that are determined largely by energy states of the d electrons of transition metals. The most important of these compounds include the disulphides and diselenides of Mo and W in addition to the disulfides of Fe, Ru, and Pt. Such materials have been shown to be somewhat stable in wet photoelectrochemical solar cells. For example, FeS2 in contact with an iodide/iodine-containing electrolyte sustained a hypothetical turnover via electrolysis of 27,000 cycles without evidencing corrosion (Ennaoui and Tributsch 1986). Nevertheless, up to now, such materials have not found technical applications owing to their complicated and unresolved transition metal chemistry, which dominates the effects of doping and interfacial behaviour. It is probable that efficient interface passivation techniques are also required in order to optimise the materials’ properties.

Photosensitive materials Zn3P4 CdP2

GaP

67

Phosphides

InP GeAs

GaAs

CdTe

MoTe2

Arsenides

ZnTe Tellurides

GaTe ZrSe2 WSe2

CdSe

GeSe2

MoSe2 InSe SnS

Selenides

GaSe

WS2

CdS Sb2S3 RuS2 ZrS2 FeS2 Bi S MoS HfS2 2 2 2

Sulfides PbO TiO 2

2

Oxides

Nb2O5

WO3 1

ZnO

3 Energy gap/eV

3.2 Examples of inorganic compound photoactive materials, arranged according to their energy gap and grouped according to their chemical classification.

It should be mentioned that the development of photoactive materials in electronics typically takes 2–3 decades of international effort. If materials exhibit promising characteristics at an early stage of development, it is likely that many researchers will study these materials. In other cases, where initial difficulties are encountered, there may be an extended time period before these can be overcome and the material gains a degree of attention, which is a pre-condition of dynamic development. The normal course of action begins with single crystals, followed by the development of polycrystalline forms, ultimately leading to thin films. However, there are exceptions to every rule. A significant challenge in establishing or enhancing the photoactivity of semiconducting materials is the achievement of a homogeneous distribution of the photosensitivity. This generally becomes evident after imaging experiments are applied to determine the distribution of the photoactivity. Figure 3.3 illustrates different photocurrent images of MoSe2 showing variation of photocurrents. Comparison with other synthetic and natural photoactive materials reveals uneven distributions of the photoactivity. Essentially all photoactive materials, save the most perfect of single crystals, show an inhomogeneous distribution of the photoactivity as well as of charge separation properties. Mixtures of polymer and fullerene, which form the photoactive materials in organic solar cells, show inhomogeneities reaching ~30% (Jeranko et al., 2004). It can be seen that the imaging of the photoactivity represents a valuable tool in the optimisation of photosensitive materials and devices (Turrion et al., 1999; Macht et al., 2002; Barkschat et al., 2003a,b).

68

Materials for energy conversion devices –300 mV (SCE)

–100 mV (SCE)

100 mV (SCE)

1 mm 0

2

4

6

8 10 Jph/mA/cm2

12

14

16

3.3 Photocurrent image of MoSe2 crystal in contact with I – /I –3 showing variation of photocurrents with electrode potential.

Thin-film semiconductors Since single-crystal materials and mechanically cut polycrystalline materials are relatively expensive, there is generally a trend toward the preparation of thin-film semiconductors for photosensitive applications. In the case of copper indium disulfide, it has never been possible to fabricate reasonably efficient solar cells from macrocrystalline materials. However, the fabrication of reasonably efficient thin-film photovoltaic cells is relatively simple because the material is quite tolerant in terms of defects and interfacial performance. When layers of copper and indium are sputtered or evaporated and subsequently sulphurised, the result is a fairly high-performance light-absorption layer for solar cells. While many issues concerning the properties of such materials have been elucidated, there remain other aspects, particularly those significant to largescale production, to be understood. The successful deposition alone of a thin film of photoactive material generally is not sufficient to yield an efficient electronic device, such as a solar cell. It is also necessary to provide suitable buffer layers so that a reasonable energy conversion efficiency (ECE) can be obtained. That is, in addition to achieving the desired bulk properties, the interfacial properties must be optimised via the buffer layers. Such an approach has been used with both copper indium disulfide/diselenide and cadmium


69

telluride. The goal of the achievement of high efficiencies has led to a similar arrangement in the form of tandem cells of photosensitive materials, as in the case of copper indium disulfide/diselenide layers combined with copper gallium chalcogenide layers. It is remarkable that, of the thin-film photosensitive materials developed for photovoltaics, only silicon is considered to be: (i) environmentally compatible and (ii) sufficiently abundant for a photovoltaic market, which is expected to become very large by mid-century. Other materials considered for solar cells contain In, Ga, Cd, Te, and/or Se, all of which are both toxic and relatively rare. Since thin-film silicon in its amorphous form is not completely stable and, at present, there is no certainty if low-cost, thin-film, nano-crystalline, silicon solar cells will become available, this is a situation that can be expected to drive research toward the development of new materials. Some work has been done on iron disulfide (pyrite) as a semiconductor for solar cells. It has an energy gap of 0.95 eV and, in theory, could reach ECEs exceeding 10%. Practical experience in the development of this material, which has an exceptionally high light absorption coefficient of 6 × 10–5 cm–1, is limited to only a few laboratories. However, the limited data available suggest thinfilm solar cells of this material are feasible (Altermatt et al., 2002). The peculiar interfacial chemistry of iron disulfide appears to be responsible for the observation of high quantum efficiencies but quite modest photopotentials. It is expected that special efforts will be required in order to improve and optimise the interfacial chemistry of this material. Some additional pioneering work will be required to increase the ECE of this or other sulphide materials before they are taken up for industrial development. The benefits of the low costs, abundances, and environmental acceptability of iron and sulphur are offset somewhat by the cost of purchasing and operating thin-film fabrication facilities. Regardless, it is almost inevitable that there will be an increasing trend toward thinner films, provided they can be produced with high qualities.

3.2.2

Nano-crystalline materials

Highly absorbing materials Owing to the high cost of thin-film fabrication, it is desirable to make the films as thin as possible. In the case of materials of low solar radiation absorption, this strategy generally can be achieved only by implementing mechanisms for the capture of solar radiation using non-imaging optics. Highly structured substrates, which trap photons by multiple scattering and confinement, are typically used. On the other hand, there are some nanomaterials that can absorb light without the use of the preceding approach. These materials, which have unusually high absorption coefficients, include transition metal dichalcogenides, such as FeS2, WS2, MoS2, WSe2, and MoSe2.

70


Their absorption coefficients exceed 10–5 cm–1 and electron excitation does not break chemical bonds since the valence band has a d electron transition metal structure. This may signal a significant advantage in photostability when thin films and nano-particles are used. A third group of materials in this category are layers of highly absorbing dye molecules. In the case of transition metal sulphide materials, the high absorption coefficients, in spite of a transition metal d → d transition, is caused by significant admixtures of sulphur states into the conduction band states. Most photons absorbed in such materials are absorbed near the film surface, with the consequence that it may dominate recombination kinetics. Therefore, special care must be taken when attempting to optimise interfaces. Since light will generate high concentrations of photogenerated carriers, this will have a positive effect on the photopotential. Further, the electronic properties of such materials may not need to be very favourable since electronic charges must be collected across only a small distance. These factors suggest that research efforts in highly absorbing photosensitive materials are a fruitful avenue to follow. Photocatalytic materials Titanium dioxide or titania can absorb the ultraviolet (UV) fraction of solar radiation, which represents ~2–3% of the solar spectrum. Titania can use the resultant photoelectric charge to: (i) react with water via an oxidation mechanism to generate OH– radicals or (ii) reduce oxygen via a reduction mechanism. In both mechanisms, radicals associated with this photocatalysis are generated, as indicated in Fig. 3.4. The photocatalytic properties of titania, which have been investigated for over 30 years (Fujishima and Honda, 1972), are used widely in coatings that represent self-cleaning surfaces (Fujishima et al., 1999), where the radicals attack organic pollutants and oxidise them. Further, photoinduced super-hydrophylicity is used to provide antifogging protection of mirrors. UV-absorbing self-cleaning surfaces are typically prepared by covering them with nano-crystalline titania films, which are deposited by the sol-gel technique and subsequently heated. It is clear that a pre-condition for efficient self-cleaning properties is a reasonably high photosensitivity for the titania. Space-resolved photocurrent measurements have been used to study titania films produced by different methods, with the result that, again, the photoactivity is not homogeneous, as shown in Fig. 3.5 (Hagen et al., 2003). This indicates that these photocatalytically self-cleaning surfaces would benefit from optimisation efforts. Recently, titania doped with nitrogen, carbon, or sulfur has been shown to extend the photosensitivity to visible light up to ~550 nm. Compared to undoped titania, up to a sevenfold increase in photocatalytic activity was

71

Energy


O2 + 2H+ CB H 2O2

Light

OH

VB OH– TiO2

Light

CO2

Organic molecules

3.4 Scheme showing the photoactivity of TiO2, which induces the formation of OH radicals, both anodically and cathodically.

0

1 mm

1 mm

1 mm

(a)

(b)

(c)

20 40 Jph /mAcm–2

60

0

5 10 15 Jph /mAcm–2

20

0

2 4 6 8 10 Jph /mAcm–2

3.5 The images compare the photocurrent distribution of TiO2 layers prepared by flame oxidation (a), thermogravimetric annealing (b) and reactive sputtering (c) (Hagen et al., 2003).

72


Energy

observed (Sakthivel and Kisch 2003) After many failed doping efforts with transition metals, these are significant discoveries since more solar radiation can be converted into photocatalytic activity. However, for energetics reasons, it is likely that the process of OH– radical formation may operate only via the reduction route involving the conduction band. On the other hand, the holes generated on impurity states above the valence band energetically should not be able to generate OH– radicals by the oxidation mechanism since the thermodynamic potential, against the normal hydrogen electrode, for this process is approximately +2.8 V. However, holes on states generated by the dopants above the valence band will be able to oxidise organic compounds by capturing electrons, as shown in Fig. 3.6.

CB

E(OH–/OH) VB

TiO2 N, C-doping levels

3.6 Energy scheme explaining photoactivity of TiO2 doped by nitrogen, carbon or sulfur. Holes on doping states may not have the energy to generate OH radicals.

3.3

Absorption and transport separated

3.3.1

Dye sensitisation materials

In contrast to the materials discussed previously, there are a group of photoactive materials in which excitation and charge transport are separated. A wellknown example is the photosynthetic membrane, where the chlorophyll reaction centre injects electrons into an electron transfer chain of proteins and macromolecules. The dye sensitisation cell has evolved by reproducing this principle (Tributsch, 1972). The electron transfer chain is replaced by oxide particles and the chlorophyll by synthetic sensitising molecules. Thus, the new photoactive material is an oxide with a large energy gap, where organic


73

or inorganic molecules are adsorbed onto the oxide. When the excitable adsorbed molecule injects electrons, this electron becomes a majority carrier in the oxide conduction band and therefore the electron cannot recombine within the oxide material. In the same way, an excited sensitiser can inject a hole into a p-type wide-gap material, so generating majority carriers there. The advantage of sensitised materials is that they can be of low-grade quality but still work well, provided their electron-conducting properties are properly adjusted. At present, dye sensitised solar cells are fabricated with nanocrystalline oxide materials, mostly titania (O’Regan and Grätzel, 1991; Hagfeldt and Grätzel, 1995; Grätzel, 2001). These sensitised nano-crystalline materials are subject to the situation that electrical fields are unable to develop owing to the penetration of the nano-crystalline material with an electrolyte or a polymeric or solid contact, so charge separation must occur via chemical kinetic mechanisms. Therefore, for kinetic reasons, this requires more efficient (i) electron injection into the oxide and (ii) charge collection compared to the reverse reaction of electrons with the redox system in the electrolyte, which is added to regenerate the oxidised sensitiser molecule. If the reverse reaction of the injected electron were rapid, the photoactivity of the sensitised material would be low. Therefore, all modern sensitised solar cells use the same redox electrolyte, iodide/tri-iodide, because the reverse reaction of the electron with the tri-iodide is complicated and kinetically inhibited. If more reversible redox systems were added instead, for example Fe2+/3+, Fe(CN)63+/4+, or hydroquinone/quinone, then the photoactivity would be reduced substantially. This indicates that kinetic irreversibility must be considered a critical factor that determines the photoactivity. More to the point, it replaces the electrical field that is generated at well crystallised semiconductor interfaces, as discussed previously. Sensitisation processes have been used widely in silver halide photography, which has a history extending over a period greater than a century. In common with today’s research efforts, the evolution of photographic chemistry initially was developed empirically, with scientific understanding following much later. It is probable that the same course will be followed in the development of modern photoactive nano-materials and nano-composites, although more rapid progress can be expected owing to the availability of modern research tools.

3.3.2

Polymer-based composite materials

Polymers were used as long ago as the 1970s as substitutes for crystallised inorganic materials in the fabrication of solar cells. However, typical ECEs did not exceed 0.5%. Much later, it was discovered that the addition of fullerene molecules to polymers could increase the ECE to 3% (Brabec and Sariciftci, 1999; Brabec et al., 2001). This occurs because the fullerene can

74


accept electrons easily but relaxation of its electronic structure suppresses the reverse reaction. By simple mixing of the fullerene and polymer components, electrons and holes find their ways to opposite electrodes that are sandwiched 150 nm apart in the composite material. The functional principle of such a photoactive material is shown in Fig. 3.7. Further, the irreversible nature of the electron exchange is a factor critical to photosensitivity and thus to solar cell or photodiode function. Remarkably, these presumably homogeneous fullerene-polymer mixtures demonstrate significantly inhomogeneous photocurrent images (Jeranko et al., 2004).

Light

ITO covered glass Polymere

Fullerene Aluminium

3.7 Scheme explaining composite materials, such as polymer/ fullerene mixtures. Charge separation occurs through kinetic irreversibility and subsequent percolation.

3.3.3

Challenges for composite solar cell materials

Sensitised nano-materials and fullerene-polymer composites still face significant difficulties. Both photodegradation and chemical degradation occur in both types of these materials. While existing sensitised solar cells appear to have operational lives of ~2 years, even the most stable sensitiser, a ruthenium complex, is subject to ongoing photoelectrochemical degradation. This phenomenon appears to be strongly dependent on the surface bonding and thus on the quality of the adsorption site for the molecules involved (Barkschat and Tributsch, 2005). That is, the molecules survive well on some sites but react rapidly and irreversibly on others. Further research is needed in order to optimise the surfaces of nano-particles and to increase the stability of attachment of sensitiser molecules. Both phases of fullerene-polymer composite materials have much more pronounced instabilities. At present, the preferred strategy to overcome these problems is to improve the sealing, which prevents access of oxygen and humidity, but this strategy may counter the advantage of inexpensive production of materials. Therefore, it is necessary to develop a new strategy, probably based on the negative experience gained from dye sensitisation and composite solar cells. Such new composite solar energy materials should not contain


75

materials with inherent photodegradation properties, which presently exclude many organic materials and many well-known semiconductors on a fine scale. For example, if cadmium sulphide or cadmium telluride are illuminated, the holes formed in the valence band are equivalent to broken bonds. When they accumulate at the interface, it is forced to disintegrate. Small particles of such materials, which have large surface areas, cannot survive for long periods because the photogenerated holes will accumulate broken chemical bonds. In contrast, if molybdenum sulphide or tungsten sulphide is used in the form of small particles as absorbers for photons, the holes formed in the valence band do not correspond to broken bonds, so the particle interfaces, which still have large surface areas, will continue to function. This has been observed for nano-particles of Mo and W chalcogenides. After the identification of suitable stable materials, the key challenge will be to modify them chemically whereby electron transfer from the absorber to the acceptor can be made much more efficient than the reverse reaction. Although complicated mechanisms, involving multiple steps, are known to reduce the probability for these reverse electron reactions, the fundamental aspects of electron transfer irreversibility still remain to be investigated. One possibility is self-organised electron transfer, which exploits the possibilities of non-equilibrium electron transfer mechanisms and so is excluded from consideration by the classical Markus theory of electron transfer. This mechanism would involve a feedback process that functions in such a way that, when electrons (or electron density) are transferred, the electronic environment changes, thereby increasing subsequent transfer of electrons (or electron density). It can be shown mathematically that a self-organised process leads not only to a significant increase in the electron exchange rate but also in the suppression of the reverse reaction (Pohlmann and Tributsch, 1997; Tributsch and Pohlmann, 1998). It is probable that some time will pass before charge separation based on kinetic irreversibility can be developed to the stage of facilitating a high standard of performance.

3.4

Property control by particle size

3.4.1

Quantum-sized materials

Certain molecules change their optical properties as their dimensions change. Some inorganic materials show similar characteristics when they reach dimensions at which quantum size effects occur. A typical consequence of this is that the energy gap widens so that the light-absorption properties shift in the direction of higher photon energies (Henglein, 1989). Characteristically, the particles must be < 3 nm so as to show quantum size phenomena. If particles of such dimensions are produced, the energy gaps of the semiconductors listed in Fig. 3.2 and others can be tailored to increase their energy gaps, as indicated in Fig. 3.8.

76


This quantum size effect results from electrons and holes being squeezed into a dimension approaching a critical quantum parameter called the ‘exciton Bohr radius’ (Brus, 1984). Both the size and geometric dimensions of the confinement influence the quantisation. Quantum wells are one-dimensional, quantum wires two-dimensional, and quantum dots three-dimensional confinement geometries for charge carriers. By changing from a wire-shape to a spherical dot, the energy gap and thus the photosensitive properties of the material can be changed. This has been demonstrated for indium phosphide. The synthesis of geometrically tailored quantum size particles is still under development but the potential to control the electronic structure of photosensitive materials is a sufficiently attractive challenge to motivate ongoing research and so it has become a rapidly growing area in nanotechnology. A major task in the fabrication and handling quantum-sized particles is their stabilisation in appropriate matrixes. Owing to the small dimensions of quantum-sized particles, the role of the surface is very important because it is extremely large compared to the volume. If the particles are not properly stabilised, they can agglomerate to form bigger particles and so lose their quantum properties. Light

Energy

Wave functions increasingly confined

Energy gap

3.8 Energy scheme explaining quantum size particles. Small size squeezes orbitals which leads to band gap widening.

The use of quantum-sized particles for band-gap-tailored solar cells and photoelectrochemical materials has already been reported (Gorer and Hodes, 1997). An aim is to tailor intermediate-band-gap states by introducing quantumsized particles in order to harvest solar energy more effectively (Luque and Marti, 1997). Another significant advantage is that many materials with energy gaps too low to be useful become applicable in solar cells in the form of quantum-sized particles. Such materials then become potentially


77

photosensitive, and solar energy materials provided agglomeration can be avoided, represent a considerable challenge.

3.4.2

Photonic materials

It is known that periodic lattice structures provide periodic potentials for electrons that can structure electronic levels into energy bands. Similarly, periodic variation in the refractive index may lead to so-called ‘photonic energy bands’ (Johnson and Joanopoulos, 2002; Sigalas et al., 1999). Since the wavelength of the photons is inversely proportional to the energy, the periodically dielectric material can block light with wavelengths in the photonic band gap while allowing other wavelengths to pass freely, as shown in Fig. 3.9. A photonic material can typically be made of a block of transparent dielectric material containing a number of minute pores, holes, or gaps arranged in a periodic lattice pattern. In this way, a dielectric interspersed with regions of low-reflecting index is generated. For the photons, this contrast in refractive index acts like the periodic potential that an electron experiences while travelling through a crystal lattice. In most circumstances, photonic band gap structures are comprised of a matrix of high refractive index material embedded in a medium of lower refractive index. A naturally occurring Low refractive index

Light

Three-dimensional periodical refractive structure

Energy

High refractive index

Photonic band gap due to periodical refractive index

3.9 Scheme explaining photonic materials. Periodic variation of the dielectric constant leads to optical band formation in photosensitive materials.

78


photonic material with a low contrast between refractive indices is the mineral opal, which consists of silica spheres dispersed in a hydrous matrix. Since the periodicity must occur on a very small scale, it has taken some time for photonic crystals for the near-infra-red or visible regions of the light spectrum to be developed. Since the required periodicity is given by the wavelength of light divided by the refractive index, then semiconductor micro-fabrication techniques are required for the synthesis of such materials. Photonic crystals promise control over photons similar to the control over electrons held by electronic materials. While present attention is focused on communication systems, photonic crystals are expected to gain gradual ground in light capture and transformation for energy conversion applications. An example of such a material is a closely packed array of spherical air voids in a titania matrix, which is produced as follows: (i) submicron-sized silica spheres are allowed to self-arrange in a colloidal suspension, (ii) the voids are filled with a titania-based suspension, (iii) the mixture is treated chemically and thermally in order to solidify it, and (iv) the silica spheres are dissolved. Processes such as this can be used to fabricate electronically conducting materials with specific optical properties. There remain many practical problems to be solved before such photosensitive materials can be of practical use.

3.5

Property control by molecular dynamics

3.5.1

Molecular electronic materials

Increasing knowledge of molecular and electronic dynamics is likely to facilitate the generation of photosensitive materials comprised of molecular electronic elements. Photoexcitation of specific centres will be linked via rectifying molecular electronic bridges to sites where electrons can be put to work, either by driving electronic circuits or by inducing luminescence, as indicated in Fig. 3.10. An interesting natural model system is the photosynthetic electron transfer chain. This photoactive macromolecular array, which mediates the transfer of electrons from water to the oxidised form (NADP+) of nicotinamide adenine dinucleotide phosphate (NADP) involves seven nonmetallic carriers, including quinones, pheophitine, the reduced form of NADP (NADPH), thyrosene, and flavine. Altogether, twenty-nine metal ions, including Fe, Mg, Mn, and Cu are involved. Parts of this photoactive electron transfer chain have already been imitated synthetically in order to enable an understanding of electron transfer mechanisms. Molecular electronic materials and devices powered by photoexcitation processes also promise the gradual development of tailored photosensitive materials for photon energy conversion and photocatalysis. A major challenge in this field will be the control of efficiency and long-term stability. Complicated macromolecules tend to engage in side reactions, which would lead to gradual deterioration of the material.


79

Molecular bridge Electron acceptor Light

Electronic feedback Suppressed reverse reaction Donor Molecular absorber Kinetically determined irreversibility

3.10 Scheme explaining function of molecular electronic photoactive materials. Kinetic (non-linear) charge separation may significantly improve photoactivity by temporarily suppressing recombination reactions.

New strategies must be adopted in order to cope with such problems. Again, natural examples can point the way toward knowledge of the mechanisms, which can be made to work against increasing entropy or disorder.

3.5.2

Biological photochemical stability

While nature is known to stabilise photoactive systems by self-organisation and complicated protection mechanisms, it has also evolved remarkably stable photoabsorbers. A well-known example is the bacteriorhodopsin molecule. Upon absorption, this molecule, which pumps protons across the membrane of salt-loving bacteria (Halobacterium halobium), performs a quite complex photoreaction cycle. It was found that, during the light-induced conformational change of this molecule, 150.84 kJ/mol of heat is released in addition to the energy content by the photon (205 kJ/mol). Therefore, the free energy stored must take the form of decreased entropy of –300 J/mol.K during the formation of the metastable state for the net energy change to be possible. This large entropy decrease implies a substantial increase in molecular order. This may be contrasted with the increased entropy that accompanies the unfolding of a protein-like lysozyme. The ability of bacteriorhodopsin to reduce its entropy upon photon absorption is reflected in its extraordinary chemical stability. Owing to its photochromic properties and its stability over many years of use, it has been considered seriously as an absorber for optoelectronics. The capacity of this molecule to export entropy upon photon absorption has been explained and linked to

80


autocatalytic self-organisation (Tributsch and Pohlmann, 1996). It behaves as an open system involved in a feedback process, which leads to a temporary build-up of order. In this way, undesired side reactions induced by the photon absorption process appear to be controllable. Although the prospect of success in handling photosensitive materials as non-equilibrium irreversible thermodynamic systems appears to be remote, these systems may be relevant because the problem of sustaining long-term stability is significant to optoelectronics, solar cells, and photocatalytic systems. Research in increasing photostability is needed urgently and can justify unconventional approaches.

3.6

Materials research challenges for photon energy conversion

3.6.1

New solar cell materials

The present century will be faced with major environmental problems if energy conversion does not include an increasing fraction of sustainable methods and materials. Experience shows that it takes typically 2–3 decades to develop a photoactive material for industrial application. Experience with silicon, copper indium diselenide/disulfide, and cadmium telluride suggests that, on average, an international effort of three years is required to improve the ECE of a solar cell by 1%. Also, it should be noted that the ECE of commercial solar cells tends to lack that of laboratory versions by ~30%. Crystalline silicon solar cells, which dominate the photovoltaic market at present, are quite expensive, with energy costs ten times higher than those required of fossil fuels. Since the learning curve is not especially steep and since thin-film silicon solar cells fabricated using economical industrial technologies are not yet in sight, then this foreshadows a bottleneck in the future. At present, thin-film solar cells, which have achieved relatively high ECEs, are comprised of comparatively rare and toxic materials, including Cd, As, In, Se, or Te. Such materials are unlikely to reach the economies of scale necessary for large-scale production and a sizeable market penetration. Therefore, there is an urgent need to develop new materials, techniques, and devices for photovoltaic energy conversion. Some materials that have the potential to compensate for the mentioned shortcomings include MoS2, WS2, FeS2, transition metal sulphides, all of which show favourable energy gaps for visible light conversion. Since these are transition metal compounds, both surface chemistry and dopant chemistry are determined by coordination chemistry, which makes the scientific approach to be used quite different from that of classical photoactive materials, such as silicon, gallium arsenide, and cadmium telluride. For any photosensitive material to be developed in the future, quality control capable of achieving homogeneity of photoactivity will be a key factor. When preparing a photoactive


81

material, a homogeneously distributed photosensitivity cannot be assumed, so this must be optimised. Another major challenge is degradation with time. The origin of photodegradation remains unclear and so it must be identified and more stable components developed. An example of this problem is thinfilm materials, especially chalcogenides, used for solar cells, which must be protected carefully against oxidation. Quantum-sized materials must avoid agglomeration of colloid particles. Dye-sensitised nano-composites and organic solar cells are subject to degradation. Interestingly, dye-sensitised solar cells in the dry condition appear to be much more susceptible to photodegradation than in the wet condition (Sirimanne et al., 2003; Sirimanne and Tributsch, 2005). The preceding is a sampling of the dimension of the research challenges for the near future. It has been shown that solar cells can be used to do more than to drive electronic currents: they may also drive protonic currents if appropriate combined electronic-ionic materials are selected as photoactive compounds (Bungs and Tributsch, 1997; Tributsch, 2000). Photoelectronic effects may drive the photoinsertion of hydrogen, which can move through materials to trigger the release of protons at the opposite surface. There is no reason why light-powered protonic photovoltaic devices should not work as efficiently as light-powered electronic ones. Further, photochargeable ion insertion devices are feasible (Betz and Tributsch, 1985). Before such systems can be developed, a new class of ion-conducting photosensitive compounds must be developed and optimised. The light-harvesting antenna chlorophylls of the photosynthetic membrane effectively use dipole-dipole energy transfer processes to photoexcite the reaction centres. A similar tailoring of energy transfer for the excitation of semiconductors is being attempted by using dye molecules in zeolite particles in contact with materials, such as silicon, that have a low absorption coefficient (Calzaferri, 2001; Maas et al., 2003). Excitation energy transferred by dipoledipole or dipole-quadrupole interaction will be subject to much higher transition probabilities, so poorly absorbing photomaterials may be intensively excitable, even as thin films. Conventional photovoltaic cells are quite expensive. This is due largely to the materials technologies required to extract and collect charge carriers. These involve, inter alia, oxide windows with high electron mobilities, optimised contacts with low resistivity, and sealants to prevent corrosion of the contacts by the environment. Titania is extremely stable in the environment, so doping in order to increase the photosensitivity to visible light is an attractive route to attempt photovoltaic water splitting. This could be done directly via tandem photoexcitation of titania in combination with an additional photoactive material. Previous experiments have shown that ~90% of generated photovoltaic energy can be converted into hydrogen energy (Licht et al., 2000). If all materials were stable in an aqueous environment, then the

82


chemical fuel hydrogen could be produced directly, as in the case of the production of the chemical energy carrier NADPH at the photosynthetic membrane. Since no intermediate photocurrent is generated, then photovoltaic tandem systems could become significantly more economical alternatives for photovoltaic fuel generation than the present projected strategy, which is to generate photovoltaic electricity to power electrolysis units to produce hydrogen. The major challenge for these tandem systems is the development of materials that are: (i) photoactive for the solar spectrum, (ii) stable in aqueous environments, and (iii) economical. This chapter represents a short introduction to photosensitive materials. It has shown that the field is at an early stage in terms of: (i) having an adequate scientific understanding of the relevant materials and processes and (ii) attaining technical control of the fabrication of the different types of photosensitive material.

3.7

References

Altermatt, P., Kiesewetter, T., Ellmer, K. and Tributsch, H., Solar Energy Mater. Solar Cells, 71 (2002) 181. Barkschat, A., Pohlmann, L., Dohrmann, J.K. and Tributsch, H., Phys. Chem. Chem. Phys., 5 (2003a) 1259. Barkschat, A., Dorhmann, J.K. and Tributsch, H., Solar Energy Mater. Solar Cells, 80 (2003b) 391. Barkschat, A. and Tributsch, H. (2005), in preparation. Betz, G. and Tributsch, H., Prog. Solid State Chem., 16 (1985) 195. Brabec, C.J. and Sariciftci, N.S., in Hadziannou, G. and van Hutten, P. (eds) Conjugated Polymers, Weiningen, Wiley-VCH, 1999. Brabec, C.J., Sariciftci, N.S. and Hummelen, J.C., Adv. Funct. Mater., 11 (2001) 15. Brus, L.E., J. Phys. Chem., 80 (1984) 1816. Bungs, M. and Tributsch, H., Ber. Bunsenges. Phys. Chem., 101 (1997) 1844. Calzaferri, G., Chimia, 55 (2001) 1009. Ennaoui, A. and Tributsch, H., Solar Cells, 13 (1984) 197–200. Ennaoui, A. and Tributsch, H., Solar Energy Mater., 14 (1986) 461. Fujishima, A. and Honda, K., Nature, 238 (1972) 37. Fujishima, A., Hashimoto, K. and Watanabe, T., TiO2 Photocatalysis: Fundamentals and Applications, BKC, Tokyo, 1999. Gorer, S. and Hodes, G., in Kamat, P.V. and Meisel, D. (eds) Semiconductor Nanoclusters: Physical, Chemical and Catalytic Aspects, London, Elsevier, 1997, p. 297. Grätzel, M., Nature, 414 (2001) 338. Hagen, A., Barkschat, A., Dohrmann, J. and Tributsch, H., Solar Energy Mater. Solar Cells, 77 (2003) 1. Hagfeldt, A. and Grätzel, M., Chem. Rev., 95 (1995) 49. Henglein, A., Chem. Rev., 89 (1989) 1861. Jaegermann, W. and Tributsch, H., Prog. Surface Sci., 29 (1988) 1. Jeranko T., Tributsch H., Sariciftci N.S. and Hummelen, J.C., Solar Energy Mater. Solar Cells, 83 (2004) 247.


83

Johnson, S.G. and Joanopoulos, J.D., Photonic Crystals: The Road from Theory to Practice. Kluwer, Boston, 2002. Licht, S., Wang B., Mukerji, S., Soga, T., Umeno, M. and Tributsch H., J. Phys. Chem. B, 104 (2000) 8920. Luque, A. and Marti, A., Phys.Rev. Lett., 78 (1997) 5014. Maas, H., Huber, S., Khatyr, A., Pfenniger, M., Meyer, M. and Calzaferri, G., in Ramamurthy, V. and Schanze, K. (eds) Molecular and Supramolecular Photochemistry 9, Marcel Dekker, New York, Basel, 2003, p. 309. Macht, B., Turrión M., Barkschat, A., Salvador P., Ellmer K. and Tributsch, H., Solar Energy Mater. Solar Cells, 73 (2002) 163. O’Reagan, B. and Grätzel, M., Nature, 353 (1991) 373. Pohlmann, L. and Tributsch, H., Electrochim. Acta, 42 (1997) 2737. Sakthivel, S. and Kisch, H., Angew. Chem., 115 (2003) 5057. Sigalas, M.M., Biswas, R., Tuttle, G., Soukoulis, C.M. and Ho, K.M., in Wiley Encyclopedia of Electrical and Electronic Engineering., Volume 16. New York, Wiley, 1999, p. 345. Sirimanne, P. and Tributsch, H., Materials Chemistry and Physics, in press (2005). Sirimanne, P., Jeranko, T., Bogdanoff, P., Fiechter S. and Tributsch, H., Semicond. Sci. Tech., 18 (2003) 708. Tributsch, H., Photochem. Photobiology, 16 (1972) 261. Tributsch, H., in Schiavello, M., Kluwer, (ed.) The Path of Electrons in Photoelectrochemistry, Photocatalysis and Environment – Trends and Applications (NATO ASI Series), Darmstadt, 1988. Tributsch, H. and Bennett, J.C., J. Electroanal. Chem., 81 (1977) 97–111. Tributsch, H. and Pohlmann, L., J. Theor. Biology, 178 (1996) 17. Tributsch, H. and Pohlmann, L., Science, 279 (1998) 1891. Tributsch, H., Int. J. Ionics, 6 (2000) 161. Turrión, M., Macht, B., Salvador, P. and Tributsch, H., Zeit. Physik. Chem., 212 (1999) 51.

4 Defect disorder, transport and photoelectrochemical properties of TiO2 J N O W O T N Y, C C S O R R E L L, T B A K and L R S H E P P A R D, The University of New South Wales, Australia

4.1

Introduction

Oxide materials have found many applications in energy conversion devices, including solid electrolytes, electrodes, and photoelectrodes. One of the most commonly used oxide materials in energy conversion is yttria-stabilised zirconia (YSZ), which has been employed as an oxygen conductor in electrochemical devices, such as solid oxide fuel cells (SOFCs); electrochemical gas sensors for greenhouse and pollution gases, such as CO2, NOx, and SOx; and electrochemical gas separators.1–3 Its applicability is the reason why substantial interest has been generated in research on YSZ.1–6 Metal oxides are also used as functional elements in SOFCs, including (Sr, La)MnO3 as cathode and LaCrO3 as interconnect,1 and as promising candidates for thermoelectrical energy convertors, including Cu2O and TiO2.7 While the technology of YSZ-based electrochemical devices is relatively well established, there is growing interest in the development of photoelectrochemical devices aiming at the conversion of solar energy into electrical and chemical energies. Similarly, semiconducting oxides, especially TiO2, have great potential in the development of these devices. The potential of TiO2 as the leading candidate for these applications has generated enormous interest by many researchers in the modification of its properties in order to impose the properties necessary for its use as a photoelectrode. Since the properties of metal oxides are determined by the defect disorder, the purpose of the present chapter is to consider the defect chemistry and defect-related properties of TiO2, including the photoelectrochemical properties. The properties of TiO2 are such that several important applications are either feasible or already realised. These include:8–11 • • • • 84

Photoelectrochemical generation of hydrogen Decontamination of water Coatings of self-cleaning materials Components of antiseptic paints

Defect disorder, transport and photoelectrochemical properties

85

• Chemical gas sensors • Coatings for non-fogging mirrors. Up to now, the applications of TiO2 have been limited largely by the properties of commercially available materials, which are usually processed at high temperatures in air. Therefore, the oxygen non-stoichiometry is related closely to the oxygen partial pressure of air. Consequently, commercial TiO2 exhibits mainly insulating properties at room temperature and these properties are not desirable for the performance of TiO2 as a photoelectrode. However, the oxygen non-stoichiometry and the related defect disorder may be modified within a wide range by the imposition of an oxygen partial pressure (p(O2)) different from that of air12–15 and by doping with aliovalent ions, which leads to the formation of donors and acceptors.16,17 It is becoming increasingly clear that the application-related properties of TiO2 are determined or influenced by defect disorder and the resultant properties, which include:18 • • • •

semiconducting charge transport electronic structure surface properties photosensitivity and photoreactivity.

Consequently, defect chemistry has been used as a framework to explain the functional properties and to modify these properties in order to achieve desired properties, which are required for specific applications.12–15,18,19 Consequently, the processing of TiO2 with desirable properties requires an increase in the present state of understanding of its defect chemistry and defect-related properties. In this regard, there have been efforts to enhance the specific properties required for its use as a photoelectrode for hydrogen generation through the decomposition of water using solar energy (solarhydrogen), including:9,20 • Maximisation of solar-energy absorption • Minimisation of charge recombination through optimised charge separation • Maximisation of charge transfer. One of the methods of increasing the solar energy absorption of TiO2 and the associated ionisation over the band gap (formation of an electron-hole pair) is through the reduction of its band gap from ~3 eV for commercial TiO2 to ~2 eV by modification of its defect disorder.9, 17 Ionisation over the band gap takes the system to an excited state. The subsequent recombination of the electron-hole pair is undesirable owing to energy losses. These losses can be minimised by the imposition of an electric field, which results in charge separation. This electric field can be imposed by an electrical potential barrier that forms at the electrolyte/TiO2 interface. In this case, the barrier is termed the ‘flat band potential’ (FBP).9 It will be shown that the FBP may be

86


modified by the imposition of segregation-induced concentration gradients and the associated electrical potential barriers. The effective performance of a photoelectrochemical cell (PEC) requires efficient charge transfer within the PEC and, in particular, within the photoelectrode. Adequate charge transfer can be achieved through the imposition of defect disorder that results in electronic charge compensation. This may be achieved through the incorporation of aliovalent ions under controlled p(O2). As will be shown, the incorporation mechanism of foreign ions depends on the oxygen non-stoichiometry and, therefore, the p(O2) during thermal treatment. The present chapter outlines the basic defect equilibria for TiO2. The gas/solid equilibria and the impact of the imposition of a well defined oxygen activity are also considered for the TiO2-O2 system. These equilibria are then used to derive defect diagrams for undoped TiO2 and its solid solutions with donors and acceptors. The use of electrical characterisation methods to verify these diagrams is considered. Finally, the importance of the defect disorder of the surface layer is discussed.

4.2

Terms and aims

TiO2 is an oxygen-deficient compound, so its formula is more correctly TiO2–x, where x is the apparent oxygen deficit. Up to now, it has been assumed widely that the semiconducting properties of TiO2 can be considered in terms of simple defect disorder, which involves the following assumptions: • The predominant defects in TiO2–x, are oxygen vacancies, which form donor-type defects. • Electrons are the predominant electronic charge carriers and so they are responsible for n-type semiconductivity. However, it is becoming increasingly clear that the defect disorder is considerably more complex. Specifically, electrons are the predominant electronic charge carriers only for reduced TiO2, while oxidised titania also may exhibit mixed conduction involving electrons and electron holes.12–15,21 It has been shown that strongly oxidised TiO2 may exhibit p-type properties, with electron holes as the predominant electronic charge carriers.15 The present chapter outlines the existing state of understanding in this area. The most common method of verifying defect disorder models is through measurements of the electrical properties, such as electrical conductivity, thermoelectric power, and work function.22 Therefore, the purpose of the subsequent material is to overview the basic literature data on the electrical properties of TiO2 and the techniques used to determine them. The main focus is on measurements of the electrical conductivity, which is the most commonly used electrical property to characterise the semiconducting properties of metal oxides.


4.3

Electrical properties

4.3.1

Experimental requirements for the determination of well-defined data

87

The defect equilibria of metal oxides are usually studied at elevated temperatures because the transport of defects is relatively rapid and so equilibrium is achieved relatively quickly. However, the defect chemistry is reflected in the experimental parameters: (i) solid composition, (ii) gas/solid equilibrium, (iii) oxygen activity, (iv) temperature, and (v) time, so it is critical to monitor these in all cases. Therefore, in determination of the defect disorder by hightemperature electrical properties, measurements should be done under welldefined conditions on well-defined specimens, bearing in mind the following considerations: • Solid composition The electrical properties data are sensitive to the composition of the solid, including the matrix material, intentionally added dopants, and incidental impurities. However, it is important to note that even trace-level impurities can cause substantial modification of the electrical properties. Aliovalent ions, even at the parts-per-million (ppm) level, can alter the electrical conductivity by several orders of magnitude.23 Consequently, it is essential to study specimens of known chemical compositions. Unfortunately, most reports of semiconducting properties do not provide chemical analyses for the spectra of the types of impurities and their concentrations. • Gas/Solid equilibrium The electrical properties of solids measured at elevated temperatures are well defined only when measured under conditions of gas/solid equilibrium. This is defined by the temperature and the composition of the gas phase surrounding the specimen. Alternatively, when the system is not at equilibrium and so is in the kinetic regime, the data for electrical properties cannot be well defined. Therefore, evidence for the achievement of the equilibrium state is essential for data to be well defined. Unfortunately, many reports do not provide evidence for this. It is generally assumed that a solid that has been densified or otherwise heat treated at high temperatures is in equilibrium. Frequently, this is not the case. However, the phenomenon of segregation results in variance between the surface and bulk compositions and so, their properties.23 Thus, it is essential to recognise this potentiality and consequently to monitor the equilibration of segregation-induced concentration gradients. Unfortunately, few reports examine this important factor and assume that the measured electrical properties are representative of the bulk chemistry. • Oxygen activity In the case of metal oxides, the most important component of the gas phase is oxygen. As shown in Fig. 4.1,24 the electrical conductivity σ and thermoelectric power S are sensitive to the oxygen activity, the


log σ(σ in Ω–1 m–1)

TiO2 (SC) 1,073 K –1.6

–1.7

–1.8

p-Type

n-p Transition

n-Type 500

S (µV/K)

88

0

–500 1

2 3 4 log p(O2) (p(O2) in Pa)

5

4.1 Electrical conductivity (upper part) and thermopower, S, (lower part) for undoped TiO2 single crystal at 1,073 K as a function of log p(O2) within the n-p transition.24

latter being symbolised by p(O2). Equivalence between the oxygen activity and the oxygen partial pressure can be considered approximately correct at relatively high oxygen partial pressures (not significantly lower than that of air – 21 kPa). However, with dilute oxygen gas solutions at very low oxygen partial pressures, there is a deviation between the two values, so this approximation is incorrect. It is important to note that the quantitative description of the defect equilibria requires the use of the oxygen activity and that this may differ significantly from the oxygen partial pressure. Therefore, there is a need to assess the defect-related properties, such as electrical conductivity, using a known oxygen activity in the reaction chamber. This requires the use of an electrochemical oxygen probe. If there is a risk of composition gradients in the gas phase (from relatively high gas flow rates), the p(O2) should be determined in the immediate vicinity of the specimen.


89

• Temperature The electrical properties depend significantly on the temperature. Since controlled-atmosphere experimental chambers often have the temperature sensor, typically a thermocouple, located remotely from the specimen, significant errors in temperature measurement can occur. Thus, it is important for the temperature to be determined in the immediate vicinity of the specimen. • Time It is essential to utilise sufficient time to achieve equilibrium during both sample fabrication and testing. Without confirmation that the time factor is adequate to ensure equilibration, specimens may be tested under the kinetic regime and so the data will not reflect equilibrium conditions. Unfortunately, it is rare for published reports to demonstrate the achievement of equilibrium. In conclusion, several specific conditions must be addressed in the determination of well defined data of electrical properties at elevated temperatures. The data for the electrical properties measured at room temperature are determined by the conditions of processing, such as temperature and p(O2), and the conditions of cooling or quenching. Therefore, this related information is needed for the assessment of data.

4.3.2

Literature data for electrical conductivity

Most studies of the effect of the p(O2) on the electrical properties of TiO2 at elevated temperatures have focused on measurements of the electrical conductivity. The electrical conductivity data vs. p(O2) have been reported by Tannhauser,25 Yahia,26 Blumenthal et al.,27 Baumard and Tani,28 Marucco et al.,29 Ballachandran and Eror,30 Son and Yu,31 and Nowotny et al.21 The experimental data can be divided approximately into three regions: highly reduced, reduced, and oxidised. Analysis of the electrical conductivity of TiO2, involving defect disorder and semiconducting properties and leading to the determination of the mobilities of electronic charge carriers, has been reported by Bak et al.12–15 Absolute values for the electrical conductivity reported by different authors may be compared only when the data were determined at the same or comparable temperatures. This is the reason why it is difficult to compare absolute values of the literature data. As shown in Fig. 4.2, the data for the isothermal effect of the p(O2) on the electrical conductivity for TiO2 single crystals, can be described by the following general dependence: 1

σ = σ o p (O 2 ) mσ

4.1

where σo = conductivity parameter independent of the p(O2) and mσ = parameter sensitive to defect disorder. The exponent for the p(O2) dependence, 1/mσ,

90

Materials for energy conversion devices 3.0 2.5 2.0

log σ (σ in S/m)

1.5 1.0 0.5

1,387 K

0.0 –0.5

–1/6

1,166 K

–1.0 985 K

–1.5 –2.0 –2.5 Undoped TiO2 (SC) –1/4 Nowotny et al., 199721 –3.0 –20 –18 –16 –14 –12 –10 –8 –6 –4 –2 0 log p(O2) (p(O2) in Pa)

1/4 2

4

6

8

4.2 Electrical conductivity for undoped TiO2 single crystal versus p(O2) in the range 985 K–1,387 K21.

conventionally is used to verify the defect disorder models. As seen in Fig. 4.2, the electrical conductivity data in the temperature range 985–1387 K exhibit several regimes of different slopes, as shown by the exponent 1/mσ:12–15 • mσ = –6 where 10–20 Pa < p(O2) >105 Pa in Fig. 4.5. These data make it clear that the characterisation of TiO2–x specimens should include determination of the concentrations of impurities that act as donors and acceptors. Figure 4.14 shows a schematic of the effect of the introduction of aliovalent ions, which results in the formation of mid-band levels. At high concentrations,


107

Conduction band EC

Energy

1.8–2.2 eV ~ 3 eV

EA Imposed mid-gap band EV Valence band

Density of states

4.14 Electronic structure of TiO2 showing schematically the effect of the acceptor band on the reduction of the effective band gap required for ionisation.

Solar energy spectrum

4.1021

3.1021

2.1021

1.10

21

∆G(H2O → H2+1/2O2) = 1.23 eV

Number of photons (s–1 m–2 eV–1)

these form local bands that result in the effective reduction of the band gap required for ionisation. The band gap reduction leads to an extension of the edge of the solar energy absorption spectrum from ~3 eV for commercial TiO2–x to ~2 eV for modified TiO2–x, as shown in Fig. 4.15.

1 E1.23

Theoretical energy range

Ai

m

of

pr

oc

es

si

ng

Undoped TiO2 range 2

3 Photon energy (eV)

4

5

4.15 Schematic illustration of solar spectrum (number of photons versus photon energy), showing photon flux available for conversion at energy ∃ energy Ei.

108

4.7


Electrical conductivity within the n-p transition regime

The electrical conductivity and its p(O2) dependence also are sensitive to the presence of aliovalent ions. Figure 4.16 shows data for σ versus p(O2) for undoped and Cr-doped TiO2 according to Carpentier et al. [16]. It can be seen that σ for undoped TiO2 is a linearly decreasing function of p(O2), although Fig. 4.2 shows that an n-p transition is expected at higher p(O2). However, the incorporation of acceptor-type ions (Cr) into TiO2–x results in a shift of the minimum in σ to lower p(O2) levels. The dependence of the electrical conductivity exponent on the p(O2) is a reflection of the disorder models discussed previously. Figure 4.16 shows that the slope of the data for undoped TiO2 is essentially 1/mσ = –1/6, which is in agreement with the model described by eqn 4.22. However, for Cr-doped TiO2, the slope of the n-type regime is essentially 1/mσ = –1/4, which is in agreement with the model described by eqn 4.25. –1.0

Cr-doped TiO2 1,273 K Carpentier et al., 198916

0

log σ (σ in Ω–1 cm–1)

m

1

–1.5

2

3 m

4

–2.0

5

–2.5

–3.0

–4

0. 1. 2. 3. 4. 5.

=– 6

=

–4

Undoped (A =3*10–5) 1 at % Cr 2 at % Cr 3 at % Cr 4 at % Cr 5 at % Cr –3

–2

–1 0 1 2 log p(O2) (p(O2) in Pa)

3

4

5

4.16 Effect of Cr on the electrical conductivity data of Cr-doped TiO2 versus p(O2) at 1,273 K according to Carpentier et al.16

The minimum in the σ at the n-p transition may be used to determine the band gap, which shows the following temperature dependence:35,36 E g = E go – βT

4.46

is determined from the minimal where β = temperature coefficient and value of the electrical conductivity (σmin) measured as a function of p(O2) at the n-p transition:36,37 E go

Defect disorder, transport and photoelectrochemical properties 1  E go   β  σ min = 2 e ( µ n µ p N n N p ) 2 exp   exp  –   2k   2 kT 

109

4.47

The determination of β is complicated, although both β and Eg may be determined from the Jonker analysis described previously. It should be noted that eqn 4.47 is derived for the electrical conductivity component (not the ionic) corresponding to electronic charge carriers. The value of Eg calculated from eqns 4.46 and 4.47 must be considered with caution owing to the complex physical meaning of the parameter σmin, which, at elevated temperatures, may include both electronic and ionic conductivity components.32 Figure 4.17 plots the total conductivity (electronic plus ionic) for undoped single-crystal TiO2 as a function of 1/T, giving a band gap of 3.13 eV.24 Figure 4.18 plots the electronic component only of σmin for the same specimen of TiO2. In this case, the band gap is much higher at 3.4 eV. 1250

T[K] 1200

log σ (σ in Ω–1 m–1)

1300

log σmin (σ in Ω–1 m–1)

–0.5

–1.0

TiO2 (SC)

E0 g

=

3.1

1150

1100

–0.6 1,198 K –0.8

–1.0

σ min 1 2 3 4 5 log p(O2) (p(O2) in Pa)

3±

0.0

–1.5

1e

V

 –E0g  σ min = const. exp    2kT 

0.75

0.80

0.85 1,000/T (K–1)

0.90

4.17 Plot of log σmin versus 1/T for TiO2 single crystal (σmin: minimum of σ measured experimentally).24

4.8

Chemical diffusion in TiO2

The concentration of defects in non-stoichiometric oxides in equilibrium is determined by the parameters describing equilibrium temperature and oxygen activity. When either temperature or p(O2) is changed for an initially equilibrated metal oxide (within a single-phase region), then the system tends to assume

110

Materials for energy conversion devices T[K] 1250

1200

log σmin,el (σ in Ω–1m–1)

–0.5

–1.0 TiO2 (SC) E0 g

–1.5

1150

log σ (σ in Ω–1 m–1)

1300

=

3.4

–1.0 1

1100

/2σ min,el

σn

–1.5

σp

1,198 K

–2.0

1 2 3 4 5 log p(O2) (p(O2) in Pa)

±

0.1

eV

 –E0g  σ min = const. exp    2kT 

–2.0 0.75

0.80

0.85 1,000/T (K–1)

0.90

4.18 Plot of log σmin.el versus 1/T for TiO2 single crystal (σmin,el: minimum of the σ component related to electronic charge carriers).24

a new equilibrium state. This new non-stoichiometry is imposed at the surface almost immediately and then it is propagated into the crystalline bulk in order to establish the new equilibrium. The rate of the propagation is determined by the chemical diffusion of the lattice defects, such as oxygen vacancies and cation vacancies, formed or annihilated as a result of the reaction between gaseous oxygen and the lattice. The process of propagation of lattice species under a chemical potential gradient, which is termed ‘equilibration’, is controlled by lattice diffusion. The latter is termed ‘chemical diffusion’ and the rate constant for this process is termed the ‘chemical diffusion coefficient’ Dchem.18 In some cases the gas/solid equilibration is completely or partially ratecontrolled by a surface reaction. This is the case when the rate of the surface reaction is lower than or equal to the rate of the lattice diffusion. The chemical diffusion coefficient is necessary in order to determine the time required to establish gas/solid equilibrium after the equilibrium is changed to a new condition. Equilibration kinetics may be monitored by measurements of changes in a defect-related property, such as weight, electrical conductivity, or colour. The equilibration kinetics data then may be used to determine the chemical diffusion coefficient using a solution of the diffusion equation that is adequate to the specific initial and boundary conditions.18 The available data, which cover a wide temperature range used during the


111

monitoring of the equilibration kinetics, for the chemical diffusion coefficient in single-crystal TiO2 as a function of reciprocal temperature are shown in Fig. 4.19.38–43 It can be seen that there is substantial scatter of the Dchem in terms of both the absolute values and their temperature dependencies. This is due largely to the use of different experimental procedures, which yield apparently conflicting data. Data for the chemical diffusion coefficient reported by different investigators should be compared only when the physical meaning of the experimental procedures is well defined. The most common issues that must be considered are: • When the Dchem depends on the p(O2), then it must be measured within small p(O2) ranges. • The p(O2) ranges used must be well defined according to the physical meaning of the p(O2), which is oxygen activity rather than oxygen partial pressure estimated from the flow rate of individual gases. In the former case, the the p(O2) represents the oxygen activity and must be determined using an electrochemical gauge. • When the Dchem depends on the non-stoichiometry, it should be determined within small non-stoichiometry ranges. • When the Dchem is determined from the gas/solid equilibration kinetics for two different equilibrium states, then these two equilibrium states must be well defined and not confused with a long-lived transient state.

1600 1400 1200 –7

600

Crosbie, 197841 Baum

–9

Bar

ard, 1 9 7 6 40

Ait-Younes et at., 1984

–10

ban

42

el a nd

Bo

Igu ch nd

–11

gom

olo

v, 1 9 7 0 37

Moser, 197138

ia ji Ya m

–12

a,

log Dchem (Dchem in m2/s)

–8

T(K) 800

1000

19 72

39

–13

Dll D⊥ –14 0.6

0.8

1.0

1.2 1.4 1,000/T (T in K)

1.6

1.8

4.19 Data for the chemical diffusion coefficient as a function of 1/T.37–43

112


In light of the variations illustrated in Fig. 4.19 and the caution that must be taken during assessment, it will be essential to verify the data for the Dchem of TiO2.

4.9

Segregation-induced effects

The surface properties of photosensitive materials, such as TiO2, are of key importance to the feasibility and effectiveness of functional applications for the following reasons:9 • Solar energy is absorbed mainly by the surface layer rather than by the bulk phase. • The reactivity and photo-reactivity of metal oxides is determined by the surface properties, including chemical potential of electrons of the outermost surface layer and surface-active centres required for water adsorption and its decomposition. While the defect disorder in the bulk phase of TiO2 is relatively well known,18 the non-stoichiometry within the surface layer is a function of the position in the surface layer and the composition exhibits segregation-induced concentration gradients such as are shown schematically in Fig. 4.20.23 At present, little is known about this gradient and its impact on the surface versus bulk properties. According to the effect shown in Fig. 4.20, the incorporation of aliovalent ions into the lattice of TiO2 results in their homogeneous distribution within the bulk phase while the surface layer exhibits a segregation-induced concentration gradient. This segregation-induced enrichment is responsible for the formation of strong electric fields within the surface layer. As can be seen in Fig. 4.21, a segregation-induced enrichment factor (surface/bulk concentration ratio) of 100 may result in an electric field

Concentration, x

Surface layer

Bulk phase

TiO2–X x

=

f(T ,p

O 2

,ξ

)

x = f(T, pO2)

Distance from the surface, ξ

4.20 Schematic representation of the effect of segregation on oxygen non-stoichiometry within the bulk and in the surface layer of TiO2.

Defect disorder, transport and photoelectrochemical properties f

L (nm)

1

10

0

10

10

6·104

100

10

1·105

Surface concentration

Cs(3)

113

F (Vcm–1)

3 f=

Cs Cb

Cs (2) 2

Cs (1)

Cb

1 Segregation-enriched surface layer

Bulk phase

L Distance from the surface

4.21 Schematic representation of the effect of the segregationinduced enrichment factor, f, on the electric field, F, within the surface layer.

of ~105 V. In effect, the surface composition and its properties can be considered to be entirely different from those of the bulk phase.23 Little is known of the segregation-induced non-stoichiometry of the surface layer of undoped or doped TiO2 and its impact on the local surface properties and reactivity. Therefore, there is an urgent need to understand the effects of segregation on the surface properties of photosensitive oxide materials. This understanding requires the accumulation of a body of empirical data that will allow the derivation of a theory of segregation in metal oxides and its impact on the functional properties. Such a theory will then allow the following: • prediction of segregation during processing • use of the phenomenon of segregation as a technology to impose desirable surface properties.

4.10

Experimental determination of electrical properties

4.10.1 Electrical conductivity The electrical conductivity may be determined using the well known fourprobe method, a schematic of which is shown in Fig. 4.22. In this method, the external (current) probes are formed of Pt plates attached to both sides of

114


a rectangular specimen. A spring mechanism, located outside the hightemperature zone, is applied to maintain adequate galvanic contact between the plates and the specimen. The internal (voltage) electrodes are formed of two electrodes wrapped around the specimen and welded to Pt connecting wires. The equipment used by the authors of the present work is described more fully elsewhere.22

4.10.2 Thermoelectric power The thermal conductivity assembly also can be used to measure thermoelectric power by incorporation in a high-temperature Seebeck probe (HTSP), which incorporates the measuring facilities for simultaneous determination of the: (i) electrical conductivity using the four-probe method, with 8.4 mm probe distance; (ii) Seebeck voltage; and (iii) oxygen activity. The HTSP incorporates a probe chamber (including a sample holder, microheaters (specific to the HTSP), and thermocouples) and probe head (including electrical outlets and circuit board), as shown in Fig. 4.22. The key elements of the sample holder are two Pt electrodes, which have the following three functions: • They act as thermocouples for the determination of the temperature gradients along the specimen. • They act as current probes for the imposition of the current required for the electrical conductivity measurements. Microheaters Current Pt Electrodes Voltage electrodes

Thermocouple

Thermocouple Specimen

V Voltage circuit

A

Current circuit

4.22 Sample holder of the high temperature Seebeck probe for the determination of both electrical conductivity and thermoelectric power.


115

• They allow the determination of the Seebeck voltage along the imposed temperature gradient (when acting as electrodes in the absence of current imposed by the external circuit). The microheaters are used to impose a temperature gradient along the specimen. The thermovoltage is measured for directionally opposing temperature gradients, ∆T and – ∆T. The thermoelectric power is determined from the slope of approximately twenty to thirty independent measurements of the thermovoltage, which is plotted against the temperature gradient. The thermoelectric power of a specimen S may be determined by adding the absolute thermoelectric power of the Pt electrode SPt to the experimentally determined value of the thermoelectric power Sexp: S = Sexp + SPt

4.48

The absolute value of the thermoelectric power of the Pt electrode in the range (100–2,000 K) was determined by Cusack and Kendall44 to show the relation SPt = –2.63–0.0145T µV/K. Considering the uncertainties in the determination of the temperature gradients ∆T and ∆T (᭙0.1 K) and the Seebeck voltage (᭙1%), the standard deviation of the individual determinations is within ᭙1%. A more comprehensive description of the HTSP is given elsewhere [22].

4.10.3 Work function The high-temperature Kelvin probe (HTKP) used by the authors of the present work is shown schematically in Fig. 4.23. As shown in Fig. 4.24, the main section of the probe is the vibrating capacitor, which is composed of a lower electrode, formed by the specimen, and an upper reference Pt electrode. The distance between the electrodes and the amplitude of vibration are ~0.1 mm and ~0.07 mm, respectively. The vibrating system, which includes a piezoelectric ceramic element on one end and a Pt reference electrode on the other, is suspended on two stainless steel membranes. The lower part of the probe is equipped with a micrometer for controlling the distance between the electrodes. Both lower and upper parts of the probe are equipped with water coolers to prevent these parts from overheating. The work function data are determined from the measured CPD data according to eqn 4.45, assuming that the work function of the Pt reference electrode, which is covered with a thin layer of PtO2,22 is the following function of p(O2):45 ∆φ  1  =1  kT  ∆ log p(O 2 ) 4

4.49

The performance principles of the HTKP used to obtain work function change measurements are reported more fully elsewhere.22

116


Piezoceramic element Gas outlet Membranes

Furnace

Water cooler Ceramic lead tube

Reference electrode Sample

Water cooler

Ceramic support

Gas inlet Distance setting mechanism

4.23 Schematic representation of the high-temperature Kelvin probe for work function measurements at elevated temperatures and under controlled p(O2).22

4.11

Conclusions

The functional properties of oxide materials used for energy conversion are determined by their defect disorder. Therefore, modification of the functional properties may be achieved by modifying the defect disorder. Consequently, much current research aims to establish the relationship between the functional properties and defect disorder. The specific objective of the research is: • Identification of the properties of key importance to performance • Processing of materials with desired properties through modification of the defect disorder. The focus of this research is TiO2 because: • TiO2 exhibits defect disorder that may be modified within a wide range • TiO2 is the main candidate for photoelectrodes for hydrogen generation by the decomposition of water using sunlight.


Vibrating system

117

Platinum reference electrode PtO2 layer R Oxide specimen

Platinum support

T, p(O2)

4.24 Sample holder of the high-temperature Kelvin probe for the determination of work function.

The effect of defect disorder in metal oxides is critical to the characteristics of several functional properties, including: • semiconducting properties • reactivity and photoreactivity • electrochemical and photoelectrochemical properties. The relationships between the defect disorder and the semiconducting properties of TiO2 may be used to modify and engineer these systems so that they exhibit desired performance parameters for electrodes and photoelectrodes to be used in photoelectrochemical cells for hydrogen generation from water using solar energy. The present work considers the defect-related properties of oxide materials that impact upon their performance as photoelectrodes. The effect of paramount importance for the surface reactivity is the effect of segregation, which leads to a substantial difference between the bulk and surface properties. This difference results in the formation of strong electric fields within the surface layer and these may be of the order of 105 V/cm.23,46,47 This field has a substantial impact on the reactivity and photoreactivity. Therefore, the establishment of suitable processing protocols for oxide semiconductors with desired photosensitivities and related properties requires improvement in state of understanding of the effect of segregation on the surface versus bulk defect-related properties. Thus, there is an urgent need to address the following questions:

118


• What is the effect of segregation on the surface versus bulk composition of undoped TiO2 and its solid solutions with aliovalent ions that form donors and acceptors? • What is the relationship between the defect disorder of TiO2 and its semiconducting properties, reactivity, and photoreactivity? • What is the effect of segregation on the surface versus bulk composition and the related electrical properties of the segregation-enriched surface layer? • How can materials of desired properties be processed using the imposition of segregation-induced concentration gradients in a controlled manner?

4.12

Acknowledgements

The authors gratefully acknowledge the finanical support of the Australian Research Council, Rio Tinto Ltd., Brickworks Ltd., Mailmasters Pty. Ltd., Sialon Ceramics Pty. Ltd., and Avtronics (Aust.) Pty. Ltd.

4.13

References

1. Badwal, S.P.S. and Foger, K., Mater. Forum, 21 (1997) 187. 2. Zhuiykov, S. and Nowotny, J., Mater. Forum, 24 (2001) 150. 3. Singhal, S.C., p. 631 in Science and Technology of Zirconia. Edited by Badwal, S.P.S., Bannister, M.J. and Hannink, R.H.J., Technomic Publishing Company, Lancaster, PA, 1993. 4. Yokokawa, H., Key Eng. Mater., 153–154 (1998) 37. 5. Schouler, E.J.L. and Kleitz, M., J. Electrochem. Soc., 134 (1987) 1445–1451. 6. Kopp, A., Nafe, H., Weppner, W., Konturous, P. and Schubert, H., p. 567 in Science and Technology of Zirconia. Edited by Badwal, S.P.S., Bannister, M.J. and Hannink, R.H.J., Technomic Publishing, Lancaster, PA 1993. 7. Mochizuki, S., Sugihara, S., Nakamura, T. and Akimoto, H., Int. J. Ionics, 7 (2001) 310. 8. Fujishima, A. and Honda, K., Nature, 238 (1972) 37. 9. Bak, T., Nowotny, J., Rekas, M. and Sorrell, C.C., Int. J. Hydrogen Energy, 27 (2002) 991. 10. Fujishima, A., Hashimoto, K. and Watanabe, T., Titania as Photocatalyst, BKC Inc., Tokyo, 1997. 11. Sharma, R.K., Bhaynagar, M.C. and Sharma, G.L., Sensors Actuators B, 45 (1997) 209. 12. Bak, T., Nowotny, J., Rekas, M. and Sorrell, C.C., J. Phys. Chem. Solids, 64 (2003) 1043. 13. Bak, T., Nowotny, J., Rekas, M. and Sorrell. C.C., J. Phys. Chem. Solids, 64 (2003) 1057. 14. Bak, T., Nowotny, J., Rekas, M. and Sorrell, C.C., J. Phys. Chem. Solids, 64 (2003) 1069. 15. Bak, T., Burg, T., Kang, S.-J.L., Nowotny, J., Rekas, M., Sheppard, L., Sorrell, C.C., Vance, E.R., Yoshida, Y. and Yamawaki, M., J. Phys. Chem. Solids, 64 (2003) 1089.


119

16. Carpentier, J.L., Lebrun, A. and Perdu, F., J. Phys. Chem. Solids, 50 (1989) 145. 17. Wilke, K. and Brauer, H.D., J. Photochem. Photobiol. A: Chemistry, 121 (1999) 49. 18. Kofstad, P., Nonstoichiometry, Electrical Conductivity and Diffusion in Binary Metal Oxides. John Wiley & Sons, New York, 1972. 19. Kröger, F.A., The Chemistry of Imperfect Crystals, Volume 3. North Holland, Amsterdam, 1974. 20. Nowotny, J., Sorrell, C.C., Bak, T. and Sheppard, L.R., Adv. Solar Energy, in press. 21. Nowotny, J., Radecka, M. and Rekas, M., J. Phys. Chem. Solids., 58 (1997) 927. 22. Nowotny, J., p. 131 in The CRC Handbook of Solid State Electrochemistry. Edited by Gellings, P.J. and Bouwmeester, H.J.M., CRC Press, Boca Raton, FL, 1997. 23. Nowotny, J., p. 79 in Science of Ceramic Interfaces. Edited by Nowotny. J. Elsevier, Amsterdam, 1991. 24. Nowotny, M.K., Semiconducting Properties of TiO2 Single Crystal. Ph.D. Thesis, University of New South Wales, in progress. 25. Tannhauser, D.S., Solid State Comm., 1 (1963) 223. 26. Yahia, J., Phys. Rev., 130 (1963) 1711–1720. 27. Blumenthal, R.N., Coburn, J., Baukus, J. and Hirthe, W.M., J. Phys. Chem. Solids, 27 (1966) 643. 28. Baumard, J.F. and Tani, E., J. Chem. Phys., 67 (1977) 857–868. 29. Balachandran, U. and Eror, N.G., J. Mater. Sci., 23 (1988) 2676–2683. 30. Marucco, J.-F., Gautron, J. and Lemasson, P., J. Phys. Chem. Solids, 42 (1981) 363. 31. Son, J. and Yu, I., Korean, J. Ceram., 2 (1996) 131–139. 32. Nowotny, J., Radecka, M., Rekas, M., Sugihara, S., Vance, E.R. and Weppner, W., Ceram. Int., 24 (1998) 553. 33. Burg, T., Microstructure and Electrical Properties of Polycrystalline TiO2. Ph.D. Thesis, University of New South Wales, in progress. 34. Jonker, G.H., Philips Res. Rep., 23 (1968) 131–152. 35. Frederikse, H.P.R., J. Appl. Phys., 33 (1962) 447. 36. Becker, J.H. and Frederikse, H.P.R., J. Appl. Phys., 32 (1961) 2211. 37. Barbanel, V.I. and Bogomolov, V.N., Sov. Phys. Solid State, 11 (1970) 2160–2168. 38. IMAALS? in Proc. Brit. Ceram. Soc.: Mass Transport in Non-Metallic Solids. Edited by Childs, P.E., Laub, L.W. and Wagner, J.B. Jr. Moser, pp. 29–38. British Ceramic Society, Stoke-on-Trent, 1971. 39. Iguchi, E. and Yajima, K., J. Phys. Soc. Jap., 32 (1972) 1415. 40. Baumard, F., Solid State Comm., 20 (1976) 859. 41. Crosby, G.M., J. Solid State Chem., 25 (1978) 367. 42. Ait-Younes, A., Millot, F. and Gerdanian, P., Solid State Ionics, 12 (1984) 437. 43. Morin, F., Solid State Comm., 58 (1986) 161. 44. Cusack, N. and Kendall, P., Proc. Phys. Soc. London, 72 (1958) 898. 45. Nowotny, J., Rekas, M. and Bak, T., Key Eng. Mater., 153–154 (1998) 211. 46. Nowotny, J., Rekas, M., Sorrell, C.C., Sheppard, L. and Bak, T., Int. J. Hydrogen Energy, 30 (2005) 521. 47. Sheppard, L.R., Semiconducting Properties of Nb-Doped TiO2. Ph.D. Thesis, University of New South Wales, in progress.

5 Polymer electrolyte fuel cells K O T A and N K A M I Y A, Yokohama National University, Japan

5.1

Introduction

The electrochemical energy conversion was originated about 2000 years ago, when the so-called ‘Baghdad Cell’ was invented as a primary cell. Since then, however, no record was found on the development of the electrochemical cell for a long time. After this hiatus, the concept of electricity was introduced by Galvani in 1791, when he made the experiment of dissection of a frog’s leg. Just after the discovery of electricity, the Volta Cell, a concept based on the Baghdad Cell, was invented in 1800. Volta dipped a copper and a zinc plate in the acidic solution and got an electromotive force of about 1V. The improvement of the Volta Cell brought about the invention of the Daniel Cell in 1836, where copper sulfate was added to the cathode electrolyte and zinc sulfate was added to the anode electrolyte. The electromotive force of the Daniel Cell was almost the same as that of the Volta Cell, but the polarization was much lower and much power was released for a longer time. The electrolysis of water was tried at around the same time and the reverse reaction must have been tried. Sir W. Grove produced the first fuel cell experiment in 1839 as shown in Fig. 5.1. He firstly electrolysed water to evolve hydrogen and oxygen in several electrolysers and then the power source was removed from the electrolysers. He showed that the electrolysers reversely generate electricity on the electrodes of each electrolyser. Then all the electrolysers were connected directly and the output power from the electrolysers was given to electrolyse water in another electrolyser. He showed that the electrolysis and generation of electricity takes place reversibly. About 130 years after Sir W. Grove’s experiments, much attention was paid to the development of fuel cells, mostly for limited purposes, i.e., space shuttles or submarines. Nowadays fuel cells are a necessary power source for the space shuttles. Recently the greenhouse effect due to the excess emission of carbon dioxide has become a major concern, and non-polluting fuels and a clean environment are clear targets. Fuel cells in which hydrogen and oxygen react 123

124

Materials for energy conversion devices ox hy

ox

hy

ox

hy

ox

hy

ox

hy

5.1 Grove’s fuel cell experiment.1

electrochemically never release hazardous materials and the theoretical efficiency of energy conversion is much higher than that of conventional thermal engines. Therefore much attention has been focused on their application to vehicles, as power sources for the home, and for mobile appliances.

5.1.1

Electrochemical energy conversion

Water molecules are easily decomposed to hydrogen and oxygen by electrolysis. The magnitude of the energy input at 25°C and 1 atm is 237 kJ mol–1 and in turn the same amount of electric energy will be generated by the fuel cell system. This process is reversible. The energy required for electrolysis is equivalent to mechanical energy and is calculated as Gibbs’ energy change. On the other hand, when hydrogen and oxygen react chemically, e.g., by combustion, the thermal energy of 286 kJ mol–1 can be obtained. However, even if we put the same amount of thermal energy into water, water will never be decomposed to hydrogen and oxygen. This process is irreversible. This thermal energy is considered as the chemical energy and represented by the enthalpy change between the reactant and the product. The relation between these energy functions is given as: ∆G = ∆H – T∆S,

5.1

where, ∆H, ∆G, ∆S are the enthalpy change, the Gibbs energy change, and the entropy change, respectively. Unless the entropy change is negative, the amount of obtainable electricity is less than that of the chemical energy and generally this rule is true. The Gibbs’ energy change at 25°C corresponds to 1.23 V as the electromotive force of the fuel cell at open circuit. When the circuit is closed, the cell

Polymer electrolyte fuel cells

125

voltage decreases due to the several resistances and polarizations. As a result the output electric energy decreases and the rest changes to heat.

5.1.2

Characteristics of fuel cells

As indicated above, the theoretical output voltage is 1.23 V at 25°C. However, it decreases with increased output current. Figure 5.2 shows the currentvoltage characteristics of a polymer electrolyte fuel cell (PEFC). If there is no internal resistance, the output voltage will always be the same as the open circuit voltage regardless of the output current, and the efficiency of the energy conversion will be kept constant, i.e., ∆G°/∆H° (= 0.83 in H2-O2 system). However, due to several resistances, the cell voltage decreases. The voltage loss is caused by the crossover of reactants through the electrolyte, cathode reaction resistance, Rc, anode reaction resistance, Ra, membrane resistance Rs, and mass transfer resistance at the higher current density. Among these, the largest resistant in the PEFC is due to the slow electron transfer in the cathode reaction. The resistance of the electrolyte membrane is also a serious problem, especially at high current density. H2 + 1/2O2

1.48

T∆S0 (17%)

Exothermic energy

∆H0 –286 kJmol–1

∆G (83%) 0

Cell voltage (V)

1.23

Voltage loss due to crossover

1.0

Cathode reaction i . Rc Anode reaction i . RA Membrane i . RS

0.5 Cell voltage

Mass transfer

H 2O 0.0

0

5.2 Voltage-current characteristics of a fuel cell.

5.1.3

Types of fuel cell

Many types of fuel cell have been investigated. According to the characteristics of the electrolytes, they are divided into roughly five types: alkaline (AFC), phosphoric (PAFC), molten carbonate (MCFC), solid oxide (SOFC), and polymer electrolyte (PEFC). Much attention has been devoted to PEFC including direct methanol fuel cell (DMFC) recently. The features of such fuel cells are listed in Table 5.1.

126


Table 5.1 Types of fuel cell and their features Type of FC

AFC

PEFC

PAFC

MCFC

SOFC

Operating temperature (°C)

50–200

60–100

170–200

600–700

900–1000

Fuel

H2

H2 (CO< 50ppm)

H2 (CO40%). The important component materials are the anode, cathode, electrolyte, and interconnect, which are mainly electronic conductors, oxygen ion (O2–) conducting ceramics, and metals. An important aspect of SOFCs is their interfaces, where different components come into contact and electrochemical reactions take place actively. Thus, it is important to clarify the properties of the component materials as well as the interfaces. The following topics are presented in relation to SOFC materials: (i) basics of SOFCs, (ii) component materials for SOFCs, (iii) operational testing and analysis for reactions at the gas/electrode/electrolyte interfaces, and (iv) current status and future developments of SOFCs.

6.2

Basics of SOFCs

Due to their high operational temperatures, the components of SOFCs are comprised mainly of refractory ceramics and metals. Compared to other types of fuel cells, SOFCs consist completely of solid-state materials for high-temperature operation (873–1273 K). Figure 6.1 shows a schematic diagram of a SOFC under operation. SOFCs consist of electrodes (anode and cathode) separated by a solid electrolyte, such as an oxygen ion conductor). At the cathode/electrolyte interfaces, oxygen molecules are reduced to oxygen ions (O2–) in order to accept electrons and the ions are transported to across the cell due to the chemical potential difference of oxygen. The conducted oxide ions react with fuel gases, such as H2 or CH4, at the anode/electrolyte interfaces to form H2O and CO2, respectively. Electricity can be extracted from the chemical energy from the oxidation of fuel gases. The Gibbs free 140

Solid oxide fuel cells Air (O2)

Cathode (air electrode)

Anode (fuel electrode)

Electrolyte

Fuel (H2, CH4, etc.)

O2–

e–

141

e–

External load H2O, CO2

6.1 Schematic diagram for solid oxide fuel cells.

energy change for the reaction (∆G) can be converted directly to electricity according to the Nernst relation as follows: E = –∆G/nF

6.1

where E = terminal voltage, ∆G = Gibbs free energy change for the oxidation reaction of fuel gases, n = number of electrons, and F = Faraday constant. In real SOFC operation, the oxygen chemical potential difference between the oxidant and fuel gases determines the terminal voltage. The extraction of current alters the terminal voltage from the open circuit condition.

6.3

Component materials for SOFCs

Up to now, significant progress in SOFC technologies during the past decade has been made, in particular in the fabrication of systems of several to hundreds of kilowatts. Concerning developments in materials, several candidates for each component have been examined. Table 6.1 summarises the materials that have been examined for SOFC applications. It can be seen that most of the components are oxide ceramics and metals. Each component must meet the requirements of each function, these being: • The cathode material should accept electrons and reduce oxygen molecules to oxygen ions (O2–).

142


Table 6.1 Component materials that have been examined for solid oxide fuel cells Components

Function

Typical materials

Form for utilisation

Cathode

Oxygen adsorption, reduction

Doped LaMnO3 Doped LaCoO3 Doped SmCoO3

Porous

Electrolyte

Conduct oxide ions, Transference number should be close to 1

Y2O3 stabilised ZrO2 Sc2O3-ZrO2 Doped LaGaO3 Doped CeO2

Dense film

Anode

Electrochemical oxidation of fuel gases, such as H2 and CH4

Ni-YSZ cermet Ru-YSZ cermet

Porous

Interconnect

Separate fuel and oxidant gases and connect single cells electrically in series

Doped LaCrO3 Ferritic alloys (Fe-Cr alloy)

Dense film

• The electrolyte should conduct oxygen ions according to the chemical potential gradient of oxygen. • The anode should oxidise the fuel electrochemically and release electrons. • The interconnect should conduct electrons and separate the fuel and oxidant gases. The important points to be considered for the application of these materials are: (i) temperature, (ii) oxygen partial pressure, and (iii) chemical compatibility with the other materials. In the following text, the physical and chemical properties, chemical stability, and compatibility with the other components are introduced for each component.

6.3.1

Electrolyte materials

Electrolyte materials should conduct oxygen ions (O2–) from the air electrode (cathode) side to the fuel electrode (anode) side due to the chemical potential gradient of oxygen. Consequently, the electrolyte material should possess the following attributes: • high ionic conductivity compared to the electronic conductivity • chemical stability under fuel and oxidant conditions • chemical compatibility with the other components, such as the cathode and anode.

Solid oxide fuel cells

143

The best known candidate electrolyte materials are stabilised ZrO2, doped CeO2, and LaGaO3-based oxygen ion conductors, which have been widely investigated not only for SOFCs but also for gas sensors. Figure 6.2 shows the temperature dependence of several candidate materials for SOFC electrolytes that have been examined. All electrolytes show an increase in oxygen ionic conductivity with increasing temperature. Therefore, it is to be expected that oxygen ion conductors will involve temperature-activated processes. 1000

Temperature (°C) 800 700 600

900

0 ZrO2 – 7.5 mol% Sc2O3 Bi 2O 3 –2 5

Ce

log (σ/Scm–1)

–1

m

ol

r0

.2 G

Y2 O

Ce

%

.8 M

g0

.11

Zr O

2–

15

m

15

ol

%

2 –1

0m ol%

O

8

m

ol

%

0.8

Y

Sr

0.2

2O 3

m

Y

ol

%

Ca

SrC

O

Gd

M

g

0.2

O

3

.95 Y

b0

2O 3

0.9

Ga

0.8

e0

85 O 3

2O 3

La

O

2–

5 Co 0.0

O

Y

–2

0.8

a0

2O 3

2–

–4 0.7

l%

O

Zr

–3

0.8 S

mo

3

2 –5

Th

La

1.0 1.1 T–1/ kK–1

.05 O 3

1.2

1.3

6.2 Temperature dependence of several oxide ionic electrolytes (reproduced from Refs. 4, 5, 8, 9 and 27).

At present, fully stabilised ZrO2-based materials are the most popular electrolytes for SOFCs. They have a cubic fluorite-type structure, which contains relatively large interstices in the lattice. ZrO2 in pure form exhibits three polymorphs: monoclinic (293–1443 K), tetragonal (1443–2643 K), and cubic (2643–2953 K). To stabilise the crystal structure in the cubic form, aliovalent oxides are added to the ZrO2. The most common stabilising oxides for ZrO2 are CaO, Y2O3, MgO, Sc2O3, and certain rare earth oxides. For example, the solubility of Y2O3 in ZrO2 to stabilise the cubic form is as shown in Fig. 6.3. However, there are some discrepancies concerning the

144


3000

0

2

mol% Y2O3 4 6

8

10

Liquid

Cubic

Temperature/°C

2000

Tet.

T+C

1000

T0c ′–t ′

Mono T0t–m′

0 0

10 ZrO2 mol% YO1.5

20

6.3 Phase diagram of a ZrO2-Y2O3 system in the low Y2O3 region (from Ref. 1.).

ZrO2-Y2O3 phase diagram in the literature.1 It can be seen from the diagram that the dissolution of Y2O3 in ZrO2 reduces the temperature of tetragonalmonoclinic transformation, with the transformation temperature decreasing with increasing Y2O3 content. In order to stabilise cubic ZrO2 down to ~1273 K (1000 °C), the minimal required amount of YO1.5 is ~15 mol%, which is equivalent to ~7.5 mol% Y2O3. The conduction of oxygen ions occurs via oxygen vacancies in the lattice. Doping with low-valence rare earths, such as Ca2+ and Y3+, increases the conductivity owing to the increase in oxygen vacancies in the lattice according to the following defect reaction: ZrO 2 Y2 O 3  → 2YZr ′ + 3O ox + Vo⋅⋅

6.2

where a lettered subscript denotes a lattice site and V denotes a vacancy (Kröger-Vink notation). Figure 6.4 shows the effect of various dopant concentrations at 1080 K on the change in ionic conductivity of cubic stabilised ZrO2.2 It can be seen that the conductivity with each dopant shows a maximum at a certain concentration. The increase in conductivity corresponds to the increase in the concentration of oxygen vacancies in the lattice, according to eqn 6.2. At the higher dopant


145

10–1 Yb2O3 4 Gd2O3

σ/Scm–1 (ionic conductivity)

2 10–2

4 Nd2O3 2 CaO

Y2 O3

10–3

4 2 10–4

4

6

8 10 12 14 16 18 C/mol% (Mole concentration of M2O3 or Mo in ZrO2)

20

6.4 Ionic conductivity change with various different dopant concentration (from Ref. 2).

concentrations, the conductivity decreases with dopant concentration. This is believed to be due to defect ordering, vacancy clustering, and/or electrostatic interaction. The electrical conductivity of Y2O3-stabilised ZrO2 (YSZ) exhibits a maximal value for 8–10 mol% Y2O3, although the conductivity is only ~0.1 S. cm–1 at 1273 K. Thus, in order to increase the electrical conductivity for use in SOFCs, the fabrication of a thin electrolyte has been considered. Several methods for fabricating thin electrolyte films on porous electrode substrates have been investigated, including sputtering, plasma spraying, electrochemical vapour deposition (EVD), electrophoretic deposition (EPD), and slurry coating method. YSZ shows a relatively high oxygen ion transference number (ti) over a wide oxygen partial pressure range (10–23–10–1 bar) and a wide temperature range (873–1273 K). Figure 6.5 shows the oxygen ionic and electronic conductivities of some solid electrolytes as a function of oxygen partial pressure at 1073 K. These data were obtained from literature references and measurements by the authors.3–7 The electronic conductivity (electron and hole) is proportional to the 1/4 power and –1/4 powers of the oxygen partial

146

Materials for energy conversion devices 800 °C 2 LSGM 0

GDC

Ion

8 YSZ

log (σ/Scm–1)

–2 GDC electron

LSGM hole GDC hole

–4 Electron –6

–8

Hole

Electron

–10 –20

8YSZ

–15

–10

–5 0 log (p(O2)/Pa)

5

10

6.5 Ionic and electronic (electron and hole) conductivities for some solid electrolytes as a function of oxygen partial pressure (1073 K).

pressure for hole and electron conductivity, respectively. The electronic conductivity is two to three orders of magnitude lower than that of the ionic conductivity. Therefore, the transference number is >0.99 over a wide range of oxygen partial pressure, so this makes it one of the most promising electrolyte materials for SOFCs. However, due to the relatively high activation energy for oxygen ion conduction (~0.7–1.0 eV), as shown in Fig. 6.2,3,4 the oxide ionic conductivity of YSZ decreases drastically with a reduction in the operation temperature. Thus, for the medium-temperature operation, another material of high ionic conductivity is needed. Doped CeO2-based materials show higher electrical conductivities than does YSZ. The electrical conductivity of Gd0.2Ce0.8O1.9 (GDC) is ~0.1 S.cm–1 at 1073 K in air, so it has been considered as a candidate for the electrolyte in intermediate-temperature SOFCs (773–1073 K).8 However, the electronic conductivity of GDC under low oxygen partial pressures is relatively high and the transference number of the oxygen ion is significantly low (ti < 0.5) under reducing atmospheres. This can affect the performance of SOFCs owing to the electrochemical leakage of oxygen, which may reduce the efficiencies of SOFCs. During the past few decades, a new electrolyte based on the perovskite structure has been investigated.9–12 Doubly doped LaGaO3 (A-site and Bsite) shows a relatively high ionic conductivity. For example, the electrical conductivity of La0.9Sr0.1Ga0.8Mg0.2O3-d is ~0.1 S.cm–1 at 1073 K, which is


147

comparable to that of Gd0.2Ce0.8O1.9, as shown in Fig. 6.2. The transference number of oxygen ions in La0.9Sr0.1Ga0.8Mg0.2O3-d at 1073 K is relatively high because the electronic conductivity is negligibly small, as indicated in Fig. 6.5. LaGaO3-based electrolytes show relatively high performance although their chemical stability under reducing atmospheres must be considered, especially in light of Ga volatilisation.13

6.3.2

Cathode materials

Cathode materials should conduct electrons and reduce oxygen to oxygen ions by consuming the electron in the following reaction:

1 O (g) + 2e – = O 2– (s) 2 2

6.3

Thus, cathode materials are deposited on the electrolyte in the porous structure. The cathode materials should have the following properties in order to operate with suitable performance: • high electronic conductivity • chemical stability and compatibility during fabrication and operation at high temperatures • thermal expansion characteristics that match those of the other components • sufficient porosity to allow the gaseous oxygen molecules to permeate the cathode/electrolyte interfaces. To satisfy the above requirements, several kinds of perovskite-based oxide materials have been investigated. The candidate materials are LaMnO3 based, LaCoO3 based, LaFeO3 based, and SmCoO3 based perovskite materials because of their high electronic conductivity as well as catalytic activity for oxygen reduction. LaMnO3-based perovskites exhibit intrinsic p-type conductivity due to changes in the Mn valence. Doping of low valence cations, such as Sr2+ Ca2+ to La3+ sites will enhance the electronic conductivity higher than 10 Scm–1 at 700 °C. When La3+ sites are replaced by Sr2+ ions, electronic holes are formed on the Mn3+ sites to maintain the electroneutrality and this leads to an increase in the electrical conductivity: 3+ 3+ (1 – x)LaMnO3 + xSrO → (La 1–x Srx2+ )(Mn 1–x Mn x4+ ) O 3

6.4

A small polaron-hopping mechanism has been considered to explain the electrical conduction in light of the temperature dependence and thermal conductivity.14 That is, the conductivity increases with increasing temperature, where the mobility is relatively low. The most important features at the cathode/electrolyte interfaces are the chemical reactions during fabrication and operation. Many investigators have

148


reported the formation of a La2Zr2O7 insulating layer at the LaMnO3-based cathode/Y2O3-stabilised ZrO2 electrolyte interfaces above 1273 K.15, 16 In order to avoid the formation of such insulating layers, control of the chemical composition of LaMnO3, including A-site-deficient compositions [(La, Sr)0.95MnO3], has been investigated. Thermodynamic calculations have been useful in suggesting optimal chemical compositions for A-site-deficient LaMnO3,17–20 as shown in Fig. 6.6. One of the interesting features of the LaMnO3-based perovskites is their oxygen non-stoichiometry, where, in addition to the oxygen-deficient region, there is an oxygen-excess region, as indicated in Fig. 6.7.21 The oxygen-excess region appears with low-Sr-doped concentrations at high oxygen partial pressures. It is believed that the oxygen excess is due to La deficiency and Mn vacancies in La1-xSrxMnO3+y. With deceasing oxygen partial pressure, the oxygen content shows a plateau in the oxygen-deficient region. In the oxygen-deficient region, the valence of Mn will be decreased from 3+ to 2+, which eventually reduces the electronic conductivity under reducing atmospheres. 10

8

M

L

0

0.

94

1

90

85

82

5

0.

YSZ

0.

0.015 0.01 0.005 0.001

0.01 0.02 0.05 0.10 0.114

2

=

3

ay

La2Zr2O7

0.001

4

y

La dissolution

log aLa/aZr

6

0.

nO

A-site deficiency

La2O3

Mn3O4

Mn dissoution –2 10

12

14

16 18 log aMn/aZr

20

22

6.6 Chemical potential diagram for the La-Mn-Zr-O system (data from Ref. 20).

For intermediate-temperature SOFCs, LaCoO3-, SmCoO3-, and (La, Sr)(Co, Fe)O3-based materials have been investigated as cathodes with CeO2-based and LaGaO3 electrolytes.22–26 The electronic conductivity of LaCoO3-based perovskites is higher than that of LaMnO3 at 973 K, so the former is considered to be appropriate for intermediate-temperature SOFCs. In addition to their high electronic conductivities, LaCoO3-based oxides show relatively high


149

3.2 3.1 LaMnO3+4

3+d

3 873K 973K 1073K 1173K 1273K 1273K(Kuo) decomp.

2.9 2.8 2.7 –30

–25

3.2

–20 –15 –10 log [P(O2)/105Pa] (a)

–5

0

3.1

3+d

3

La0.1Sr0.5MnO3+4 873K 973K 1073K 1173K 1273K 1273K(Kuo) decomp.

2.9 2.8 2.7 –30

–25

–20 –15 –10 log [P(O2)/105Pa] (b)

–5

0

3.1 3

3+d

2.9

La0.1Sr0.5MnO3+4 873K 973K 1073K 1173K 1273K decomp.

2.8 2.7 2.6 –30

–25

–20 –15 –10 log [P(O2)/105Pa] (c)

–5

0

6.7 Oxygen nonstoichiometry of LaMnO3-based perovskite as a function of oxygen partial pressure (data from Ref. 21).

oxygen ionic conductivities. The oxygen ionic conductivity of LaCoO3 is estimated to be 0.05 S.cm–1 at 1073 K, which is comparable to the conductivity of YSZ. Thus, electronic and mixed ionic conductions occur in LaCoO3 based materials. However, due to their reactivity with YSZ and consequent

150


formation of La2Zr2O7 at the interfaces, LaCoO3-based oxides are not suitable for cathodes when using a YSZ electrolyte.

6.3.3

Anode materials

Anode materials usually are a mixture of metals and oxide ceramics, often referred to as ‘cermets’. Anode materials should be stable under reducing atmospheres and they must have sufficient porosity to allow the diffusion of fuel gases and to transport the product gases. The anode reaction can be expressed as follows when H2 is utilised as the fuel: O2–(s) + H2(g) = H2O(g) + 2e–

6.5

Under reducing atmospheres, some metals can be utilised together with YSZ powders in porous structures. The cermet structure has been used for the anode owing to the following reasons: (i) the metal acts as a catalyst for fuel oxidation and for electronic conduction and (ii) the oxide retains the porous (skeletal) structure of the cermet and it supplies the oxygen ions for diffusion through the electrolyte. Nickel (Ni) is one of the most promising materials for anode cermets owing to its relatively high catalytic activity for fuel oxidation and reasonable cost. In the cermet structure, ~40 vol% Ni is required for adequate electronic conductivity, as shown in Fig. 6.8.27 The percolation 104

Conductivity, Ω–1 cm–1

10

Lower surface area YSZ

3

102

Higher surface area YSZ

101

100

10–1

10–2 0

10

20

30 40 Vol% nickel

50

60

6.8 Electronic conductivity of Ni-YSZ cermet anode as a function Ni volume content (data from Ref. 27).


151

theory has been adopted to explain the drastic change in the electronic conduction, where the electronic paths (Ni-to-Ni) are connected continuously with >40 vol% Ni. The minimal value for necessary Ni concentration varies with particle size of both Ni and YSZ and their dispersion states in the cermet. There have been many studies aiming to optimise the Ni-YSZ cermet structure. For example, Fukui et al.28 prepared Ni particles surrounded by small YSZ particles and this morphology showed relatively high performance with a YSZ electrolyte, which had an area specific resistance < 0.1 Ω.cm2 at 1273 K. Long-term stability also is important for practical Ni-YSZ cermet anodes. Itoh et al.29 fabricated an Ni-YSZ cermet structure to increase longterm stability, as shown in Fig. 6.9. Two different sizes of YSZ particles were dispersed, where the large YSZ grains retained the skeletal structure while the small YSZ grains suppressed sintering of the Ni particles. This sophisticated Ni-YSZ cermet structure showed relatively stable performance in the range of 1000 hours in addition to stability against redox cycles.

SEl

Zr

Ni

6.9 Microstructures of Ni-YSZ cermet anode prepared for long-term stability (after long-term stability test) (data from Ref. 29).

Another important aspect of anode materials is in the catalytic activity of CH4-reforming. Since the operation temperature is high in SOFCs (>873 K), it is possible to reform hydrocarbon fuels internally. Thus, Ni-YSZ cermets can be used as catalysts for the internal steam reforming reaction: CH4 + H2O = CO + 3H2

(steam reforming)

6.6

CO + H2O = CO2 + H2

(shift reaction)

6.7

Since the steam reforming reaction is endothermic, heat must be supplied from outside the system. However, the fuel cell reaction is exothermic and so an optimal thermal balance can increase efficiency of the SOFC system. Some precious metals (Ru, Pd, Pt, etc.), together with Ni, are catalytically active for the preceding reactions.30 At 1273 K, under high-level-steam conditions, these reactions proceed rapidly enough to produce sufficient H2

152


for the electrochemical reactions. In the stacks under development (10–250 kW class stacks), internal steam reforming was demonstrated successfully and relatively good performances have been reported. In natural gas systems, the steam/carbon (S/C) ratio usually is set >2 in order to avoid carbon deposition at the Ni-YSZ cermet anode surface. However, under high-level-steam conditions, Ni can be oxidised, which drastically decreases the electronic conductivity. The degradation of Ni-YSZ anode performance is one of the technological issues of the greatest interest. The main reasons for this degradation are sintering of Ni particles in the cermet, which disconnects the electronic pathway, and carbon deposition on the Ni, which may alter the microstructure. Another method to utilise hydrocarbon fuels is partial oxidation (POX): CH4 + 1/2O2 = CO + 2H2

6.8

Since large amounts of water in the SOFC system can reduce the performance due to endothermic reaction, POX is one of the more promising methods for fuel processing, although the energy conversion efficiency is lower than that of the steam reforming process. Some SOFC developers, such as SulzerHexis AG and Versa Power Systems (formerly Global Thermoelectric), have adopted the POX method to utilise natural gas in SOFCs. However, in order to promote the preceding reaction, an appropriate catalyist is required.

6.3.4

Interconnect materials

Interconnect materials should separate the oxidant and fuel gases and they should be electrically conductive under operational conditions at high temperatures. The requirements of interconnects are as follows: • chemical and physical stability under both oxidising and reducing atmospheres and at operating temperatures • gas tightness so as to separate fuel and oxidant gases without leakage of either gas • high electronic conductivity so as to connect the single cells electrically in series • thermal expansion match with other component materials • almost no reactivity with the cathode or anode materials, which would result in the formation of insulating layers at the interfaces. In order to satisfy the above requirements, only a few candidate materials are available, these currently being doped LaCrO3 and some metallic alloys. Up to now, doped LaCrO3 has emerged as the only viable candidate for oxide interconnects. It is very difficult to sinter, which must be done at high temperatures and under reducing atmospheres so as to achieve sufficiently high densities. Since ~1990, densification has been enhanced through liquid-


153

phase formation associated with the use of (La, Ca)CrO3. A Ca-rich liquid forms along the grain boundaries in air >1473 K, as shown in Fig. 6.10,31 and it is associated with the formation of low-temperature-melting CaCrO4 around the grain boundaries. > 1273 K

1273 K

CaCrO4 La2CrO6

Cam(CrO4)n [m > n] 1573 K

Ca rich region

> 1573 K

Cam(CrO4)n (M > n)

1873 K

CaO

6.10 Schematic diagram for the liquid phase assisted sintering in Ca-rich (La,Ca)Cr3 (data from Ref. 31).

Doped LaCrO3 is a p-type conductor and its electronic conductivity increases with the concentration of low-valence cations, such as Sr2+ or Ca2+ in the La3+ sites according to the following reaction: LaCrO3 + xAEO + 0.5xO2 → La1–xAExCr3+1–xCr4+xO3 + 0.5xLa2O3

6.9

Electron holes will be formed on the Cr4+ sites and the conduction mechanism consists of a small polaron hopping process. The electronic conductivity is ~10–100 S.cm–1 at 1273 K in air. The electronic conductivity decreases with decreasing oxygen partial pressure owing to the decrease in the Cr4+ concentration under reducing atmospheres as follows: 3+ La 1–x AE x Cr1–x Crx4+ O 3 → La1–xAExCr3+ O x + x O 2 3– 4 2

6.10

In LaCrO3-based ceramics, a small amount of oxygen can permeate through the dense LaCrO3 interconnect via oxygen vacancies in the material. The permeated oxygen flux has been estimated using electrochemical concentration

154


cells and the isotope labelling method.32–34 The oxygen permeation flux was 873 K is the surface diffusion of adsorbed O2 on the Pt surface to and from the TPB. At 2), these reactions proceed without any carbon deposition. However, if insufficient steam is present, carbon may be deposited according to the following reactions: 2CO = CO2 + C

6.20

CH4 = 2H2 + C

6.21

The carbon deposition should be minimised because the carbon deposited on Ni will degrade the anode performance by blocking the active sites. Also, the deposited carbon can block gas flow by physical occlusion. In order to reduce or prevent carbon deposition, other metal catalysts, such as Fe, Ru, and Cu, have been examined in cermet anodes. Further, the oxides in cermet anodes also have an effect on the carbon deposition. It has been reported that the addition of CeO2-based oxides can enhance the electrode reaction and the internal reforming.59–65 Recently, hydrocarbons of higher carbon numbers have been investigated for direct oxidation in SOFCs. Liquid hydrocarbons, such as decane, toluene, and diesel, have been tested in SOFCs using Cu-based cermet anodes, as shown in (Fig. 6.16).66–69 Cu-YSZ cermet anodes were relatively stable and showed good performance for 12 hours. The measured power density was >0.1 W.cm–2 at 973 K with direct feeding of liquid fuels. However, Cu is not highly active in the reforming of hydrocarbons, although it does prevent carbon deposition in the cermet anode. The addition of CeO2-based materials has also been shown to be effective in the oxidation of liquid fuels.

Materials for energy conversion devices 0.6 Decane

0.6

0.4

0.4

0.2

0.2

0.0

0.0 Toluene

Voltage (V)

0.6

0.6

0.4

0.4

0.2

0.2

0.0

0.0

Current density (A/cm2)

162

Diesel

0.6

0.6

0.4

0.4

0.2

0.2

0.0 0

2

4

6 8 Time (hr)

10

0.0 12

6.16 Plots of cell potential and current density as a function of time for n-decane, toluene and diesel fuel. Each of the fuels was fed to the cell with N2 at a concentration of 40 wt% hydrocarbon (data from Ref. 67).

In recent decades, the porous microstructures of Ni-YSZ cermets have been controlled in order to increase the anodic reaction rates. To these materials were dispersed CeO2 particles and nano-sized Ru particles (3 wt%)64. These cermets showed excellent performance at medium temperatures (at an overpotential of 0.1 V, at 0.5 A.cm–2 at 1073 K). Since one of the technologically important issues in these cermets is their long-term stability, the sintering of the Ni particles during long-term operation at high temperatures must be prevented or retarded. In order to avoid such sintering, microstructural control has been examined by some authors.29

6.5

Current status and future development of SOFCs

6.5.1

Current status

Significant recent developments have been reported for both SOFC materials and systems. Several developers have formulated different stack designs using different component materials. Table 6.2 summarises the cell designs

Materials

LaMnO3-based cathode support tube, YSZ film with LaCrO3 stripe interconnect

LaMnO3-based cathode support tube, YSZ film with LaCrO3 stripe interconnect

Porous ceramic supported tube, LaMnO3, YSZ and CaTiO3 base interconnect

YSZ self-support plate with LaMnO3 cathode, Ni-YSZ anode, LaCrO3

YSZ self-support plate with LaMnO3 cathode, Ni-YSZ anode, metallic interconnect

Doped LaGaO3 self support electrolyte, (Sm,Sr)CoO3 cathode, Ni-CeO2 based anode, metallic interconnect

Ni-YSZ anode-supported, LaMnO3-cathode, YSZ based electrolyte, metallic interconnects

Flat-tube design, operation at 950 °C, with YSZ, Ni-YSZ, LaMnO3

Design type

Tubular

Tubular

Tubular

Planar

Planar

Planar

Planar

Planar

Slurry coating and firing

Compressive seal, operation around 700 °C, slurry coating and firing

Disk type seal-less design, slurry coating and firing

Disk type, seal-less design, slurry coating and firing

MOLB-type, slurry coating and firing

Slurry coating and firing

Extrusion of porous LaMnO3 support tube, slurry coating of YSZ and LaCrO3 and firing

EVD process

Fabrication method

Rolls-Royce Fuel Cell Systems Ltd.

Versa Power Systems (formerly Global Thermoelectric)

1.6 W cm–2 at 1023 K, 2 kW system testing

SOFC-gas turbine hybrid system

Mitsubishi Materials Corporation and Kansai Electric Power Corporation

Sulzer-Hexsis

1 kW class has been achieved, 43% LHV

1 kW class stack for residential application

Mitsubishi Heavy Industry (Kobe) & Chubu Electric Power Company

0.24 W cm–2 at 1273 K, 10 kW stack testing

TOTO Corporation Ltd.

0.18 W cm–2 at 1273 K, 15 kW testing

Mitsubishi Heavy Industry (Nagasaki) & Electric Power Development Company

Siemens-Westinghouse Power Corporation

Higher than 0.2 W cm–2 at 1273 K, 250 kW field test

High pressure 10 kW stack

Developer

Reported performance

Table 6.2 Cell design and materials for SOFC stacks developed by several companies

164


and component materials for several developers. There are two general types of SOFC designs, these being tubular- and planer-types. The tubular-type has relatively high mechanical strength due to its cell structure, although a relatively long electrical path is necessitated for one unit cell. The planertype has a relatively simple configuration, although this comes at the cost of the requirement of a large area for sealing. Among the developers, Siemens-Westinghouse Power Corporation has been a leading company in the development of high-temperature large-scale stack systems. Figure 6.17 illustrates the Siemens-Westinghouse tubular cell design, which features a LaMnO3-based porous support tube, a LaCrO3based interconnect (9 mm wide and 85 µm thickness strip), and YSZ electrolyte (40 µm thickness), the latter two of which are fabricated by EVD.70 This tubular cell shows a relatively good mechanical and electrical performance using a natural gas fuel at 1273 K. In operation: (i) an oxidant (air or oxygen) is introduced through a ceramic injector tube positioned inside the cell; (ii) the oxidant is discharged near the closed end of the cell and flows through an annular space formed by the cell and a coaxial injector tube; and (iii) fuel flows on the outside of the cell from the closed end and is oxidised electrochemically while flowing to the open end of the cell, so generating electricity. Stacks of this type have been operated more than 20,000 hours without significant degradation. The stacks perform satisfactorily under a variety of operating conditions with less than 0.1% performance degradation per 1000 hours. Some typical performances are shown in Fig. 6.18. The voltage-current characteristics show more than 0.2 W.cm–2 at 1273 K (1000 °C), which is considered to be relatively good performance for the SOFC stacks.70

Interconnection

Electrolyte Air electrode

Fuel flow

Air flow Fuel electrode

6.17 Schematic diagram for Siemens-Westinghouse type tubular SOFC (data from Ref 70).


165

1.000 800 °C 900 °C 1000 °C

Voltage, volt

0.800

0.600

0.400 Fuel: 89% H2 + 11% H2O (85% Fuel utilisation)

0.200

Oxidant: air (4 stoichs) 0.000 0

100

200 300 400 Current density (mA/cm2)

500

600

6.18 Typical performance data for Siemens-Westinghouse SOFC single cell (data from ref. 70).

Since EVD involves relatively high fabrication costs, some developers have attempted to reduce these costs by using alternative fabrication methods. TOTO Company Ltd. (Japan) has succeeded in fabricating a tubular cell by a wet method involving a ceramic slurry. A support tube of LaMnO3 cathode is made by extrusion. After pre-firing the support tube, dense LaCrO3 interconnects and thin YSZ electrolyte films are coated with the ceramic slurry and these forms are co-fired. This method yields a low-cost, highperformance, and stable tubular stack. The stack demonstrates >0.18 W.cm–2 at 1273 K with internal natural gas reforming and C[PO(OH)2]2 show higher conductivity due to higher carrier concentration, though the latter’s conductivity depends considerably on humidity.24,25

7.5

Lithium ion conductors

Fast lithium ion conductors usable at ambient temperature would be especially useful, as they would enable the development of a high performance solid-

Fast ionic conductors

183

1.E–01

σT/KS · cm–1

1.E–02

1.E–03

1.E–04

1.E–05 2.5

3.0

3.5 1000/T(K)

4.0

4.5

7.9 Ionic conductivity of hydroxyfullerene C60(OH)12 as a function of temperature.

state lithium battery with improved stability and safety over conventional ones. Thus a variety of materials has so far been investigated as a Li+ conductor. To date, however, no fast conductor to realize a solid-state battery for highrate use has been found. One of the few Li ion conductors which have been put into practice is a composite of LiI and a dielectric substance like Al2O3 usually obtained by heating a well-blended mixture of anhydrous LiI and activated alumina with high specific surface areas at temperatures around 600°C.26, 27 The conductivity of LiI itself is relatively low (~5 × 10–7 S cm–1, 25°C). It is enhanced by two or three orders of magnitude with incorporation of dielectric particles, because defects ( VLi′ and or Li ⋅i ) are generated by the space charge effect at the interface between an ionic conductor (LiI) and an insulator (Al2O3). Such composite solid electrolytes are used in lithium batteries for a cardiac pacemaker, which requires very high reliability, while the necessary current is sufficiently small. In applications for high-rate use batteries, further enhancement of conductivity, for example through nano-structure designing of composites, is essential. Composites based on zeolite matrices (as an insulator phase) having a bicontinuous nano-structure are being investigated.28 As far as only conductivity is concerned, Li3N, a layered compound built up of Li2N layers interspersed with the rest of the Li, is a good Li solid electrolyte (~10–3 S cm–1 at 25°C).29 However, there is a disadvantage in that its decomposition voltage is as low as 0.45 V, which means that, thermodynamically, a battery having an emf higher than this voltage could not be constructed with this solid electrolyte. Around 1980, some efforts were made to synthesize Li3N derivatives (in the Li3N-LiCl and Li3N-LiI-

184


LiOH systems, for example) having a higher decomposition voltage. A compound in the latter system with conductivity comparable to Li3N shows higher decomposition voltage of 1.6–1.8 V.30 Recently a series of compounds, (La2/3–xLi3x)TiO3, has attracted much attention due to the high Li+ conductivity as well as because of a fundamental interest in the ion dynamics in the compound.31–34 They take a defective perovskite-type structure where part of the A-sites are vacant ((AV)1/3–2x; 0 < x < 1/6) and Li ions can migrate via those vacancies. The conductivity (at 25°C) depends on the Li concentration in terms of 3x as shown in Fig. 7.10, in which there is a sharp maximum (1.5 × 10–3 S cm–1) at about x = 0.3. Assuming the Li can jump only when the nearest A-site is vacant, a conductivity maximum may appear at a Li concentration at which the number of the LiAv bonds, given as 3x(1/3 – 2x) by a simple statistic consideration, is maximized. This parabolic function takes a maximum at 3x = 0.25, which is not far from the observed value. Figure 7.11 shows the temperature dependence of the conductivity observed with a single crystal sample of (La2/3–xLi3x)TiO3 (where 3x ~ 0.27), showing the activation energy of about 0.35 eV almost the same as that for a polycrystalline specimen. The activation energy is associated with the energy barrier at the window 3c-site connecting two adjacent A-sites, which is surrounded by four oxygens. A slight anisotropy in the conductivity between the parallel and anti-parallel directions with the caxis is caused by an ordered distribution of La along the c-axis. Clear super× 10–3 1.6 1.4

1.0

σ/arb.unit

σ/S · cm–1

1.2

0.8 0.6 0.4 0.2 0

0

0.1

0.2

0.3

0.4

0.5

3x

7.10 Ionic conductivity of (La2/3–xLi3x)TiO3 as a function of lithium concentration 3x.


185

0

log(σ/S · cm–1)

–1

–2

–3

–4 ⊥c //c –5 1

2

3 T –1/10–3K–1

4

5

7.11 Temperature dependence of the ionic conductivity of (La2/3–xLi3x)TiO3.

lattice reflections are observed in the X-ray pattern. The origin of non-linear dependencies in the high temperature region has not been clarified. From a practical point of view, these TiO2-based compounds have a problem in that they are unstable to metallic Li.

7.6

Sodium ion conductors

Sodium ion conductors are practically useful for advanced batteries represented by the sodium/sulfur (NAS) batteries, which are being widely developed for load-leveling or automotive use. One of the most important groups is the βalumina family. Although the crystal structure of β-alumina had already been solved as early as in the 1930s, it was not until 1967 that its high Na+ conductivity was found, and the principle of the NAS battery based on this solid electrolyte was proposed by Yao and Kummer.35 The ideal composition of β-alumina is NaAl11O17, whereas it is usually nonstoichiometric and represented as Na1+xAl11O17+x, where x is typically 0.2. Its structure is based on a hexagonal unit cell in which the Al11O17 units in a spinel-like arrangement (spinel block) and the NaO layers are stacked alternately along the c-axis. The Na+ conduction takes place within the NaO layer. Thus it shows strong anisotropy and the conductivity of a single crystal sample is as high as 0.01 S cm–1 (25°C) in the a–b plane, while several orders of magnitude lower along the c-axis. The Na sites in the conduction

186


layer takes a honeycomb arrangement, as shown in Fig. 7.12, where the BR and a-BR sites are nonequivalent. Regular sodium ions sit on the former site, whereas the excess sodium ions (x Na) are located on the mid oxygen (mO) sites, forming a pair. Migration of Na+ occurs according to an interstitial mechanism, i.e., Na(mO) → Na(BR) → Na(mO). 2-BR

BR

mO

O2– Ion pair

7.12 Position of Na ions in honeycomb plane of β-alumina.

Another principal member of this family is β″-alumina, the ideal composition of which is Na2O-5.33Al2O3. However, this binary phase is unstable, and incorporation of divalent cations such as Mg2+ is needed for stabilization; the stabilized β″-phase is thus represented, for example, as Na1+xMgxAl11–xO17, where x is typically 2/3. Its crystal structure, belonging to the rhombohedral system, is constructed of a similar alternation of the spinel blocks and NaO layers, but the sequence of the oxygen layers in the spinel block are different. In this phase the BR and a-BR sites are equivalent, and sodium ions are statistically distributed on both sites, leaving part of these vacant. It is, therefore, believed that the ionic conduction takes place by the vacancy mechanism. The conductivity of the β″-phase is about one order of magnitude higher than that of the β-phase. Sintered polycrystalline β″-alumina exhibits conductivity as high as 0.5 S cm–1 at 400°C. Thus this phase is usually used for NAS batteries. Sodium ions in the β-alumina family can be exchanged easily with various cations like Li+, Ag+, NH4+. Ion exchange is usually carried out by immersing a crystal of β or β″-alumina in a fused salt containing the cation to replace Na+. Exchanged β family can serve as a fast ionic conductor of each exchanged cation.

7.7

Silver and copper ion conductors

Silver iodide (AgI) shows a superionic transition at 147°C when it transforms from a wurzite to a cubic high temperature form (α-AgI). The Ag+ conductivity


187

of α-AgI is comparable to that of an H2SO4 aqueous solution. Since Tubandt first observed such high conductivity in the 1910s, numerous studies were carried out on this typical superionic material. As a result, a variety of fast ionic conductors like RbAg4I5, which shows almost the same conductivity at room temperature, were obtained. Copper-versions of AgI and its derivatives such as Rb4Cu16I7Cl13 also show similarly high conductivity. Exceptionally high ionic conductivity of those compounds is primarily due to the fact that they are composed of extremely polarizable ions (Ag+, Cu+) and their chemical bonds, therefore, show a strong covalent nature. Although silver- and copper-based superionic conductors remain very interesting as a subject of basic studies, they are less important from a practical point of view, mainly because their decomposition voltage is not high enough to construct a battery for practical use.

7.8

Amorphous ionic conductors for energy applications

Ionic conduction in alkali oxide glasses was known even in the nineteenth century by Warburg,36 and used in the field of the vacuum tube industry and semiconductor production in relation to the breakdown of the insulator characteristics of oxide glasses. Since the discovery of fast silver ion transport in silver oxyhalide glasses at the end of the 1960s, many glasses showing large ionic conductivity up to 10–4 ~ 10–2 S cm–1 at room temperature have been developed, chiefly silver and copper ion conductors.37 Some of these glasses were tried as an electrolyte for all solid-state batteries.38–40 However, their single cell voltage of about 0.65 V was too small to be used widely. In the 1980s, lithium ion conduction in glasses was investigated in relation to lithium ion batteries. Initially, research was conducted on oxide glasses.41,42 The next step grew out of the discovery of very high ionic conductivity in lithium sulfide glasses in the 1980s,43 whose lithium ion conductivity is close to 10–3 S cm–1. Since then a lot of work has been devoted to improving the thermal, mechanical and electrical properties of lithium sulfide glasses.44 On the other hand, the low lithium ion conductivity of oxide glasses can be compensated for by reducing the thickness of the electrolyte, which results in so-called ‘thin film batteries’.45 A large number of ionic conductor glasses have now been developed, in which ionic species such as silver (37, 46), copper (47, 49), lithium (50–53), sodium (54) fluorine (55, 56) and proton (57) are major carriers of the electric current. Some typical examples are shown in Table 7.1. In the 1970s, lithium ion conduction was found in solid polymer electrolytes such as polyethylene oxide (PEO) dissolving lithium perchlorate (LiClO4), and was proposed as the electrolyte in lithium batteries.58,59 Since then, a great deal of work has been devoted to improve the lithium ion

188


Table 7.1 Ionic conductivity and glass transition temperatures of typical ionic conductor glasses. Carrier ion Examples

Conductivity (20°C) S cm–1

Reference

Ag

AgI-AgPO3 Ag-Ge-S AgI-(CH3)3N(CH2)3(CH3)4NI2

~10–2 3 ×10–4 10–2

46 50 95

Cu

CuI-CuPO3 CuI-CuO-MoO3

10–4 10–4

47 48

Li

Li2O-B2O3 LiCl-Li2O-B2O3 Li2S-SiS2 Li2S-SiS2-LiI LiI-(C2H5)4NI-(C3H7)4NI

7 ×10–8 3 ×10–6 5 ×10–4 2 ×10–3 10–5 (at –40°C)

Na

Na2O-B2O3 NaF-NaCl-Na2O-B2O3 Na2S-SiS2

6 ×10–10 10–6 2 ×10–7

H

SiO2-P2O5-ZrO2-H2O SrO-BaO-PbO-P2O5

10–2 10–8

F

ZrF4-BaF2-CsF PbF2-MnF2-Al(PO4)3

5 ×10–6 (at 200°C) 10–4 (at 200°C)

50 50 53 53 120 50 50 50 226 57 56 56

conductivity of the polymers, which has led to 10–4 S cm–1 at room temperature in graft polymers of PEO units.60 The structure of the solid polymer electrolyte is a mixture of crystalline and amorphous phases. However, the high ionic conductivity has been believed to be confined to the amorphous region.58,59 Some of the polymers are now used as the electrolyte in lithium ion secondary batteries. However, the relatively low conductivity of the polymer electrolyte especially at low temperatures (~ –40°C) restricts its applicability. For now so-called ‘gel electrolytes’ have been used as a compromise,61 which contains organic solvent and electrolyte in a polymer matrix. Also, ‘room temperature ionic liquids’ (RTIL) are blended with polymers instead of flammable organic solvents.62–65 Since the 1960s, a large amount of work has been reported into the use of a perfluorinated polymer membrane (Nafion®) as an electrolyte and a separator of fuel cells.66,67 In the 1990s, proton conductors applicable above 100 °C have been strongly demanded for the fuel cells as was discussed in Section 7.4. For this purpose, along with the improvement of perfluorinated polymer membranes, new proton conductor polymers68 and inorganic proton conductor glasses57 have been proposed and investigated.


7.9

189

Ionic conduction mechanism of amorphous materials

Ionic transport in ‘amorphous materials’ has not been well understood in comparison with the crystalline materials due to their random structures. In particular, the well developed defect chemistry concept is not always applicable to amorphous ionic conductors. For example, the doping of a different valence cation to oxide glass results in only a small change in ionic conductivity, which is in contrast to the strong variation in oxide crystals or ceramics due to the creation of defects. This difference is strongly related to the structure of liquid, where every constituent ion tends to fulfill the local chemical bond requirement. Thus, even if a heterovalent ion is doped in the melt, it will be surrounded by counter ions to fulfill the chemical bond requirement to minimize the local free energy due to the fast structural relaxation in the melt. The local structure is frozen at glass transition temperature to form macroscopically random structure. Consequently, only a little variation is seen in the conductivity of glass. This local structural relaxation is forbidden in crystals due to the long-range periodicity constraint. Thus the ionic transport in glass and polymers is often described using the theory of liquids.69–71 Conductivity of ionic liquid is inversely proportional to the local viscosity of the liquid, which is known as the Stokes–Einstein law: σ = (ze)2N/(4πη)

7.1

Temperature dependence of the viscosity is expressed by the following Vogel– Tummann–Fulcher equation: η = η0 exp (DT0/(T – T0).

7.2 –1

Here, D is a parameter relating to the fragility~ D , and T0 is the Kauzmann temperature related to the glass transition temperature Tg72 by: Ig = (1 + 0.0255D)T0.

7.3

Thus, the first guiding principle for seeking high ionic conductivity is to find the liquid of low viscosity, which can be realized by lowering the glass transition temperature Tg and/or increase the fragility; see Fig. 7.13. This strategy is actually useful in the search for good ionic solvents or polymers for electrolytes. However, Eq. (7.1) has been found to break more than ten orders of magnitude in so-called superionic conductor glasses such as AgI-oxide systems etc.46,73–78 The deviation from Eq. (7.1) is often discussed by using another key concept of ‘coupling-decoupling’, which was proposed by Moynihan79 and widely applied by Angell.72–74,80 The so-called ‘decoupling index’, R = τσ /τs, which is the ratio of the electrical relaxation time τσ and the

190


Ionic conductivity (S cm–1)

100

10–5

Fragile Strong

10–10

10–15 0

0.25

0.5 Tg / T

0.75

1

7.13 Effect of glass transition temperature and fragility on ionic conductivity of liquid and polymers; increasing the fragility and decreasing the Tg lead to high conductivity.

mechanical relaxation time τs, was found to be a good measure of the ion dynamics in supercooled liquids and glasses. By using this index, Angell exhibited the Rτ ~ 1012 for superionic conductor glasses such as silver oxyhalide glass and even a sodium silicate glass, and called them ‘decoupling systems’. On the other hand, some glass forming systems such as LiCl-6H2O or a solid polymer electrolyte such as PEO-LiClO4 has only below Rτ ~ 104 at Tg; they are called coupling systems. This phenomenological concept has been confined by statistical models by using an excess-free-volume theory46,75–77,80,81 and a unified free energy theory,82 where the decoupling index is predicted to be enhanced by an increase in the remaining free volume for mobile ions, and by a decrease of binding force between the mobile ion and its surroundings. Another factor relates to the inhomogeneous structure as will be discussed later. The effect of the decoupling on the ionic conductivity is shown in Fig. 7.14. Although some deviations have been discussed, the ionic conductivity obeys almost Arrhenius-type temperature dependence below the glass transition temperature as: σ = σ0T –1 exp (–∆E/kT),

7.4

where ∆E is the apparent activation energy of conductivity. The apparent activation energy relates to the binding energy of mobile ions to the surrounding


Ionic conductivity (S cm–1)

Glass

Liquid

100

191

10–5

Decoupling

Coupling 10–10

10–15

0

0.5

1 Tg / T

1.5

2

7.14 Effect of decoupling on the ionic conductivity of liquids and glasses; increasing the decoupling leads to high conductivity.

anion and a mechanical strain energy to enlarge a portal for ion jump.50,82 The pre-exponential factor σ0 is expressed in a simple hopping model as: σ0 = γn(Ze)2a2ν0 /k

7.5

where n is the carrier density, Z is its charge, a is the jump distance and ν0 is the attempt frequency5 the geometrical factor γ is 1/6 for isotropic uncorrelated system, however it also depends on the percolation probability of the diffusion channels.94,95 Overall temperature dependence of the ionic conductivity throughout the liquid and glass regions can be expressed by the combination of Eqs (7.1, 7.2) and (7.4).46,75,76,126 One important conclusion to be drown from these fundamental considerations is that the ionic transport mechanism for inorganic glass is completely different from that for solid polymer electrolyte; the former is a decoupling system and the latter is a coupling one.78,80,83,84 It means that the ionic diffusion in the inorganic glass is decoupled from the oxide glass framework, wereas the diffusion of ions in polymer electrolyte is strongly coupled with the motion of polymer chains. From this difference, the inorganic glass can be used as a solid electrolyte below Tg, despite the ‘solid’ polymer electrolyte being useful only above Tg, in rubber state where the macroscopic rigidity is maintained by network entanglement. Also, most of the inorganic glass shows single ion conduction, but both cations and anions are usually mobile in the polymer electrolytes. Another important conclusion from the fundamental studies on glass and

192


polymer electrolyte is to recognize their ‘dual structure’, composed of a framework to keep rigidity and conduction channel of flexible region.85–90 In the organic polymers, the framework is the main polymer chain and the flexible part is the side chains or doped plasticizers and salts. In inorganic glasses, the framework is made of oxide, sulfide, oxynitride, etc., maintained by strong covalent bonds. The soft part is composed of non-bridging oxygen (NBO), or doped halide, sulfide, etc. These structures fulfill the dual requirement of mechanical stability seeking a strong chemical bond, and high ionic mobility preferring weak bonding to release the ions to move. The ‘double structure’ relates to the ‘cluster’ models87,88 of the glasses and percolation theory for ion conduction.91,95 A similar idea is also used in proton conducting polymer membranes, where the nanosized ‘cluster’ of water is assumed to be phase separated from perfluorinated polymer chains, and the proton transfer is also discussed with the free volume and percolation concept.96 Electronic conduction also affects amorphous materials as in the case of crystals. Electronic conduction is mainly observed in the inorganic glasses containing transition metals such as vanadium, tungsten, etc., or heavy metals such as tellurium, etc.97,98 and by conjugated bond chain in organic polymers.99 Mixed conduction, which is the combination of the electronic and ionic conductions, is very important for the application to cathode or anode materials for batteries.97,98 However, the electronic property of the amorphous ionic conductor requires further study.

7.10

Amorphous materials used for lithium batteries

Since the lithium ion battery was commercialized in 1991 by Sony Co., much effort has been devoted to improve the performance of the cell.101 The first task is to replace the LiCoO2 cathode with such other materials as LiMoO4, LiNiCoO2, etc. In this study some amorphous materials such as V2O5 were investigated,100,102–104 although it is difficult to use them for conventional batteries since their OCV was found to depend strongly on the composition. A second approach is to find new anode materials to replace carbon, which revealed such amorphous candidates as SnO,105 Li2O-SnOP2O5-Al2O3 glasses.106 A third approach involved replacing the organic liquid electrolyte by a solid polymer electrolyte, a gel electrolyte, an inorganic glassy electrolyte and their composites. The problems due to the use of organic liquid electrolyte in conventional lithium ion batteries (such as leakage of the electrolyte, possibility of burning or even an explosion of the flammable solvent, and dendritic growth of lithium metal at the anode) can be suppressed by using these solid electrolytes.


193

7.10.1 Polymer lithium ion electrolytes The research and development of solid polymer electrolyte (SPE) began when Wright reported ion conductivity of ionic complexes of polyethylene oxide (PEO) in 1975, which has been widely used as an electrolyte for lithium ion batteries.58,59,107–110 Ionic transport in PEO-based polymer electrolyte has been intensively studied and it has been concluded that the lithium ions are bounded by five oxygens of the ether group in the polymer chain and can be mobile by the fast segment motion of the polymer chains.60 Thus the enhancement of the rapid motion of the polymer chain will result in faster lithium ionic diffusion. The acceleration of the chain motion has been achieved by using highly branched polymers as a dendritic polyether 111 or graft-polymerized polyethers.112,113 A very high ionic conductivity up to 10-4 S cm–1 at room temperature is achieved in the dendritic polymers,111 although the transport number of the lithium ion is rather small at ~0.2. Some typical lithium ion conductor polymers are shown in Fig. 7.15. The small transport number induces concentration polarization of electrolytes during the charge-discharge process to result in a lowering of the power density. This is a natural consequence of the rather strong coordination bond between the oxygen in the PEO structure and lithium ions. Much effort has been made in recent years to improve the lithium transport number with the ultimate aim of single ion conduction. The first idea is to increase the dissociation of lithium by introducing fluorine in the network to reduce the oxygen charge114 or to trap the anions to the immobile polymer chains. Fujinami and some other groups developed organic–inorganic hybrid polymers containing boroxine rings, 115–118 borosiloxane,115 aluminate116 and oxalate anion capped borate.119 Large localized negative charges on Lewis base of boron attract anions around boroxine ring to result in higher lithium transport number. Some typical organic–inorganic hybrid polymers are shown in Fig. 7.15. Another idea is to decouple the lithium from the polymer chain by adding plenty of salt, which is called a ‘polymer-in-salt’ system.119 This approach has been successful in the gel electrolytes and room temperature ionic liquid (RTIL) as shown later. A reverse approach is to replace a network oxide of an inorganic fast ion conductor glass by organic ions, which results in fast single ion motion even below Tg.95,120 Ionic conductivity in solid polymer electrolytes has long been viewed as confined to the amorphous phase above Tg, where polymer chain motion creates a dynamic, disordered environment that plays a critical role in facilitating ion transport. On the other hand, Bruce et al. demonstrated the possibility of higher conductivity in ‘crystalline’ polymer electrolytes than amorphous ones, which seems a contradiction to the well-known belief of high conductivity

194

Materials for energy conversion devices (1) Polyether

(CH2

CH2

O )n

(2) Dendritic polymer (MEEGE) (Watanabe 02) Hyper-branched macromonomer

O O

O O

O

O

y

O

O

O

O

xn

O O

O

O

O

Ox

Oy

O

O

O

O

O

y

O

Network polymer

xn

O

O

O y

x

O

n

O

n

O

(3) Borosiloxane polymer (Kurono 01) CH2CH2CH2(OCH2CH2)3OCH3 Si

O

O B

O

O

a

b

Borosiloxane polymer

(4) Boroxine ring containing polymer (Mehta 99)

O

(CH2CH2O ) m

B

(OCH2CH2C)m

O

O

O

B

B O

O

(CH2CH2O )m y

(5) Mono-oxalato-capped orthoborate containing polymer (XuO2) poly [lithium oxalate ologo(ethylene glycolato) orthoborate (P(LiOEGnB)

O

O P(LiOEGnB)

O

O B–

H

O

O

CH2CH2 ( OCH2CH2 )n–1

m

OH

7.15 Examples of the recent lithium ion conducting polymers for lithium ion batteries.

in amorphous state of polymer electrolytes.121,122 They carefully compared the conductivity and NMR spectra of crystalline and amorphous state of PEO6:LiXF6 (X = P, As, Sb). The observed conductivity of the crystal is 10–6 S cm–1 at 30°C, which is considerably higher than the value of the corresponding amorphous material of 10–7 S cm–1. Also they claimed a


195

possible single lithium ion motion in the crystalline phase based on the NMR data. The structure analyses revealed that the lithium ions are located in the one-dimensional tunnels created by the PEO chains. The enhancement of the ionic conductivity by ordered structure suggests a new possibility of highly conducting polymer electrolytes; by applying uniaxial stress123 or synthesizing liquid crystalline polymers.124

7.10.2 Gel electrolytes In spite of large efforts devoted to increasing the ionic conductivity of polymer electrolytes, the maximum value is up to 10–4 S cm–1 at room temperature which is considerably smaller than conventional liquid electrolytes whose conductivity is 10–2 S cm–1 the difference is much larger at lower temperatures. Thus, a compromise is chosen for engineering purposes to blend the polymer with organic liquid electrolyte as plasticizer to result in so-called ‘gel electrolytes’, which are used for lithium batteries and also for capacitors.59,61,125 Most widely used for this purpose are polyviniridenfluoride (PVdF), its copolymer with hexafluropropylene (PVDF-HFP), polyaclironitril (PAN) and plymethylmethacrylate (PMMA); they are gellated with ethylencarbonate (EC), propyrencarbonate (PC) or dimethylcarbonate (EMC); see Fig. 7.16. By using the gel electrolyte instead of liquids, the leakage of the electrolyte and the dendritic growth of lithium anode are suppressed. In order to increase the ionic conductivity of the gel electrolytes. Further work has been carried out using room temperature ionic liquid (RTIL) to blend inorganic fillers.59,62,65 PEO (polyethylenoxide) ( CH2 – CH2—O )n PAN (polyacrylonitril) ( CH2 – CH )n CN PMMA (polymethylmetacrylate) CH3 ( CH2 – C ) n COOCH3 PVdF (polyvinylidenfluoride) ( CH2 – CF2 ) n PVdF-HFP (polyvinylidenfluoride-hexafluoropropyren) ( CF2 – CH2 ) ( CF2 – CF ) n

m

CF3

7.16 Matrix polymers of gel electrolytes for lithium ion batteries.

196


7.10.3 Inorganic oxide glasses Solid-state batteries using inorganic solid electrolyte have been studied for many years, initially using silver or copper conductors38 followed by lithium oxide,42,126 oxyhalide,127,128 oxynitride129,130 and sulfide glasses131–134 in recent years. Relatively low ionic conductivity of lithium oxide glasses (about 10–6 S cm–1 ) restricts their application to only thin film batteries, where the thickness of the electrolyte is less than 1 µm corresponding the internal resistance of at most 100 µ/cm2, which still limits the available current density to less than 1 mA/cm allowing for a voltage drop of 0.1 V. The history of the thin film battery started with the announcement by Hitachi Co., Japan in 1982 of an all solid-state thin film battery comprising a TiS2 cathode prepared by CVD, a Li3.6Si0.6P0.4O4 glass electrolyte by RF sputtering and metallic lithium as anode deposited by vacuum evaporation.45 Also, NTT Co. Group in Japan135–139 developed thin film batteries using Li3.4V0.6Si0.4O4 glass for electrolyte and LiCoO2136 and LiMn2O4138 for cathodes by using RF sputtering. Recently, Kuwata et al. prepared a thin film secondary lithium battery of LiCoO2 cathode, Li-V-Si-O glassy electrolyte and SnO anode, by means of PLD technique.140,141 Thin film batteries using LiBO2 were also announced in France by the Balkanski group.142 Thin film batteries were also developed by Ever-ready Battery Co. and Bellcore Co., USA, in the 1980s using sulfide glass of Li4P2S7 or Li3PO4P2S5 for electrolytes, TiS2 cathode and Li and LiI for anode.143–146 Bellcore Co. also announced the lithium cell consisted of LiMn2O4 cathode, a lithium borophosphate (LiBP) or lithium phosphorus oxynitride (LiPON) electrolyte and metallic lithium anode.

7.10.4 Nitride and oxynitride glasses As shown in Section 7.5, lithium nitride (Li3N) is a good crystalline lithium conductor, although it is not useful in lithium batteries because of the selfdischarge problem.28,29,147 Lithium oxynitride glasses were investigated in detail in the 1990s. They were found to be chemically more stable than oxides.148,149 However, the importance of the oxynitride glass was recognized after the announcement of lithium phosphonitroxide glass (LiPON) from the Oak Ridge National Laboratory (ORNL) in the USA.129,130,150 The LiPON can be easily prepared by RF sputtering of Li3PO4 target in nitrogen gas, which is quite stable in comparison with other lithium oxide or sulfidecontaining glasses in spite of rather poor ionic conductivity of 10–6 S cm–1 at room temperature. The LiPON is also proposed as a protective film in conventional lithium batteries.151


197

7.10.5 Sulfide and oxysulfide glasses Unusual high lithium ion conductivity up to 10–3 S cm–1 at room temperature was found in sulfide-based glasses in the 1980s;43,52,152–154 (See refs 53 and 155 for a review). It was modified to oxysulfide as Li3PO4-Li2S-SiS2156,157 to improve the stability and conductivity. All solid-state battery was fabricated by Matsushita Battery Co. using this oxysulfide glass as solid electrolyte.133 It has been investigated in detail by Minami and Tatsumisago’s group,158,159 where, besides a melt quenching method, a mechanochemical milling technique was found useful for preparation. A slow degradation of sulfide glass in contact with lithium anode has been known as a critical issue in battery application. However, a unique construction was reported to overcome this problem.162 These batteries contain two kinds of lithium ion-conductive solid electrolytes, Lil-Li2S-P2S5 glass contacted with the anode material and Li3PO4-Li2S-SiS2 glass or Li2S-GeS2P2S5 crystals contacted with the cathode. The former electrolyte was stable against electrochemical reduction, and the latter two against oxidation.162 This construction made it possible to use lithium graphite as the anode and LiCoO2 as the cathode. Lithium sulfide or oxysulfide glasses can also be formed into thin films by thermal vacuum evaporation,131,163 and RF sputtering.164 The lithium sulfide glass has been used for electrolyte and cathode materials in thin film batteries especially by French groups.165–169 In particular, the use of glassy electrolyte for thin film batteries is reviewed by Duclot and Souquet.170,171

7.10.6 Glass ceramics A recent topic is the use of partially crystallized glass or glass ceramics (158,160–173) for ionic conductors. It was first demonstrated in 1991 by Tatsumisago et al.174 that αAgI nanocrystals were stabilized in a glassy matrix by a rapid quenching method, which showed silver ionic conductivity up to 10–1 S cm–1 at room temperature. An oxide glass of Li2O-Al2O3-TiO2P2O5 was devitrified by thermal treatment to form a glass ceramic whose lithium ion conductivity increased to 10–3 S cm–1 at room temperature;172 similar glass ceramics have been commercialized by OHARA Co., Japan. Moreover, a good lithium conductor glass of Li2S-SiS2-P2S5 is devitrified to precipitate microcrystals whose conductivity increased to 10–2 S cm–1 at room temperature.158,160,161,173 It is interesting to note that the structure of the precipitated crystal is very close to the recently found thio-LISICON crystals.175 The ionic conductivity of the typical lithium ion conductors including polymers, gels, ionic liquids, inorganic glasses and crystals are shown in Fig. 7.17 as a function of temperature.

198

Materials for energy conversion devices 200


150

20

0

101 Crystal Glass Polymer Gel, liquid

100

LiCl-H

Conductivity (S cm–1)

10–1

0.15

Li

Li

3N

LiAlC

2 S-

(C)

–2

10

Li

2O

-V

2O 5 -S

10–3

Li

Cl

10–4

iO

2 (G

-L

i2 O

10–5

-2

5A

l4 -0.8

)

P(

-B

(G

Li

2O

PE

)

O-

LiB

F4

(P

EO

/M

EE

G

E)

)

l2 O

3 -5

0S

55

40

(P

)

iO

1 (G

10–6 2

2 O(L)

MC-L

iPF (L 5 ) 5 CH 5 SO C SiS 2 l/PM MA(G 2 -L E) i4 S Mlm IO /TFS P l(IL) 4 (G AN ) -PC /EC -LiA Li sF 3.2 5 Ge 8 (G E) 0.2 5P 0.7 5S ( 4 C )

2O 3

25

EC/D

2.5

)

3 1000/T(K–1)

3.5

4

7.17 Temperature dependence of typical lithium ion conductor glasses and polymers, where (L) denotes liquid, (C) crystal, (G) glass, (GE) gel, respectively. Data sources are: EC/DMC-LiPF6 (L),176 0.15LiAlCl4-0.85CH3SO2Cl/PMMA (GE),80 MIm/TFSI(IL),212 PAN-PC/ECLiAsF6 (GE),61 Li2S-SiS2-Li4SiO4 (G),157 PEO-LiBF4 (P),205 La0.51Li0.34TiO2.94 (C),32 Li3N (C),29 Li2O-V2O5-SiO2 (LVSO) (G),136 P(EO/ MEEGE)5540 (P),111 LiCl-Li2O-B2O3 (G),50 25Li2O-25Al2O3-50SiO2 (G).50

7.10.7 Amorphous electrode materials Modern lithium ion batteries owe their existence to the use of lithium carbon anode instead of metallic lithium.176,177 In order to improve the capacity, some oxides and nitride compounds have been investigated besides lithium alloys. Amorphous tin oxide-based glasses,106 phosphates,178 borate,179–183 and phosphoborate178 have been investigated for new anode materials. The film formation was also reported by RF sputtering178,179 and PLD.105 The mechanism of the anode reaction is starting from the reduction of the tin


199

oxides to metallic tin and lithium oxide, which causes the irreversible capacity followed by nano-sized lithium thin alloy formation180,181,185. Transition metal oxides and sulfides such as V2O5, MoO3, LiMn2O4, LiNiVO4 and MoS2 are used as cathode materials for lithium batteries. These can be formed into amorphous or disordered crystalline phases especially by thin film formation. They showed mixed conductivity and often work even better as a cathode material than the crystals.186 Vanadium pentaoxide V2O5 has been investigated in detail as an amorphous cathode for lithium ion batteries, which can be modified by adding TeO2,131 P2O5,100 Fe2O3, etc.102,187 The battery using amorphous V2O5-based cathode has good capacity performance, although the output voltage monotonically varies depending on the lithium concentration, and no plateau is seen.102,132,187

7.10.8 Composites with polymers There have been reported many trials of organic–inorganic composites based on the polymer electrolytes. Even non-reactive inorganic ceramics such as Al2O3, SiO2, MgO, etc., can improve the properties of the electrolytes (increase the conductivity and lithium transport number, electrode-electrolyte interfacial stability and also increase the glass transition temperature.188–192) This improvement is probably relating to the decoupling of the lithium ion motion from the polymer matrix at the interface.190 Hybridizing good inorganic ionic conductor with good polymer electrolyte will result in better composite electrolyte, which have been tried by using lithium sulfide-based glasses and PEO-based polymer electrolytes193–195 or an ethylmethylimidazolium tetrafluoroborate (ATMS) and P2S5 PEO and LiTFSI.196,197 The lack of a percolation threshold in the conductivity composition curve suggests that a part of the inorganic glass may dissolved in the polymers.193–195

7.11

Amorphous proton conductors

Proton conductors for fuel cell applications were discussed in Section 7.5 mainly for crystalline materials. Here, we look at other aspects for amorphous materials. The ionic conductivity of typical proton conductors are shown in Fig. 7.18 as a function of temperature.

7.11.1 Proton conductor polymers Recent advances in polymer electrolyte fuel cells (PEFC) are partly supported by the use of perfluorosulfonate proton exchange membranes (PEM) such as Nafion® (Dupont Co.), Flemion® (ASAHI Glass), ACIPLEX® (ASAHI Chemical), Daw membrane (DOW Chemical Co.), etc.67 The polymer structure

200



150

20

0

101

100

(a)

H3 P

O4 (b) mesoporous silica gel 5 MH2SO4

10–1

(h)

10–2

CsH

(e) S-PPBP SO

(i)

4

(i)

PE

55

O-

Sr O-

10–3

15 Ba O10

Conductivity (S cm–1)

(c) PEEK (d) Nafion

Pb

O-

(f) polysilses quioxane-P WA (g) NH SNO 4C 2 -H lO 2O 4 b le nd (k )P EI

70

10–4

O5 P2 (l) m eso p

10–5 (m

orou

:S )Y

ssili

ca g el

rZ rO 8

10–6 2

2.5

3 1000/T(K–1)

3.5

4

7.18 Temperature dependence of ionic conductivity of typical proton conductors near room temperature. (a) H3PO4,21 (b) mesoporous silica gel containing 5M H2SO4,229 (c) sulfonated polyetherketones (PEEK),21 (d) Nafion 100% water,51 (e) poly(4-phenoxybenzoyl-1,4phenylene, Poly-X 2000 (S-PPBP),207 (f) polysilsesquioxane-PWA complex polymer,217 (g) SnO2-H2O,15 (h) CsHSO4 crystal,21 (i) PEONH4ClO4 blend,51 (j) 55SrO-15BaO-10PbO-70P2O5 glass,57 (k) polyethylenimine hydrate(PEI),205 (l) dry mesoporous silica gel,229 (m) Yittrium doped SrZrO3.51

of the membrane comprises the PTFE (polytetrafluoroethylene) backbones and perfluorocarbon sulfonates (-OCF2-CF2-SO3H) as pendant groups.66,198–202 Highly dissociated protons can transfer through the membrane with the aid of the water molecules absorbed in the membrane. The structure of the membrane and the nature of the absorbed water molecules have been investigated by neutron scattering, XSAS, IR absorption, ESR and NMR.67,201 Results from these experiments led to the acceptance of an ionic cluster model for the membranes. The water and acid sites phase separately from the


201

fluorocarbon matrix to form inverted micellar structures. A schematic model proposed by Ogumi et al.198 is shown in Fig. 7.19.

(CF)n (CF)n

H2O SO–3

CF2

H2O

SO–3

Na+

F3C-CF O CF2 Na+ CF2

H2O

SO–3

SO–3

H 2O

H2O SO–3 H 2O A

(CF)n

(CF)n

B C (CF)n

7.19 Ionic cluster model of perfluorocarbon proton conductor membrane by Ogumi et al.198 Region A: rigid hydrophobic backbone. Region B: flexible perfluorocarbon. Region C: ionic cluster region containing water similar to bulk water.

Besides the perfluorosulfonate polymer, much research have been devoted to synthesize new proton conducting polymers mainly based on sulfonated or phosphated hydrocarbons,68 ex. poly(ethylene imine),203 polyetherketones (PEEK), 204 polyimin, 205 polyimide, 206 poly-phenoxybenzoyphenylene(PPBP), 207 polysulphone (PSU), 208 polyethylenimid (PI), polybenzoimidasole (PBI) and polyalylether, etc. (see Fig. 7.20). The merit of these hydrocarbon-based membranes is the lower cost than perfluorinated polymers and stability above 100°C where Nafion membrane does not work, although it is less stable under strong oxidation conditions. In normal PEFC operating conditions the protons diffuse mainly with the aid of the water molecules by proton transfer between rotating water molecules (Grotthus mechanism) or by molecular diffusion of hydronium ions (Vehicle mechanism). In consequence, the proton conductivity decreases at lower

202

Materials for energy conversion devices (a) Perfluorosulfonate —(CF2CF2)x — (CF2CF)y— | (OCF2CF)mO(CF2)nSO3H | CF3 ® Nafion 117 (m ≥ 1, n = 2, x = 5 ~ 13.5, y = 1000) Flemion® (m = 0, 1: n = 1 ~ 5) Aciplex® (m = 0, 3; n = 2 ~ 5, x = 1.5 ~ 14) (b) S-PPBP207 SO3H

(c) S-PEEK204 SO3H O

O

O

C

O

m

C

(d) S-PSU(208) O

O

m

S

O

O

SO3H

O

(e) Sulfonated polyimides S-PIH206 O N O

O N O

x O

HO3S

N SO3H

100-x

O

O N O

7.20 Some typical proton conductor sulfonated polymers. (a) perfluorosulfonates, (b) sulphonated poly-phenoxybenzoyphenylene(PPBP),207 (c) polyetherketones (PEEK),204 (d) polysulphone(PSU)208 and (e) sulfonated polyimide.206

water content region where only ‘hopping’ of the proton between the trapped sulfonate sites can contribute to the ionic diffusion.67,209 In order to increase the proton diffusivity at low water content region, some new ideas have been proposed • Introduce highly dissociative groups into polymer chains, often achieved by hybridizing inorganic element (B,Si, Al, etc.) into the polymer chains. • Gel formation with non aqueous low molecular weight solvent or room temperature ionic liquid (RTIL) to increase the mobility of protons. • Composite formation with inorganic fillers of oxides or sulfides. The first approach is an ‘acid-in-chain, base-in-chain’ concept,208,210 where the introduction of Lewis acid into the polymer chain enhances the proton mobility. The ultimate idea is to use the acid and base couples with optimized stability.211–216 Instead of sulfate or phosphate groups, inorganic superacid is also hybridized with organic polymer networks. Honma et al.217–221 combined


203

silicate, phosphate and 12-phosphotungstate (PWA) with PEO and other polymers. Resultant organic–inorganic hybrid polymer showed stable proton conductivity of ca. 10–3 S cm–1 up to 120°C.217–221 The idea of the gel electrolyte for lithium conductor is similarly applicable to the proton conductors. For instance, LiPF6-EC-PC-PVdF is replaced by H3PO4-EC-PC-PvDF-SiO2 for proton conductor gels, where SiO2 is a fumed silica.62 By using non-aqueous liquid, the problem relating to the water can be overcome. The proton conducting gel electrolytes are used not only for fuel cells215,216,222 but also for electric double-layer capacitors (EDLC)223 and nickel-metal hydride batteries.224

7.11.2 Proton conductor glasses In the 1990s proton transport in oxide glasses was investigated in detail by Abe et al. who synthesized good proton conductors based on alkali earth lead phosphate glasses such as 55SrO-15BaO-10PbO-70P2O5 prepared by the melt quenching method.57,225 Following this work, many efforts have been made to optimize the proton conductivity and stability of the glasses. Recent approaches have focused mainly on phosphosilicate glasses with pore structures226,229–231 Durability to water is improved by using silicabased glass, proton dissociation is enhanced by acidic phosphate units and the proton conduction channel is provided by the adsorbed water in the pores. Matsuda et al. also prepared phosphosilicate gel proton conductors with proton conductivity of 10–1 S cm–1 at room temperature, which is found to stabilize by addition of borons in the matrix.231 They also hybridized organic polyimide to the phosphosilicate gels.232,233 These proton conducting glasses were tested for fuel cells operating above 100°C. Nogami et al. prepared proton conductor glasses by sol-gel method based on silica doped with P2O5 and ZrO 2, TiO2, etc. The proton conductivity increases to 10–2 S cm–1 at room temperature and the glass is stable above 100°C with conductivity 170 mS cm–1 at 150°C.226–228

7.11.3 Fluoride and possible oxide ion conductor glasses Fluoride ion conducting glasses were found during the development of optical fibre glasses in the 1980s and rather high F– ion conductivity of 10– 4 S cm–1 (at 200°C) was recorded in PbF2-MnF2-Pb(PO3) oxyfluoride glasses.56 Although no application for energy devices has been reported, the transport mechanism of fluoride anion in glass is very interesting from a basic point of view. Since the fluoride anion is the network forming ion to construct the glass framework, it seems unlikely to move in glass. However, the large coordination number around the cations allows the fluoride anion to move in glasses. This information is useful in developing the oxide ion conductor glass described below.

204


Oxide ion conductor is a key material for Solid Oxide Fuel Cell (SOFC) devices. Despite the discovery of a large number of oxide ion conductors in crystals or ceramics such as ZrO2-Y2O3 (YSZ), LaGaO3(LGO), etc., as discussed in Section 7.2, no oxide ion conductors have been reported in glassy materials except in some recent research by Angell’s group.234 Possible reasons why no oxide ion conductor glass has been available include: (i) oxide anion is an integral part of the glass network, (ii) doubly negative charges, (iii) relatively high melting point of oxide ion conductor candidate.234 However, in comparison with the structure of fluorine anion conductor glasses it might be possible to overcome the first restriction employing cations with high coordination numbers of oxides such as ZrO2. Starting from this idea a eutectic melt of ZrO2 and WO3 were vitrified by rapid quenching from the high temperature melt in a xenon arc image furnace and also by a pressure amorphization. The observed conductivity is up to 10–3 S cm–1 at 500°C surpassing YSZ and close to the value of LGO234 However, the reported glasses are unstable to heating and a large electronic conduction is expected to overlap. Anyway, it is worth investigating this unexplored field, where a new concept is necessary to control defects in glasses.

7.12

References

1. Nernst, W., Z. Elektrochem., 6, 41 (1900). 2. Mizutani, Y., Tamura, T., Kawai, M. and Yamamoto, O., Solid State Ionics, 72, 271 (1994). 3. Kudo, T. and Obayashi, H., J. Electrochem. Soc., 123, 416 (1976). 4. Ishihara, T., Furutani, H., Honda, M., Yamada, T., Shibayama, T., Akbay, T., Sakai, N., Yokokawa, H. and Takai, Y., Chem. Mat., 11, 2081 (1999). 5. Benz, R., Z. Phys. Chem. N. F., 95, 25 (1975). 6. Dickens, M.H., Hayes, W., Hutchings, M.T. and Smith, J., J. Phys. C15, 4043 (1982). 7. Kummer, J. and Milberg, M.E., Chem. and Eng. News, 47, 90 (1969). 8. Frant, M.S. and Ross, J.W., Science, 154, 1553 (1966). 9. Okuda, T., Fukutomi, S., Takesako, K., Ohki, H. and Yamada, K., Solid State Ionics: Trends in the New Millennium (ed. by B.V.R. Chowdari et al.), pp. 569–74, World Scientific Publishing Co. (2002). 10. Berastegui, P., Hull, S. and Erikson, S-G., J. Phys. Condens. Matter, 13, 5077 (2001). 11. Vilminot, S. and Schulz, H., Acta Cryst. B 44, 233 (1988). 12. Yamada, K., Ahmad, M.M., Ohki, H., Okuda, T., Ehrenberg, H. and Fuess, H., Solid State Ionics, (in press, 2005). 13. Nakamura, O., Kodama, T., Ogino, I. and Miyake, Y., Chem. Lett., 17 (1979). 14. Howe, A.T. and Shilton, M.G., J. Solid State Chem., 28, 345 (1979). 15. Li, Y-M., Hibino, M., Miyayama, M. and Kudo, T., Electrochemistry, 69, 2 (2001). 16. Hara, S., Sakamoto, H., Miyayama, M. and Kudo, T., Solid State Ionics, 154/155, 679 (2002).


205

17. Colomban, P. and Novak, A., Proton Conductors (ed. by Colomban, P.), Chap. 3, Cambridge University Press (1992). 18. Baranov, A.I., Merinov, B.V., Tregubchenko, A.V., Khiznichenko, V.P., Shunalov, L.A. and Schagina, N.M., Solid State Ionics, 36, 279 (1989). 19. Yamada, K., Yamane, Y., Sagara, T., Ohki, H. and Okuda, T., Proc. 29th Symposium on Solid State Ionics in Japan (2003, Matsushima), p. 148. 20. Bonowska, W., J. Chem. Phys., 114, 611 (2001). 21. Kreuer, K-D., Chem. Mater., 8, 610 (1996). 22. Boysen, D.A., Chisholm, C.R.I., Haile, S.M. and Narayanan, S.R., J. Electrochem. Soc., 147, 3610 (2000). 23. Hinokuma, K. and Ata, M., Chem. Phys. Lett., 341, 442 (2001). 24. Hinokuma, K. and Ata, J. Electrochem. Soc., 150, A 112, (2003). 25. Li, Y-M. and Hinokuma, K., Solid State Ionics, 150, 309 (2002). 26. Liang, C.C., J. Electrochem. Soc., 120, 1289 (1973). 27. Poulsen, F., Andersen, N.H., Kindl, B. and Schoonman, J., Solid State Ionics, 9/10, 119 (1983). 28. Yamada, H., Karasuyama, Y., Moriguchi, I. and Kudo, T., Proc. 29th Symposium on Solid State Ionics in Japan (2003, Matsushima), p. 8. 29. Alpen, U.V., Rabenau, A. and Talat, G.H., Appl. Phys. Lett., 30, 621 (1977). 30. Obayashi, H., Nagai, R., Goto, A. and Kudo, T., Mater. Res. Bull., 16, 587 (1981). 31. Latie, L., Villeneuve, G., Conte, D. and Flem, G.L., Solid State Chem., 51, 293 (1984). 32. Inaguma, Y., Chen, L., Itoh, M., Nakamura, T., Uchida, Y., Ikuta, M. and Wakihara, M., Solid State Commun., 86, 689 (1993). 33. Inaguma, Y. and Itoh, M., Solid State Ionics, 86/88, 257 (1996). 34. Katsumata, T., Inaguma, Y., Itoh, M. and Kawamura, K., J. Ceram. Soc. Japan, 107, 615 (1999). 35. Yao, Y.F. and Kummer, J.T., J. Inorg. Nucl. Chem., 29, 2453 (1967). 36. Tomozawa, M., Treatise on Materials Science and Technology, 12, 283 (1977). 37. Minami, T., J. Non-Cryst. Solids, 56, 15 (1983). 38. Takada, K., Kanbara, T., Yamamura, Y. and Kondo, A.S., Solid State Ionics, 40/41, 988 (1990). 39. Baskaran, N., Govindaraj, G. and Narayanasamy, A., Journal of Power Sources, 55, 153 (1995). 40. Frumar, M. and Wagner, T., Current Opinion in Solid State & Materials Science, 7, 117 (2003). 41. Ravaine, D. and Souquet, J.L., In Solid Electrolytes: General Principles, Characterization, Materials, Application, pp. 277. Academic Press (1978). 42. Takahashi, T. and Yamamoto, O., Chem. Lett., 135 (1979). 43. Mercier, R., Malugani, J.-P., Fahys, B. and Robert, G., Solid State Ionics, 5, 663 (1981). 44. Tatsumisago, M., Journal of the Ceramic Society of Japan, 109, 809 (2001). 45. Kanehori, K., Matsumoto, K., Miyauchi, K. and Kudo, T., Solid State Ionics, 9&10, 1445 (1983). 46. Kawamura, J. and Shimoji, M., Mater. Chem. Phys. 23, 99 (1989). 47. Liu, C. and Angell, C.A., Solid State Ionics, 13, 105 (1984). 48. Machida, N., Chusho, M. and Minami, T., J. Non-Cryst. Solids, 101, 70 (1988). 49. Suzuki, K., Shibata, K., Tsurui, T. and Kawamura, J., J. Non-Cryst. Solids, 232, 278 (1998).

206


50. Fusco, F.A. and Tuller, A.H.L., Superionic Solids and Solid Electrolytes, Academic Press, Boston, 43 (1989). 51. Levasseur, A. and Menetrier, M., Mater. Chem. Phys. 23, 1 (1989). 52. Kennedy, J.H., Mater. Chem. Phys., 23, 29 (1989). 53. Pradel, A. and Ribes, M., Mater. Chem. Phys. 23, 121 (1989). 54. Hunter, C.C. and Ingram, M.D., Solid State Ionics, 14, 31 (1984). 55. Kawamoto, Y., Nohara, I., Fujiwara, J. and Umetani, A.Y., Solid State Ionics, 24, 327 (1987). 56. Reau, J.M., and Poulain, A.M., Mater. Chem. Phys, 23, 189 (1989). 57. Abe, Y., Hosono, H., Akita, O. and Hench, L.L., J. Electrochem. Soc., 141, L64 (1994). 58. Wright, P.V., Mrs Bulletin, 27, 597 (2002). 59. Murata, K., Izuchi, S. and Yoshihisa, Y., Electrochim. Acta, 45, 1501 (2000). 60. Nishimoto, A., Watanabe, M., Ikeda, Y. and Kohjiya, S., Electrochim. Acta, 43, 1177 (1998). 61. Song, J.Y., Wang, Y.Y. and Wan, C.C., Journal of Power Sources, 77, 183 (1999). 62. Ciuffa, F., Croce, F., D’Epifanio, A., Panero, S. and Scrosati, B., Journal of Power Sources, 127, 53 (2004). 63. Morita, M., Tanaka, A., Yoshimoto , N. and Ishikawa, M., Soldi State Ionics., 152, 161 (2002). 64. Kono, M., Hayashi, E., Nishiura, M. and Watanabe, M., J. Electrochem. Soc., 147, 2517 (2000). 65. Yoshizawa, M., Ogihara, W. and Ohno, H., Electrochemical and Solid State Letters, 4, E25 (2001). 66. Kreuer, K.D., Solid State Ionics, 97, 1 (1997). 67. HeitnerWirguin, C., Journal of Membrane Science, 120, (1996). 68. Rikukawa, M. and Sanui, K., Progress in Polymer Science, 25, 1463 (2000). 69. Smedley, S.I., The interpretation of ionic conductivity in liquids, Plenum, New York (1980). 70. Grest, G.S. and Cohen, M.H., Adv. Chem. Phys. 48, 455 (1981). 71. Cusack, N., The physics of structurally disordered matter: an introduction, Adam Hilger, (1987). 72. Angell, C.A., Annual Review of Physical Chemistry, 43, 693 (1992). 73. Angell, C.A., Solid State Ionics, 9&10, 3 (1983). 74. Angell, C.A., Solid State Ionics, 18&19, 72 (1986). 75. Kawamura, J. and Shimoji, M., J. Non-Cryst. Solids, 88, 295 (1986). 76. Kawamura, J. and Shimoji, M., J. Non-Cryst. Solids, 88, 281 (1986). 77. Kawamura, J. and Shimoji, M., J. Non-Cryst. Solids, 79, 367 (1986). 78. McLin, M. and Angell, C.A., J. Phys. Chem, 92, 2083 (1988). 79. Macedo, P.B., Moynihan, C.T. and Bose, R., Phys. Chem. Glasses, 13, 171 (1972). 80. Angell, C.A., Xu, K., Zhang, S.-S. and Videa, M., Solid State Ionics, 86–88; 17 (1996). 81. Swenson, J. and Borjesson, L., Phys. Rev. Lett., 77, 3569 (1996). 82. Souquet, J.L., Levy, M. and Duclot, M., Solid State Ionics, 70, 337 (1994). 83. Angell, C.A., Chemical Reviews, 90, 523 (1990). 84. Angell, C.A., Solid State Ionics, 105, 15 (1998). 85. Malugani, J.P., Mercier, R. and Tachez, M., Solid State Ionics, 21, 131 (1986). 86. Martin, S.W., J. Am. Ceram. Soc., 74, 1767 (1991). 87. Ingram, M.D., Mackenzie, M.A., Muller, W. and Torge, M., Solid State Ionics,. 28–30, 677 (1988).


207

88. Ingram, M.D., Mackenzie, M.A., Muller, W. and Torge, M., Solid State Ionics, 40/ 41, 671 (1990). 89. Wicks, J.D., Borjesson, L., Bushnell-Wye, G. and Howells, W.S., Phys. Rev. Lett., 74, 726 (1995). 90. Borjesson, L., McGreevy, R.L. and Wicks, J., Journal de Physique IV, 2, 107 (1992). 91. Bunde, A., Solid State Ionics. 75, 147 (1995). 92. Bunde, A., Funke, K. and Ingram, M.D., Solid State Ionics, 105, (1998). 93. Ratner, M.A., A., N., Solid State Ionics, 28–30, 3 (1988). 94. Kawamura, J., Kuwata, N. and Nakamura, Y., Solid State Ionics, 113–15, 703 (1998). 95. Kuwata, N., Kawamura, J. and Nakamura, Y., Solid State Commun, 124, 221 (2002). 96. Edmondson, C.A. and Fontanella, J.J., Solid State Ionics, 152–153, 355 (2002). 97. Levy, M. and Souquet, J.L., Mater. Chem. Phys., 23, 171 (1989). 98. Lebrun, N., Levy, M. and Souquet, J.L., Solid State Ionics, 40–41, 718 (1990). 99. Varela, H., Huguenin, F., Malta, M. and Torresi, R.M., Quimica Nova, 25, 287 (2002). 100. Sakurai, Y. and Yamaki, J., J. Electrochem. Soc., 132, 512 (1985). 101. Takehara, Z. and Kanamura, K., Electrochimica Acta, 38, 1169 (1993). 102. Machida, N., Fuchida, R. and Minami, T., Solid State Ionics, 35, 295 (1989). 103. Machida, N., Fuchida, R. and Minami, T., J. Electrochem. Soc., 137, 1315 (1990). 104. Machida, N., Fuchida, R. and Minami, A.T., Solid State Ionics, 37, 299 (1990). 105. Ding, F., Fu, Z., Zhou , M. and Qin, Q., J. Electrochem. Soc., 146, 3554 (1999). 106. Idota, Y., Kubota, T., Matsufuji, A., Maekawa, Y. and Miyasaka, T., Science, 276, 1395 (1997). 107. Scrosati, B., Polymer International, 47, 50 (1998). 108. Giridhar, P., Prasad, K.A., Kalaiselvi, N., Gopalakrishnan, K., Ganesan, M. and Veluchamy, A., Bulletin of Electrochemistry, 15, 414 (1999). 109. Croce, F., D’Epifanio, A., Hassoun, J., Reale, P. and Scrosati, B., Journal of Power Sources, 119, 399 (2003). 110. Zhao, F., Qian, X.M., Wang, E.K. and Dong, S.J., Progress in Chemistry, 14, 374 (2002). 111. Watanabe, M., Hirakimoto, T., Mutoh, S. and Nishimoto, A., Solid State Ionics, 148, 399 (2002). 112. Ikeda, Y., Kobunshi Ronbunshu, 57, 761 (2000). 113. Matsui, S., Muranaga, T., Higobashi, H., Inoue, S. and Sakai, T., Journal of Power Sources, 97–98, 772 (2001). 114. Watanabe, M., Suzuki, Y. and Nishimoto, A., Electrochimica Acta, 45, 1187 (2000). 115. Kurono, R., Mehta, M.A., Inoue, T. and Fujinami, T., Electrochimica Acta, 47, 483 (2001). 116. Matsushita, K., Shimazaki, Y., Mehta, M.A. and Fujinami, T., Solid State Ionics, 133, 295 (2000). 117. Matsumi, N., Mizumo, T. and Ohno, H., OLYM BULL, P51, 389 (2004). 118. Xu, W., Williams, M.D. and Angell, C.A., Chem. Mater. 14, 401 (2002). 119. Angell, C.A., Liu, C. and Sanchez, E., Nature, 363 137 (1993). 120. Imanaka, N. and Kawamura, J., Japanese patent, 3081922 (2000). 121. MacGlashan, G.S., Andreev, Y.G. and Bruce, P.G., Nature, 398, 792 (1999). 122. Gadjourova, Z., Andreev, Y.G., Tunstall, D.P. and Bruce, P.G., Nature, 412, 520 (2001).

208


123. Chung, S.H., Wang, Y., Greenbaum, S.G., Golodnitsky, D. and Peled, E., Electrocheml. Solid-State Lett., 2, 533 (1999). 124. Wright, P.V., Zheng, Y., Bhatt, D., Richardson T. and Ungar, G., Polym. Int., 47, 34 (1998). 125. Wang, Q.W. and Xie, D.M., Progress in Chemistry, 14, 167 (2002). 126. Souquet, J.L., Solid State Ionics, 28–30, 693 (1988). 127. Malugani, J.P. and Robert, G., Mat. Res. Bull. 14, 1075 (1979). 128. Glass, A.M. and Nassau, K., J. Appl. Phys., 51, 3756 (1980). 129. Bates, J.B., Gruzalski, G.R., Dudney, N.J., Luck, C.F. and Yu, X.H., Solid State Ionics, 70, 619 (1994). 130. Bates, J.B., Dudney, N.J., Neudecker, B., Ueda, A. and Evans, C.D., Solid State Ionics, 135, 33 (2000). 131. Creus, R., Sarradin, J., Astier, R., Pradel, A. and Ribes, A.M., Materials Science and Engineering, B3, 109 (1989). 132. Creus, R., Solid State Ionics, 53–6, 641 (1992). 133. Iwamoto, K., Aotani, N., Takada, K., and Kondo, S., Solid State Ionics, 70, 658 (1994). 134. Takada, K., Aotani, N., Iwamoto, K. and Kondo, S., Solid State Ionics, 86–8, 877 (1996). 135. Ohtsuka, H. and Yamaji, A., Solid State Ionics, 8, 43 (1983). 136. Ohtsuka, H. and Yamaki, J., Jpan. J. Appl. Phys., 28, 2264 (1989). 137. Ohtsuka, H. and Yamaki, J., Solid State Ionics, 35, 201 (1989). 138. Ohtsuka, H., Okada, S. and Yamaki, J., Solid State Ionics, 40–1, 964 (1990). 139. Yamaki, J., Ohtsuka, H. and Shodai, T., Solid State Ionics, 86–88, 1279 (1996). 140. Kuwata, N., Kawamura, J., Toribami, K. and Hattori, T., In Solid State Ionics, The Science and Technology of Ions in Motion (Proc. 9th Asian Conf. Solid State Ionics), ed. B.V.R. Chowdari et al., pp. 637–44. Singapore, World Sci. Publ. Co. (2004). 141. Kuwata, N., Kawamura, J., Toribami, K., Hattori, T. and Sata, N., Electrochem. Commun. 6, 417 (2004). 142. Balkanski, M. Julien, C. and Emery, J.Y., J. Power Sources, 26, 615 (1984). 143. Akridge, J.R. and Vourlis, H., Solid State Ionics, 18–19, 1082 (1986). 144. Akridge, J.R. and Vourlis, H., Solid State Ionics, 28–30, 841 (1988). 145. Jones, S.D. and Akridge, J.R., Solid State Ionics, 53–6, 628 (1992). 146. Jones, S.D., Akridge, J.R. and Shokoohi, F.K., Solid State Ionics, 69, 357 (1994). 147. Bittihn, R., Solid State Ionics, 8, 83 (1983). 148. Unuma, H., Komori, K. and Sakka, S., J. Non-Cryst. Solids, 95–6, 913 (1987). 149. Reau, J.M., Kahnt, H., Rocherulle, J., Verdier, P. and Laurent, Y., Journal of NonCrystalline Solids, 155, 185 (1993). 150. Yu, X., Bates, J.B., Jellison, G.E. and Hart, F.X., J. Electrochem. Soc., 144, 524 (1997). 151. Dudney, N.J., Journal of Power Sources, 89, 176 (2000). 152. Carette, B., Glass Technol. (GB), 24, 157 (1983). 153. Burckhardt, W., Makyta, M., Levasseur, A. and Hagenmuller, P., Mater. Res. Bull., 19, 1083 (1984). 154. Kondo, S., Takada, K. and Yamamura, Y., Solid State Ionics, 53–6, 1183 (1992). 155. Pradel, A. and Ribes, M., Mater. Sci. Eng., B3, 45 (1989). 156. Aotani, N., Iwamoto, K., Takada, K. and Kondo, S., Solid State Ionics,. 68, 35 (1994).


209

157. Minami, T., Hayashi, A. and Tatsumisago, M., Solid State Ionics, 136, 1015 (2000). 158. Tatsumisago, M., Hama, S., Hayashi, A., Morimoto, H. and Minami, T., Solid State Ionics., 154, 635 (2002). 159. Hayashi, A., Komiya, R., Tatsumisago, M. and Minami, T., Solid State Ionics, 152, 285 (2002). 160. Hayashi, A., Ohtomo, T., Mizuno, F., Tadanaga, K. and Tatsumisago, M., Electrochem. Com., 5, 70 (2003). 161. Mizuno, F., Hama, S., Hayashi, A., Tadanaga, K., Minami, T. and Tatsumisago, M., Chem. Lett., 1244 (2002). 162. Takada, K., Inada, T., Kajiyama, A., Sasaki, H., Kondo, S., Watanabe, M., Murayama, M. and Kanno, R., Solid State Ionics, 158, 269 (2003). 163. Kbala, M., Makyta, M., Levasseur, A. and Hagenmuller, P., Solid State Ionics, 15, 163 (1985). 164. Joo, K.H., Vinatier, P., Pecquenard, B., Levasseur, A. and Sohn, H.J., Solid State Ionics, 160, 51 (2003). 165. Menetrier, M., Levasseur, A., Delmas, C., Audevert, J.F. and Hagenmuller, P., Solid State Ionics, 14, 257 (1984). 166. Levasseur, A. and Vanatier, P., in Solid State Ionics, Science & Technology (ed. B.V.R. Chowdari et al.) World Scientific Publ.Co., pp. 421–9 (1998). 167. Jourdaine, L., Souquet, J.L., Delord, V. and Ribes, M., Solid State Ionics, 28–30, 1490 (1988). 168. Ribes, M., Technology of Thin Films: Fabrication and Characterization. Singapore, World Scientific Publ. (1988). 169. Ribes, M. and Delord, V., Thin Films of Conductive Glasses, Applications in Microionics, Singapore, World Scientific Publ. (1988). 170. Duclot, M. and Souquet, J.L., Journal of Power Sources, 97–8, 610 (2001). 171. Souquet, J.L. and Duclot, M., Solid State Ionics, 148, 375 (2002). 172. Fu, J., Solid State Ionics., 96, 195 (1997). 173. Mizuno, F., Hayashi, A., Tadanaga, K., Minami, T. and Tatsumisago, M., Electrochemistry, 71, 1196 (2003). 174. Tatsumisago, M., Shinkuma, Y. and Minami, T., Nature, 354, 217 (1991). 175. Kanno, R., and Maruyama, M., J. Electrochem. Soc, 148, A742 (2001). 176. Tarascon, J.M. and Armand, M., Nature, 414, 359 (2001). 177. Wu, Y.P., Rahm, E. and Holze, R., Journal of Power Sources, 114, 228 (2003). 178. Lee, J.Y., Xiao, Y.W. and Liu, Z.L., Solid State Ionics, 133, 25 (2000). 179. Nakai, M., Hayashi, A., Morimoto, H., Tatsumisago, M. and Minami, T., J. Ceram. Soc. of Japan, 109, 1010 (2001). 180. Hayashi, A., Nakai, M., Tatsumisago, M., Minami, T. and Katadab, M., J. of the Electrochem. Soc., 150, A582 (2003). 181. Hayashi, A., Nakai, M., Tatsumisago, M., Minami, T., Himei, Y., Miura, Y. and Katada, M., Journal of Non-Crystalline Solids, 306, 227 (2002). 182. Gejke, C., Zanghellini, E., Swenson, J. and Borjesson, L., Journal of Power Sources, 119, 576 (2003). 183. Ahn, J.H., Wang, G.X., Yao, J., Liu, H.K. and Dou, S.X., Journal of Power Sources, 119, 45 (2003). 184. Lee, W.H., Son, H.C., Moon, H.S., Kim, Y.I., Sung, S.H., Kim, J.Y., Lee, J.G. and Park, J.W., Journal of Power Sources, 89, 102 (2000). 185. Courtney, I.A. and Dahn, J.R., J. Electrochem. Soc., 144, 2045 (1997). 186. Julien, C.M., Mater. Sci. Eng. R., 283, 1 (2002).

210


187. Machida, N., Fuchida, R. and Minami, T., Solid State Ionics, 37, 299 (1990). 188. Tarascon, J.M., Gozdz, A.S., Schmutz, C.N., Shokoohi, F. and Warren, P.C., Solid State Ionics, 86–88, 49 (1996). 189. Koksbang, R., Olsen, II, and Shackle, D., Solid State Ionics, 69, 320 (1994). 190. Kumar, B. and Scanlon, L.G., Journal of Electroceramics, 5, 127 (2000). 191. Nishio, K., Okubo, K., Watanabe, Y. and Tsuchiya, T., Journal of Sol-Gel Science and Technology, 19, 187 (2000). 192. Zhang, X.W., Wang, C.S., Appleby, A.J., Little, F.E., Journal of Power Sources, 112, 209 (2002). 193. Hayashi, A., Kitade, T., Ikeda, Y., Kohjiya, S., Matsuda, A., Tatsumisago, M. and Minami, T., Chem. Lett., 814. (2001). 194. Ikeda, Y., Kitade, T., Kohjiya, S., Hayashi, A., Matsuda, A., Tatsumisago, M. and Minami, T., Polymer, 42, 7225 (2001). 195. Kohjiya, S., Kitade, T., Ikeda, Y., Hayashi, A., Matsuda, A., Tatsumisago, M. and Minami, T., Solid State Ionics, 154, (2002). 196. Hayashi, A., Wang, L.M. and Angell, C.A., Electrochimica Acta, 48, 2003 (2003). 197. Hayashi, A., Yoshizawa, M., Angell, C.A., Mizuno, F., Minami, T. and Tatsumisago, M., Electrochemical and Solid State Letters, 6, E19 (2003). 198. Ogumi, Z., Kuroe, T. and Takehara, Z., J. Electrochem. Soc., 132, 2601 (1985). 199. Kreuer, K.D., Solid State Ionics, 136, 149 (2000). 200. Kreuer, K.D., Ise, M., Fuchs, A. and Maier, J., Journal De Physique IV, 10, 279 (2000). 201. McBrierty, V.J., Martin, S.J. and Karasz, F.E., Journal of Molecular Liquids, 80, 179 (1999). 202. Gierke, T.D. and Hsu, W.S., in Perfluorinated Ionomer Membranes, ed. A. Eisenberg and H.L. Yeager ACS Symp. Ser., 180, Amer. Chem. Soc., Washington, DC, (1982). 203. Daniel, M.F., Desbat, B., Cruege, F., Trinquet, O. and Lassegues, J.C., Solid State Ionics, 28–30, 637 (1988). 204. Bishop, M.T., Karasz, F.E., Russo, P.S., Langley, K.H., Macromolecules, 18, 86 (1998). 205. Watanabe, M., Ikezawa, R., Sanui, K. and Ogata, N., Macromolecules, 20, 968 (1987). 206. Miyatake, K., Asano, N. and Watanabe, M., Journal of Polymer Science Part A – Polymer Chemistry, 41, 3901 (2003). 207. Kobayashi, T., Rikukawa, M., Sanui, K. and Ogata, N., Solid State Ionics, 106, 219 (1998). 208. Kerres, J.A., Journal of Membrane Science, 185, 3 (2001). 209. Bouchet, R., Miller, S., Duclot, M. and Souquet, J.L., Solid State Ionics, 145, 69 (2001). 210. Sun, X.G. and Angell, C.A., Electrochimica Acta, 46, 1467. (2001). 211. Xu, W., Wang, L.M., Nieman, R.A. and Angell, C.A., J. Phys. Chem. B, 107, 11749 (2003). 212. Yoshizawa, M., Xu, W. and Angell, C.A., J. Am. Chem. Soc. 125, 15411 (2003). 213. Xu, W. and Angell, C.A., Science, 302, 422 (2003). 214. Xu, W., Cooper, E.I. and Angell, C.A., J. Phys. Chem. B, 107, 6170 (2003). 215. Noda, A., Susan, A.B., Kudo, K., Mitsushima, S., Hayamizu, K. and Watanabe, M., J. Phys. Chem. B, 107, 4024 (2003). 216. Susan, A.B.H., Yoo, M.Y., Nakamoto, H. and Watanabe, M., Chem. Lett. 32, 836 (2003).


211

217. Honma, I., Nakajima, H., Nishikawa, O., Sugimoto, T. and Nomura, S., Solid State Ionics, 162, 237 (2003). 218. Honma, I., Hirakawa, S., Yamada, K. and Bae, J.M., Solid State Ionics, 118, 29 (1999). 219. Honma, I., Takeda, Y. and Bae, J.M., Solid State Ionics, 120, 255 (1999). 220. Nakajima, H. and Honma, I., Solid State Ionics, 148, 607 (2002). 221. Nakajima, H., Nomura, S., Sugimoto, T., Nishikawa, S. and Honma, I., J. Electrochem. Soc, 149, A953 (2002). 222. Qiao, J.L., Yoshimoto, N., Ishikawa, M., Morita, M., Electrochim. Acta., 47, 3441 (2002). 223. Qiao, J.L., Yoshimoto, N. and Morita, M., J. Power Sources, 105, 45 (2002). 224. Iwakura, C., Nohara, S., Furukawa, N. and Inoue, H., Solid State Ionics, 148, 487 (2002). 225. Abe, Y., Li, G., Nogami, M., Kasuga, T. and Hench, L.L., J. Electrochem. Soc., 143, 144 (1996). 226. Daiko, Y., Akai, T., Kasuga, T. and Nogami, M., J. Ceram. Soc. Jpn, 109, 815 (2001). 227. Daiko, Y., Kasuga, T. and Nogami, M., Chemistry of Materials, 14, 4624 (2002). 228. Nogami, M., Daiko, Y., Goto, Y., Usui, Y. and Kasuga, T., Journal of Sol-Gel Science and Technology, 26, 1041 (2003). 229. Matsuda, A., Kanazaki, T., Tadanaga, T., Tatsumisago, T. and Minami, T., J. Electrochem. Soc., 149, E292 (2002). 230. Matsuda, A., Kanazaki, T., Tadanaga, K., Tatsumisago, M. and Minami, T., J. Ceram. Soc. Jpn, 110, 131 (2002). 231. Matsuda, A., Nono, Y., Tadanaga, K., Minami, T. and Tatsumisago, M., Solid State Ionics, 162, 253 (2003). 232. Matsuda, A., Nakamoto, N., Tadanaga, K., Minami, T. and Tatsumisago, M., Solid State Ionics, 162, 247 (2003). 233. Tadanaga, K., Yoshida, H., Matsuda, A., Minami, T. and Tatsumisago, M., Chemistry of Materials, 15, 1910 (2003). 234. Jacob, S., Javornizky, J., Wolf, G.H. and Angell, C.A., Int. J. Inorg. Mat., 3, 241 (2001).

8 Oxygen ionic conductor K Y A M A J I and H Y O K O K A W A, National Institute of Advanced Industrial Science and Technology (AIST), Japan

8.1

Introduction

The history of oxygen ionic conductors dates from the end of the nineteenth century when Nernst invented a lamp (‘Nernst Lamp’) with a stabilized zirconia.1 Since then a series of stabilized zirconias such as calcia-stabilized zirconia (CSZ) and yttria stabilized zirconia (YSZ) has been investigated as representative oxygen ionic conductors. At present, YSZs are widely applied as oxygen ionic conductors for sensing such devices as oxygen sensors and for energy conversion devices such as water electrolysers and solid oxide fuel cells (SOFCs). Particularly, recent technological advance and success in SOFCs has led to more extensive and intensive investigations into oxygen ionic conductors; in this sense, a new era of oxygen ionic conductors has now opened up. In SOFCs, the oxygen ionic conductor is the most fundamental and important material as the electrolyte, so that many kinds of oxide ionic conductors have been intensively proposed and investigated. Among them, YSZ electrolytes have been recognized as a well-developed electrolyte for SOFCs to be operated around 1000°C because of its excellent electrical, chemical, thermodynamic and mechanical stabilities under operational and manufacturing conditions. On the other hand, there has been growing interest in recent years in reducing the operating temperature of SOFCs in order to utilize metal interconnectors to improve the anti-thermal-shock performance and to lower materials and manufacturing costs. 2,3 When the operating temperature of SOFCs decreases, YSZ electrolytes are no longer a good candidate because the oxygen ionic conductivity of YSZ decreases rapidly with lowering temperature. Thus, alternative materials, which have higher conductivity than YSZ, have attracted considerable interest in recent years; some materials have been developed successfully and have already been tested in practical SOFC stacks. In particular, doped-lanthanum gallates exhibit excellent characteristics for intermediate temperature SOFCs as will be shown later. Accordingly, we describe here 212

Oxygen ionic conductor

213

those fast oxygen ionic conductors which have higher oxygen ionic conductivity than YSZ electrolytes. This chapter is structured as follows. Section 8.2 deals with fundamental features of oxygen ionic conductors; in particular, the importance of the ratio of the ionic to the electronic conductivity is noted. Section 8.3 deals with the current status of oxygen ionic conductors; typical oxygen ionic conductors are briefly compared from the viewpoint of utilization in energy conversion applications. Finally, Section 8.4 describes the physico-chemical properties and related topics of oxygen ionic conductors in more detail.

8.2

Fundamental features of oxygen ionic conductor

A solid electrolyte should be defined as an ionic conductor having essentially no electronic conductivity. When the main charge carrier is oxide ions (O–2), this is called the oxygen ionic conductor. The electric current flow is originated from diffusion of oxide ions via oxide ion vacancies. The vacancy is generally formed by the substitution of aliovalent ions to the host cation lattice. For example, in the YSZ system, oxide ion vacancies are created by the substitution of Zr4+ sited in the fluorite lattice with Y3+; this can be written as the following equation using the Kröger-Vink notation:

⋅⋅ Z rO 2 Y2 O 3  → 2YZr ′ + 3O Ox + VO

8.1

where Y′Zr means Y in the Zr sites with the apparent single negative charge, O Ox means oxygen in the oxygen sites with net charge of zero, and VO⋅⋅ means the vacancy in the oxygen sites with double positive charge. Oxide ions are transported by hopping through the vacancy sites. The concentration of oxygen vacancy is determined by the dopant concentration, whereas the ionic conductivity does not necessarily increase linearly with the dopant concentration. Thus, there appears a conductivity maximum as a function of dopant concentration. In the case of YSZ, this is around 10 mol% of Y2O3 doping.4 When an oxygen ionic conductor plate is separately exposed to two gases at different partial oxygen pressures with appropriate electrodes, an electromotive force (emf) is generated due to the difference in the oxygen potential between two electrodes. If there are no other charge carriers, the emf, E, can be described as follows, E=

RT  p (O 2 ) I  ln   4F  p (O 2 ) II 

8.2

where R, T and F are the gas constant, temperature and the Faraday constant, respectively, and p(O2)I and p(O2)II are the higher and the lower oxygen

214


partial pressures on each electrode. This is called Nernst’s equation for the theoretical emf value. When effects of other charge carriers such as electrons are not neglected, the terminal voltage deviates from the theoretical emf value derived from Eq. 8.2. Since electrons and holes can be regarded as minor defects in the oxygen ionic conductor, their concentration and conductivity can easily be discussed in terms of the formation or the extinction of the oxygen vacancies as a function of the oxygen partial pressure (p(O2)). For example, the electronic and electronic hole conduction in YSZ can be written as the following equations in the reducing and the oxidizing atmospheres, respectively: O Ox + 2 h ⇔ VO⋅⋅ + 1 O 2 , p ∝ [VO⋅⋅ ]1/2 p (O 2 )1/4 2

in the oxidizing atmosphere

8.3

O Ox ⇔ VO⋅ ⋅ + 1 O 2 + 2 e , n ∝ [VO⋅⋅ ]1/2 p (O 2 )1/4 2 in the reducing atmosphere

8.4

where p and n mean the concentration of electronic hole and electron, respectively, [VO⋅⋅ ] means the concentration of oxygen vacancies, and p (O2) is the oxygen partial pressure. Thus, the electronic hole and electron conductivity are proportional to p(O2)1/4 and p(O2)–1/4, respectively. When the valence of some cations in an oxygen ionic conductor is varied with oxygen partial pressure, the relation between the electronic conduction and the oxygen partial pressure becomes more complicated. In order to evaluate oxygen ionic conductors, the transference number of ion, ti, is usually defined as follows, ti =

σi σ total

8.5

where σi is an ionic conductivity and σtotal is the total electric conductivity; total conductivity is the sum of all ionic and electronic conductivity.

8.2.1

Electronic conductivity measurement with a polarization cell

For utilization of the oxygen ionic conductors in energy conversion devices such as SOFCs, the oxygen ionic conductors should meet the following requirements over a wide oxygen potential range covering from air to humidified hydrogen: • The oxygen ionic conductivity should be high. The low conductivity inevitably makes the energy conversion efficiency in SOFCs low due to a large joule loss.


215

• The electronic conductivity should be small. The high electronic conductivity also makes the energy conversion efficiency low because the transported oxide ions will partly consume without power generation. • The chemical, thermodynamic, and mechanical stabilities should be high during the fabrication processes and the long-term (more than 50,000 h) operation of SOFCs. Among the requirements as electrolytes for SOFCs we focus here on the relation between the oxygen ionic conduction and the electronic conduction. In order to evaluate oxygen ionic conductors as electrolytes for SOFCs from the viewpoint of energy conversion efficiency, it is desirable to measure the electronic conductivity as functions of p(O2) and temperature. From the detailed conductivity data, the fundamental characteristic features associated with the energy conversion efficiency can be evaluated in terms of the optimum thickness (or the optimum current density); these features can be used to judge whether candidate materials of the oxygen ionic conductor are suitable in view of the efficiency loss due to the oxygen permeation. Here, the method of the electronic conductivity measurement in oxygen ionic conductors is described and in Section 8.2.3, detailed features of energy conversion will be described. When the transference number of oxygen ion is almost unity, the oxygen ionic conductivity can usually be measured by simple methods such as a DC four-probe method or an AC impedance method; for the measurement of the electronic conductivity, however, some special experimental techniques will be required because the electronic conduction is hidden behind oxygen ionic conduction. In order directly to measure the electronic conduction in those ionic conductors, the most appropriate method is an ion-blocking method with a polarization cell having an ion blocking electrode; this was first proposed by Hebb5 and Wagner.6 For a typical oxygen ionic conductor of YSZ, the electronic conduction has already been reported by some authors using Hebb–Wagner type polarization cells.7–9 Recently, the electronic conductivity of noble fast oxygen ionic conductors, such as doped ceria and lanthanum gallate, has also been evaluated.10,11 The essential point of measurement is to ensure that the ionic current is blocked, whereas the electronic current is allowed to flow. Figure 8.1 shows a schematic view of our ion-blocking cell. A cell is made of a polished electrolyte having a diameter and thickness of about 17 mm and 1 mm, respectively. Porous Pt electrodes at a diameter of 10 mm were attached on both surfaces of the sample with Pt meshes and wires as current collectors. The blocking electrode exists in a closed space surrounded by an alumina spacer and glass seals. The atmosphere around the reversible electrode was usually kept in a constant flow of 1%O2–99%Ar mixture gas. When a voltage is applied between the reversible and blocking electrodes by a potentiostat, the electric current between the electrodes changes

216

Materials for energy conversion devices Pt wire Porous Pt electrode (reversible electrode) Sample pellet (17 mmφ) Glass seal Porous Pt electrode (blocking electrode) Alumina spacer Potentio/galvano state

8.1 Schematic view of ion-blocking cell.

continuously until the steady state is established. Then, the oxygen partial pressure at the blocking electrode/oxygen ionic conductor interface in the closed space is reduced against the reversible electrode to cancel the applied voltage. In the steady state, the oxygen ion flow through the blocking electrode is blocked, and the measured current is the electronic current originated from electrons and electronic holes. When the electronic current was measured as a function of the applied voltage, the electronic conductivity (σe) was determined by the following equation:

 ∂I  σe = L ⋅  A  ∂E 

8.6

where A and L are the electrode area and the thickness of the sample, respectively. The oxygen partial pressure at the blocking electrode is calculated from the applied voltage using the Nernst’s equation as follows: E=

RT  p (O 2 ) reversible  ln 4 F  p (O 2 ) blocking 

8.7

where p(O2)reversible and p(O2)blocking are the oxygen partial pressure at the reversible and blocking electrodes, respectively. As a result, the electronic conductivity can be evaluated as a function of the oxygen partial pressure. Figure 8.2 shows a typical DC polarization (I-V) curve obtained at 1073 K for a doped lanthanum gallate having a composition of La0.9Sr0.1Ga0.8Mg0.2O2.85 (denoted as LSGM9182); LSGM9182 is a fast oxygen ionic conductor as will be described later. Figure 8.3 shows the electronic conductivity determined from the I-V curve shown in Figure 8.2 using Eqs 8.6 and 8.7, in which the total conductivity was also given for comparison. The electronic conductivity is proportional to p(O2)1/4 in oxidizing atmospheres, which indicates the electronic hole conduction is dominant in higher p(O2) region according to Eq. 8.3. The electronic conductivity is proportional to p(O2)–1/4 in reducing atmospheres, which indicates that the electron conduction


217

4

I/mA

3

2

1

0 0

–200

–400

–600 E /mV

–800

–1000

8.2 Typical I-V curve for LSGM9182. 0 –1

σ ion

log(σ/Scm–1)

–2 –3 –4

–1/4

1/4 σ electron

–5

σ hole

–6 –7 –8 –30

–20

–10 log(p(O2/atm))

0

8.3 Electronic and ionic conductivity for LSGM9182.

is dominant in lower p(O2) region according to Eq. 8.4. With using the electronic conductivity determined as a function of oxygen partial pressure, the efficiency of the oxygen ionic conductor as an electrolyte for SOFCs can be evaluated as described in the following subsection.

8.2.2

Characteristic feature as electrolyte for energy converter

The large electronic conduction in an SOFC electrolyte gives rise to a significant problem, because oxygen can permeate without generating electricity. By taking the energy loss from oxygen permeation into account, energy conversion efficiency due to the electrolyte for SOFC, εelectrolyte, is simply evaluated as follows. The total energy conversion factor of a SOFC is defined by the following equation:

218


ε= =

J ext Vterm J O2 Vth

J ext ( ∆ E electrolyte – ∆ E electrode – ∆ Eseparator ) RT p (O 2 ) I ( J O 2– ,electrolyte + J O 2– ,separator + J 2– ) ln O ,leak 4 F p (O 2 ) II

8.8

Here, Jext and Vterm are the current density and the terminal voltage, respectively. JO2 is the total oxygen flux from cathode to anode, and Vth is the theoretical voltage calculated from the oxygen partial pressures at the cathode and the anode written as p(O2)I and p(O2)II, respectively. The efficiency due to the electrolyte, εelectrolyte, can be extracted as follows:

ε electrolyte =

J ext ⋅ ∆ E electrolyte RT p (O 2 ) I ln J O 2– ,electrolyte ⋅ 4F p(O 2 ) II

8.9

where the first factor Jext /JO2– originates from the utilization efficiency and the second factor ∆Eelectrolyte/Vth is the voltage efficiency inside the electrolyte. Choudhury and Patterson have reported that ∆Eelectrolyte, ∆ J O 2– ,electrolyte and Jext can be written as follows:12 ∆ E electrolyte = –

J O 2 , electrolyte =

RT 2F

r RT 2 FL

∫

p (O 2 ) II

p (O 2 ) I

∫

p (O 2 ) II

p (O 2 ) I

J ext =  1 + 1  J O 2– r 

σ O 2– d ln p(O 2 ) rσ e – σ O 2–

8.10

σ e σ O 2– d ln p (O 2 ) rσ e – σ O 2–

8.11

8.12

Here, the parameter r is a ratio of the ionic to the electronic current densities, which is constant throughout the electrolyte layer under a steady state condition. Energy conversion efficiency due to the electrolyte can be evaluated by calculating these equations over the operating oxygen pressure range. Figure 8.4 shows the evaluated efficiency of LSGM9182 at different temperatures as functions of current density and thickness of the sample. In the evaluation process, p(O2) values of the oxidizing sides are fixed at 0.21 atm (as ambient air), and p(O2) values of the reducing side are 10–18, 10–22 and 10–26 atm (as a typical fuel of 3%H2O-97%H2) at 1000, 800 and 600°C, respectively. By increasing the thickness of the sample, the efficiency loss due to the joule loss cannot be neglected, and by decreasing the sample thickness, the efficiency loss due to the oxygen permeation cannot be neglected. Therefore, the efficiency plotted as a function of the sample thickness shows


219

L /µm (Jext = 0.5 Acm–2) 10 100 1000 10000

1 1.0

1000 °C 800 °C 600 °C

ε

0.9

0.8

0.7 –5

–4

–3 –2 –1 log (JextL/Acm–1)

0

1

8.4 Efficiency of LSGM9182 as functions of Jext and L.

a maximum value as shown in Fig. 8.4. For LSGM9182, the highest efficiency increases with decreasing operating temperature, and the maximum value is about 96% at 600°C. This figure also shows information about the optimum thickness of the oxygen ionic conductor as the SOFC electrolyte when the current density is fixed. The optimum thickness of LSGM9182 at 600°C was about 10 µm, which is still within a limit determined by the current fabrication technology for SOFCs. From these estimates, it is indicated that LSGM9182 is preferable as a candidate for the electrolyte of intermediate temperature (IT) SOFCs.

8.3

Current status of oxygen ionic conductors

8.3.1

Kinds and properties of fast oxygen ionic conductors

Figure 8.5 shows the oxygen ionic conductivity of YSZ and typical fast oxygen ionic conductors. The electrolytes compared in Fig. 8.5 are 8mol%Y2O 3-stabilized zirconia (8YSZ),8 10mol%GdO 1.5-doped ceria (10GDC),13 10mol%Sc2O3-stabilized zirconia doped with 1mol% CeO2 (1Ce10ScSZ), 14 LSGM9182, 15 LSGM doped with cobalt of La0.8Sr0.2Ga0.8Mg0.115Co0.085O2.8 (LSGMC),16 copper-doped bismuth vanadium oxide of Bi2V0.9Cu0.1O5.35 (BICUVOX),17 and calcia-doped lanthanum germanium oxide of La1.6Ca0.2GeO5–δ (LCGO).18 In this figure, BICUVOX shows the highest conductivity, so that the electrolyte might be a good candidate for SOFCs. However, a series of bismuth-based oxides are not structurally stable and are reduced in reducing atmospheres. BICUVOX is also reduced in an anode atmosphere of SOFCs so that it could not be utilized for SOFCs. Rare-earth doped ceria electrolytes, such as 10GDC in Fig. 8.5, also have high oxygen ionic conductivity especially at lower temperatures. However,

220


T /°C 700

600

500

–0.5

log(σ/Scm–1)

–1.0

–1.5 Bicuvox 8YSZ 10GDC LCGO LSGMC LSGM 9182 1C e10ScSZ

–2.0

–2.5

–3.0 0.8

0.9

1.0 1.1 1000K/T

1.2

1.3

8.5 Comparison of electric conductivity between typical oxygen ionic conductors.

they also show high electronic conductivity in reducing atmospheres. Figure 8.6 shows a comparison of energy conversion efficiency estimated from the ionic and electronic conductivity of LSGM9182,10 20mol%GdO1.5–doped ceria (20GDC)19 and 8YSZ8 as a function of temperature. In this estimation, the current density was fixed at 500 mA cm–2, and the atmospheric gases of the cathode and the anode were assumed as air and 3%H2O-H2, respectively. The thicknesses of the electrolytes were selected as 5, 50 and 500 µm, for respective electrolytes. At higher temperatures, 20GDC shows a poor efficiency because of the significant high electronic conductivity in reducing atmospheres. With decreasing temperature, the effect of the electronic conduction becomes moderate. The efficiency increases over 60% below 600°C, and reaches about 80% at 400°C at the thickness of 5 µm. If a rare-earth doped ceria is independently used as an SOFC electrolyte, the lower operating temperature should be required from a viewpoint of efficiency. As shown in Fig. 8.5, the efficiency loss in 8YSZ increases rapidly with lowering temperature due to the poor ionic conductivity at lower temperatures. In order to use 8YSZ at intermediate temperatures around 700°C, a thin film must be fabricated to avoid the severe efficiency loss. Recent investigations on fabrications of electrode supported thin electrolytes have shown successfully that high performance is achieved by using a thin 8YSZ film at intermediate temperatures between 600°C and 800°C. Lanthanum gallate doped with strontium and magnesium (LSGM) shows more excellent electrical performance at lower temperatures than that of

Oxygen ionic conductor 700 °C

1.0

221

500 °C I = 300 mA/cm

0.9

ε

0.8

0.7

0.6

0.5

YSZ 500 µm YSZ 50 µm YSZ 5 µm GDC 500 µm GDC 50 µm GDC 5 µm LSGM 500 µm LSGM 50 µm LSGM 5 µm 0.6

0.8

1.0 1000K/T

1.2

1.4

1.6

8.6 Comparison between YSZ, GDC and LSGM in energy conversion efficiency as SOFC electrolytes at the current density of 0.3 A cm–1. The efficiency was evaluated as a function of temperature, and the electrolyte thicknesses are selected as 500, 50 and 5 µm.

YSZ because of the superior ionic conductivity and the small electronic conductivity. At the thickness of 50 µm, where the electrolyte can be applied as a self-supported structure, the efficiency was high at around 700°C as shown in Fig. 8.6. Furthermore at a thickness of 5 µm, where an anode supported structure is required, the efficiency was over 90% above 450°C, which is the great advantage as compared with the other electrolytes. Doping with a transition metal for the B-site of LSGM significantly increases the ionic conductivity. Cobalt is the most effective dopant for enhancing the conductivity, and the typical one is shown in Fig. 8.5 as LSGMC; however, it is noted that the electronic conductivity increases with increasing Codoping and with decreasing operating temperature, which results in a significant loss of efficiency. Scandia-stabilized zirconia electrolytes show the highest conductivity among all stabilized zirconia electrolytes. Even so, there was a disadvantage on ScSZ due to cubic-rhombohedral phase transition occurring at 600 to 700°C when Sc2O3 was doped with over 9 mol%. Recently, it has been found that adding a small amount of oxide such as CeO2 suppressed the phase transition. As shown in Figure 8.5, 1mol%CeO2-doped ScSZ shows a good oxygen ionic conductivity above 600°C. A series of La2GeO5-based oxides is one of the latest oxygen ionic conductors which has higher ionic conductivity than 8YSZ. Although the oxygen ionic

222


conductivity is not that much greater than 8YSZ, finding a new family of compounds will open a new era for the investigation of new electrolytes in the intermediate temperature range.

8.3.2

Application to solid oxide fuel cells

Recently, some attractive attempts have been made to use fast oxygen ionic conductors as SOFC electrolytes instead of conventional YSZ. Here, such attempts on doped-ceria, doped-lanthanum gallate, and scandia stabilized zirconia will be described briefly. With a doped-ceria electrolyte, CERES Power Ltd., teamed with Imperial College, has recently started to fabricate low-temperature SOFCs to be operated below 600°C.20 They are based on a criterion that the electrolyte component should not contribute more than 0.15 Ωcm2 to the total cell area specific resistance; for the electrolyte with thickness of 15 µm, the associated specific ionic conductivity should exceed 10–2 S cm–1. In order to achieve this criterion, they have selected gadolinia doped-ceria (GDC) at a thickness of 10 to 30 µm. As a low-cost technique for fabricating dense GDC films on steelsupported structure, they have tried adopting a electrophoretic deposition (EPD) technique.21 A layer of Ce0.9Gd0.1O0.95 (GDC10, Rhohdia) was deposited on a steel substrate by EPD and followed by pressing with a Cold Isostatic Pressing machine. After sintering the pressed sample at 1000°C, a thin (around 10 µm) and dense film of GDC10 was fabricated successfully. In their report, a high power density of more than 200 mW cm–2 at 550°C on moist H2/air was obtained for a cell with 16 cm2 active area.20 With a LSGM electrolyte, Mitsubishi Materials Corp. in collaboration with the Kansai Electric Power Co., Inc., has developed intermediatetemperature SOFCs.22 Cobalt-added LSGM in the chemical formula of La0.8Sr0.2Ga0.8Mg0.15Co0.5O3-δ (LSGMC) was selected and fabricated by a conventional solid state reaction method. The electrolyte supported design can be adopted even below 800°C for the SOFCs using LSGMC electrolyte because of a low joule loss for 200 µm thick electrolytes. In fabrication, a calcined mixture of La2O3, SrCO3, Ga2O3, MgO and CoO was mixed with organic binder and tape-casted to a green sheet; after disk-shape green sheets were fired at 1400 to 1500°C, 200 µm thick LSGMC electrolytes were obtained. Its relative density is greater than 98%. Recently, a seal-less planartype SOFC module of 1 kW class was constructed successfully using 25 cells in a diameter of 154 mm. Its operation was also successful in obtaining an output power of 1 kW without heating system below 800°C. For scandia stabilized zirconia (ScSZ), Toho Gas Co., Ltd. has been investigating the use of electrolyte and the electrode in planar SOFCs for more than ten years.23 They adopted a self supported planar-type cell; ScSZ-electrolyte plates are more than 100 µm thick. This is to be operated


223

below 800°C. Furthermore, tetragonal scandia-doped zirconia (Sc-TZP) has better characteristic features in a sense that Sc-TZP has a higher mechanical strength than ScSZ, although the electrical conductivity is lower than that of ScSZ. With a Sc-TZP of 6mol%Sc2O3-doped zirconia, they have recently constructed a planar type SOFC system to be operated at 800°C at a power of 1 kW.24

8.4

Recent topics of typical oxygen ionic conductors

This section describes the fundamental properties of typical fast oxygen ionic conductors together with some recent topics. Focus will be placed on electrolyte materials for SOFCs and therefore we do not discuss those materials which are not stable in SOFC environments. Rare earth doped ceria, doped lanthanum gallate, scandia stabilized zirconia and the other new oxygen ionic conductors are selected from the viewpoint of having a superior oxygen ionic conductivity in comparison with YSZ.

8.4.1

Rare earth doped ceria

Doped cerias are not noble oxygen ionic conductors and they have been actively studied since the 1970s; therefore, a large number of investigations are available in the literature. Recently, Mogensen et al. have published an extensive review, which covers available data on the physical, chemical, electrochemical and mechanical properties of pure and doped ceria.25 As dopants to ceria, the divalent alkaline earth ions and the trivalent rare earth ions have been investigated thoroughly because those materials show excellent oxygen ionic conductivity due to the formation of oxygen vacancies, although the oxygen ionic conductivity in pure ceria is not high. The oxygen ionic conductivity of alkaline earth doped ceria is smaller than that of rare earth doped ceria except for a calcia-doped ceria. This is mainly because of the poor solubility of alkaline earth into ceria. Yahiro et al. investigated the ionic conductivity of 20 mol% rare earth oxide doped ceria electrolytes (Ce0.8Ln0.2O1.9, Ln = Yb, Y, Ho, Dy, Gd, Sm, Nd, La) in air at 1073 K, as shown in Fig. 8.7.26 The ionic conductivity increases with ionic radius from Y to Sm, whereas from Sm to La it decreases. As a result, Ce0.8Sm0.2O1.9 showed the highest conductivity in the Ce0.8Ln0.2O1.9 system, followed by Ce0.8Gd0.2O1.9. According to Steele,13 gadlinia doped ceria (GDC) shows the highest conductivity in the Ce0.8Ln0.2O1.9 system, in contradiction to the above results. Among GDC electrolytes, Ce0.9Gd0.1O1.95 has the highest oxygen ionic conductivity of 0.054, 0.025 and 0.0095 at 700, 600 and 500°C, respectively. This apparent disagreement may be due to the well-known fact that the total oxygen ionic conductivity in a doped ceria system is affected

224


Sm

log (σ.T/S.cm–1.K)

2.0

Gd Dy Y

1.8

Yb

Ho Nd

La 1.6

0.10 0.11 0.12 Radius of dopant cation/nm

8.7 Dependence of ionic conductivity for (CeO2)0.8(LnO1.5)0.2 at 1073 K on radius of dopant cation (reprinted from Ref. 25 with permission from Elsevier).

greatly by the poor grain boundary conductivity that is caused directly by the fabrication process. This grain boundary effect increases with decreasing temperature. At present, samaria- and gadolinia-doped ceria electrolytes can be selected from the conductivity point of view, whereas yttria-doped ceria will be selected from the cost point of view. The most crucial phenomenon in doped ceria is the partial reduction of Ce4+ ions into Ce3+ ions under reducing atmospheres. The reduction of cerium ions significantly increases the electronic conductivity, and also leads to the expansion of crystal lattice due to the formation of oxygen vacancies and the creation of Ce3+ having larger ionic size. For 10GDC and 20GDC, Yasuda and Hishinuma precisely reported the electronic conductivity and the relative lattice expansion as a function of oxygen partial pressure in a temperature region of 600 to 1000°C.19 The ratio of the electronic conductivity to the total conductivity in Ce0.9Gd0.1O1.95 is larger than that in Ce0.8Gd0.2O1.9 in the whole temperature range. This means that the cerium ions in Ce0.9Gd0.1O1.95 can be reduced more easily than in Ce0.8Gd0.2O1.9. This tendency becomes weak with decreasing temperature. This predominantly electronic conductivity under a practical fuel environment of SOFCs shown in Fig. 8.8 makes the use of gadlinia-doped ceria inappropriate for SOFCs above 600°C. Note also that Fig. 8.6 shows the significant decrease of efficiency due to the oxygen permeation (in other words, electron conduction) through ceria. The chemical volume expansion due to reduction shown in Fig. 8.9 increases with increasing temperature and decreasing oxygen partial pressure. Above 700°C, the volume significantly expands under reducing atmospheres. However, the critical p (O2) where the volume starts to expand decreases with decreasing temperature, and eventually at 600°C, no measurable volume expansion takes place under a practical fuel environment of SOFCs.

Oxygen ionic conductor 0.5

1000 °C

900 °C

0.0 800 °C

log (σ/Scm–1)

–0.5

225

Closed: CGO10 Open: CGO20

700 °C

–1.0 –1.5 –2.0 –2.5 600 °C –3.0 –24

–20

–16 –12 –8 log (Po2/atm)

–4

0

8.8 Electrical conductivity of 10GDC and 20GDC as a function of oxygen partial pressure at 600 to 1000°C (reproduced from Ref. 19 by permission of The Electrochemical Society, Inc.).

2.0 CGO20 1000°C CGO20 900°C CGO20 800°C CGO20 700°C CGO20 600°C CGO10 1000°C CGO10 900°C CGO10 800°C CGO10 700°C CGO10 600°C

∆L/L [%]

1.5

1.0

0.5

1000

900

800

700

600

0.0 log PO2 In fuel (H2O/CH4 = 2)

–0.5 –25

–20

–15 –10 log (PO2/atm)

–5

0

8.9 Relative expansion of 10GDC and 20GDC as a function of oxygen partial pressure at 600 to 1000°C (reproduced from Ref. 19 by permission of The Electrochemical Society, Inc.).

226


In order to mitigate the effect of the electronic conductivity, it is proposed to reduce the operating temperature of SOFCs below 600°C. From recent thermodynamic and electrical conductivity data, Steele evaluated the most appropriate electrolyte composition for IT-SOFC operation at 500°C and selected Ce0.9Gd0.1O1.95.13 According to him, Ce0.9Gd0.1O1.95 has an ionic lattice conductivity of 10–2 S cm–1 at 500°C and the Gd3+ ion is the preferred dopant compared to Sm3+ and Y3+ at this temperature. The I-V characteristics of single cells incorporating 25 µm thick electrolytes were modelled, and the requirements for composite electrode discussed briefly. From these considerations, a power density of 400 mW cm–2 at 500°C could be evaluated. Even so, the effect of lowering the energy conversion efficiency due to the high electronic conduction cannot be neglected even below 600°C as shown in Fig. 8.6. As described above, the utilization of doped ceria alone as the electrolytes for SOFCs is inappropriate from the viewpoint of efficiency. Instead, doped ceria can be used with other electrolyte in oxidative atmospheres. The most successful application is an interlayer between cathode materials and stabilized zirconia electrolytes. The high potential cathode materials of (La,Sr)MO3 (M = Co, Fe) tend to react easily with stabilized zirconias during hightemperature fabrication processes, leading to the formation of undesired secondary phases at the interfaces. Chen et al. fabricated a buffer layer with a samaria-doped ceria (Ce0.8Sm0.2O1.9, SDC) between a YSZ electrolyte and a (La,Sr)(Co,Fe)O3 (LSCF) cathode.27 Undesired secondary phases of La2Zr2O7 and SrZrO3 are formed at the interface of YSZ/LSCF without the buffer layer. On the other hand, SDC does not react with LSCF; therefore, the buffer layer worked as a protective layer to prohibit the reaction between YSZ and LSCF. In recent IT-SOFCs, stabilized zirconia electrolytes are used with a doped-ceria interlayer between the electrolyte and the cathode. Ceria-based oxides have attracted attention because of their high catalytic activities enhancing the electrochemical reactions. Some research implies that the high activity depends on an interaction between proton and ceriabased electrolyte. In order to identify the relation between protons and ceria, Sakai et al. evaluated hydrogen solubility in rare earth doped cerias Ce0.8M0.2O1.9 (M = Yb, Y, Gd, Sm, Nd and La) in combination with isotope exchange technique using deuterium oxide (D2O) and subsequent analysis by secondary ion mass spectrometry (SIMS).28 According to them, hydrogen is soluble in doped cerias compared with YSZ or pure ceria, and the solubility in doped ceria increases drastically with decreasing dopant size.

8.4.2

Doped lanthanum gallate

Perovskite-type oxides can be described as ABO3 which consists of the Asite, the B-site and the oxygen sites. A large number of oxygen vacancies can


227

be formed by doping lower valence cations into the A- and/or the B-sites. Takahashi et al. developed a doped lanthanum alminate as an oxygen ionic conductor about 30 years ago.29 In their report, LaAlO3 doped with Ca into the A-site exhibits a pure oxygen ionic conductivity over a wide range of p(O2); however, the ionic conductivity is lower than that of stabilized zirconia. About twenty years later, in 1994, Ishihara et al. found that a doped lanthanum gallate is a pure oxygen ionic conductor with extremely high electrical conductivity over a wide range of p(O2).15 Hence, a series of doped lanthanum gallate electrolytes are only ten years old, and therefore investigations on those materials and their performances and characteristics are still continuing most extensively and intensively among oxygen ionic conductors. According to Ishihara, effects of alkaline earth cations added for the La site in LaGaO3 on the electrical conductivity were first investigated. The electrical conductivity depends strongly on the alkaline earth cations and increases in the following order: Sr > Ba > Ca. The electrical conductivity increases with the amount of Sr added and attained the maximum value at x = 0.1 in La1–xSrxGaO3–δ; because the solubility of Sr into the La-site was limited at x = 0.1 when strontium alone was added to LaGaO3. As a next step, the effect of additives at the Ga-site of La0.9Sr0.1GaO3 on the electrical conductivity was studied, and the electrical conductivity was improved by doping with Mg, Al and In. The addition of Mg was the most effective method to increase the electrical conductivity among three dopants, and La0.9Sr0.1Ga0.8Mg0.2O3 (denoted to LSGM9182) was recognized as the most desirable candidate. In his subsequent reports, it was clarified that the solubility limit of Sr into the La site was enhanced by doping with Mg to Ga site, and that La0.8Sr0.2Ga0.8Mg0.2O2.8 shows the highest conductivity in a family of LaxSr1–xGayMg1–yO3–δ (denoted to LSGM) electrolytes.30 Furthermore, in a detailed investigation by Huang and Gooderough, the highest values of the oxygen ionic conductivity were found for La0.8Sr0.2Ga0.83Mg0.17O2.815 at 600– 800°C; the values are 0.17, 0.08 and 0.03 S cm–1 at 800, 700 and 600°C, respectively.31 Further improvement of the oxygen ionic conductivity for LSGM electrolytes was attained by doping with transition metal oxides. Ishihara et al. chose a strategy of double doping for the Ga site. The electrical conductivity of La0.8Sr0.2Ga0.8Mg0.1M0.1O3–δ (M = Ni, Co, Mn, Fe and Cu) increases by doping with Co, Fe and Ni, but decreases by doping with Cu and Mn in comparison with non-doped LSGM of La0.8Sr0.2Ga0.8Mg0.2O2.8 as shown in Fig. 8.10. Among transition metal cations examined, Co and Fe were found to be the most effective in increasing the oxygen ionic conductivity. The effect of a doped amount of Co for La0.8Sr0.2Ga0.8Mg0.2O2.8 was subsequently investigated by them. The electronic conductivity of La0.8Sr0.2Ga0.8Mg0.2–xCoxO2.8 (denoted LSGMC) increased with the amount of Co; however, the electronic contribution also increases. Therefore, an

228


log (σ/S cm–1)

0.0

–1.0

La0.8Sr0.2Ga0.8Mg0.2-xMxO3 –2.0

–3.0 0.7

Ni Co Mn Fe Cu None 0.9

1.1 1000/T/K–1

1.3

8.10 Arrhenius plots of electrical conductivity of La0.8Sr0.2Ga0.8Mg0.1M0.1O3–δ (M = Fe, Co, Ni, Cu and Mn) and La0.8Sr0.2Ga0.8Mg0.2O2.8 at p(O2) = 10–5 atm (reprinted from Ref. 16 with permission from Elsevier).

excess amount of Co is not desirable for SOFC electrolytes, and it was concluded that the most promising composition is La0.8Sr0.2Ga0.8Mg0.115Co0.085O3–δ above 800°C. However, it is noted that the transference number of oxygen ion in La0.8Sr0.2Ga0.8Mg0.115Co0.085O3–δ rapidly decreases below 750°C so that the amount of doped Co in LSGMC should be reduced with decreasing applied temperature. The electronic conductivity of a typical LSGM electrolyte without transition metal dopants is lower than the oxygen ionic conductivity as shown in Fig. 8.3. The transference number of oxygen ion is over 99% in air below 900°C and increases with decreasing temperature. In fuel atmospheres of SOFCs, the transference number is over 99% at 800–1000°C. Accordingly, LSGM electrolytes have high energy conversion efficiency and are the major candidate electrolyte for IT-SOFCs due to the superior oxygen ionic conductivity, as shown in Figs 8.4 and 8.6. The most critical characteristic feature for LSGM electrolytes is their chemical stability in reducing atmospheres.32,33 We confirmed that holding LSGM electrolytes in reducing atmospheres at high temperatures gave rise to Ga depletion from the electrolyte surface because of the vaporization of the Ga component in gaseous forms such as Ga2O. After annealing LSGM9182 electrolyte in a humidified hydrogen atmosphere of 3%H2O-97%H2 at 1000°C, the surface of the electrolyte became porous and some secondary phases such as La2O3 and LaSrGaO4 were formed, followed by the Ga depletion. After annealing even at 800°C for 10 h, a slight amount of Ga was depleted


229

from the electrolyte surface. Subsequently, the amount of vaporized Ga component in LSGM electrolyte was investigated in detail as functions of composition, atmosphere and temperature. The Ga depletion was moderated with increasing p(O2). Doping with Sr for La site causes serious Ga depletion from the electrolyte, but Co doping with Mg for Ga site weakened the enhancement of Ga depletion. The amount of Ga depletion decreased with decreasing temperature, and no Ga depletion was observed after annealing at 750°C for 10 h. These results suggest that the LSGM electrolyte is applicable to SOFCs at least below 750°C and the operating atmosphere should be carefully controlled. At an earlier stage in the development of SOFCs with LSGM electrolytes, the reactivity between the electrolyte and the cathode/anode materials was well recognized. Perovskite-type cathodes such as (La,Sr)MnO3–δ easily react or interdiffuse with LSGM electrolytes at high temperatures; however, such phenomena were weakened by decreasing the preparation temperatures with using a cobaltite-based electrode with lower sintering temperatures. Ni-cermet anode was also known to react with LSGM electrolytes when they are cosintered around 1400°C in air, although the sintering temperatures above 1400°C are required to fabricate LSGM electrolytes. Consequently, LSGM electrolytes are applied to electrolyte supported SOFCs now; Elangovan et al. have recently made some progress in preparing electrode supported SOFCs with LSGM electrolyte and NiO-based anode.34

8.4.3

Scandia stabilized zirconia

Scandia (Sc2O3) doped zirconia has the highest oxygen ionic conductivity in a family of doped-zirconia electrolytes.35,36 Similarly to other dopant zirconia systems, those having the cubic structure show the highest conductivity, and are usually called scandia stabilized zirconia (denoted as ScSZ). On the other hand, the compositional range where the cubic phase exists stably is narrow, that is about 8–9 mol% scandia in the scandia–zirconia binary system. In addition, the decrease in conductivity on ageing is significant for materials in those compositions. The electric conductivity did not decrease on ageing for 11 mol% Sc2O3-ZrO2; however, the conductivity decreases rapidly below 600–700°C because of the phase transition from the cubic to the rhombohedral phases. In addition, existence of the phase transition is not suitable for SOFC electrolytes because many thermal cycles between room and operating temperatures are required through the phase transition temperature. In view of this, a doping strategy was adopted to stabilize the cubic phase of 10–12 mol% Sc2O3 doped zirconia at lower temperatures. Doping with a large amount of oxides such as Al2O3 can stabilize the cubic phase, but the electric conductivity severely decreases by such doping.36 Ishii et al. successfully stabilized the cubic zirconia phase at lower temperatures by

230


adding a small amount of Al2O3.37 In a 88 mol% ZrO2–11.5 mol% Sc2O3– 0.5 mol% Al2O3 system, the phase transition was not observed above 500°C. Ukai et al. investigated the electrical and mechanical properties of the three types of ScSZ (88 mol% ZrO2–11 mol% Sc2O3–1 mol% Al2O3, 89 mol% ZrO2–10 mol% Sc2O3–1 mol % CeO2 and 89 mol% ZrO2–10 mol% Sc2O31 mol% Y2O3).14 Among them, CeO2-doped ScSZ has the highest conductivity and the highest mechanical property (bending strength is 336 MPa and fracture toughness is 2.0 MPa m0.5). Recently, Hirano et al. have reported the effect of BiO2 additives in 10ScSZ electrolyte; adding 1 mol% Bi2O3 decreased the sintering temperature of 10ScSZ to 1200°C and inhibited the cubic-rombohedral phase transition.38 The electrical conductivity of 0.12 S cm–2 at 1073 K was comparable to LSGM and doped ceria electrolytes. On the other hand, Badwal et al. have shown that some ScSZ electrolytes with Sc2O3 content above 9 mol% show relatively low degradation of the electronic conductivity on ageing at 850 and 1000 °C.39 The ScSZ with 9.3 mol% Sc2O3 has a maximum oxygen ionic conductivity and has no phase transition from the cubic to the rhombohedral phase. This result suggests that pure ScSZ also offers a good alternative for IT-SOFC electrolytes. A major problem often associated with the utilization of ScSZ electrolytes is the high material cost. The price of Sc2O3 was relatively high; however, the cost has gradually reduced, and was about US$1000/kg as of 2000.40 If electrode-supported structures are applied to ScSZ electrolytes, the material cost will decrease drastically by reducing the amount of material related to the thin electrolyte thickness. On the other hand, the mechanical strength of 11ScSZ is comparable to that of a conventional 8YSZ electrolyte. The high mechanical stability of ScSZ is the most attractive characteristic as compared with the other high oxygen ionic conductivity such as LSGM and dopedceria electrolytes.

8.4.4

Other candidates

As described above, some oxides having the fluorite-type structure (stabilized zirconia, doped-ceria, etc.) and the perovskite structure (LSGM, etc.) are known to have a relatively high oxygen ionic conductivity. Meanwhile, Nakayama and Sakamato have developed a new family of oxygen ionic conductors, RExSi6O12+1.5x (RE = La, Nd, Sm, Gd and Dy, x = 8–11), having a hexagonal apatite structure.41 Among them, La10Si6O27 showed the highest conductivity. The oxygen ionic conductivity measured in ambient air was relatively low at higher temperatures, but was higher than that of 8YSZ below 600°C because of the low activation energy. While the oxygen ionic conductivity is not relatively high in comparison with LSGM and doped ceria, it was very attractive that those oxides exhibited high oxygen ionic conductivity in spite of the low symmetry of the crystal lattice.


231

Ishihara et al. were highly interested in the above apatite structure oxides with high oxygen ionic conductivity and started to investigate La10Si6O27and La10Ge6O27-based electrolytes.42 The electrical conductivity increased by doping La site with Sr reaching a maximum value at x = 0.25 in La10–xSrxSi6O27–x/2 and was higher than that of YSZ in all investigated temperature ranges. La10Ge6O27 showed high oxygen ionic conductivity which was comparable to those of LSGM electrolytes above 750°C, but the conductivity decreased rapidly around 700°C and was lower than that of La10Si6O27 below 600°C. By doping St for La site of La10Ge6O27, the electric conductivity at low temperatures was improved, and the electric conductivity of La9Sr1Ge6O26.5 was also higher than that of 8YSZ in all temperature ranges. The electrical conductivity of La10Si6O27- and La10Ge6O27-based electrolytes was independent of the oxygen partial pressure between 1 to 10–21 atm at 900°C and the transference number of oxide ion was nearly unity. Subsequently, Ishihara et al. noted La2GeO5-based oxides which have the monoclinic crystal structure.18,43 They indicated that La10Ge6O27 can be included in a family of La-deficient La2GeO5 electrolytes. The highest oxygen ionic conductivity at high temperatures was obtained at the composition of La1.61Ge6O4.41. The transference number of oxide ion was almost unity and the oxygen ionic conductivity was as high as 0.2 S cm–1 at 950°C, which is a similar value to that of LSGM9182. On the other hand, the Arrhenius plots of the electrical conductivity exhibited a large knee around 725°C, and the electrical conductivity decreased rapidly below that temperature. However, the knee disappeared by doping Ca, Sr, or Ba for La site in La-deficient La2GeO5, and in particular, it became clear that Ca is most effective in increasing the electrical conductivity in the low temperature range. Although conductivity at a temperature higher than 750°C slightly decreased by Ca doping, La 1.5Ca 0.2GeO 4.45 exhibited a high conductivity over all the temperatures. Consequently, La10Si6O- and La2GeO5-based new oxygen ionic conductors having the highly anisotropic crystal structures are a new family of the fast oxygen ionic conductors. Although the crystal structure is not identified well and the ionic conductivity is lower than those of optimized LSGM-based and doped-ceria electrolytes, the development of those electrolytes is presently at an initial stage. Therefore, further improvement of the oxygen ionic conductivity by optimizing the composition will be expected for La10Si6Oand La2GeO5-based electrolytes in the near future.

8.5

Conclusion

Recently, the fast oxygen ionic conductors, having a higher oxygen ionic conductivity than YSZ, have been required in the field of SOFCs to lower

232


the operating temperature below 800°C. With this mind, some candidates for intermediate temperature SOFC electrolytes were described in this chapter. In order to evaluate those electrolytes on energy conversion efficiency in SOFCs, we focused on the relation between the oxygen ionic conduction and the electronic conduction. At around 700°C, doped lanthanum gallates exhibit the most excellent characteristics of all oxygen ionic conductors because of the suitable oxygen ionic conduction and relatively low electronic conductivity. However, the chemical, thermodynamic, and mechanical stabilities of LSGM electrolytes are lower than those of stabilized zirconia electrolytes. Therefore, scandia stabilized zirconias are also attractive as SOFC electrolytes around 700°C. It is difficult to utilize doped-ceria electrolytes solely for SOFCs above 600°C because of the low energy conversion efficiency. However, those electrolytes will be candidates for SOFC electrolytes with doped lanthanum gallate electrolytes if SOFCs can be operated at temperatures below 500°C.

8.6

References

1. Nernst, W., ‘Demonstration Eines Neuen Widerstandmaterials’, Z. Electrochem., 1889 6 41. 2. Steele, B.C.H. and Heinzel, A., ‘Materials for Fuel-cell Technologies’, Nature, 2001 414 345–52. 3. Yokokawa, H., ‘Recent Developments for Solid Oxide Fuel Cell Materials’, Fuel Cells, 2001 1 1–15. 4. Baumard, J.F. and Abelard, P., ‘Defect Structure and Transport Properties of ZrO2Based Electrolytes’, Science and Technology of Zirconia II, ed. Claussen, N., Rühle, M. and Heuer, A.H., American Ceramic Society, Inc., Columbus, Ohio, pp. 555–71 (1984). 5. Hebb, M.H., ‘Electric Conductivity of Silver Sulfide’, J. Chem. Phys., 1952 20(1/12) 185–90. 6. Wagner, C., ‘Galvanic Cells with Solid Electrolytes Involving Ionic and Electronic Conduction’, Proc. Int. Comm. Electrochemical Thermodynamics and Kinetics, 7th Meeting, Butterworth, London, 1957 361–77. 7. Weppner, W., ‘Electronic Transport Properties and Electrically Induced p-n Junction in ZrO2 + 10 m/o Y2O3’, J. Solid State Chem., 1977 20 305–14. 8. Park, J.H. and Blumenthal, R.N., ‘Electric Transport in 8 Mole Percent Y2O3-ZrO2’, J. Electrochem. Soc., 1989 136(10/12) 2867–76. 9. Kawada, T., Sakai, N., Yokokawa, H. and Dokiya, M., ‘Electrical Properties of Transition-metal-doped YSZ’, Solid State Ionics, 1992 53–56 418–25. 10. Yamaji, K., Horita, T., Ishikawa, M., Sakai, N., Yokokawa, H. and Dokiya, M., ‘Some Characteristics for Fabrication of LaGaO3-based Electrolyte’, Proc. of Solid Oxide Fuel Cells V, Aachen, Germany, 1997, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV97-24, 1041–50. 11. Xiong, Y., Yamaji, K., Sakai, N., Horita, T. and Yokokawa, H., ‘Electronic Conductivity of 20 mol % YO1.5 Doped CeO2’, J. Electrochem. Soc., 2002 149(11/12) E450–4. 12. Choudhury, N.S. and Patterson, J.W., ‘Solid-State Chemical Potential Profiles in


233

Solid Electrolytes’, J. Electrochem. Soc., 1970 117(11/12) 1384–88. 13. Steele, B.C.H., ‘Appraisal of Ce1–yGdyO2–y/2 Electrolytes for IT-SOFC Operation at 500 °C’, Solid State Ionics, 2000 129 95–110. 14. Ukai, K., Mizutani, Y., Kawai, M. and Nakamura, Y., ‘Solid Oxide Fuel Cell Using Sc-Doped Zirconia Electrolytes’, Proc. of the 3rd Int. Fuel Cell Conf., Nagoya, Japan, Fuel Cell Development and Information Centor (FCDIC), Tokyo, Japan, 1999 441–4. 15. Ishihara, T., Matsuda, H. and Takita, Y., ‘Doped LaGaO3 Perovskite Type Oxide as a New Oxide Ionic Conductor’, J. Am. Chem. Soc., 1994 116(9/12) 3801–3. 16. Ishihara, T., Akbey, T., Furutani, H. and Takita, Y., ‘Improved Oxide Ion Conductivity of Co Doped La0.8Sr0.2Ga0.8Mg0.2O3 Perovskite Type Oxide’, Solid State Ionics, 1998 113–115 585–91. 17. Abraham, F., Boivin, J.C., Mairesse, G. and Nowogrocki, G., ‘The BIMEOX Series: A New Family of High Performances Oxide Ion Conductors’, Solid State Ionics, 1990 40/41 934–7. 18. Ishihara, T., Arikawa, H., Nishiguchi, H. and Takita, Y., ‘Fast Oxide Ion Conductivity and Oxygen Tracer Diffusion in Doped La2GeO5–δ’, Solid State Ionics, 2002 154– 155 455–60. 19. Yasuda, I. and Hishinuma, M., ‘Electrical Conductivity, Dimensional Instability and Internal Stresses of CeO2-Gd2O3 Solid Solutions’, Proc. of Solid Oxide Fuel Cells V, Aachen, Germany, 1997, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV9724, 178–87. 20. Lewis, G., Brandon, N., O’Dea S. and Steele, B.C.H., ‘Metal Supported IT-SOFCs for Operation at 500–600C’, Extended Abstract of Fuel Cell Seminar 2003, Miami Beach, FL, U.S. DOE, 2003 431–4. 21. Oishi, N., Rudkin, R., Steele, B.C.H., Atkinson, A., Brandon, N.P. and Kilner, J.A., ‘Stainless Steel Supported Thick Film IT-SOFCs for Operation at 500–600°C’, Proc. of Int. Conf. of Electrophoretic Deposition: Fundamentals and Applications, Banff, Canada, 2002, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV2002-21, 230–7. 22. Yamada, T., Akikusa, J., Murakami, N., Akbey, T., Miyazawa, T., Adachi, K., Hasegawa, A., Yamada, M., Hoshino, K., Hosoi, K. and Komada, N., ‘Development of Intermediate-Temperature SOFC Module Using Doped Lanthanum Gallate’, Proc. of Solid Oxide Fuel Cells VIII, Paris, France 2003, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV2003-07, 113–18. 23. Sumi, H., Ukai, K., Hisada, K. and Mizutani, Y., ‘High Performance Cell Development Using Scandia Doped Zirconia Electrolyte for Low Temperature Operating SOFCs’, Proc. of Solid Oxide Fuel Cells VIII, Paris, France, 2003, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV2003-07, 995–1002. 24. Ukai, K. and Hirakawa, M., ‘Development of 1 kW SOFC System with Cubic Scandia Stabilized Zirconia’, Nenryoudenchi, 2004, 3 (3) 41–43, in Japanese. 25. Mogensen, M., Sammes, N.M. and Tompsett, G.A., ‘Physical, Chemical and Electrochemical Properties of Pure and Doped Ceria’, Solid State Ionics, 2000 129 63–94. 26. Yahiro, H., Eguchi, K. and Arai, H., ‘Electrical Properties and Reducibilities of Ceria-Rare Earth Oxide Systems and Their Application to Solid Oxide Fuel Cell’, Solid State Ionics, 1989 36 71–5. 27. Chen, C.C., Nasrallah, M.M. and Anderson, H.U., ‘Cathode Electrolyte Interactions and Their Expected Impact on SOFC Performance, Proc. of Solid Oxide Fuel Cells

234

28.

29.

30.

31.

32.

33.

34.

35. 36.

37.

38.

39.

40. 41.

42. 43.

Materials for energy conversion devices III, Honolulu, HI, 1993, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV934, 598–612. Sakai, N., Yamaji, K., Horita, T., Yokokawa, H., Hirata, Y., Samashima, S., Nigara, Y. and Mizusaki, J., ‘Determination of Hydrogen Solubility in Oxide Ceramics by Using SIMS Analyses’, Solid State Ionics, 1999 125 325–31. Ishihara, T., Minami, H., Matsuda, H. and Takita, Y., ‘Application of the New Oxide Ionic Conductor, LaGaO3, to the Solid Electrolyte of Fuel Cells’, Proc. of Solid Oxide Fuel Cells IV, Yokohama, Japan, 1995, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV95-1, 344–53. Takahashi T. and Iwahara H., ‘Ionic Conduction in Perovskite-type, Oxide Solid Solution and its Application to the Solid Electrolyte Fuel Cell, Energy Conversion, 1971 11 105–11. Huang, K. and Goodenough, J.B., ‘A Solid Oxide Fuel Cell Based on Sr- and Mgdoped LaGaO3 Electrolyte: The Role of a Rare-Earth Oxide buffer’, J. Alloys and Compounds, 2000 303–304 454–64. Yamaji, K., Horita, T., Ishikawa, M., Sakai, N. and Yokokawa, H., ‘Chemical Stability of the La0.9Sr0.1Ga0.8Mg0.2O3 Electrolyte in a Reducing Atmosphere’, Solid State Ionics, 1999 121 217–24. Yamaji, K., Negishi, H., Horita, T., Sakai, N. and Yokokawa, H., ‘Vaporization Process of Ga from Doped LaGaO3 Electrolytes in Reducing Atmospheres’, Solid State Ionics, 2000 135 389–96. Elangovan, S., Balagopal, S., Larsen, D., Timper, M., Pike, J. and Heck, B., ‘Lanthanum Gallate Electrolyte for Intermediate Temperature Operation’, Proc. of Solid Oxide Fuel Cells VIII, Paris, France, 2003, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV2003-07, 299–303. Stricker, D.W. and Carlson, W.G., ‘Electrical Conductivity in the ZrO2-rich Region of Several M2O3–ZrO2 Systems’, J. Am. Ceram. Soc., 1965 48 286–89. Yamamoto, O., Kawahara, T., Takeda, Y., Imanishi, N. and Sakaki, Y., ‘Zirconia Based Oxide Ion Conductors in Solid Oxide Fuel Cells’, in Science and Technology of Zirconia V, Technomic Publishing Co. Inc., 1993 733–41. Ishii, T., Iwata, T. and Tajima, Y., ‘Cubic Phase Stabilization and High Ion Conductivity in ZrO2-Sc2O3-Al2O3 System’, Proc. of Solid Oxide Fuel Cells III, Honolulu, HI, 1993, The Electrochem. Soc. Proc. Series, Pennington, NJ, PV93-4, 59-64. Hirano, M., Oda, T., Ukai, K. and Mizutani, Y., ‘Effect of Bi2O3 Additives in Sc Stabilized Zirconia Electrolyte on a Stability of Crystal Phase and Electrolyte Properties’, Solid State Ionics, 2003 158 215–23. Badwal, S.P.S., Ciacchi, F.T. and Milosevic, D., ‘Scandia-Zirconia Electrolytes for Intermediate Temperature Solid Oxide Fuel Cell Operation’, Solid State Ionics, 2000 136–137 91–99. Badwal, S.P.S. and Ciacchi, F.T., Oxygen-Ion Conducting Electrolyte Materials for Solid Oxide Fuel Cells’, Ionics, 2000 6 1–21. Nakayama, S. and Sakamoto, M., ‘Electrical Properties of New Type High Oxide Ionic Conductor RE10Si6O27 (RE = La, Pr, Nd, Sm, Gd, Dy)’, J. Euro. Ceram. Soc., 1998 18 1413–18. Akikusa, H., Nishiguchi, H., Ishihara, T. and Takita, Y., ‘Oxide Ion Conductivity in Sr-doped La10Ge6O27 Apatite Oxide’, Solid State Ionics, 2000 136–137 31–37. Ishihara, T., Arikawa, H., Akbey, T., Nishiguchi, H. and Takita, Y., ‘Nonstoichiometric La2–xGeO5–d Monoclinic Oxide as a New Fast Oxide Ion Conductor’, J. Am. Chem. Soc, 2001 123 203–9.

9 Defect chemistry of ternary oxides X - D Z H O U and H U A N D E R S O N, University of Missouri-Rolla, USA

9.1

Introduction

The perovskite or pseudo-perovskite structure class of oxides is very important since many of them are utilized in electrochemical processes. The structure is basically cubic with the general formula ABO3, in which A, the large cation site, may be an alkali, alkaline earth, or rare earth ion, and B, the small cation site, a transition metal cation. The large cations are in 12-fold coordination with oxygen while the small cations fit into octahedral positions. Since these two sites are very different in size, the occupancy of these sites is determined primarily by ionic size rather than valency, so it is possible to substitute selectively for either the A or B ion by introducing isovalent or aliovalent cations. This gives the materials scientist an opportunity to alter the properties of a given oxide by substituting different cations onto either the A or B site. The main criterion which must be followed is that the ionic radius of a substitution cation must be close (with ~15%) to a cation for which it substitutes without regard to valency. In the early 1950s, Verwey et al.1 observed that when substitutions are made on to the perovskite lattice, if the valence of the substituting ion is different (i.e., it is aliovalent) from that required by the site, then the charge imbalance and overall charge neutrality will be maintained by the formation of electrons, holes or charged point defects. The compensation process creates carriers which can take part in electrical conductivity. Thus, depending upon whether the basic oxide is a p- or n-type conductor, the substitution of either acceptors or donors, can increase or decrease the carrier concentration. This is very important because it has lead to a number of devices, such as NTC resistors, PTC resistors, electrodes for electrochemical devices, etc. When a perovskite which contains transition metal ions on the B site is heated to a sufficiently high temperature that it can equilibrate with the ambient oxygen activity, reversible changes in the oxygen content occur as the oxygen activity is varied. This behaviour occurs with both p- and n-type perovskites and represents a compensation mechanism in addition to that 235

236


presented by Verwey et al.1 This can cause the neutrality condition to change from electronic to ionic or vice versa. Thus to understand completely the defect behaviour in the perovskite oxides which contain transition metal cations, it is necessary to include the equilibration reaction with the ambient temperature and oxygen activity in addition to the influence of aliovalent effects. This occurs with all perovskites, but for the sake of brevity, in this review only acceptor doped p-type oxides will be considered with appropriate examples included.

9.2

Defect chemistry background

The defect chemistry of oxides has been studied for several decades, with significant success on binary compounds such as zirconia (ZrO2) and ceria (CeO2) based materials.2 The ternary compounds (ABO3), also known as the perovskite family of oxides, have been extensively studied from the early 1950s, particularly on BaTiO3,3–6 LaCrO3,7–10 LaMnO3,9, 11–16 LaFeO317,18 and LaCoO3.19–22 Perovskite type oxides are of great interest in energy conversion devices because • site occupancy is determined mainly by ionic size; so the site location of a particular cation is fairly certain • electronic conductivity (σ) is determined by the B site ion • ionic conductivity results from the motion of oxygen vacancies. Thus, this family of oxides has been tailored to be used as dielectrics, mixed ionic and electronic conductors superionic conductors and superconductors. Defect chemistry is the most important technique involved in gaining an understanding of the mass and charge transfer properties because it determines the defect type, density, defect associations and carrier mobility. Understanding the defect chemistry in these systems allows us to search for the novel materials used in fossil energy conversion systems, which requires a sufficient vacancy density, a mixed ionic and electronic conductivity, catalytic activity and thermodynamic stability. Examples include: • cathode components in solid oxide fuel cells (SOFCs), which require high oxygen vacancy density and electronic conductivity; • anode components in SOFCs, which require stability at a very reducing atmosphere, electronic conductivity and high oxygen vacancy density; • components in coal-gasfire steam electrolyzers, which require stability at a very reducing atmosphere, catalytic activity to dissociate H2O and oxygen ion conductivity; • novel sensors in fossil energy conversion systems, which require stability at a high temperature, very reducing, corrosive and harsh atmospheres; • membranes for oxygen separation which require stability at a high temperature, very reducing, corrosive and harsh atmospheres.

Defect chemistry of ternary oxides

237

As noted previously, the perovskite oxides can be represented by ABO3 where the charge related to the A and B sites is +6 with the valence of the B site cation ranging from +3 to +5. In the discussion here, both the A and B sites will be +3 which covers a number of the rare earth perovskites which are important for electrical conductivity and magnetic applications. For simplicity, the following assumptions are made 1. The A to B site ratio is one, with the A site occupied by a trivalent rare earth and the B site by trivalent transition metal ions: Cr, Fe, Mn, Co or mixtures thereof. 2. Only fully ionized oxygen vacancies are present. 3. No defect association. 4. No interstitial defects. 5. A divalent acceptor ion, I, can substitute on either the A or B site with the A to B site ratio remaining unity. These assumptions can be quite restrictive, but generally nonconformity to them does not affect the general predicted behaviour, for example, when defect association occurs, carrier concentration will be altered, but the overall predicted behaviour is still valid.

9.3

Determination of stoichiometry

Quantitative determination of the stoichiometry of non-stoichiometric oxides enables a validation study of defect chemistry models and thus lies at the heart of understanding the defect chemistry. Stoichiometry can be studied and understood through the direct and indirect techniques: 1. Direct study is mainly by thermogravimetric analysis. 2. Indirect studies can be chemical titration, electrical conductivity, Mössbauer spectrometry, neutron diffraction, x-ray diffraction or magnetic moments. This section will focus on various studies used in our laboratory, including thermogravimetric analysis, neutron diffraction, and Mössbauer spectrometry.

9.3.1

Thermogravimetric analysis

Examples of the rare earth chromites are contained in the studies by Flandermeyer23 who studied La(Mg,Cr)O3. Figure 9.1 gives an example of data which show that the simplified models fit their results quite well. In Flandermeyer’s data, evidence of defect association was noted by Van Roosmalen et al.24 who improved the fit of the experimental data to the model by including association. This indicates the importance of defect association. However, the simplified model described the overall behaviour quite well.

238


0.100 LaMg0.20Cr0.80Q3 LaMg0.10Cr0.90Q3 LaMg0.05Cr0.95Q3 LaMg0.02Cr0.98Q3

0.090

Moles oxygen deficit

0.080 0.070

temp = 1255°C

0.060 0.050 0.040 0.030 0.020 0.010 0.000 –13

–11

–9

–7

–5 –3 Log Po2 (Pascal)

–1

1

3

5

9.1 Moles oxygen lost per mole sample as a function of log PO2 and dopant content at 1255°C.23

Figure 9.2 shows results for (La0.8Sr0.2)MnO325 which are typical for the Mn-containing perovskites. The simplified model appears to fit the experimental data quite well. However, the model predicts a constant oxygen stoichiometry of 2.9 which does not occur. In order to account for the observed behaviour,

Oxygen content (mole)

3.10

3.00

2.90 1000 °C LaMnO3 La.99Sr.01MnO3 La.95Sr.05MnO3 La.90Sr.10MnO3 La.80Sr.20MnO3

2.80

2.70 –18

–16

–14

–12

–10 –8 Log(PO2/atm)

–6

–4

–2

0

9.2 Moles oxygen weight loss per mole sample vs. Log PO2 for various Sr-dopant levels. The solid lines are calculated from model.25


239

both Kuo et al.25 and Stevenson et al.26 had to invoke thermally excited disproportionation of Mn+3 to Mn+4 and Mn+2. Compositions within the (La,Sr)(Fe,Co)O3 family have been studied extensively because of their mixed electronic and ionic conductivity. Typical behaviour of this system is shown in Fig. 9.3.27 This family of compositions also follows the simplified model quite well, but as was the case with manganites, a region of constant stoichiometry was not observed. This is probably due to continuous reduction of the cations on the B site. It is interesting to note that for most of the compositions, dissociation does not occur until the oxygen stoichiometry reaches the 2.4–2.7 range. This suggests that for the 40% Sr composition, oxygen vacancy content as high as 20% can be expected. This is the reason that high oxygen ion conductivity is observed.

Oxygen content

3.0

2.5

2.0

1.5

1.0 –16

x=0 x = 0.2 x = 0.4 –14

–12

–10 –8 –6 Log oxygen activity

–4

–2

0

9.3 Oxygen content (moles) of La1–xSrxCo0.2Fe0.8O3–δ as a function of oxygen activity and Sr content (moles) at 1200°C.27

9.3.2

Neutron diffraction

Neutron diffraction is a powerful tool to characterize these oxides because it resolves not only the crystal structure, but also the magnetic properties and the oxygen vacancy concentration. Compared to x-ray diffraction, neutron diffraction possesses several significant advantages, including • the sensitivity of neutron scattering to such light atoms as oxygen is far greater than that of x-ray scattering because the coherent scattering of neutrons is only determined by the nucleus and is independent of the number of electrons • the neutron has a magnetic moment so it can probe the magnetic structures and excitations through a strong interaction. Many of the perovskite type oxides (ABO3) are magnetic oxides because of the unpaired electron(s) of the B site ions, such as Mn3+, Fe3+ and Co3+.28, 29

240


Specimens of La0.60Sr0.40FeO3 quenched from T < 1100°C temperature show a weak first peak and large rhombohedral distortion, while those quenched from T > 1100°C show a strong magnetic peak and nearly cubic structure.29 Figure 9.4 shows the neutron diffraction pattern of the sample quenched from 1500°C and Fig. 9.5 shows the neutron diffraction pattern for the sample without quenching. The first Bragg peak is purely magnetic and the change in its intensity at room temperature reflects both the increase in Curie temperature with increasing oxygen vacancy concentration and the change in average valence of the Fe atoms. The neutron refinements also yield those concentrations, which are shown in Fig. 9.6 as a function of refined vacancy concentration. It is clear that either the moment or the volume may be used to reliably determine the vacancy content, and these two track each other well. 19000 r–3c_mpl_1859.prf: Yobs Ycalc Yobo-Ycalc Bragg_position

16000

Intensity (a.u.)

13000 Magnetic contribution only

10000

7000

4000 1000

–2000 –5000 0

20

40

60 2θ (°)

80

100

120

9.4 Neutron diffraction pattern of La0.60Sr0.40FeO3–δ quenched at 1500°C to room temperature.

Figure 9.6 shows a plot of 3-δ vs. quenching temperature for La0.60Sr0.40FeO3–δ. A datum of ~3 for the specimen without quenching is shown in Fig. 9.6 as well. Oxygen content was determined directly from refinements of neutron diffraction results. A value around 2.8 was observed for La0.60Sr0.40FeO3–δ quenched from 1500°C, whereas full stoichiometry (δ ~ 0) was determined for La0.60Sr0.40FeO3–δ without quenching. From charge neutrality, it is evident that Fe is in the valence state of 3+ for La0.60Sr0.40FeO2.8


241

15000 13000

La0.60Sr0.40FeO3–δ without quenching

1: Nucleus Bragg position 2: Magnetic position

11000

Intensity (a.u.)

9000 Magnetic 7000 congtribution only 5000 3000 1000 –1000

2

–3000 –5000 0

20

40

60 2θ (°)

80

100

120

9.5 Neutron diffraction pattern of La0.60Sr0.40FeO3–δ without quenching. 3.00

3-δ

2.95

2.90

2.85 Quenching data from ND 2.80 600

Datum without quenching Quenching data from Mossbauer 800

1000 1200 1400 Quenching temperature (°C)

1600

9.6 Oxygen occupancy (3–δ) as a function of temperature for La0.60Sr0.40FeO3–δ.

and exhibits an average valence state of 3.4 for La0.60Sr0.40FeO3.0. Therefore, the magnetic moments for La0.60Sr0.40FeO3–δ are expected to be a function of δ. The magnetic moment and oxygen content can be determined independently by Rietveld refinement of neutron diffraction data. A strong correlation between oxygen deficiency and magnetic moment has been observed, which indicates that this technique can be used to resolve the oxygen content in perovskite

242


type ferrites. The saturation moments for La0.60Sr0.40FeO3–δ were determined from neutron powder diffraction measurements at 10K, as a function of 3–δ (Fig. 9.7).28 In this study, the highest deficiency (δ = 0.2) corresponds to a nearly pure Fe3+ state while for a stoichiometric composition, the fraction Fe3+ is 60% and that of Fe4+ is 40%. The magnetic moment for La0.60Sr0.40FeO2.8 is ~ 3.8 µB, which is a typical moment for Fe3+ in LaFeO3 system. The magnetic moment for La0.60Sr0.40FeO3.0 is ~ 2.3µB (~ 3.8µB × 60%). The magnetic moment is as expected, linear with vacancy concentration and can be used to determine oxygen content by direct crystallographic refinement. The use of the magnetic moment as a measure of the vacancy concentration has the advantage that it is quite precise and the uncertainty in magnetic moment is 2% at low vacancy concentration, decreasing to less than 1% when the moment is large. 3.8

La0.6Sr0.4FeO3–δ

Saturation moment (µ B)

3.6 3.4 3.2 3.0 2.8 2.6 2.4 2.2 0.00

0.05

0.10 δ

0.15

0.20

9.7 Correlation between the saturation magnetic moment and oxygen deficiency for La0.60Sr0.40FeO3–δ.28

9.3.3

Mössbauer spectrometer

Mössbauer spectroscopy can reveal information on chemical bonding, valence state and magnetic properties of Fe-containing systems (ferrites). For example, the isomer shift of Mössbauer spectra provides unequivocal information on the valence state of Fe and bonding between Fe and O, which relates directly to the electronic conductivity and the reaction between oxygen and the ferrites. The average isomer shift and hyperfine field values were used to study the valence state and hyperfine interaction in these compounds, from which the average Fe valence was obtained for each specimen. Oxygen content was then calculated, as shown in Fig. 9.6. The relative ratio of Fe3+ and Fe4+ ions


243

for unquenched La0.60Sr0.40FeO3–δ obtained from relative areas of the Mössbauer spectra is 64:36, indicating nearly zero oxygen vacancy content. This ratio changes to 70:30 for the specimen quenched from 800°C, showing an increase of oxygen vacancy concentration. As the quench temperature becomes higher than 900°C, the Fe4+ spectrum disappears (Fe4+ normally is non-magnetic with a single line) and the magnetic sextets become dominant. The spectra of the specimens quenched from T > 1200°C are particularly sharp, which represents an increase in the Fe magnetic ordering temperature and suggests a structural transformation in the sample. It is found that the valence state of Fe changes from 3.36 to 3.04, suggesting that the Fe valence states change from a mixture of Fe3+ and Fe4+ to about 96% Fe3+ as quenched at 1500°C. The change in the valence state of Fe results in an increase in both the hyperfine field and magnetic moment for the quenched samples. The oxygen content changes from 0.02 to 0.18 per formula after quenching at 1500°C. The oxygen content obtained from Mössbauer spectra are again consistent with those obtained from the neutron diffraction refinements. Nearly fully stoichiometry oxygen occupancy has been observed by Tai et al.30 and Stevenson et al.31 on (La,Sr)(Fe,Co)O3–δ. The accuracy for determination of nonstoichiometry lies at the heart of understanding the defect chemistry. Exact oxygen occupancy and valence state of B site cation are the most crucial parameters that allow development of the correct defect chemistry model and then the possibility to tailor the materials properties.

9.4

Defect chemistry modelling

9.4.1

Basic defect chemistry equilibria

Based on these assumptions and that p-type disorder prevails in nonstoichiometric ABO3 allows the development of a defect chemistry model. The procedure which is used is as follows. 1. List the basic defect reactions which can occur: intrinsic, stoichiometric, oxygen excess and oxygen deficient. 2. Write the overall neutrality relation. 3. Combine the resulting equations to yield a relationship which can be solved for particular temperature and oxygen activity regimes. Point defects in ternary oxide systems (ABO3) (Kroger-Vink notation is used)32 Transition metal cations from Cr, Mn, Fe, to Co on the B site will be emphasized. Thus, the occupants at A site can be A xA , VA′′′ and SrA′ (for simplicity, assume Sr is the low valence element substituting at the A site). The occupants at the

244


B site are B Bx , B ′B , B⋅B and VB′′′, where B Bx indicates a majority of B cations are in valence of 3+, B ′B represents that some B site cations are in 2+ state and B⋅B shows some B site cations are in 4+ valence state. The oxygen site can have two type of occupants, VO⋅ ⋅ and VOx . Nine equations are therefore needed in order to determine these nine defect concentrations. The equilibria of interest The Schottky defect reaction: nil = VA′′′ + VB′′′ + 3VO⋅⋅

9.1

[VA′′′][VB′′′][VO⋅⋅ ]3

9.2

KS =

The intrinsic electronic defect reaction: 2B Bx = B ′B + B⋅B Ki =

[B ′B ][B⋅B ]/[B Bx ] 2

9.3 9.4

The oxygen deficient reaction, a redox reaction, is the most important reaction in nonstoichiometric oxides which involves releasing and absorbing oxygen. For the ternary oxides, this reaction can be written as:

2B⋅B + O Ox → 1 O 2 + 2B Bx + VO⋅⋅ 2 K

⋅⋅ VO

=

[VO⋅⋅ ][B Bx ] 2 PO1/22 ([B⋅B ] 2 [O Ox ])

9.5

9.6

The oxygen excess reaction takes place at high oxygen activity, particularly in manganites. Cation vacancies are considered as the main defects for oxygen excess oxides. However, compensation of cation vacancies can be in the form of either electron holes or oxygen vacancies. The formation of electron holes, where B⋅B = h ⋅ :

3/2 O 2 = VA′′′ + VB′′′ + 3O Ox + 6h ⋅

9.7

K = [VC′′′] 2 p 6 PO–3/2 2

9.8

Mass balance of A site cations:

[A xA ] + [VA′′′] + [SrA′ ] = 1

9.9

Mass balance of B site cations: [B Bx ] + [B ′B ] + [B⋅B ] + [VB′′′] = 1

Oxygen site mass balance:

9.10


[O Ox ] + [VO⋅ ⋅ ] = 3

245

9.11

Low valence element substitution level in A1–xSrxBO3: [SrA′ ] = x

9.12

The electroneutrality relation is then: 3[VA′′′] + 3[VB′′′] + [B ′B ] + [SrA′ ] = [B⋅B ] + 2[VO⋅⋅ ]

9.4.2

9.13

Modelling

These basis equations can be used to give the overall behaviour of the defects in ABO3 as a function of temperature, oxygen activity and acceptor concentration. Two methods can be used to solve the resulting equations: 1. Divide into regions of particular neutrality conditions and solve for that particular region. 2. Do not use limiting conditions, but allow a computer to calculate a numerical solution to the overall equation using the total neutrality condition. In this discussion, both the particular solutions and the global solutions will be considered. Simple solution Since the details using the first method to develop the expressions for the defect equations have been reported previously, only the results are given here.33 Figure 9.8 illustrates how the defect concentrations change with oxygen activity over six regions of limited neutrality conditions (assuming the temperature is high enough for the attainment of thermodynamic equilibrium). Table 9.1 shows the predicted oxygen activity dependence for the six regions. Global solution – 1 An extension of this treatment in which the overall electron neutrality expression is used results in a term called the ability for oxygen vacancy generation (AOG). This method is considered as a model based on delocalized electron holes. Method 2 was originally given by Nowotny and Rekas15 and later by Poulsen,16 both treated LSM based on a model of localized electron holes. The treatment which is shown here is an extension of these two reports with an effort for the first time to analytically solve the defect concentration, and then to simulate and compare the results of chromites, manganites, ferrites and cobaltites. The total electrical conductivity, σ, is given by σ = Nµq,

9.14

246


Log ( [Vo••], n, p)

VI Decomposition

[VO••]

V Reduction (B3+ to B2+)

IV

III

[VO• •] =[I′]+[B′]/2 [VO••] =[I′]/2

II

[VO• •]≈pO2–1/2

I

[VO• •]∝pO2–1/8

σp ∝ p

[I′] n

σc ∝ p p

Oxygen content

•• High [VO ] Regions

3 3–[I′]/2

Log (pO2)

9.8 Defect concentration in acceptor substituted ABO3 as a function of oxygen activity at constant temperature. The B site is occupied by a transition metal ion. ⋅⋅

Table 9.1 Table of constant ‘m’ in ([VO ], n, p ∝ pOm ) 33 2

p n ⋅⋅ [VO ] Neutrality condition

VI

V

IV

III

II

I

Oxide decomposition

1/6 –1/6 –1/6 [B B′ ] = n = ⋅⋅ 2 [VO ]

1/4 –1/4 0 ⋅⋅ 2 [VO ] = [I RE ′ ]

1/4 –1/4 ~–1/2 p = [I RE ′ ] ⋅⋅ –2 [VO ]

0 0 –1/2 p = [I RE ′ ]

3/16 –3/16 –1/8 p = 3 [VI′′′] + 3 [VB′′′]

⋅⋅

High [VO ] regions

where µ is the mobility, q is the carrier charge and N is the carrier concentration. Because the mobility of either the electrons or holes is much higher than that of oxygen ions, the total conductivity in ferrites is dominated by hole conduction. The carrier concentration, N is: N = [SrLa ′ ] – 2[VO⋅⋅ ],

9.15

where [SrLa ′ ] is the acceptor content and [VO⋅⋅ ] is the oxygen vacancy concentration. For simple analysis, it is assumed that all doping centers are dissociated. In the (La, Sr)FeO3 system, generation of oxygen vacancy, VO⋅⋅ , follows:


O Ox → VO∞ + 1 O 2 + 2e ′ 2

247

9.16

The reaction constant of reaction in Eq. (9.16) is 2 K V ⋅⋅ = [VO⋅⋅ ]pO 1/2 2 n ,

9.17

O

where, pO2 is oxygen partial pressure and n is the electron concentration. Assuming this reaction follows Arrhenius law:

K V ⋅⋅ = K V ⋅⋅ ,0 exp (– E V ⋅⋅ /kT), O

O

O

9.18

where K V ⋅⋅ is the activation energy for oxygen vacancy generation and k is O Planck’s constant. Considering np = Ki and using the neutrality condition as: [SrLa ′ ] = [h˙] + 2[VO⋅⋅ ] = p + 2[VO⋅⋅ ]

9.19

where p is the hole concentration, also represented as [h˙], yields, n=

Ki Ki = p [SrLa ′ ] – 2[VO⋅ ⋅ ]

9.20

The intrinsic equilibrium constant, Ki, is defined by Eq. 9.4. Substituting (9.18) and (9.20) into (9.17) results in: 2

[VO⋅ ⋅ ]

  Ki –1/2  [Sr ] – 2[V ⋅⋅ ]  pO 2 = K VO⋅ ⋅ ,0 exp (– E VO⋅⋅ / kT) ′  La O 

9.21

and,  K V ⋅⋅ ,0  E V ⋅⋅   [VO⋅⋅ ] 1 O O ln  = ln   + 2 ln (pO 2 ) – kT ⋅⋅ 2  2   K ′ ] – 2[VO ])   ([SrLa i  

9.22

which shows the oxygen vacancy concentration, [VO⋅⋅ ], as a function of oxygen partial pressure (pO2) and temperature (T). Therefore the total carrier concentration, N, changes with pO2 and T and can be determined by combining Eq. (9.15) and Eq. (9.21). The mobility term (µ) in Eq. (9.14) is determined by the diffusion of the majority carriers in the lattice, which can be expressed as: µ=

µ0 E exp  – h  , Τ  kT 

9.23

where µ0 is the pre-exponential term, and Eh is hole mobility activation

248


energy. The total conductivity, σ, in Eq. (9.11) will then be achieved by substituting Eqs (9.15), (9.22) and (9.23) into (9.14), which yields:

E  µ σ = N ⋅ µ ⋅ q = f1 (pO 2 , T, composition)  0 exp  – h   ⋅ q 9.24  kT    T The solution to this equation yields:  1 + 8MN – 1   µ 0  Eh   σ=N⋅µ⋅q=   ⋅  T exp  – kT   ⋅ q 4M    

9.25

where M=

K VO⋅⋅ ,0 K 2i

 E V ⋅⋅ ⋅ exp  – O  kT 

  ⋅ PO–1/2 2  

9.26

and N = [SrLa ′ ]

9.27

At a specific temperature, T, ‘M’ is the only term in Eq. 4.25 that is a function of oxygen activity. Therefore,

K V ⋅⋅ ,0 O

K 2i

 E V ⋅⋅ ⋅ exp  – O  kT 

  represents the  

ability for oxygen vacancy generation (AOG) at a specific temperature and oxygen activity, which can be achieved by simulating a plot of σ vs. PO 2 . Carrier concentration and oxygen vacancy concentration can then be calculated from Eqs 9.15 and 9.21 by applying the value of AOG. Figures 9.9 and 9.10 show a plot of carrier concentration (N) and oxygen vacancy concentration vs. oxygen activity. One can see a flat region of carrier concentration in Fig. 9.9 when the value of AOG is very small (< 10–23). A nearly flat isothermal conductivity vs. oxygen activity has been observed in chromites and manganites, whereas this type of flat region was not observed in either ferrites or cobaltites. Figure 9.11 is a plot of conductivity vs. pO2 for La0.8Sr0.2MnO3,25 La0.60Sr0.40FeO3, and La0.6Sr0.4Co0.2Fe0.8O330 at 1000°C. Simulation of the data shown in Fig. 9.12 can yield AOG values for those compounds. Figure 9.12 illustrates the simulated results for La0.8Sr0.2MnO3, La0.60Sr0.40FeO3 and La0.6Sr0.4Co0.2Fe0.8O3 with AOG of 2 × 10–27, 3 × 10–21and 3 × 10–20, respectively. The ability for oxygen vacancy generation is a function of temperature and oxygen activity. Thus, the reaction to generate oxygen vacancies by Eq. 9.16 is a thermally activated process. Tables 9.234 and 9.3 35 list a series AOG values for manganites and chromites, respectively, as a function of Sr content and temperature.


249

22

log(N/(#/cm3))

21 20 10–34 10–30 10–26 10–23 10–22 10–21

19 18 17 16

AOG =

15

K V •• ,0 O K i2

 E •• ⋅ exp  – VO   kT

   

–26 –24 –22–20 –18 –16 –14 –12 –10 –8 –6 –4 –2

0

log(pO2/atm)

9.9 A plot of carrier concentration ( N ) as a function of oxygen activity with various AOG values.

22

log( [VO⋅⋅ ] /(#/cm3))

20 18 16 14 12 10 8

10–34 10–30 10–26 10–23 K •• 10–22 AOG = VO ,0 ⋅ exp –21 10 K i2

 E ••  – VO   kT

   

–26 –24 –22 –20 –18 –16 –14 –12 –10 –8 –6 –4 –2

0

log(pO2/atm)

9.10 A plot of oxygen vacancy concentration as a function of oxygen activity with various AOG values.

Global Solution – 2 Another model considers the valence state changes of the transition metal ions. This model can also be described as either the localized charge or small polaron system or narrow band structure. The main defect reaction for generation of oxygen ion vacancies is 2B⋅B + O Ox → 1 O 2 + 2B Bx + VO⋅ ⋅ 2 K V ⋅⋅ = O

PO1/22 [B Bx ] 2 [VO⋅⋅ ] [B⋅B ] 2 [O Ox ]

9.5

9.6

250

Materials for energy conversion devices 2.0

La0.80Sr0.20MnO3–δ

log (σ/(S·cm–1)

1.5

Sr 0 La 0.60

O 3–δ Fe 0.80 o 0.20 C .40

1.0 0.5 La0.60Sr0.40FeO3–δ 0.0

–0.5

–20 –18 –16 –14 –12 –10

–8

–6

–4

–2

0

log(pO2/atm)

9.11 A plot of conductivity as a function of oxygen activity for various p-type conductor perovskites at 1000°C. 2.5 La0.80Sr0.20MnO

log (σ/(S·cm–1)

2.0 La0.60Sr0.40Co0.20Fe0.80O3–δ

1.5 1.0 La0.60Sr0.40FeO3–δ 0.5 0.0

–0.5 –1.0 –14

–12

–10

–8

–6

–4

–2

0

log(pO2/atm)

9.12 A plot of simulated conductivity as a function of oxygen activity. The experimental data are shown in Fig. 9.11. Table 9.2 AOG values for La1–xSrxMnO3–δ34 x

0

800°C 900°C 1000°C

2.9 × 10 2.5 × 10–29 –30

0.10

0.20

0.30

0.40

9.6 × 10–32 5.0 × 10–30 3.1 × 10–29

1.2 × 10–30 3.8 × 10–29 1.6 × 10–29

5.8 × 10–30 1.6 × 10–28 8.1 × 10–28

5.5 × 10–30 5.2 × 10–28 2.9 × 10–27

For the compounds in this study, i.e. ABO3, we emphasize the regime where the cation vacancies are minor, thus the electroneutrality condition (Eq. 9.13) becomes: p = [SrLa ′ ] – 2[VO⋅⋅ ] + n

4.28


251

Table 9.3 AOG values for La1–xCaxCrO3–δ35 X

0.30

900°C 950°C 1000°C 1050°C

3.6 1.1 4.3 1.3

× × × ×

10–29 10–28 10–28 10–27

0.20

0.10

3.3 × 10–28

6.7 × 10–29

Three assumptions must be made to analytically solve the equations: 1. Assume that a higher valence transition metal ion, B⋅B , functions as an electron hole; and a lower valence ion, B ′B , functions as an electron, thus:

[B⋅B ] = [h ⋅ ] = p and [B ′B ] = [e ′] = n

9.29

2. Assume electron concentration, n, is much smaller than electron hole concentration, p. That is: n TLSF (~600°C)39 > TLSCo (~550°C) 40 > TLSFCu (~400°C).41 As discussed previously this maximum in conductivity represents the temperature at which the oxygen vacancy concentration starts to influence the carrier concentration. It does not mean


255

450 La0.20Sr0.80Fe0.80Cu0.20O3–δ

400 350

La0.60Sr0.40Co0.20Fe0.80O3–δ

σ(Ω–1cm–1)

300 250 La0.60Sr0.40FeO3–δ

200 150 100

La0.80Sr0.20MnO3–δ

50 0

La0.60Sr0.40CrO3–δ 200

400

600

800

1000

1200

T(°C)

9.15 A plot of conductivity measured in air as a function of temperature for five types of perovskites.

that the oxygen vacancy concentration is negligible at this temperature, but on the other hand, the influence of oxygen vacancy concentration on total carrier concentration is negligible below this temperature and the concentration of oxygen vacancies is so small that their contribution to transport processes becomes minimal. The values of AOG and understanding of the maximum conductivity are of particular importance in the search for materials for energy conversion devices. Since oxygen vacancies are required for lower cathodic overpotentials, this temperature also represents the temperature below which a cathode can be expected to have high overpotentials. Therefore, when pure LSM is used as the cathode, it will be expect to work well when the operating temperature is ~ 1000°C because the oxygen vacancy concentrations are sufficiently high to support the required transport processes of the cathode. Below this temperature range, overpotential problems are commonly encountered. It has been found that the addition of a second ionic conducting phase to the cathode enhances the ionic transport processes, which allows lower temperature operation. Thus the use of cathodes consisting of mixtures of LSM and YSZ or LSM and CGO has become common practice in the SOFC industry, but these mixtures will not extend the temperature much below 750–800°C before cathodic overpotentials become too large because the concentration of oxygen vacancies is insufficient to support the transport processes which are required for oxygen reduction and transport of the oxygen ions to the electrolyte. Questions still remain as to where reduction takes place, how we can quantitatively understand the relationship between oxygen vacancy concentration (and ionic transport number) and cathodic overpotential, and

256


how oxygen ions transport during cell operation? These are important questions whose answers will lead us to improved cathode materials.

9.6

Future trends

Two questions on defect chemistry in p-type perovskite compounds are not yet well understood: 1. Excess oxygen in LSM compounds at high oxygen activity has been observed. Conductivity increases are not observed with the addition of excess oxygen. Therefore, it seems that the excess oxygen is not compensated by electron holes. Oxygen interstitials have been ruled out because the large ionic radii of oxygen ions have problems in fitting interstitially into the close-packed perovskite structure. Hence, the question is what are the negative charges? If they are cation vacancies, the question as to what the positive charges are which compensate the cation vacancies needs to be answered. 2. Hysteresis: The accuracy of both defect chemistry models in this work is determined by the conductivity measurements, in particular in the region where oxygen vacancies start to influence the overall carrier concentration significantly. The oxygen activity regime corresponding to this transition is related to the ability for oxygen vacancy generation. However, from the literature these values vary a great deal. An interesting ‘hysteresis’ behaviour of the conductivity as a function of oxygen activity has been observed by several groups.17,38 This behaviour is yet not well understood. Some classical equilibrium defect chemistry concepts derived from binary compounds can be applied to ternary compounds. The main difference is the redox reaction of ternary compounds that arises from the ability for the valence state of the B site ions to change. Currently, the defect chemistry is still focused on bulk ceramics binary, ternary or multicomponent compounds. It is necessary to conduct research on the atomic scale studies of defect formation and the associated properties for nanocrystalline materials (in the form of thin film and bulk), with an emphasis on the role of size on surface and interface phenomena. Novel techniques to accurately determine stoichiometry of the ternary oxides are of particular interest and importance in elucidating defect characteristics and then designing the new materials for use in energy conversion systems, such as the electrodes for SOFCs and gas separation membranes. Oxygen ion conductivity is a key parameter in ternary oxides used either as ionic conductors or MIECs. Therefore, it is of particular interest to design reliable experiments to measure σi and to determine oxygen vacancy concentration.


9.7

257

Acknowledgements

The authors wish to thank the Department of Energy and the Gas Research Institute who provided financial support for part of this research.

9.8

References

1. Verwey, E.J.W., Haaijam, P.W., Romeijh, F.C. and Van Oosterhout, G.W. ‘Controlledvalency semiconductors’, Philips Res. Report (1950), 5, 173–87. 2. Kosacki, I. and Anderson, H.U. ‘Microstructure-property relationships in nanocrystalline oxide thin films’, Ionics, (2000) 6(3/4), 294–311. 3. Nowotny, J. and Rekas, M. ‘Defect structure, electrical properties and transport in barium titanate. III. Electrical conductivity, thermopower and transport in single crystalline BaTiO3’, Ceramics International (1994), 20(4), 225–35. 4. Nowotny, J. and Rekas, M. ‘Defect structure, electrical properties, and transport in barium titanate. II. Consistency requirements between defectt models and crystal properties’, Ceramics International (1994), 20(4), 217–24. 5. Nowotny, J. ‘Defect structure, electrical properties, and transport in barium titanate. I. Introductory remarks’, Ceramics (1994), 20(4), 213–15. 6. Huebner, W. ‘Thermally-stimulated current and dielectric properties of doped and undoped barium titanate’, Dissertation, University of Missouri-Rolla, 1987. 7. Boroomand, F., Wessel, E., Bausinger, H. and Hilpert, K. ‘Correlation between defect chemistry and expansion during reduction of doped LaCrO3 interconnects for SOFCs’, Solid State Ionics (2000), 129(1–4), 251–8. 8. Ling, S. ‘Statistical thermodynamic formulation of high concentration point defect chemistry in perovskite crystalline systems: application to strontium doped lanthanum chromite’, J. Phys. Chem. Solids (1994), 55(12), 1445–60. 9. Anderson, H.U., Kuo, J.H. and Sparlin, D.M. ‘Review of defect chemistry of lanthanum manganese oxide (LaMnO3) and lanthanum chromium oxide (LaCrO3)’, Proceedings Electrochemical Society (1989), 89–11 (Proc. Int. Symp. Solid Oxide Fuel Cells, 1st, 1989), 111–28. 10. Larsen, P.H., Hendriksen, P.V. and Morgensen, M. ‘Dimensional stability and defect chemistry of doped lanthanum chromites’, J. Thermal Analysis (1997), 49(3), 1263– 75. 11. Anderson, H.U. ‘Review of p-type doped perovskite materials for SOFC and other applications’, Solid State Ionics (1992), 52, 33–41. 12. van Roosmalen, J.A.M. and Cordfunke, E.H.P. ‘The defect chemistry of LaMnO3±δ 4. Defect model for LaMnO3+δ’, J. Solid State Chem. (1994), 110(1), 109–12. 13. van Roosmalen, J.A.M. and Cordfunke, E.H.P. ‘The defect chemistry of LaMnO3±δ 5. Thermodynamics’, J. Solid State Chem. (1994), 110(1), 113–17. 14. van Roosmalen, J.A.M., Cordfunke, E.H.P., Helmholdt, R.B. and Zandbergen, H.W. ‘The defect chemistry of LaMnO3±δ 2. Structural aspects of LaMnO3+δ’, J. Solid State Chem. (1994), 110(1), 100–5. 15. Nowotny, J. and Rekas, M. ‘Defect chemistry of (La,Sr)MnO3,’ J. Am. Ceram. Soc. (1998), 81 (1), 67–80. 16. Poulsen, F.W. ‘Defect chemistry modeling of oxygen-stoichiometry, vacancy concentrations, and conductivity of (La1–xSrx)yMnO3±δ’, Solid State Ionics, (2000), 129, 145–62.

258


17. Patrakeev, M.V., Bahteeva, J.A., Mitberg, E.B., Leonidov, I.A., Kozhevnikov, V.L. and Poeppelmeier, K.R. ‘Electron/hole and ion transport in La1–xSrxFeO3–δ’, J Solid State Chem. (2003), 172, 219–31. 18. Mizusaki, J., Yoshihiro, M., Yamauchi, S. and Fueki, K. ‘Nonstoichiometry and defect structure of the perovskite-type oxides La1–xSrxFeO3–δ’, J. Solid State Chem. (1985), 58(2), 257–66. 19. Bak, T., Nowotny, J., Rekas, M., Ringer, S. and Sorrell, C.C. ‘Defect chemistry and electrical properties of La1–xSrxCoO3–δ IV. Electrical properties’, Ionics (2001), 7(4, 5 & 6), 388–93. 20. Bak, T., Nowotny, J., Rekas, M., Ringer, S. and Sorrell, C.C. ‘Defect chemistry and electrical properties of La1–xSrxCoO3–δ III. Oxygen nonstoichiometry’, Ionics (2001), 7(4, 5 & 6), 380–7. 21. Bak, T., Nowotny, J., Rekas, M., Ringer, S. and Sorrell, C.C. ‘Defect chemistry and electrical properties of La1–xSrxCoO3–δ II. Defect diagrams’, Ionics (2001), 7(4, 5 & 6), 370–9. 22. Bak, T., Nowotny, J., Rekas, M., Ringer, S. and Sorrell, C.C. ‘Defect chemistry and electrical properties of La1–xSrxCoO3–δ I. Defect equilibria’, Ionics (2001), 7(4, 5 & 6), 360–9. 23. Flandermeyer, B.K., Nasrallah, M.M., Agarwal, A.K. and Anderson, H.U., ‘Defect structure of magnesium-doped lanthanum chromate(III) model and thermogravimetric measurements’, J. Am. Ceram. Soc. (1984) 67, 195–8. 24. Van Roosmalen, J.A.M., Huiusmans, J.P.P. and Cordfunke, E.H.P. ‘A new defect model to describe the oxygen deficiency in perovskite-type oxides’, J. Solid State Chem. (1991), 93, 212–19. 25. Kuo, J.H., Anderson, H.U. and Sparlin, D.M. ‘Oxidation-reduction behavior of undoped and strontium-doped lanthanum manganese oxide (LaMnO3): defect structure, electrical conductivity, and thermoelectric power’, J. Solid State Chem. (1990), 87, 55–63. 26. Stevenson, J.W., Nasrallah, M.M., Anderson, H.U. and Sparlin, D.M. ‘Defect structure of yttrium calcium manganate (Y1–yCayCrO3) and lanthanum calcium manganate (La1–yCayCrO3). I. Electrical properties’, J Solid State Chem. (1993), 102, 175–84. 27. Anderson, H.U., Tai, L-W., Chen, C.C., Nasrallah, M.M. and Huebner, W. ‘Review of the structural and electrical properties of the (La,Sr)(Co,Fe)O3 system’. In Proc. of the 4th Int’l Symp. on Solid Oxide Fuel Cells, ed S. Singhal, The Electrochemical Society (1994) 95-1 (Solid Oxide Fuel Cells (SOFC-IV)), 375–84. 28. Yelon, W.B., Cai, Q., James, W.J., Anderson, H.U., Yang, J.B., Zhou, X.D and Blackstead, H.A. ‘Neutron diffraction studies of magnetic and superconducting compounds’, Physica Status Solidi (2004), 201, 1428–35. 29. Zhou, X.D., Cai, Q., Chu, Z., Yang, J., Yelon, W.B., James W.J. and Anderson, H.U. ‘Utilization of neutron diffraction and Mössbauer spectroscopy in the studies of the cathode for SOFCs’, Solid State Ionics (2005), 175, 83–6. 30. Tai, L.W., Nasrallah, M.M. and Anderson, H.U. ‘Thermochemical stability, electrical conductivity, and Seebeck coefficient of Sr-doped LaCo0.2Fe0.8O3–δ’, J. Solid State Chem. (1995), 118, 117–24. 31. Stevenson, J.W., Armstrong, T.R., Carneim, R.D., Pederson, L.R. and Weber, W.J. ‘Electrochemical properties of mixed conducting perovskites La1–xMxCo1–yFeyO3–δ (M = Sr, Ba, Ca)’, J. Electrochem. Soc. (1996), 143(9), 2722–9. 32. Kröger, F.A. The Chemistry of Imperfect Crystals, John Wiley and Sons, New York, 1962.


259

33. Anderson, H.U. ‘Defect chemistry of p-type perovskites’, Proceedings of the 14th Riso Inter. Symp. On Mater. Sci: High Temp. Electrochm. Behavior of Fast Ion and Mixed Conductors. Poulsen, F.W., Bentzen, J.J., Jacobsen, T., Skou, E. and Ostergard, M.J.L (eds). 1–188, 1993. 34 Mizusaki, J., Yonemura, Y., Kamata, H., Ohyama, K., Mori, N., Takai, H., Tagawa, H., Dokiya, M., Naraya, K., Sasamoto, T., Inaba, H. and Hashimoto, T. ‘Electronic conductivity, Seebeck coefficient, defect and electronic structure of nonstoichiometric La1–xSrxMnO3’, Solid State Ionics (2000), 132(3,4), 167–80. 35 Yasuda, I. and Hikita, T. ‘Electrical conductivity and defect structure of calciumdoped lanthanum chromites’, J. Electrochem. Soc. (1993), 140(6), 1699–704. 36 Carini, G.F., II, Anderson, H.U., Sparlin, D.M. and Nasrallah, M.M. ‘Electrical conductivity, Seebeck coefficient and defect chemistry of calcium-doped yttrium chromium oxide (YCrO3)’, Solid State Ionics (1991), 49, 233–43. 37. Mizusaki, J., Sasamoto, T., Cannon, W.R. and Bowen, H.K. ‘Electronic conductivity, Seebeck coefficient, and defect structure of lanthanum strontium iron oxide La1–xSrxFeO3 (x = 0.1, 0.25)’, J. Am. Ceram. Soc. (1983), 66(4), 247–52. 38. Karim, D.P. and Aldred, A.T. ‘Localized Level Hopping Transport in La(Sr)CrO3’, Phys. Rev. B (1979), 22, 2255–63. 39. Zhou, X.D., Shin, Y.W. and Anderson, H.U. unpublished. 40. Tai, L.W., Nasrallah, M.M., Anderson, H.U., Sparlin, D.M. and Sehlin, S.R. ‘Structure and electrical properties of La 1–x Sr x Co 1–y Fe y O 3 , part 2. The system of La1–xSrxCo0.2Fe0.8O3’, Solid State Ionics (1995), 76(3,4), 273–83. 41. Kaus, I. and Anderson, H.U. ‘Electrical and thermal properties of La0.2Sr0.8Cu0.1Fe0.9O3–d and La0.2Sr0.8Cu0.2Fe0.8O3–δ’, Solid State Ionics (2000), 129(1–4), 189–200.

10 Surface properties of ionic conductors H - D W I E M H Ö F E R, University of Münster, Germany

10.1

Surfaces, segregation and nanoscaling in solid electrolytes

10.1.1 Surface analysis on solid electrolytes Most solid electrolytes are used in electrochemical applications as compact polycrystalline materials, thick films or thin films. Accordingly, surfaces and interfaces not only occur in contact with the gas phase or at the electrodes, but also among the grains. These external and internal interfaces may show surface charges, space charge layers, deviations from the bulk with respect to elemental composition or segregation. All these phenomena may modulate ion transport through or along these interfaces and electrode kinetics as illustrated in Fig. 10.1. Hence, knowledge on segregation, surface composition and changes with temperature or chemical pretreatment is important for the optimization of electrochemical properties. Table 10.1 gives examples for some typical surface analytical techniques and the information which can be obtained by these. X-ray photoelectron spectroscopy (XPS or ESCA) is useful for analysing the elemental composition and bonding at a surface. ISS and SIMS probe the composition of the first atomic layer, whereas electronic spectroscopies such as XPS and AES give information on the first 3–8 monolayers. Much larger information depths are attainable by combination of SIMS and SNMS with sputtering or by using characteristic X-ray emission induced by electron microprobe techniques. Comparing the results of techniques with different information depths can yield valuable information about details of the surface properties. As surface analytical techniques are mostly based on excitation or analysis with charged particles, overcharging effects are to be expected, if insulating materials are measured. Although solid electrolytes are not insulating, the missing electronic conductivity usually leads to polarization voltages and compositional changes in such samples. These can cause considerable shifts 260

Surface properties of ionic conductors

Pt

YSZ

Pt

pO ′2

Gas phase

261

pO ′′ 2

YSZ

O2 4e –

O2– 2O2–

Pt

Double layer

3-phase line

Internal interfaces

10.1 Significance of surfaces and interfaces for ion transport and electrode kinetics of solid electrolytes.

of the energy of emitted electrons or ions. For instance, during an XPS analysis of an yttria stabilized zirconia sample (YSZ), the back of the sample becomes dark (‘blackening of zirconia’) due to strong chemical reduction by the electron flux from the back contact into the sample. Alkali ion conducting materials show metal deposition during electron spectroscopic analysis comparable to electrolysis. The conventional approach to circumvent these difficulties for insulating samples is to use an additional low energy electron beam that reduces the overcharging effects. But in the case of solid electrolyte samples, a better way is to replace the usually inert back contact by materials which act as reversible or reference electrodes in the electrochemistry of the solid electrolytes (Wiemhöfer and Göpel, 1991; Wiemhöfer et al., 1990a). At the interface of a solid electrolyte to a suitable reference electrode, the chemical potential of the key component (e.g. oxygen in YSZ) and with it the Fermi level will then be fixed. This is a precondition for well-defined energy scales for emitted electrons in photoelectron spectroscopy. This approach allows the measurement of surfaces of solid ionic conductors with a comparable precision as for metals and semiconductors. For instance, a suitable reference electrode for YSZ is an electronically conducting metal/metal oxide mixture with low oxygen partial pressure (compatible with UHV conditions). The mixture must be able to exchange not only electrons but also oxygen ions in contact with YSZ. Good results

262


Table 10.1 Examples of useful techniques for surface analysis applicable to solid electrolytes and electrodes (Egelhoff Jr., 1987; Ertl and Küppers, 1985) Surface analytical methods

Information available

XPS (or ESCA): X-ray Photoelectron Spectroscopy (Electron Spectroscopy for Chemical Analysis)

•

UPS: UV Photoelectron Spectroscopy

•

• • •

• •

Energy of core level electrons and valence band referred to εF Chemical shifts due to different chemical environments Elemental surface composition Information depth: 3–10 monolayers Surface density of occupied states in the valence band including surface states as a function of energy referred to εF Occupied electronic states of adsorbed species Work function, ionization energy, position of the surface Fermi level εF with regard to the valence band edge εC

SPEM: Scanning Photoelectron Microscopy PEEM: Photoelectron Emission Microscopy

•

Local surface composition

•

Local work function analysis

EELS: Electron Energy Loss Spectroscopy

•

Energy difference between occupied and unoccupied states at the surface (electronic states, surface plasmons) Vibrational states of adsorbed particles (HREELS)

• (S)AES: (Scanning) Auger Electron Spectroscopy

• •

SIMS/SNMS: Secondary Ion Mass Spectrometry (SNMS - secondary neutral particles)

•

ISS (or LEIS): Low Energy Ion Scattering (Ion Scattering Spectroscopy)

•

TDS: Thermal Desorption Spectrometry

•

•

•

• •

Characteristic energy differences of core levels, chemical shifts due to different chemical environments Information depth similar to XPS Static: elemental composition of the first atomic layer Dynamic: depth profile of elemental composition Analysis of kinetic energy of elastically scattered ions Elemental composition of the first atomic layer (bad resolution for elements with similar mass) Mass spectrometry of thermally desorbed atoms and molecules Information obtained about adsorbed species and about number and energy of adsorption sites Detection of decomposition and changes of bulk stoichiometry


263

have been achieved, for example, with thick films of Fe, FeO at the backside of a YSZ sample instead of a direct Pt contact. This prevents any overcharging during photoemission of electrons (Wiemhöfer and Vohrer, 1992). The same is true for oxygen ion conducting ceramics based on stabilized Bi2O3 (Shuk et al., 1997). Similarly, silver ion conductors should be contacted at the back with silver (Wiemhöfer et al., 1990b). Fluoride ion conducting materials such as LaF3 are easily measurable with XPS and UPS, if one uses an Ag, AgF or a Sn, SnF2 mixture as the back contact (Wiemhöfer et al., 1990a). Figure 10.2 shows the set-up employed for temperature dependent XPS and UPS investigations of solid electrolytes such as YSZ and doped Bi2O3 in the temperature range between 300 K and 1250 K (Wiemhöfer and Vohrer, 1992). Ceramic pellets or single crystals can be analysed in this way with respect to electronic structure, bonding and composition of solid electrolyte as well as electrode surfaces as a function of temperature and oxygen (Wiemhöfer et al., 1990a; Wiemhöfer, 1993; Schindler et al., 1989; Vohrer et al., 1993; Zipprich et al., 1995). Before reproducible results are obtained by surface analytical methods, a defined, reproducible state and composition of the solid electrolyte surface YSZ (single crystal or ceramic) Pt foil (= resistive heater)

Sample Back contact (reference: e.g. Fe, FeO)

Titanium holder

Ceramic insulator

(a)

RE: Fe, FeO

Electrical contacts, thermocouple

CE: Fe, FeO

RE: Pt (UHV)

CE: Pt (UHV)

(b)

WE: Pt, Ag, …

(c)

WE: Pt, Ag, …

10.2 Experimental set-up for surface analysis of single crystal and ceramic solid electrolytes. On the right, examples of typical planar YSZ samples: (a) YSZ with reference back contact for analysis of the free YSZ surface (Wiemhöfer and Vohrer, 1992), (b) galvanic cell with thin film electrode on YSZ for analysis of the polarized electrode surface (Zipprich et al., 1995), (c) galvanic cell with microstructured working electrode as used, for example, by Luerssen et al. (2002); (RE = reference electrode, CE = counter electrode, WE = working electrode).

264


have to be ascertained. Ceramic or single crystal samples of solid electrolytes that are stable under UHV up to high temperatures are usually ‘prepared’ by an appropriate sputter cleaning followed by annealing which often has to be repeated several times until all surface impurities and mobile bulk impurities are removed completely. For instance, polycrystalline and single crystal samples of YSZ normally show impurities such as Na, K, Ca, Mg, Al, Si, F and Cl as detectable by SIMS and ISS (Schindler et al., 1989; de Ridder et al., 2002). Some of these above impurities arise from sintering aids or vessel materials that are used during the synthesis of ceramic and single crystal samples. They are enriched at the grain boundaries or dissolved in the bulk and segregate to the surface during a high temperature annealing at about 1000°C under UHV conditions. The increased grain boundary resistivity of stabilized zirconia below 500°C is often attributed to segregation effects and formation of glassy phases due to traces of silica and alumina (Bäuerle, 1969; Badwal et al., 1998). Elements like Fe and Ti which often occur as trace impurities or additions in stabilized zirconia do not segregate to the surface (cf. Fig. 10.3) except in the case when higher concentrations above the solubility limit are produced by surface ion implantation (Vohrer et al., 1991, 1993). 0.7 Zirconium

Relative intensity

0.6

Pore Single grain: YSZ doped with 10.7 mole % titania

0.5 0.4

Bulk

Grain boundary

0.3 0.2

Yttrium

0.1

Titanium

Grain boundary

Bulk 0.0 0

4

8

12 µm

16

20

24

10.3 Results of a microprobe analysis on a YSZ grain: plot of the relative concentration changes of Zr, Y and Ti perpendicular to the grain boundary (Vohrer et al., 1991).

10.1.2 Segregation of dopants A more important surface segregation is that due to dopants such as gadolinia in ceria or yttria in zirconia. This has been investigated for YSZ by a number of groups (Schindler et al., 1989; Winnubst et al., 1983; Theunissen et al., 1992; Steele and Butler, 1985; de Ridder et al., 2002). All these results of


265

surface analytical studies on YSZ give clear evidence for a surface enrichment of yttria in YSZ for all dopant concentrations. Typical surface concentrations of Y as obtained by AES for YSZ samples after high temperature annealing of freshly prepared surfaces were up to 34 mole %, while the bulk concentration was 17 mole % and lower (Burggraaf et al., 1985; Burggraaf and Winnubst, 1988). This corresponds to an yttrium enrichment by at least a factor of two. A comparison of XPS and SIMS results (see Fig. 10.4) indicates that yttria segregation is predominantly concentrated in the first monolayer (Wiemhöfer, 1996). The XPS technique probes a substantially larger depth and shows much lower yttria enrichment (cf. Fig. 10.4) as compared to the surface sensitive static SIMS which probes the first monolayer. Recent experiments with low energy ion scattering (LEIS) which also probes the first monolayer confirmed this (de Ridder et al., 2002). SIMS and XPS experiments on single crystals as well as on ceramic samples also showed that a further reversible increase of the Y segregation occurred for increasing temperatures above 500°C (Schindler et al., 1989). 2.4

Y/Zr (relative change)

2.2

Y/Zr (SIMS)

2.0 1.8 1.6 1.4 Y/Zr (XPS)

1.2 1.0 400

600

800 Temperature [K]

1000

1200

10.4 Results of XPS and SIMS analysis for the temperature dependent yttrium segregation at the surface of an YSZ single crystal. Y/Zr denotes the relative ratio of the corresponding signal intensities for each curve; the ratio was set equal to 1 at 400 K. SIMS probes the first atomic layer whereas XPS probes the first 3–10 monolayers. It is evident that the reversible changes of the yttrium segregation at these temperatures are concentrated in the first monolayer (Wiemhöfer, 1996).

The Y segregation which is most probably driven by strain relaxation (higher ionic radius of Y3+ compared to Zr4+) may be accompanied by a surface charge on the grains and a corresponding space charge region (Steele and Butler, 1985; Burggraaf and Winnubst, 1988). Model calculations give

266


a good agreement between the calculated space charge layer thickness and the observed thickness of the Y enriched zone at the surface (Theunissen et al., 1992; Guo and Maier, 2001). Segregation-induced concentration gradients also lower the charge transfer at electrode interfaces on YSZ surfaces and thus act on the surface functionality of corresponding sensors or fuel cells and, in particular, on the oxygen exchange kinetics (Mizutani and Nowotny, 1998). Many further examples are now known for segregation in solid oxide electrolytes. It also occurs with calcia stabilized zirconia (Aoki et al., 1996). A similar finding is reported from LEIS investigations on Gd doped ceria which shows strong Gd enrichment in the first five monolayers (Scanlon et al., 1998). From EELS investigations on single crystal YSZ and on Gd doped ceria ceramics, another group derives a general trend of fluorite structured ceramics towards dopant segregation coupled with surface enrichment of vacancies and electrons (Lei et al., 2002). However, space charge effects have also been found for undoped ceria which gives a hint that space charge and segregation may not be coupled in every case (Guo et al., 2003). Apart from surface analytical techniques, STM and its variants are now developing as a rich source for direct information on surface defects such as single vacancies and surface atoms. Point defects and defect clusters as well as changes during reduction have recently been studied for surfaces of pure and doped ceria (Norenberg and Briggs, 1997; Schierbaum, 1998; Berner and Schierbaum, 2002; Namai et al., 2003). With AFM, it was shown that defects and defect clusters can be studied in detail and that oxygen mobility and reaction of oxygen with a surface vacancy are even observable at room temperature (Namai et al., 2003).

10.1.3 Nanoionics The particular properties of grain boundaries and segregation layers with respect to ionic and electronic defect concentrations offer many possibilities to modify the transport behaviour and electrochemical characteristics of solid electrolytes and their interfaces. Decreasing the crystallite size into the nanometer range is a straightforward way to prepare materials with properties that are dominated by their internal interfaces, i.e. by surface and space charges. A great deal of the present interest lies in the development of techniques for preparation, analysis and fine tuning of nanocrystalline materials. A number of recent reviews are available for the special context of solid electrolytes (Knauth, 2002; Tuller, 1997, 2000; Maier, 2003). Ceria-based nanocrystalline ceramics and films have been found to exhibit anomalous expansion due to increased subsurface concentrations of vacancies (Nair et al., 2003). However, space charge effects can also decrease the majority charge carrier concentrations as has been shown for acceptor doped nanocrystalline ceria (Guo et al., 2003). In that case, the grain boundaries


267

were characterized by a depletion of mobile oxygen vacancies together with an increase of the electron concentration. Accordingly, the conductivity of Gd-doped ceria changes from predominantly ionic to electronic, if the crystallite size is reduced drastically (Tschöpe et al., 2002). Nanosize particles of YSZ can lead to higher density ceramics as well as to a reduced resistivity with values that approximately become equal to that of a YSZ single crystal (Muccillo and Muccillo, 2002). Nanocrystalline YSZ also shows enhanced oxygen diffusivity at the grain boundaries which also leads to enhanced oxygen surface exchange as found by tracer experiments (Knöner et al., 2003). Effects due to nanoscaling have also been analysed on thin film solid electrolyte systems. Thin films and corresponding layer structures make possible a precise control of the space charges between the layers or with respect to the substrate interface. Very detailed studies of the ionic conductivity of thin film structures made of fluoride or silver ion conductors were possible in this way showing a strong dependence of the electrical properties on the layer thickness (Lee et al., 2000; Maier, 1999; Jin-Phillipp et al., 2004; Sata et al., 2002). Thin nanolayers of ionic conductors are also of interest as dielectric films in contact with typical semiconductors such as Si and InP. Epitaxial CaF2 and SrF2 on InP(001) were demonstrated to be suitable dielectric protecting layers which minimize the surface defect concentration on the InP semiconductor and form sharp interfaces (Weiss et al., 1992). Figure 10.5 shows the conditions at the interface CaF2/InP. In this case, mobile fluoride ions help to minimize surface charges. Recently, wide band gap materials χe = 0.8 eV

εvac Φe = 4.1 χe = 3.86 eV eV

∆εC = 4.56 eV

εC εvac Φe = 2.7 eV

εC

εF

εV

InP

∆εV = 6.2 eV CaF2 1.1 eV

εV

2 nm

10.5 Band scheme for a dielectric thin film of CaF2 on a semiconducting InP single crystal (Weiss et al., 1992). The results have been measured using XPS, UPS, AES and LEED (= low energy electron diffraction).

268


such as ZrO2 and HfO 2 have also been tested as dielectric layers on semiconductors (Zhu and Liu, 2004; Robertson, 2000).

10.2

Electronic properties of solid electrolyte surfaces

10.2.1 Techniques for the study of electronic properties Electronic properties of solid electrolyte surfaces are concerned in the adsorption of gaseous molecules, in surface gas reactions as well as during oxygen exchange with the gas phase. Some well-known oxide ion conductors such as ceria and zirconia are also widely used in heterogeneous catalysis of redox reactions. For instance, ceria and the increasingly important ceriazirconia system serve as fast oxygen storage compounds for exhaust gas catalysts. Accordingly, it is of interest to measure electronic properties such as electronic state densities, work functions, electron affinities and Fermi levels of solid electrolyte surfaces. Information on these is helpful to explain or predict the minor electronic conductivity of solid electrolytes. The latter is one of several factors that limit the efficiency in corresponding fuel cells. The electronic conductivity and its dependence on temperature and chemical potential determine the limits of the electrolytic domain, i.e. the range of chemical potentials where the ionic conductivity predominates the electronic one. All these issues are strongly related to the band gap and the electronic state density of solid electrolytes. Electronic spectroscopies such as XPS, UPS, EELS and optical spectroscopy are excellent tools for the analysis of electronic properties of solid electrolyte surfaces. UV photoelectron spectroscopy (UPS), in particular, gives direct access to an experimental determination of absolute work functions, surface potentials, electron affinities, and the position of the Fermi level (Egelhoff Jr., 1987; Ertl and Küppers, 1985). Figure 10.6 shows the information available with different techniques for the example of an YSZ surface. Photoelectron spectroscopic techniques have been greatly improved with respect to lateral resolution. Photoelectron emission microscopy (PEEM) yields information on the local work function (Casalis et al., 1995) and scanning photoelectron microscopy (SPEM) gives the local surface composition (Von Oertzen et al., 1991). Another surface sensitive method is the Kelvin probe technique for measuring changes of the work function. Kelvin probes have the advantage to be applicable under gas pressure whereas UHV conditions are necessary for photoelectron spectroscopy. Kelvin probe techniques have been developed especially for studies of solid electrolyte surfaces as a function of gas interaction (Nowotny and Sloma, 1985; Bak et al., 2001a,b, Badwal et al., 2001).


269

ε [eV] εvac 6

χe

Conduction band

εC

5 4 3 2

Φe

e′ EELS

UPS

Absorption spectroscopy

εgap Photoluminescence h·

εF

1 εV

0 Valence band

Density of states

10.6 Determination of the energies of band edges and the density of states by electronic and optical spectroscopy. Work function, electron affinity, and the position of the surface Fermi level can be determined with respect to the band scheme.

10.2.2 Band gap and density of states in solid electrolytes The concentration of electrons and holes are basically coupled to each other by the equilibrium of electron-hole pair formation. The general thermodynamic relation for the equilibrium formation of electron-hole pairs is given by: °  – ∆ Geh  [ e ′ ][ h˙] = N e° N h° exp    kT 

10.1

[e′] and [h˙] denote the concentrations of electrons and holes and the prefactors on the right-hand side denote the standard concentrations of electrons and ° corresponds to the standard free energy of electron-hole pair holes. ∆ Geh formation. If the electrons and holes can be treated as quasi-free particles, the prefactors can be replaced by the effective state densities in the conduction and valence band and the vibrational entropy of pair formation can be set to zero. One obtains: °  ∆ H eh  ° with ∆ H eh ⬇ εgap [e′][h˙] = N Ceff N Veff exp  – k T  

10.2

° is usually derived from temperature-dependent The thermal band gap ∆ H eh electron and hole concentrations. These are calculated from independent measurements of the electronic conductivities and mobilities.

270


The optical band gap, εgap, is in general not exactly the same as the thermal band gap. It is obtained from spectroscopic transitions and thus represents an energy difference between optically determined band edge energies. In addition, care has to be taken with respect to the Franck–Condon principle according to which optical energies derived from absorption may be higher than the minimum difference between the two involved electronic energy levels. If the entropy of electron-hole formation is negligible, the optical band ° . The gap εgap can be well approximated by the thermal (enthalpy) gap ∆ H eh mobilities of electrons and holes are usually rather small in solid electrolytes making evident that electrons and holes are strongly localized and have to be considered as polarons. The interaction of localized electrons with the lattice, then, will become comparable to that of ionic point defects leading to a net vibrational entropy contribution for the electron-hole pair formation. If experimental results for YSZ for temperature dependent electron and hole concentrations are evaluated with Eq. 10.2 (Sasaki and Maier, 2000), one observes a strong deviation of the resulting effective state densities from values expected for free electrons and holes (by 2 to 4 orders of magnitude). Accordingly, a perfect equivalence of thermal and optical band gap cannot be expected. On the other hand, it is difficult to measure mobilities and conductivities of electrons and holes with sufficient accuracy. This is also ° for YSZ. reflected in the large scatter of reported thermal gap energies ∆ H eh ° Values of ∆ H eh between 2.61 eV (Park and Blumenthal, 1989) and 4.72 eV (Sasaki and Maier, 2000) have been reported. But also the reported optical band gaps may scatter depending on the applied experimental methods. For cubic YSZ, the value of the optical band gap, εgap, most probably lies at 5.1 ± 0.1 eV (Wiemhöfer and Vohrer, 1992), which is supported by theoretical caculations (Kobayashi et al., 2003). An important consequence of the high ionic defect concentrations of solid electrolytes can be found for the electronic density of states near the band edges. As predicted by Anderson for highly disordered electronic semiconductors (Anderson, 1958; Mott et al., 1975), the band edges become less sharp and show a broadening of the electronic state densities into the band gap (band tailing). This is strongly supported by the results of UPS and EELS for YSZ. Figure 10.7 shows a considerable band tailing at the valence band edge which is further enhanced by sputtering (Wiemhöfer et al., 1990a). The band gap of stabilized zirconia (thermal and optical) can be changed by the dissolution of certain metal oxides. Dissolving the isovalent titanium dioxide in YSZ leads to a considerable decrease of the optical band gap by about 1.5 eV to a value of εgap = 3.7 eV after dissolving 12 mole % TiO2 as Fig. 10.8 shows (Vohrer et al., 1991). The 3d levels of titanium ions form a new impurity band in a range of 2 eV below the conduction band edge of stabilized zirconia. These electronic states shorten the energy difference to

Surface properties of ionic conductors UPS (He I) Annealed (+O2) After sputtering

Secondary electron emission

Intensity [arb. units]

271

Valence band emission

εV

Band gap emission εF

20

15

10 5 Binding energy [eV]

0

10.7 UPS on a single crystal of the composition (ZrO2)0.87(YO1.5)0.13 (Wiemhöfer and Vohrer, 1992). It is evident that the electron density of states in the band tail between 0 eV and 5 eV below εF increases after sputtering due to the increase of sputter induced surface defects. An annealing with oxygen gas restores the previous situation and lowers the density of states in the band tailing range. A remarkable feature is the non-zero density of states up to the Fermi level. (a)

Band gap εgap [eV]

5.2

Intensity (arbitrary units)

4.8 4.4 4.0

(b) (c) (d)

(e)

2 4 6 8 10 12 mol % TiO2

e d c b a 25

20

15

10 5 Energy loss [eV]

0

10.8 EELS analysis of the band gap of YSZ doped with different concentrations of TiO2 (Vohrer et al., 1991).

the occupied localized states above the valence band edge. It is remarkable that, for a given oxygen partial pressure, the position of the Fermi level remains unchanged with respect to the valence band edge for increasing

272


titania content. One can therefore expect that the addition of titania enhances the electronic concentration under reducing conditions where the Fermi level approaches the empty 3d band of Ti4+ ions which is indeed the case (Worrell and Weppner, 1984; Duran et al., 1999). Dissolved Fe2O3 generates additional electronic state density slightly above the valence band as well as in the middle of the band gap (Vohrer et al., 1993). Hence, a high concentration of dissolved iron is also likely to increase the electronic conductivity. A recent optical investigation on YSZ confirms these results for Ti and Fe in YSZ and also gives data for a number of further transition metal ions dissolved in YSZ (Sasaki and Maier, 2000).

10.2.3 Work function and Fermi level at solid electrolyte surfaces A particular useful feature of UPS is the possibility to determine the absolute value of the work function and the position of the Fermi level with respect to the band edges at the surface, although the limitation to UHV conditions does not allow for gas interaction and equilibrium with a gas phase. The work function consists of two contributions according to Φ = (εvac – εC) + (εC – εF) = χe + (εC – εF)

10.3

The first term is the electronic affinity χe corresponding to the energy difference between the energy εvac of an electron at rest in the vacuum and the conduction band edge εC. The second contribution to the work function is the difference between conduction band edge and the Fermi level εF. For YSZ, the Fermi level is fixed, if the oxygen activity is held constant. In order to get welldefined spectra with a defined Fermi level, one has to use reference contacts as described in section 10.1.1. At the contact between YSZ and a Fe,FeO mixture, the oxygen activity is constant and, accordingly, the following equilibrium is valid: 2– – s O (electrolyte) O(Fe,FeO) + 2e (electrode)

10.4

The thermodynamic equilibrium condition for the reaction 10.4 is given by: ε F = 1 ⋅ ( µ˜ O 2– – µ O ) 2

10.5

In a good oxide ion conductor such as YSZ, the chemical potential µ˜ O 2– of oxide ions is virtually constant due to the high and constant vacancy concentration. Then, according to Eq. 10.5, the Fermi level is a definite function of the chemical potential µO of neutral oxygen which can also be expressed in terms of the oxygen partial pressure:


∆ ε F = – 1 ⋅ ∆µ O = – 2.303 kT ⋅ ∆ log pO 2 2 4e

273

10.6

Equation 10.6 and the assumption of a constant electrochemical potential of oxygen holds as long as the electron and hole concentrations are negligible as compared to the oxygen defect concentration. With respect to the band scheme, the Fermi level is far from the band edges under these conditions. For YSZ, the corresponding range of oxygen partial pressures where Eq. 10.6 is valid covers at least 30 decades at 800°C. The electron affinity, on the other hand, is not a thermodynamically controlled quantity. It depends on the polarity of the surface and, hence, on its orientation. Furthermore, it is strongly influenced by the surface dipole moment of adsorbed gas molecules. The first results for the electron affinity of YSZ as a function of the yttria content were obtained in experiments based on thermal emission of electrons (Odier and Loup, 1982). Reversible changes of the work function as a function of the oxygen partial pressure were determined for YSZ with Kelvin’s method (Nowotny and Sloma, 1986). With UPS, values for the absolute work function, the electron affinity and the position of the Fermi level were obtained (Wiemhöfer and Vohrer, 1992). For a single crystal of YSZ (10 mol% yttria) at 600°C in equilibrium with a Fe,FeO mixture, the work function was Φe = 5.2 eV and the electron affinity χe = 3.2 eV. If the measurement is carried out with a non-reversible platinum contact, the work function decreases and the reproducibility becomes poor. The Kelvin probe technique as well as UPS yielded the expected partial pressure dependence of the work function at constant temperature, namely ∆Φe = –(kT/4)∆log[p(O2)] (Schindler et al., 1989; Nowotny and Sloma, 1986). With UPS, it could be verified that this is due to the reversible shift of the position of the Fermi level with respect to εV as predicted by Eq. 10.6.

10.2.4 Examples for other solid electrolytes There are also examples of surface studies for other solid electrolytes. Figure 10.9 gives surface spectroscopic results for the band scheme of yttria stabilized bismuth oxide (Bi0.75Y0.25O1.5) which has the δ-Bi2O3 structure and shows a higher oxygen ion conductivity than YSZ (Shuk et al., 1997). The position of the Fermi level was determined in equilibrium with a Fe,FeO reference contact. As can be seen, the drawback of Bi2O3-based electrolytes is the narrow band gap leading to a much smaller electrolytic range. The onset of a significant electronic conductivity occurs even at moderately reducing conditions, i.e. at the low oxygen partial pressure of Fe,FeO mixtures. The band gap of bismuth oxide increases with the amount of dissolved yttria and zirconia. Therefore, attempts have been made to increase the band gap using different dissolved metal oxides as dopants. For Bi0.75Y0.25O1.5,

274

Materials for energy conversion devices ε – εV [eV]

εvac 4

Conduction band εC

Φe εF

2 εV

0 Valence band –2 –4

Density of states

10.9 Results for the band scheme of Bi0.75Y0.25O1.5 as obtained from EELS and UPS (Shuk et al., 1997). The Fermi level is drawn for equilibrium with a Fe,FeO reference contact at 600°C.

the band gap was εgap = 2.8 eV (Shuk et al., 1997). Further addition of yttria up to a composition of Bi0.6Y0.4O1.5, however, yielded only a moderate increase to εgap = 3.2 eV. This was not enough to depress the electronic conductivity sufficiently. Ceria-based solid electrolyte systems have become increasingly interesting in the past ten years due to the catalytic role of ceria on redox reactions, cf. Trovarelli (1996). In this context, a wealth of surface properties has been studied on pure and doped ceria and derived materials. UPS, XPS and EELS were used extensively on pure ceria films to analyse the changes during the Ce4+ → Ce3+ reduction which lead to a filling of the Ce(4f) state (Pfau and Schierbaum, 1994; Pfau et al., 1996; Mullins et al., 1998). The energy of the transition O(2p) → Ce (5d) in ceria is almost the same as for O(2p) → Zr(3d) in YSZ. But the effective band gap of pure and doped ceria is determined by the transition O(2p) → Ce(4f) which is centered at energies of 2.6 eV to 2.8 eV. As 4f electrons are strongly localized, the Ce(4f) states form a narrow band and the effect of band tailing is negligible. For CeO2, the Fermi level usually lies between the valence band edge and the Ce(4f) band. Figure 10.10 shows a compilation of data from the literature for the band diagram of doped ceria (Lübke and Wiemhöfer, 1999). The partial pressure scale reflects results derived from electronic conductivity measurements. Thus, considering the distance between valence band edge and Ce(4f), the effective band gap of ceria is far less than that of zirconia.


275

ε – εV [eV] Conduction band

Electrolytic domain in terms of εF (pO2)

5

εC

E/Volt

Ce(4f)

–1

0

εV

log [pO2/bar] –20

0

0

+1

+20

Valence band Density of states

10.10 Band scheme for doped ceria using available data from the literature and results of electrical measurements (Pfau and Schierbaum, 1994; Lübke and Wiemhöfer, 1999). The two scales on the right refer to the electrochemical interpretation in terms of electrode potential of electrodes that are applied to doped ceria as a solid electrolyte (cf. Section 10.3).

10.3

Electrode interfaces and electrode potential scale

10.3.1 Galvanic cells and electrode potential Figure 10.11 shows the principle of a typical oxygen concentration cell as used for solid oxide fuel cells or exhaust gas sensors. It illustrates the thermodynamic conditions and their relation to the band scheme of the solid electrolyte (Levy et al., 1988; Kleitz et al., 1991). The Fermi level and the chemical potential of oxygen are drawn assuming electrochemical equilibrium at the two electrode interfaces so that the values are fixed at both sides of each electrolyte/electrode interface. The cell voltage U of the cell is given by: pO′′ 2 U = – 1 ( ε ′′F – ε ′F ) = 2.303 kT log e 4e pO′ 2

10.7

The difference of oxygen chemical potentials at the two electrodes maintains a corresponding gradient of the electronic Fermi level in the solid electrolyte. This gradient is responsible for a gradient of the concentrations of electrons and holes as well as of the oxygen nonstoichiometry. In principle, this is a steady-state with a non-zero, but normally very small flux of neutral oxygen (permeation current). On the other hand, the distance between the Fermi level and the band edges or other electronic energy levels allows an estimate of the voltage limits at which the electronic conductivity begins to predominate the charge flow.

276

Materials for energy conversion devices Cell voltage U Pt″ (porous)

Pt′ (porous) ZrO2(+Y2O3) “YSZ”

pO ′2

pO ′′ 2

µO ′2 µO ′′ 2

µ˜ O2– Conduction band

ε F′′

ε F′ Valence band

10.11 Oxygen concentration cell with YSZ as solid electrolyte: the thermodynamic relations between the chemical and electrochemical potentials and the relation to the band scheme of YSZ are shown.

The band edges will remain virtually flat for the currentless case (space charges at the electrode interfaces can be neglected at higher temperatures), because any electric field will be cancelled out due to the high ionic conductivity. Hence, if the position of the Fermi level has been determined with respect to the band edges of the electrolyte for a given reference electrode potential, the entire electrode potential scale can be related unambiguously to the electronic band scheme of the solid electrolyte (Wiemhöfer and Vohrer, 1992). This approach relating the electrode potentials to a fixed energy scale of electrolyte is sometimes termed an absolute electrode potential scale and was first developed for aqueous electrolytes (Trasatti, 1977). Figure 10.12 illustrates this and makes evident why YSZ is a suitable electrolyte material for fuel cells. The partial pressure range from 1 bar to 10–30 bar oxygen corresponds to Fermi levels varying over a range of 1.5 eV around the centre of the band gap far from the band edges. The derivation of absolute electrode potential scales has been discussed extensively in recent years by other authors (Riess and Vayenas, 2003; Leiva and Sanchez, 2003; Tsiplakides and Vayenas, 2002). They take the vacuum level εvac of an electron at rest directly above the solid electrolyte surface as the definition of the zero level ϕ = – εvac/e = 0 for the electrode potential scale. Then, as Fig. 10.13 illustrates for a Pt/YSZ interface, the accuracy of measuring the work function of YSZ is decisive, because the absolute electrode potential corresponds to ϕabs = – Φe/e. A problem with this definition is that the electron affinity χe as one part in the work function (cf. Eq. 10.3) depends


277

ε [eV] εvac εF – εV Conduction band

= const (T) – εC

5 4

5.2 eV

3

ZrO2 → Zr ZrAg ZrAu4 ZrPt3

(εF – εV)Fe/FeO

2.3 kT log pO2 4

–40 –20

–1

0

0

20

1

2 (εF – εV)Pd/PdO

1 0

Valence band

–2

pO2 = 1.013 bar

εV log PO2 εO2 [V] [bar] ZrO2 (+ 10 mole % Y2O3), 800°C

Density of states

10.12 Construction of an absolute electrode potential scale referred to the experimentally determined band diagram of YSZ at 800°C (Wiemhöfer and Vohrer, 1992). The zirconium alloys at the oxygen partial pressure scale denote the limits where Pt, Au and Ag begin to react with the electrolyte under high cathodic polarization (formation of alloys with Zr at the interface between YSZ and the metal electrode). YSZ ε vac

ε Pt vac

εF

Conduction band (YSZ)

Conduction band (Pt)

∆εC

∆µ eYSZ εF

∆µ Pt e

εC Pt

χe Φ e εC

εV Valence band (YSZ) YSZ

10.13 Energy relations at a Pt/YSZ interface which can be used for the definition of reference levels in the context of an absolute electrode potential scale.

on the surface preparation (adsorbed molecules, surface orientation, etc.) and thus is not a well-controlled quantity. A much better reference point is one of the band edges which can be determined with good accuracy. Furthermore, the position of the Fermi level is thermodynamically controlled with respect to the band edges.

278


10.3.2 Structure and composition of electrode interfaces In general, the electrolyte/electrode interface is not directly accessible to surface spectroscopic techniques. In view of the small information depth of these techniques, interface studies have to be carried out with thin film electrodes on an electrolyte. The system Pt/YSZ is taken as an example illustrating some atomistic properties of the two-phase boundary that were studied with surface analytical methods. Figure 10.14 summarizes some of the main results from XPS and ISS as obtained on thin platinum films evaporated on single crystals of YSZ (Schindler et al., 1989). Reversible changes of the geometry of the Pt film were observed as a function of the pre-treatment at high and low oxygen partial pressures. High oxygen partial pressures lead to the formation of small isolated platinum clusters, whereas after a treatment under low oxygen partial pressures the Pt clusters spread out covering the electrolyte surface almost entirely. Oxygen adsorption 970 K, PO2 = 10–2 Pa Pt

“Pt-O”

Pt-Zr

YSZ

Oxygen desorption

YSZ

970 K, UHV: PO2 < 10–8 Pa XPS: Pt-Zr alloy formed XPS: Pt-O ISS: Pt only, very few O ISS: Pt, O, Zr, Y

10.14 Reversible geometrical changes observed on thin Pt films on a YSZ single crystal after annealing at different oxygen partial pressures (Schindler et al., 1989).

With XPS, at decreasing oxygen partial pressures, i.e. p O 2 < 10–10 mbar and 850°C, an additional shoulder of the Zr3d peak at lower binding energies was found, the intensity of which increases with time. This observation indicated a higher electron density of the zirconium ions directly at the Pt/ YSZ interface as a result of the decreasing interface concentration of oxygen. It is the first step to the formation of a PtZr3 alloy which is known to form at higher cathodic potentials (Lu et al., 1995). The reduced zirconium atoms at an oxygen deficient Pt/YSZ interface lead to a stronger bonding to the platinum metal. This explains the observed better adhesion of platinum on YSZ after a preceding reduction of the YSZ surface. Thus, the primary electrochemical effect of a changing electrode potential at the YSZ/electrode interface is a strong and reversible change of the oxygen-to-metal ratio and an accompanying change of bonding type from ionic to metallic at the interface.


279

10.3.3 In situ studies on electrodes on solid electrolytes For information on the electrochemistry of electrode/electrolyte systems, in situ studies of the electrode interface as a function of an applied electrode potential are quite attractive. Of course, a serious restriction of surface analytical techniques based on photoelectron emission is the necessity of ultra-high vacuum conditions. However, this does not exclude in situ investigations of non-equilibrium gas evolution at polarized electrodes. First examples of such experiments were published by Arakawa et al. who studied oxygen evolution at silver electrodes on stabilized zirconia with XPS (Arakawa et al., 1983a,b). In the past ten years, a series of further in situ experiments was published with regard to electrodes on YSZ (Wiemhöfer and Vohrer, 1992; Zipprich et al., 1995; Schindler et al., 1989; Rösch et al., 2000; Luerssen et al., 2000, 2001, 2002; Poppe et al., 1998, 1999; Hong et al., 1995). UPS as well as XPS was used to analyse polarized platinum, silver and gold electrodes. Various chemisorbed oxygen species could be distinguished in the analysis of the XPS O(1s) peaks. Subsurface as well as hydrogen-containing oxygen species were detected. Evaporated silver electrodes on YSZ show similar morphological changes after cathodic and anodic polarization as compared to the Pt/YSZ interface. Anodic polarization of Ag electrodes on YSZ at 500°C under 10–2 bar–1 bar led to the formation of a surface oxide layer which was accompanied by surface roughening. In open pores of an evaporated silver film on YSZ, small spherical silver particles appear after some cycles of cathodic and anodic polarization (Zipprich et al., 1995). No silver particles were detected in the pores after merely heating up the galvanic cell without polarization. A larger amount of very small silver particles and a more uniform distribution over the pores was found for anodic polarization. The appearance of the silver particles differed between anodic and cathodic experiments. The in situ experiments of various authors differed with respect to the applied counter and reference electrodes. A well-defined electrode potential makes necessary the use of a reference electrode with constant oxygen activity. Two approaches are available. One can apply the solid oxide ion conductor in the form of a tube closed at one end together with an inner Pt electrode and air as reference gas (Arakawa et al., 1983b; Rösch et al., 2000). The solid electrolyte tube separates the UHV at the outer working electrode from the air pressure at the inner counter electrode. This set-up has the advantage that high anodic current densities are available due to the large oxygen buffer at the counter electrode. In other studies, single crystal YSZ samples were applied with evaporated or platinum paste electrodes according to the principle shown in Fig. 10.2(b) (Zipprich et al., 1995). A disadvantage of using Me, MeO contacts as an oxygen buffer is that a slow irreversible loss of oxygen from the metal/metal

280


oxide mixtures occurs during the experiments in the UHV. Therefore, one has to reload the reference and counter electrodes from time to time by anodic polarization under increased oxygen pressure (in a special preparation chamber of the UHV equipment). In situ experiments have been done with XPS on a cell according to Fig. 10.2(b) and 10.2(c) with a YSZ single crystal and all three electrodes made of Pt or Au (Vayenas et al., 1997; Neophytides et al., 2000). Standard UHV conditions correspond to a reducing atmosphere. Thus the reference potential may be far in the cathodic region and may also shift with time during polarization. The authors analysed the oxygen species appearing at the surface of the working electrode during polarization. Experiments with lateral resolution were carried out recently using PEEM (photoelectron emission spectroscopy) and SPEM (scanning photoelectron microscopy) (Luerssen et al., 2000, 2001, 2002). These techniques allow the study of lateral gradients on and between microstructured electrodes. Recent interesting results obtained with mass spectrometric techniques showed that bare YSZ surfaces and polarized Ag and Pt electrodes on YSZ emit O– anions from the surface (Torimoto et al., 2002; Fujiwara et al., 2003b). Even a free YSZ surface shows field-induced emission of O– and e– resulting from dissociation of surface ions O2– (Nishioka et al., 2003). This observation may be used for an electrochemical oxygen ion source (Fujiwara et al., 2003a).

10.3.4 Work function of polarized electrodes on YSZ The work function of an electrode is a suitable indicator for changes of composition and for adsorbed species that occur at the electrode surface as a function of the electrode potential. As demonstrated by Vayenas and coworkers in many experiments on catalytically active metal electrodes with YSZ, the work function of real electrodes is often proportional to the electrode potential and, thus, the catalytic activity of a corresponding electrode surface can be strongly enhanced by electrochemical polarization (Tsiplakides and Vayenas, 2001, 2002). This is usually termed the NEMCA effect. The common interpretation uses the assumption that a complete surface equilibrium of the active species is obtained by a ‘spill-over’, i.e. a surface diffusion of active particles starting from the three-phase boundary line at dispersed metal contacts to the outer free surface of the metal. For metal electrodes that do not dissolve gases, the primary cause for a changing work function must be attributed to adsorbed gas molecules and their effect on the surface dipole moment. The work function of a mixed conducting electrode will depend in addition on the stoichiometric composition which can influence the electron affinity as well as the position of the Fermi level. Such a behaviour can be expected for oxides like La1–xSrxCoO3 on


281

YSZ, but also for metals like silver or palladium which can dissolve oxygen or hydrogen atoms. There are indeed results from UPS on thin Ag film electrodes on YSZ that showed a nearly linear dependence of the work function of Ag on the applied electrode potential (vs. a reference electrode) (Zipprich et al., 1995; Rösch et al., 2000). Figure 10.15 shows typical results for a porous Ag film. However, a complete equivalence according to ∆Φ = e ∆U (as postulated in Tsiplakides and Vayenas (2002)) was not found by Rösch et al. (2000). This may be due to the competition between electrochemical formation of adsorbed species at the interface to YSZ and their desorption at the outer surface under the rather low pressures of the UHV which prevents a clear observation of the spillover of volatile, adsorbed species such as oxygen on thicker electrodes. 6.0 820 K

Ag(evap.) | YSZ | Pd, PdO

Φe(Ag) (eV)

5.5 Anodic 5.0

4.5

4.0

0.0

Cathodic

0.5 1.0 1.5 Electrode potential vs. reference (V)

2.0

10.15 Work function of silver electrodes on YSZ as a function of the applied electrode potential under UHV conditions (Zipprich et al., 1995).

10.4

Outlook

The combination of the atomistic resolution of surface analytical techniques with a high lateral resolution offers a great chance for the understanding and modelling of the electrochemistry of solid electrolyte interfaces. This concerns the modern scanning photoelectron emission techniques as well as the near field techniques based on STM and AFM. The use of these possibilities for investigation of solid electrolyte surfaces and electrodes can deliver new insight into the atomistic kinetics of surface defects. Surfaces and interfaces in solid electrolyte devices can be analysed with the same accuracy as metals and semiconductors. Indeed, as was shown, the electronic properties of solid electrolytes play a decisive role in their electrochemistry.

282

10.5


Abbreviations and symbols

SOFC solid oxide fuel cell YSZ yttria stabilized zirconia (Zr1–xYxO2–x/2 with x usually between 0.1 and 0.2) e elementary charge k Boltzmann’s constant partial pressure of oxygen pO 2 U cell voltage energy of the conduction band minimum (lower conduction band εC edge) Fermi level, Fermi potential (partial free energy per electron) εF band gap energy (εgap = εC – εV) εgap energy of the valence band maximum (upper valence band edge) εV energy of an electron at rest in the vacuum just outside the solid εvac ˜µ e electrochemical potential of electrons (identical to εF) chemical potential of oxygen µO chemical potential of oxygen molecules µ O2 µ˜ O 2– electrochemical potential of oxygen ions ϕ electrical potential work function of electrons (Φe = εvac – εF) Φe electron affinity (χe = εvac – εC) χe

10.6

References

Anderson, P.W. (1958), Phys. Rev., 109, 1492–1505. Aoki, M., Chiang, Y.M., Kosacki, I., Lee, I.J.R., Tuller, H. and Liu, Y.P. (1996), J. Am. Ceram. Soc., 79, 1169–80. Arakawa, T., Saito, A. and Shiokawa, J. (1983a), Appl. Surf. Sci., 16, 365–72. Arakawa, T., Saito, A. and Shiokawa, J. (1983b), Chem. Phys. Lett., 94, 250–2. Badwal, S.P.S., Bak, T., Jiang, S.P., Love, J., Nowotny, J., Rekas, M., Sorrell, C.C. and Vance, E.R. (2001), J. Phys. Chem. Solids, 62, 723–9. Badwal, S.P.S., Ciacchi, F.T., Rajendran, S. and Drennan, J. (1998), Solid State Ionics, 109, 167–86. Bak, T., Nowotny, J., Rekas, M. and Sorrell, C.C. (2001a), Bull. Pol. Acad. Sci.-Chem., 49, 67–73. Bak, T., Nowotny, J., Rekas, M. and Sorrell, C.C. (2001b), Bull. Pol. Acad. Sci.-Chem., 49, 75–83. Bäuerle, J.E. (1969), J. Phys. Chem. Solids, 30, 2657. Berner, U. and Schierbaum, K.D. (2002), Phys. Rev. B, 65. Burggraaf, A.J. and Winnubst, A.J.A. (1988), In Surface and Near-Surface Chemistry of Oxide Materials (Eds, Nowotny, J. and Dufour, L.-C.), Elsevier, Amsterdam, pp. 449– 77. Burggraaf, A.J., Van Hemert, M., Scholten, D. and Winnubst, A.J.A. (1985), In Reactivity of Solids (Eds, Barret, P. and Dufour, L.C.), Elsevier, Dijon, pp. 797–802. Casalis, L., Jark, W., Kiskinova, M., Lonza, D., Melpignano, P., Morris, D., Rosel, R.,


283

Savoia, A., Abrami, A., Fava, C., Furlan, P., Pugliese, R., Vivoda, D., Sandrin, G., Wei, F.Q., Contarini, S., Deangelis, L., Gariazzo, C., Nataletti, P. and Morrison, G.R. (1995), Rev. Sci. Instrum., 66, 4870–5. de Ridder, M., van Welzenis, R.G., van der Gon, A.W.D., Brongersma, H.H., Wulff, S., Chu, W.F. and Weppner, W. (2002), J. Appl. Phys., 92, 3056–64. Duran, P., Capel, F., Moure, C., Gonzalez-Elipe, A.R., Caballero, A. and Banares, M.A. (1999), J. Electrochem. Soc., 146, 2425–34. Egelhoff Jr., W.F. (1987), Surf. Sci. Rep., 6, 253–415. Ertl, G. and Küppers, J. (1985), Low Energy Electrons and Surface Chemistry, WileyVCH, Weinheim. Fujiwara, Y., Kaimai, A., Hong, J.O., Yashiro, K., Nigara, Y., Kawada, T. and Mizusaki, J. (2003a), J. Electrochem. Soc., 150, E117–E124. Fujiwara, Y., Sakai, T., Kaimai, A., Yashiro, K., Nigara, Y., Kawada, T. and Mizusaki, J. (2003b), J. Electrochem. Soc., 150, E543–E547. Guo, X. and Maier, J. (2001), J. Electrochem. Soc., 148, E121–E126. Guo, X., Sigle, W. and Maier, J. (2003), J. Am. Ceram. Soc., 86, 77–87. Hong, J.K., Oh, I.H., Hong, S.A. and Lee, W.Y. (1995), Appl. Surf. Sci., 89, 229–35. Jin-Phillipp, N.Y., Sata, N., Maier, J., Scheu, C., Hahn, K., Kelsch, M. and Ruhle, M. (2004), J. Chem. Phys., 120, 2375–81. Kleitz, M., Siebert, E., Fabry, P. and Fouletier, J. (1991), In Chemical and Biochemical Sensors Part I, Vol. 2 (Eds, Göpel, W., Hesse, J. and Zemel, J.N.), Wiley VCH, Weinheim, pp. 341–428. Knauth, P. (2002), J. Solid State Electrochem., 6, 165–71. Knöner, G., Reimann, K., Röwer, R., Sodervall, U. and Schaefer, H.E. (2003), Proc. Natl. Acad. Sci. U.S.A., 100, 3870–3. Kobayashi, S., Yamasaki, A. and Fujiwara, T. (2003), Jpn. J. Appl. Phys. Part 1 – Regul. Pap. Short Notes Rev. Pap., 42, 6946–50. Lee, J.S., Adams, S. and Maier, J. (2000), Solid State Ionics, 136, 1261–6. Lei, Y.Y., Ito, Y., Browning, N.D. and Mazanec, T.J. (2002), J. Am. Ceram. Soc., 85, 2359–63. Leiva, E.P.M. and Sanchez, C.G. (2003), J. Solid State Electrochem., 7, 588–92. Levy, M., Fouletier, J. and Kleitz, M. (1988), J. Electrochem. Soc., 135, 1584–9. Lu, F.H., Newhouse, M.L., Dieckmann, R. and Xue, J. (1995), Solid State Ionics, 75, 187–92. Lübke, S. and Wiemhöfer, H.D. (1999), Solid State Ionics, 117, 229–43. Luerssen, B., Gunther, S., Marbach, H., Kiskinova, M., Janek, J. and Imbihl, R. (2000), Chem. Phys. Lett., 316, 331–5. Luerssen, B., Janek, J. and Imbihl, R. (2001), Solid State Ionics, 141, 701–7. Luerssen, B., Janek, J., Gunther, S., Kiskinova, M. and Imbihl, R. (2002), Phys. Chem. Chem. Phys., 4, 2673–9. Maier, J. (1999), J. Eur. Ceram. Soc., 19, 675–81. Maier, J. (2003), Solid State Ionics, 157, 327–34. Mizutani, N. and Nowotny, J. (1998), J. Mater. Synth. Process, 6, 401–5. Mott, S.N., Pepper, M., Pollitt, S., Wallis, R.H. and Adkins, C.J. (1975), Proc. Roy. Soc. Lond. A, 345, 169–205. Muccillo, E.N.S. and Muccillo, R. (2002), Br. Ceram. Trans., 101, 259–62. Mullins, D.R., Overbury, S.H. and Huntley, D.R. (1998), Surf. Science, 409, 307–19. Nair, J.P., Wachtel, E., Lubomirsky, I., Fleig, J. and Maier, J. (2003), Adv. Mater., 15, 2077–2081.

284


Namai, Y., Fukui, K. and Iwasawa, Y. (2003), J. Phys. Chem. B, 107, 11666–73. Neophytides, S.G., Zafeiratos, S. and Kennou, S. (2000), Solid State Ionics, 136, 801–6. Nishioka, M., Torimoto, Y., Kashiwagi, H., Li, Q.X. and Sadakata, M. (2003), J. Catal., 215, 1–6. Norenberg, H. and Briggs, G.A.D. (1997), Phys. Rev. Lett., 79, 4222–5. Nowotny, J. and Sloma, M. (1985), Bull. Pol. Acad. Sci.-Chem., 33, 337–45. Nowotny, J. and Sloma, M. (1986), J. Physique, 47, 807–11. Odier, P. and Loup, J.P. (1982), In 9th Int. Symp. Reactivity of Solids (Eds, Dyrek, K., Haber, J. and Nowotny, J.), Elsevier, Amsterdam, pp. 458–73. Park, J.H. and Blumenthal, R.N. (1989), J. Electrochem. Soc., 136, 2867–76. Pfau, A. and Schierbaum, K.D. (1994), Surf. Science, 321, 71–80. Pfau, A., Sanz, J., Schierbaum, K.D., Göpel, W., Belzunegui, J.P. and Rojo, J.M. (1996), In 11th International Congress on Catalysis – 40th Anniversary, Pts A and B, Vol. 101, pp. 931–40. Poppe, J., Schaak, A., Janek, J. and Imbihl, R. (1998), Ber. Bunsenges. Phys. Chem., 102, 1019–22. Poppe, J., Volkening, S., Schaak, A., Schutz, E., Janek, J. and Imbihl, R. (1999), Phys. Chem. Chem. Phys., 1, 5241–9. Riess, I. and Vayenas, C.G. (2003), Solid State Ionics, 159, 313–29. Robertson, J. (2000), J. Vac. Sci. Technol. B, 18, 1785–91. Rösch, M., Reinhardt, G. and Göpel, W. (2000), Solid State Ionics, 136, 791–9. Sasaki, K. and Maier, J. (2000), Solid State Ionics, 134, 303–21. Sata, N., Jin-Phillipp, N.Y., Eberl, K. and Maier, J. (2002), Solid State Ionics, 154, 497– 502. Scanlon, P.J., Bink, R.A.M., van Berkel, F.P.F., Christie, G.M., van Ijzendoorn, L.J., Brongersma, H.H. and Van Welzenis, R.G. (1998), Solid State Ionics, 112, 123–30. Schierbaum, K.D. (1998), Surf. Science, 399, 29–38. Schindler, K., Schmeisser, D., Vohrer, U., Wiemhöfer, H.D. and Göpel, W. (1989), Sens. Actuator B-Chem., 17, 555–68. Shuk, P., Wiemhöfer, H.D. and Göpel, W. (1997), Z. Anorgan. Allg. Chem., 623, 892–6. Steele, B.C.H. and Butler, E.P. (1985), Brit. Ceram. Proc., 36, 45–55. Theunissen, G., Winnubst, A.J.A. and Burggraaf, A.J. (1992), J. Mat. Sci., 27, 5057–66. Torimoto, Y., Nishioka, M. and Sadakata, M. (2002), J. Catal., 209, 256–9. Trasatti, S. (1977), In Advances in Electrochemical and Electrical Engineering, Vol. 10, pp. 213–321. Trovarelli, A. (1996), Catal. Rev.-Sci. Eng., 38, 439–520. Tschöpe, A., Kilassonia, S., Zapp, B. and Birringer, R. (2002), Solid State Ionics, 149, 261–73. Tsiplakides, D. and Vayenas, C.G. (2001), J. Electrochem. Soc., 148, E189–E202. Tsiplakides, D. and Vayenas, C.G. (2002), Solid State Ionics, 152, 625–39. Tuller, H.L. (1997), J. Electroceramics, 1, 211–18. Tuller, H.L. (2000), Solid State Ionics, 131, 143–57. Vayenas, C.G., Lambert, R.M., Ladas, S., Bebelis, S., Neophytides, S., Tikhov, M.S., Filkin, N.C., Makri, M., Tsiplakides, D., Cavalca, C. and Besocke, K. (1997), In Spillover and Migration of Surface Species on Catalysts, Vol. 112, pp. 39–47. Vohrer, U., Wiemhöfer, H.D., Göpel, W., Schilling, F. and Arndt, J. (1991), Sens. Actuator B – Chem., 4, 411–16. Vohrer, U., Wiemhöfer, H.D., Göpel, W., Van Hassel, B.A. and Burggraaf, A.J. (1993), Solid State Ionics, 59, 141–9.


285

Von Oertzen, A., Rotermund, H.H., Jakubith, S. and Ertl, G. (1991), Ultramicroscopy, 36, 107–16. Weiss, W., Wiemhöfer, H.D. and Göpel, W. (1992), Phys. Rev. B, 45, 8478–89. Wiemhöfer, H.D. (1993), Ber. Bunsenges. Phys. Chem., 97, 461–9. Wiemhöfer, H.D. (1996), In Ceramic Oxygen Ion Conductors and Their Technological Applications, Vol. 56 (Ed, Steele, B.C.H.), The Institute of Materials, London, pp. 1– 23. Wiemhöfer, H.D. and Göpel, W. (1991), Fresenius J. Anal. Chem., 341, 106–11. Wiemhöfer, H.D. and Vohrer, U. (1992), Ber. Bunsenges. Phys. Chem., 96, 1646–52. Wiemhöfer, H.D., Harke, S. and Vohrer, U. (1990a), Solid State Ionics, 40–1, 433–9. Wiemhöfer, H.D., Schmeisser, D. and Göpel, W. (1990b), Solid State Ionics, 40–1, 421– 7. Winnubst, A.J.A., Kroot, P.J.M. and Burggraaf, A.J. (1983), J. Phys. Chem. Solids, 44, 955–60. Worrell, W.L. and Weppner, W. (1984), Am. Ceram. Soc. Bull., 63, 998–9. Zhu, J. and Liu, Z.G. (2004), Appl. Phys. A – Mater. Sci. Process., 78, 741–4. Zipprich, W., Wiemhöfer, H.D., Vohrer, U. and Göpel, W. (1995), Ber. Bunsenges. Phys. Chem., 99, 1406–13.

11 Interface mass transport in oxide materials E G G O N T I E R - M O Y A, A S I A H M E D and F M O Y A, Université Paul Cezanne, France

11.1

Introduction

The ‘life’ of a functional material (fabrication, service and degradation) involves several diffusion-related processes, in which interfaces play a key role (Gupta, 2003), as they spawn additional equilibria and diffusion paths. The knowledge of the defect structures and diffusion effects of grain boundaries are necessary to optimize the properties of ceramics and, furthermore, of the nanocrystalline oxides. In the first section of this chapter, we describe a novel approach to characterize defects induced by impurities by calling for Positron Annihilation Lifetime Spectroscopy. In the second section, we outline the present state-of-the-art concerning grain boundary diffusion in oxides. We refer to the available literature, as a comprehensive basis, and we focus on some relevant aspects encountered during our research in this field, which shed light on the complexity of oxides compared to metals.

11.2

Characterization of defects in oxide ceramics by Positron Annihilation Lifetime Spectroscopy

Positron Annihilation Lifetime Spectroscopy (PALS) is sensitive to vacancylike defects in material (Hautojärvi, 1979). In oxide ceramics, the dissolution of aliovalent cations requires that charged point defects be created in order to maintain electric neutrality. In alumina, these extrinsic defects dominate by far the intrinsic ones at any temperature (Dörre and Hübner, 1984). Positively charged defects repel positrons whereas the neutral or negatively charged ones can act as positron traps. In practice, the materials contain various impurities greater or lesser in valence than the host cation. Hence, interactions between extrinsic defects are expected during the elaboration process (Grimes, 1994; Lagerlöf and Grimes, 1998), as well as during the material use. If impurities that are able to give birth to positron traps through their dissolution mechanism are present, one would expect PALS to somewhat reflect the defect structure. However, 286

Interface mass transport in oxide materials

287

this is possible only if the lifetime components, which characterize the most important annihilation processes, are resolved. Recently (Kansy et al., 2001; Moya et al., 2003; Si Ahmed et al., 2004), three annihilation processes have been resolved from spectra performed in sintered alumina at room temperature (i.e., annihilation in the bulk-free defects, in defects within the grains and in defects located at grain boundaries). One of the novelties of these contributions is an attempt to characterize the defects at grain boundaries. The objective of this section is to describe the model and define its scope in a way which allows further utilisation in Al2O3 and other oxides.

11.2.1 The model and method of analysis The experimental spectra obtained with a standard NaCl positron source (a typical spectrum is shown in Fig. 11.1) can be fitted by using a recent version of the LT program (Kansy, 1996), in which the three-state trapping model (Krause-Rehberg and Leipner, 1999) was introduced into the source code. An experimental spectrum, S(t), can be decomposed as 3

S(t) = Σ I j exp (–t/τj)

11.1

j =1

a

No. of counts

100,000

b 10,000 c

–1

0

1 2 kTime (ns)

3

4

11.1 PALS spectrum measured for sintered alumina of average grain diameter 1.7 µm. The points represent the measured data. The lines show the fitted components of the spectrum originating from: (a) positron annihilation in defects located within the grain, (b) in the bulk-free defects and (c) in defects at grain boundaries.

288


where τj are the lifetimes and Ij the intensities. These parameters are expressed via the following relations: τ1 = 1/[(1/τb) + κg + κgb] τ2 = τg τ3 = τgb

11.2

    I1 = 1 – (I2 + I3) I 2 = κ g  1 – 1  I 3 = κ gb  1 – 1  τ τ τ τ  1  1 g  gb  11.3 In these relations, τb is the bulk lifetime (the positron lifetime in the bulkfree defects), τg the lifetime in the defects within the grain, τgb the lifetime in defects at grain boundaries, while κg and κgb are the corresponding trapping rates. The fitting parameters are the trapping rates (κg and κgb), the trapped positron lifetimes (κg and κgb) and the bulk lifetime τb. It must be pointed out that, in Al2O3, the bulk lifetime, τb, is no longer a fitting parameter, as it has been unambiguously derived from measurement in Al2O3 single crystal of high purity where τb was found equal to 122 ± 2 ps (Moya et al., 2003). Figure 11.1 shows the lifetime components resulting from the deconvolution of an experimental spectrum. The fitted trapping rates κg and κgb can be expressed as κ = µC

11.4

where µ is the specific trapping rate (a constant for each type of defects) and C is the concentration of positron traps. Therefore, the value of C (and hence the concentration of impurities that induce the traps) can be reached.

11.2.2 The scope of the method The application of the model to a particular oxide must be preceded by a survey of the possible dissolution mechanisms of aliovalent impurities. Cations greater in valence than the host are the only ones which are expected to create negatively charged cationic vacancies. Such isolated defects would act as positron traps. They could also associate with other extrinsic defects to form neutral or negatively charged clusters, which could also trap positrons. In the case of Al2O3, the possible dissolution mechanisms of tetravalent oxides, such as SiO2 and TiO2, written using the Kröger-Vink notation, are: ⋅ 2 MO 2 ⇔ 2 M Al + O ′′i + 3O O×

11.5

⋅ 3 MO 2 ⇔ 3 M Al + VA1 ′′′ + 6O O×

11.6

Simulation studies (Grimes, 1994) and some experimental evidence (Moon and Phillips, 1991) have shown that the process described by Eq. 11.6, which leads to the creation of negatively charged aluminium vacancies, is more likely than the oxygen interstitial compensation mode (Eq. 11.5).


289

If Si (or Ti) is the sole impurity, the defect structure likely to be achieved ⋅ during the sintering process comprises isolated VA1 : VA1 ′′′ and [Si A1 ′′′ ]′′ , ⋅ ⋅ x [Si A1 : VA1 ′′′ ]′′ , [3Si A1 :VA1 ′′′ ] clusters. However, the situation can be further complicated by the presence of elements lesser in valence than Al such as Mg and Ca. There, mutual compensation of the substitutional Ca ′A1 (or ⋅ Mg ′A1 ) and Si A1 is likely (Gavrilov et al., 1999a, b). Hence, by this means, some fractions of Si could be prevented from inducing negatively charged cationic vacancies. Indeed, mass action calculations (Lagerlöf and Grimes, 1998) have predicted that the concentration of VA1 ′′′ depends strongly on the presence of Mg and Ca and could even become negligible if Mg and Ca dominate. The method, therefore, applies to materials containing cations of valence greater than the host. In alumina, Si is often present and its effect on the microstructural development during sintering has been recognized (Bae and Baik, 1993). In previous investigations (Kansy et al., 2001; Moya et al., 2003; Si Ahmed et al., 2004) concerning sintered alumina where Si was the dominant foreign element, the natures of defects responsible for positron trapping were identified (i.e., negatively charged cationic vacancies inside the grain and clusters containing these vacancies and defects induced by segregated elements at grain boundaries). The model provides further characterizations such as: • the extent of Si segregation at grain boundaries via the ratio of the positron trapping rates κgb/κg • the assessment of the mutual compensation effects when elements lesser in valence than the host are present • the determination of the solubility limits. For the purpose of illustrating these possibilities, we report in Fig. 11.2 recent results (Si Ahmed et al., 2004) concerning sintering of alumina, where in addition to Si, Mg and Ca are present. In this figure, the ratio (κgb /κg) is plotted as a function of the specific area of grains (3/R). The grain radii R, were reached by controlling the firing schedule. From the straight line, it can be written: (κgb /κg) ∝ (µgb /µg) × (Cgb /Cg)

11.7

Since a positron trap (within the grain or at grain boundaries) contains a cationic vacancy that is induced by Si, the ratio (Cgb /Cg) obtained from the fitted trapping rates is also the enrichment ratio of silicon. All these characterizations could provide information for a better understanding of the transport mechanism in sintered Al2O3, through the identification of defects.

290

Materials for energy conversion devices Trapping rate ratio κgb/κg (%)

1.2

0.8

0.4

0 0

2 4 Specific area of grains 3/R (µm–1)

6

11.2 Trapping rate ratio, κgb/κg, as a function of specific surface of grains, 3/R.

11.3

Mass transport in polycrystalline oxides

Grain boundary diffusion studies require the knowledge of bulk diffusion, since the two processes are complementary. Therefore, we approach the subject in two parts, i.e. bulk and grain boundary diffusion, to present the findings and the particular problems concerning oxides.

11.3.1 Bulk diffusion in oxides To describe the concentration profiles in semi-infinite substrates, one usually refers to two simple cases, namely the ‘instantaneous source’ and the ‘constant source’ (Philibert, 1991). The corresponding solutions are a Gaussian curve and an erfc function. In practice, the source geometry may fall between these extremes (instantaneous source = thin deposit and high solubility, constant source = thick deposit and low solubility, or element provided by a gas phase), and the form of the profile may approximate the erfc type for short times, and the Gaussian type for longer times. In all cases, the depth of the concentration profiles is defined by the diffusion length Dt , where D is the bulk diffusion coefficient and t the time. For ionic crystals, the bulk diffusion process may not be so simple. The segregation of defects at interfaces gives rise to an electric field (space charge), and the transport of charged defects throughout this near-surface or interface layer may be partially rate controlling (Adamczyk and Nowotny, 1986; Nowotny, 1988a, b). Consequently, the classical analysis of the diffusion equations lead to apparent diffusion data which cannot simply be ascribed to bulk transport.


291

Self-diffusion in oxides For metals, simple correlations have been found between the self-diffusion coefficients and the melting temperature Tm. The Arrhenius curves of self diffusion in fcc metals, plotted as a function of the reciprocal reduced temperature Tm/T, fall close to a common line, described by D = D0 exp (–Q/RT), with average values Q/RTm = 18.41 (Brown and Ashby, 1980) (see Fig. 11.3). 10–10

D(m2 s–1)

10–12

10–14

10–16

CoO

10–18

MgO Fcc metals NiO

–20

10

1

1.5

2 Tm /T

2.5

3

11.3 Bulk diffusion coefficients as a function of the reciprocal reduced temperature. Common line for fcc metals (Brown and Ashby, 1980), cationic self diffusion in MgO (Wuensch et al., 1973), NiO (Atkinson and Taylor, 1978), CoO (Hoshino and Peterson, 1985).

The case of oxides appears much more complicated. Diffusion data can be found in a book (Kofstad, 1972) and review papers (Harrop, 1968; Freer, 1980; Brown and Ashby, 1980; Peterson, 1984a; Matzke, 1986, 1987, 1991; Nowick, 1989; Monty, 1992). The ratios Q/RTm span from about 10 to 25. The chemistry of the compound, the concentration of intrinsic defects, the extent of nonstoichiometry and the presence of impurities are factors which influence the defect population. Consequently, for similar structures, the self-diffusion coefficients can be very different. We report in Fig. 11.3 the Arrhenius curves of cation self diffusion coefficients in MgO (Wuensch, et al. 1973), NiO (Atkinson and Taylor, 1978) and CoO (Hoshino and Peterson, 1985). Although these oxides have in common the rocksalt structure, no common curve can be plotted to describe these kinetics. In addition, large discrepancies are observed for data given by different authors. It has been noted (Harrop, 1968; Freer, 1980) that such differences should be attributed to varying extrinsic factors rather than to experimental errors.

292


Heterodiffusion in oxides Heterodiffusion in oxides is more complicated than self-diffusion. The interactions of impurities with defects modify both the concentration of defects and their mobility, with diffusion results depending on the relative modifications induced by the two effects. As an example, let us consider diffusion in cobalt oxide. Incorporation of chromium atoms in this nonstoichiometric oxide is assumed to increase the concentration of cobalt vacancies, and therefore the nonstoichiometry (Mrowec and Grzesik, 2003). However, it has been observed that bulk diffusion coefficients of Cr in CoO are 103 to 104 lower than cation self-diffusion ones. This result contradicts what is expected from the increase of cation vacancy concentration. It indicates that the mobility of Cr ions is substantially reduced near cobalt vacancies (Peterson, 1984b). It has been suggested that the trivalent Cr ions, the host cobalt ions and the charged cobalt vacancies associate in a spinel-like complex of low mobility (Kowalski et al., 1996). As a guideline, it would be interesting to relate the bulk diffusion coefficients in a given type of oxide to the size of the diffusing ion. These attempts to rationalize the data yield contradictory conclusions. In MgO, the impurity diffusion coefficients of divalent cations decrease when their ionic radius increase (Wuensch, 1983; Matzke, 1986). In contrast, in NiO and CoO, the activation energy for transition-metal impurity-ion decreases with increasing ionic radius (Hoshino and Peterson, 1984). In alumina, the yttrium diffusion coefficient is close to that of chromium (Moya et al., 1998), in spite of the large difference of the ionic radii (r(Y3+) = 0.093 nm and r(Cr3+) = 0.063 nm). Near-surface/interface diffusion In addition to the near-surface diffusive resistance resulting from segregation of charged species (Adamczyk and Nowotny, 1986; Nowotny, 1988a, b), another effect, which has received less attention but may have significant practical consequences, should be considered. Indeed, after thermal treatments at high temperatures, such as in the sintering process, ceramics are allowed to come back to room temperature without quenching, to avoid thermal shocks. During a ‘slow’ cooling, a redistribution of impurities may occur in the near surface or interface region. The shape of the profile depends on the reaction which takes place at the interfaces, the heat treatment temperature and the cooling rate. For instance, Fig. 11.4 illustrates the distribution of diffused Ti3+ in sapphire, followed by fluorescence, after ‘fast’ and slow coolings from 1950°C to room temperature (Hickey, 1998). The same effect occurs when starting with an initially doped material, as in the case of Cr-doped CoO (Bernasik et al., 1997). The surface segregationinduced depth profiles determined by SIMS are very different when the


Fluorescence yield (equivalent wt% Ti2O3 in Al2O3)

(a) 2 hour diffusion

293

(b) 8 hour diffusion

0.15

Cooled slowly Cooled rapidly

S137 0.10

S139

0.05

S130

S133

0.00 0

10

20 30 40 Depth (µm)

50

0

10

20 30 40 Depth (µm)

50

11.4 Comparison of Ti3+ distribution in sapphire for rapidly cooled samples (dotted lines) and slowly cooled samples (solid lines) (Hickey, 1998).

annealings (1373–1673 K) are followed by a quenching or a slow cooling. After quenching at 1000°C/h, a marked enrichment of Cr is observed within a surface layer a few nm thickness. On the other hand, a slow cooling at 500°C/h results in substantial impoverishment of a wider layer. The authors suggest that the segregating species are still mobile during the cooling stage, where a decomposition of the surface structure takes place.

11.3.2 Grain boundary diffusion Grain boundaries are paths which allow a transport of matter in deep regions of the solids. It must be reminded that the boundary thickness δ is small (a few interatomic distances). Assuming that the grains are spherical of radius R, one can estimate roughly the fraction f of atoms ‘in the boundaries’ as given by: f = atoms in the boundaries = 3δ R atoms in the grains

11.8

For example, with small grains (R = 5 µm), assuming the usual value δ = 1 nm, this ratio is equal to 6 × 10–4. From this estimation, one can see that the number of atoms in the grain boundaries is so small that the increase of concentration resulting from grain boundary diffusion is, in fact, the result of bulk diffusion from the boundaries inside the grains. Depending on the bulk penetration distance Dt , and the distance d between two boundaries, three diffusion regimes A, B and C can be distinguished, as schematized in Fig. 11.5. Fundamentals on grain boundary diffusion can be found in a book (Kaur et al. 1995) and review papers (Peterson, 1983; Mishin et al., 1997; Mishin

294

Materials for energy conversion devices x d

d

d

Dt ≈ 0

Dt Dt δ→

→

y

Regime A:

D ′t

Dt >> d

Regime B: 100δ < Dt < d/20

Regime C:

Dt < δ/20

11.5 Schematic representation of the three diffusion kinetics regime A, B and C in a polycrystalline body D = bulk diffusion coefficient, D′ = grain boundary diffusion coefficient, t = diffusion time, d = distance between the boundaries, δ = grain boundary thickness.

and Herzig, 1999; Mishin, 2001). The main parameters to take into account are the increase in diffusivity D′/D, where D′ is the grain boundary diffusion coefficient, the segregation coefficient α, the bulk diffusion length Dt and the grain boundary thickness δ, which are contained in the dimensionless parameter β: β = (D′αδ /D)/2 Dt

11.9

To measure the grain boundary diffusion coefficients, experiments are generally carried out on polycrystals in the conditions of the B regime (100 δ < Dt < d/20). From sectioning experiments, the average concentration c is determined as a function of the penetration y, or as a function of the reduced depth η = y/ Dt . Characteristic diffusion profiles exhibit two distinct parts: a steep part corresponding to bulk diffusion, and beyond this range, a long tail of small slope resulting from short circuit paths. Most diffusion data are based on such profiles, and analysed by Le Claire relation (Le Claire, 1963), which gives the gradient of the linear plot log c = f (y6/5) as a function of the diffusion parameters. It is also possible to use a log c = f (y) plot, which practically appears as linear in the sectioning range, and to compare the slope p = d log10 c/dη with those given by the following expressions (Moya and Moya, 1986): 5 < β < 10, p = 0.527β–0.540, 10 < β < 100, p = 0.503β–0.522, 100 < β < 1000, p = 0.483β–0.511

11.10

Whatever the chosen method, the experiments give access to a triple product, D′αδ. Some works have indicated large δ values for oxides, however it seems now that δ in oxides and metals are comparable, i.e. δ = 0.5 to 1 nm. Assuming δ = 1 nm, experiments in the B regimes yield the product D′α.


295

A survey on grain boundary diffusion in oxides Several articles (Kingery, 1974; Atkinson, 1984; Monty and Atkinson, 1989; Déchamps and Barbier, 1989, 1991; Matzke, 1991; Moya et al., 1991; Monty 1992; Lesage, 1994; Harding, 2003) have summarized the progress in research on interface diffusion in oxides. Alumina has been the subject of separate reviews (Dörre and Hübner, 1984; Moya and Moya, 1988, 1989; Harding, et al., 2003), which can be completed by two recent publications (GontierMoya et al., 2001; Vallasek et al., 2001). All available data on dislocation and grain boundary diffusion in ceramics (oxide and non-oxide) up to 1999 are collected in a data book (Erdélyi and Beke, 1999). To complete this survey, it is worth mentioning some recent measurements of grain boundary diffusion in materials of interest for solid oxide fuel cells (Matsuma et al., 1998; Horita et al., 1998; Kowalski et al., 2000; Bak et al., 2002). Conditions of observation of type B kinetics regime The enhancement of diffusion along oxide interfaces is generally recognized. However, experiments on nickel oxide have yielded conflicting results. Some authors (Chen and Peterson, 1980; Atkinson and Taylor, 1981, 1982, 1986) measured grain boundary diffusion, whereas others (Barbier et al., 1987; Barbier and Déchamps, 1988) did not observe this enhancement of diffusivity. As pointed out elsewhere for diffusion from an ‘instantaneous source’ (Moya, et al. 1990) and from a ‘constant source’ (Fielitz et al., 2003) experimental conditions allowing a clear observation of a B kinetics profile can be difficult to realize. In a given material, the only adjustable parameter is the bulk diffusion length Dt . The range of its appropriate values, depending on the grain size d, the ratio D′α/D and the sensitivity of the detection technique, may be very narrow. Schematic profiles derived from the ‘instantaneous source’ case are plotted in Fig. 11.6. This figure shows that, when D′α/D = 104, the grain boundary diffusion profile slope is high. With the same experimental conditions, but D′α/D = 105 or 106, the longer second part of the profile becomes easier to detect. Experiments carried out on polycrystalline nickel oxide illustrate this difference (Amalhay and Moya, 1991; Moya et al., 1991), as can be seen in Fig. 11.7. When the penetration of nickel via grain boundaries is about 30 µm, that of calcium extends over 120 µm. This difference can be attibuted to a segregation of calcium along the grain boundaries in nickel oxide, which increases the ratio D′α/D. This effect is developed in the following section. Influence of segregation The influence of segregation on grain boundary diffusion is a complex problem, which has been studied intensively for metals (Divinski, et al., 2001, Bernardini

296


Concentration (a.u.)

1

Bulk diffusion

10–1 (106) 10–2 (104)

Grain boundary diffusion

(105)

10–3 0

5

10

15 20 Depth (µm)

25

30

11.6 Schematic profiles for diffusion in a polycrystalline body from an ‘instantaneous source’. The concentrations are plotted in logarithmic scale as a function of depth. The first part corresponds to bulk diffusion. The linear parts for grain boundary diffusion are calculated using the relations given in Moya and Moya (1986) and Moya et al. (1988). The grain size is d = 10 µm and the bulk diffusion distance is Dt = 1 µm. Three cases are considered: D′α/D = 104, D′α/D = 105 and D′α/D = 106 (values indicated on the curves).

Activity (cnts/S)

104 103 102 (b)

10 (a) 1 0

50 100 Penetration depth (µm)

150

11.7 Penetration profiles of Ni and Ca in polycrystalline NiO (d = 40 µm), using the radiotracer technique. Empty circles: Ca/NiO, 1223 K, 1 hour: the following parameters have been obtained: Dt = 3 µm, D′/D = 8 × 104. Crosses: Ni/NiO, 1223 K, 140 hours: the following parameters have been obtained: Dt = 1.5 µm, D′α /D = 2 × 106.

et al., 2003). It is often argued that impurities of low solubility segregate along grain boundaries, where they are less mobile than the solvent atoms. Reactive elements, used to reduce the oxidation rates of metals, are generally those which segregate at grain boundaries. Although the mechanism of their action is not clearly understood, it is supposed that they ‘block’ cation diffusion as they bind strongly with vacancies. Computer simulation studies on nickel oxide (Harris et al., 2000) give evidence of the increase of the vacancy


297

migration energies when segregated impurities are present. Consequently, as α increases with segregation, D′/D decreases. The resulting effect on D′α/D will depend on the relative importance of these opposite variations. In our investigations on grain boundary diffusion in NiO and CoO, we observed that D′α/D values are higher for segregating impurities than for non-segregating ones and host cations. This appears in Table 11.1 where some data on grain boundary diffusion in CoO (Kowalski et al., 1996) are reported. Considering the last column of this table, we can see that the enhancement of diffusivity along the grain boundaries, characterized by D′α/ D, is almost the same for cobalt and nickel and is significantly larger for calcium and chromium. Indeed, nickel ions are very similar to cobalt ions, and no segregation is expected. In contrast, calcium ion, of larger size, exhibit the same diffusion rate as nickel and cobalt ions in the lattice. However, the parameter D′α/D is one order of magnitude higher. This can be explained by a strong interface segregation of calcium. The case of chromium is particularly interesting. In spite of its very low diffusion coefficient in bulk CoO, the parameter D′α/D is comparable to that of calcium. This can also be the result of a strong segregation, which reflects the observed segregation of chromium at surfaces in CoO (Nowotny, 1988a). Table 11.1 Diffusion of Co, Ni, Ca and Cr in bulk and along grain boundaries in CoO at 953°C. The data reported in Kowalski, Moya and Nowotny (1996) for a given temperature have been averaged. The value δ = 10–9 m has been used to calculate the D′α/D parameter Radiotracer

D(m2s–1)

60

3.5 1.5 2.2 1.2

Co Ni 45 Ca 51 Cr 63

× × × ×

10–14 10–14 10–14 10–17

D′α/D 2.2 3.1 1.5 1.2

× × × ×

105 105 106 106

Similar results have been obtained for grain boundary diffusion in NiO (Amalhay and Moya, 1991). They seem to indicate that the ‘segregation effect’ dominates the ‘retardating effect’. Dissolution in the type A kinetics regime Very few quantitative works have been reported for diffusion in the A regime. However, this model is of interest for many practical cases, where diffusion occurs in polycrystalline bodies (or in single crystals with a high dislocation density). When the bulk diffusion distance Dt is much larger than the distance d between two boundaries, a type A profile, similar to a diffusion profile in an homogeneous medium, is obtained. The kinetics are then described by an effective diffusion coefficient Deff expressed as:

298


Deff = (1 – f ) D + f D′

11.11

where f is the volume fraction of the grain boundary sites, given by Eq. 11.8. When the grain size decreases, the relative importance of the second term increases. If we suppose δ = 1 nm and D′/D = 104, the second term equals the first one when R ≈ 30 µm, and is 10 times higher when R ≈ 3 µm. It has been calculated that a pure type A kinetics regime requires Dt > 150 d (Kaur et al. 1995). However, apparent diffusion coefficients, higher than the real bulk ones, can be measured in polycrystalline samples, when the diffusion fields from adjacent boundaries overlap. This analysis has been used to explain the diffusion rate in oxidation scales (Huntz et al., 1997; Balmain et al., 1997). In recent years, considerable interest in nanocrystalline oxides has emerged due to their unique properties (Chadwick, 2003; Hahn, 2003). Among them, the transport properties should obviously be controlled by the high density of interfaces. In addition, the space charge layers along the boundaries can also strongly modify the diffusive properties. Practically, in these materials, a type B kinetics transport cannot be observed, since it would require Dt < d/20, which for d 600 °C in the very harsh environments of vehicle exhausts. Up to now, analytical instruments, such as those based on the Saltzman colour reactions, chemical luminescence, or infra-red (IR) absorption, have been developed and used to measure these pollutants. However, while these and other instruments provide high-precision measurements of desired gases, some cannot monitor precisely rapid changes in gas concentrations owing to the lengthy time required for data acquisition. This disadvantage, combined with their large sizes, high power consumptions and high cost, represent some of the serious problems facing real-time in-situ monitoring of pollutant gases and feedback control of combustion processes.4,5 On the other hand, in-situ, solid-state, electrochemical gas sensors offer the attractive features of rapid response, compactness and low cost. Consequently, if highly sensitive and reliable gas sensors are developed, then they can be used to control combustion processes on-line. Further, they may be used for the continuous 303

304


monitoring of atmospheric gaseous pollutants at fixed, e.g., industrial, and multiple, e.g., urban and regional, sites. During the last few years, an ultra-lean-burn (or direct-injection type) engine system has been developed in order to improve fuel efficiency and to reduce CO2 emissions from the engine. In this new engine system, as shown in Fig. 12.1, a newly developed NOx-storage catalyst must be used in order to compensate the low NOx-removal ability of conventional three-way catalysts under lean-burn (air-rich) conditions.6 Thus, it is vital to have high-performance NOx sensors installed behind (or both before and behind) the NOx-storage catalyst. The NOx sensor should be capable of operating in the exhaust gas at a temperature range of 600–700 °C so as to optimise the catalyst performance. The NOx concentration in the gas flow from the NOx-storage catalyst increases gradually with time owing to the saturation of NOx-storage capacity of the catalyst, as shown in Fig. 12.2. In order to regenerate the storage capacity, a fuel-rich gas containing a large amount of hydrocarbons is allowed to flow through the catalyst. Consequently, the NOx concentration in the gas emitted from the catalyst decreases rapidly to zero, followed by a gradual increase. Thus, the on-board NOx sensor must monitor the NOx concentration in the gas flow from the catalyst and the data it produces allow determination of the timing for regenerating the catalyst. NOx-storage catalyst Three way catalyst

Oxidation catalyst Oxygen sensor

Direct-injection type engine A/F sensor

NOx sensor NOx sensor

12.1 Catalytic converter system equipped with NOx sensors for the exhaust gas emitted from a new-type car engine (reprinted from Ref. 6 with permission from Elsevier Science).

There have been many types of solid-state gas sensor developed over the last two decades. 7–32 These include resistive-type sensors that use semiconducting oxides or metal phthalocyanines and chemical capacitortype sensors that use a mixture of oxides and sensors based on solid electrolytes. The former type of sensors operate on the basis of changes in materials properties that take place upon adsorption-desorption and/or surface reaction with the analysed gas. With such sensing mechanisms, the selectivity for a

NOx conc. after NOx storage catalyst

Solid-state electrochemical gas sensors for emission control

Time Lean

A/F

305

Lean HC

Lean

Generation

Rich

Rich Time

12.2 Regeneration pattern of NOx storage catalyst (reprinted from Ref. 6 with permission from Elsevier Science).

particular gas is not always adequate, particularly since the sensitivities of these sensors decrease sharply at the high temperatures where gas adsorption diminishes. As a result, these types of solid-state sensor usually are unable to detect gaseous pollutants at temperatures >600 °C. On the other hand, sensors based on solid electrolytes generally operate at very high temperatures and so they are sufficiently sensitive and selective for specific gases. These advantages derive from the nature of the sensing mechanism, in which the output signal is determined solely by the electrode processes or electrochemical equilibrium. Therefore, these types of the solid-electrolyte sensors have great promise for in-situ monitoring of gaseous pollutants in high-temperature combustion exhausts and in other environments.

12.2

Stabilised zirconia-based gas sensors for emission control

12.2.1 Potentiometric gas sensors Although chemical gas sensors based on solid electrolytes have been under development for the past two decades, they have received considerable attention recently owing to the introduction and commercialisation of in-situ λ-sensors based on an yttria-stabilised zirconia (YSZ) electrolyte for the detection of the equilibrium oxygen partial pressure in automotive exhausts. The signal of the λ-sensor is used to regulate the air/fuel ratio in a narrow range that approximates stoichiometric combustion, which is critical to the successful operation of the three-way-catalyst behind the λ-sensor. The air/fuel ratio must be maintained with a precision of 1–2% since carburettors have been replaced largely by fuel injection systems. The main attraction of solid-

306


electrolyte YSZ-based λ-sensors results from the thermodynamically controlled detection principle of these so-called ‘potentiometric devices’, where the equilibrium oxygen partial pressure in the exhaust gas is monitored relative to a known oxygen partial pressure of a reference system, typically air. Figure 12.3 shows the principle of such an electrochemical cell, which is based on a dense, oxygen-ion-conducting, closed tube of YSZ. This electrolyte is covered with an outer sensing electrode (SE) and inner reference electrode (RE), both made of platinum. The YSZ tube separates the sensing or measuring side from the reference side, where the oxygen partial pressure is known. At sufficiently high temperatures, the gaseous oxygen, the mobile oxygen ions in the zirconia, and the electrons of the electrodes are in thermodynamic equilibrium. At each electrode, the following equilibrium occurs: O2 + 4e– = 2O2–

12.1

SE (Pt)

Measuring gas)

Air

Es

RE (Pt)

12.3 Cross-sectional view of the tubular YSZ-based potentiometric sensor.

The electrochemical potential of the oxygen ions must be constant throughout the entire inter-electrode cross-section of the zirconia, particularly at the interfaces with both electrodes. Thus, YSZ-based electrochemical potentiometric cells can be described as follows: O2 [PO2 (gas)], Pt | YSZ (O2– mobile ions) | Pt, O2 [PO2 (reference)]

12.2

If the measuring and reference sides of the cell are exposed to different oxygen partial pressures, where PO2 (gas) PO2 (reference), then this induces different chemical potentials for the oxygen ions in zirconia at the interfaces with gas phases. Since the electrochemical potential remains constant, then the electrical potential must be different. Therefore, the output signal E (emf) of the electrochemical cell can be described according to Nernst’s law:


PO2 (gas) E = RT ln , 4F PO2 (reference)

307

12.3

where R = gas constant, T = absolute temperature, and F = Faraday constant. It is clear that knowledge of the output signal E and the oxygen partial pressure of the reference gas PO2 (reference) allows calculation of the unknown oxygen partial pressure on the measuring side PO2 (gas). This equation contains only thermodynamic quantities and does not require any information about the microstructure of the system. Hence, ageing effects on the microstructure of typical YSZ/Pt electrodes do not influence the sensor signal a priori and current λ-sensors have lifetimes of more than 160,000 km7. It should be noted that the surface chemistry of a λ-sensor under normal operating conditions is considerably more complicated than would be expected from the simplicity of eqn 12.3. This equation implies that oxygen alone is involved in the potential-forming electrode reaction. Consideration of eqn 12.1 and the composition of air and combustion gases leads to the possibility of a series of potential reactions at the SE: 2NO + 2O2– → 2NO2 + 4e– 2–

CO + O

→ CO2 + 2e

–

12.4 12.5

CH4 + 4O2– → CO2 + 2H2O + 8e–

12.6

H2 + O2– → H2O + 2e–

12.7

These reactions determine the apparent potential of the λ-sensor. Since the raw exhaust gas constitutes a non-equilibrium mixture, thermodynamic equilibrium must be achieved at the active SE surface of the λ-sensor before monitoring the potential. Consequently, λ-sensors contain catalytically active materials, which are operated at >600 °C. For less active materials or temperatures 700 °C. Since the lifetimes of probes are very important for practical and economic reasons, this sensor was subjected to extended-lifetime tests. Figure 12.6 shows results for one of these tests, of length 120 days, in which the SO2 concentration in the gas at 720 °C was changed every 20–30 days.18 The measured emf was within ±5 mV of the calculated values and stability was observed throughout the entire tenure of


311

600 10,000 ppm SO2

550

5720 ppm SO2

Sensor emf, mV

5720 ppm SO2 500

450 43 ppm SO2

98 ppm SO2

400 18 ppm SO2 350

300

BaSO4-K2SO4-SiO2 Temperature 720 °C 0

20

40

60 80 Test time, days

100

120

12.6 Long-term stability test of the emf response of the SOx measuring electrochemical cell of the dual SOx/O2 sensor (reprinted from Ref. 18 with permission from Elsevier Science).

the test. No phase transition was observed in the sulphate electrochemical cell when the SO2 concentration was changed from 18 ppm to 10,000 ppm during the test. After these long-term tests, the other probe components showed no observable chemical or mechanical degradation that would limit the probe’s lifetime. It appears that the dual, solid-electrolyte, potentiometric SOx/O 2 sensor demonstrates: (i) a high level of reliability, (ii) good SO2-sensing characteristics, (iii) selectivity over a long period of time, and (iv) long-term chemical stability. Another example of a potentiometric YSZ-based sensor using an auxiliary phase for gas determination is a NOx sensor with a Ba(NO3)2-CaCO3 SE.25,26 The design of this tubular sensor for NOx measurement at high temperatures is quite similar to the sensor shown in Fig. 12.3. However, instead of a Pt SE, mixtures of 60–90 mol% Ba(NO3)2 and 0–40 mol% CaCO3 were attached directly to the surface of YSZ and covered with Au meshes, which had attached Au leads for current conduction. This device detected NO x concentrations in the range 5–1000 ppm over the temperature range 400– 450 °C. Response transients upon exposure of the 90 mol% Ba(NO3)2 + 10 mol% CaCO3 by 500 ppm NO and 100 ppm NO2 in air at 450 °C were sharp and reasonably stable, with 90% response and 90% recovery times being~ 30 s and 60 s, respectively. However, the low melting point of 590 °C of Ba(NO3)2 restricted NOx measurements to 450 °C. It may be noted that the NO-sensing properties at 400 °C deviated significantly from Nernst-type behaviour, so this device is not suitable for general NO measurement. Figure

312


∆E/mV

250

200

150

100

0

10

20 30 40 CaCO3 content/mol.%

50

12.7 Sensitivity to 500 ppm NO in air for tubular device employing Ba(NO3)2-CaCO3 as a function of CaCO3 content (450 °C) (reprinted from Ref. 26 with permission from Elsevier Science).

12.7 shows the sensitivity to NO as a function of the CaCO3 content, which demonstrates that higher levels of CaCO3 substantially decreased the NO sensitivity of the sensor. In light of these results, it is concluded that, despite the ability of binary auxiliary systems to detect measuring gas concentration as low as the ppm level in air at relatively high temperatures, devices based on these auxiliary phases still appear to be unsatisfactory for combustion applications because the working temperatures are higher than acceptable to the auxiliary phases.26 This situation prompted the exploration of other possibilities for YSZ-based gas sensors using different oxide materials as SEs.

12.2.2 Mixed-potential gas sensors Many researchers33–36 have reported anomalous emf values at relatively low temperatures under oxidising or reducing gases when there are at least two simultaneous oxidation/reduction reactions that occur at the SE of a YSZbased oxygen electrochemical cell with Pt electrodes. This type of anomalous emf value represents a mixed potential at the SE and it is generated as a consequence of the coupling between electrochemical oxidation and reduction reactions.37–40 The sensing properties of these mixed-potential sensors can be improved substantially by replacing the Pt SE with a suitable semiconducting oxide electrode.32 In terms of practical implications, this replacement is favourable because the oxide SE is inexpensive and it can be used at relatively high operating temperatures, which makes it suitable for use as a practical NOx sensor in vehicle exhausts. The search for oxide electrode materials


313

during the late 1990s revealed that mixed-metal oxides exhibit better sensing characteristics for both NO and NO2 compared to single-metal oxides in the temperature range of 500–600 °C.29,30 This observation suggested the suitability of further research in this area since vehicle exhaust temperatures are in the range 650–700 °C. Consequently, this research led to the observation of large emf responses to both NO and NO2 by twelve spinel-type oxides, which are derived from trivalent (Co, Fe, Mn, and Cr) and divalent (Cu, Zn, and Cd) transition metals.40–45 The best results were obtained with ZnCr2O4 and ZnFe2O4, the latter of which is relatively stable and shows the highest sensitivity to NOx at 700 °C.40 It may be noted that this material is environmentally innocuous. Another factor important to the potential success of in-situ NO x measurements at high temperatures is the sensor design, which still requires improvement. Figure 12.8 shows a recent development in the design of a NOx sensor. This design incorporates: (i) a planar structure, (ii) an inner cavity, and (iii) an electrode for conversion of NO to NO2.46,47 Consequently, this sensor design is suitable for sensitivity to NO2 but not to NO at high temperatures. The combination of a superior sensor design and a highperformance oxide SE is likely represent the path to significantly improved NOx sensors in vehicle exhausts.

Exhaust gas diffusion path

NOx sensing electrode (Oxide)

NOx conversion electrode (Pt-Rh)

Air duct (O2-pumping) YSZ electrolyte

Inner cavity Pt heater

V

Reference electrode

12.8 Cross sectional view of planar YSZ-based NOx sensor (reprinted with permission from SAE paper number 2000-01-1203 ©2000 Society of Automotive Engineers, Inc.).

Figure 12.9 compares the emf responses obtained for exposure of tubular YSZ-based sensors (ZnFe2O4; NiCr2O4; ZnCr2O4, and CrMn2O4 SEs) to 200 ppm NO in air and to 200 ppm NO2 in air at 700 °C.44 In the carrier gas, which was dry synthetic air, the emf value was nearly nil, so the measured emf values can be considered to indicate the sensitivities to NO and NO2. Pure Pt was not sensitive to either NO or NO2 at this temperature, while all of the spinel-type oxides showed considerably superior sensitivities, with

314


ZnFe2O4

NiCr2O4 (T = 700 °C) ZnCr2O4

CrMn2O4

200 ppm NO –20

–10

200 ppm NO2

Pt 0

10

20 emf/mV

30

40

50

60

12.9 Comparison of the emf responses at 700 °C for the devices attached with each of various spinel-type oxide SEs (reprinted from Ref. 44 with permission from Elsevier Science).

ZnFe2O4 showing the greatest sensitivity to NO2 in the temperature range 600–700 °C. Although the NOx sensitivity of the ZnFe2O4 SE was relatively stable at 700 °C, other spinel-type oxides showed degradation of the sensitivity to NO and NO2 after one month at 700 °C. Interestingly, the emf value for the ZnFe2O4 SE upon exposure for one month to 100 ppm NO2 in dry synthetic air showed a gradual increase instead of degradation. Further, the change in emf value upon exposure for 3–8 months at 700 °C to 100 ppm NO2 in dry synthetic air varied no more than ~20% from the initial value and the emf of the base air remained nearly stable. The reason why the ZnFe2O4 SE provides the highest NO2 sensitivity in the temperature range 600–700 °C has been clarified by correlating the NO2 sensitivity with the various properties of the oxides tested, including gas adsorption-desorption behaviour, oxygen-sensing characteristics, and catalytic activity of the gas-phase reaction of NO2.45 Examination of the temperatureprogrammed-desorption (TPD) profiles of NO2 for various spinel-type oxides showed that: the amount of NO2 desorption from ZnFe2O4 was larger than those from the other oxides (NiCr2O4; ZnCr2O4, and CrMn2O4) and the desorption peak for ZnFe2O4 occurred at the highest temperature (~350 °C).44 These results suggest that NO2 can be adsorbed relatively strongly on the surface of ZnFe2O4. Interestingly, the amount of NO2 desorbed and the temperature of the NO2 desorption peak correlated roughly with the NO2 sensitivity at 700 °C. That is, the higher the amount of desorption and peak temperature for NO2 desorption, the higher the NO2 sensitivity. This suggests


315

that the NO2 gas adsorbed at the YSZ/SE interface promotes the rate of the cathodic reaction of NO2 at high temperatures: NO2 + 2e– → NO + O2–

12.8

The other spinel-type oxides did not demonstrate desorption of NO2 in the temperature range 500–800 °C. Investigation of the O2 desorption from these materials40,42,44 revealed that the oxygen adsorbed on the oxide SE plays a significant role in the sensing mechanism, which involves a mixed potential. In the case of ZnFe2O4 at ~700 °C, the amount of O2 desorbed and the temperature of the O2 desorption peak were higher than those for the other spinel-type oxides (NiCr2O4; ZnCr2O4, and CrMn2O4.45 This observation supports the suggestion of strong O2 adsorption on the ZnFe2O4 SE and so it can be speculated that its O2 adsorptiondesorption process will be less reversible, even at 700 °C, so catalytic activity for the electrochemical reaction involving O2 will less than those for the other oxides studied. The merit of this speculation was examined by investigation of the oxygensensing properties of the four spinel-type oxides. The emf values of the SEs were measured when the oxygen concentration in the gas mixture (N2 + O2) was increased from 1 to 100 vol% at 700 °C. Figure 12.10 shows the Nernstian plots for these SEs and a Pt paste electrode at 700 °C.40 All of the oxides except ZnFe2O4 yielded theoretical Nernstian plots for electron number n = 4.0 as did the Pt electrode. The slope, 24 mV/decade, of the plot for ZnFe2O4 was much lower than theoretical, 48 mV/decade. These results suggest that 40 (T = 700 °C) 20

emf/mV

0 –20 ZnFe2O4 ZnCr2O4 NiCr2O4 CrMn2O4 Pt

–40 –60 –80 1

10 Oxygen concentration/vol.%

100

12.10 Dependence of emf on the logarithm of O2 concentration for the YSZ sensor using spinel-type oxide SE (reprinted from Ref. 44 with permission from Elsevier Science).

316


the ZnFe2O4-SE acts as an irreversible oxygen electrode, where the catalytic activity for the electrochemical reaction of O2 is low: O2– → 1/2 O2 + 2 e–

12.9

Figure 12.11 shows the NO2 conversions for the four spinel-type oxides exposed to a gas mixture of 100 ppm NO2 + 21 vol% O2 + and He balance.44 This reaction is not an electrochemical gas-phase reaction: NO2 → NO + 1/2 O2

12.10

The NO2 conversion on the ZnFe2O4 SE is relatively low in the temperature range 500–700 °C compared to the other three oxides. In fact, the catalytic activities at 550 °C of the three oxides correlate roughly with the NO2 sensitivity at 700 °C. Here, the lower the catalytic activity of the oxide, the higher the NO2 sensitivity of the sensor. 100 90

NO2 conversion/%

80 70 60

Equilibrium ZnCr2O4 CrMn2O4 NiCr2O4 ZnFe2O4 Blank

50 40 30 20 10 0 200

300

400 500 Temperature/°C

600

700

12.11 Temperature dependence of NO2 conversion to NO on the various oxides tested (reprinted from Ref. 44 with permission from Elsevier Science).

Since NO dominates the equilibrium gas mixture at temperatures >500 °C, the conversion of NO2 to NO usually is high when these catalysts are used. If the catalytic activity of the SE is reasonably high at high temperatures, most of the NO2 can be converted easily to NO according to the gas-phase reaction given in eqn 12.10. Since this reaction takes place readily on the surface or in the bulk of the SE layer, then it is essentially impossible for NO2 to reach to the YSZ/SE interface. Consequently, the NO2


317

sensitivity is low for such a sensor. Conversely, if the catalytic activity of the SE is relatively low at high temperatures and the SE is relatively permeable, then NO2 can diffuse through the SE layer and reach the YSZ/SE interface, resulting in a high NO2 sensitivity. Figure 12.12 schematically shows the influence of catalytic activity, for both cathodic (1) and anodic (2) reactions, on the polarisation curves and the mixed potential area.44 According to the mixed-potential theory, the reactions (eqns 12.1 and 12.4) take place at or very near the YSZ/SE/gas three-phase boundary (TPB). The mixed potential (Em) can be given by the intersection of the cathodic and anodic polarisation curves where the anodic current is equal to the absolute value of the cathodic current. At this potential, electrochemical reactions (1) and (2) proceed simultaneously at equal rates. If the catalytic activity for cathodic reaction (1) of NO2 is high, the cathodic polarisation curve for NO2 shifts upward. This brings about a change in the mixed potential in the direction of positive potential (Em1), which yields an increase (∆E1) in the NO2 sensitivity. On the other hand, if the catalytic activity for anodic reaction (2) of O2 is low, the anodic polarisation curve for O2 shifts downward. In this case, the mixed potential also changes in the direction of positive potential (Em2), which again yields an increase (∆E2) in the NO2 sensitivity.

Cathodic reaction (1) for NO2

Current

Anodic reaction (2) for O2

∆E2 ∆E2 Em Em2 Em1 Potential

12.12 Schematic polarization curves for the cathodic reaction (1) of NO2 and the anodic reaction (2) of oxygen (reprinted from Ref. 44 with permission from Elsevier Science).

Another example of a potentiometric mixed-potential-type NOx sensor based on YSZ is shown in Fig. 12.13.29 The cross-sectional view of this sensor shows that it has three electrodes. The RE and counter electrode (CE) consisted of Pt paste and were attached to the inner and outer surfaces,

318

Materials for energy conversion devices SE (Oxide)

CE (Pt) VS-C

Air

NO or NO2

Es

RE (Pt)

12.13 Cross-sectional view of the potentiometric NOx sensor based on YSZ tube (reprinted from Ref. 29 with permission from Elsevier Science).

respectively, of the YSZ tube. The SE consisted of NiCr2O4 and it was attached to the outer surface of the YSZ tube in the vicinity of the CE. The emf output characteristics of this NOx sensor are shown in Fig. 12.14. The semilog plots of the emf values as a function of NO and NO2 concentrations were essentially linear at all temperatures. The 90% response times to NO and NO2 at concentrations of 200 ppm each in air at 550 °C were ~3 min and ~2 min, respectively. Since practical applications of sensors require operation in humid atmospheres, the NiCr2O4 SE was tested in air containing 2 vol% 120

550 °C

100 80

NO2

emf/mV

60

600 °C 650 °C

40 20 0 650 °C

–20 –40

NO

–60 –80 10

600 °C 550 °C

100 NOx concentration/ppm

1000

12.14 Dependence of emf on the logarithm of NO or NO2 concentration for the YSZ-based device using the NiCr2O4 sensing electrode (reprinted from Ref. 29 with permission from Elsevier Science).


319

H2O vapour at 550 °C. The influence of water vapour on NO2 detection was negligible. Although, in the case of NO sensing, the emf decreased slightly after the introduction of water vapour into the sample gas, the emf gradually recovered and reached its initial level after 5 h, after which it remained stable. It can be noted that the NO and NO2 reactions can be enhanced or suppressed, respectively, if the SE potential is polarised anodically (positively) or if the reverse effect occurs with cathodic polarisation.32 This indicates that the selectivity to NO or NO2 can be controlled by polarising the SE anodically or cathodically. Such a sensor may be considered to be a certain type of amperometric sensor, where a positive or negative potential is applied to the SE in order to provide exclusive sensitivity to the measuring gas. Since the output signal in this case is still emf, then the three-electrode sensor discussed previously may be considered as a type of potentiometric NOx sensor. In order to attain dominant selectivity to the measured gas, a tubular device using an NiCr2O4 SE was biased by a voltage (Vs–c) relative to the Pt CE and then the SE potential (Es) relative to the Pt RE was measured upon exposure to various atmospheres. The shift in Es from the value in base air to those in air containing NO or NO2 (the sensitivity Vs) was dependent on the Vs–c. Figure 12.15 shows the influence of Vs–c on the Es during exposure to 200 ppm NO in air, 200 ppm NO2 in air, and base air at 550 °C. Analysis of this figure suggests that, if a suitable polarisation is selected, it is possible to identify the sensitivity increment due to NO while suppressing that of 120

80

T = 550 °C 200 ppm NO2

Es /mV

40

0 Air 200 ppm NO –40

–80 –250 –200 –150 –100

–50 0 Vs–c /mV

50

100

150

200

12.15 Es versus Vs–sc correlation for the device using the threeelectrode structure in air, 200 ppm NO2 and 200 ppm NO, at 550 °C (reprinted from Ref. 29 with permission from Elsevier Science).

320


NO2 and the converse. This allows the selective detection of NO (NO2) over NO2 (NO) using this sensor design. Thermodynamically, NO dominates the NO + NO2 mixture at high temperatures. For example, the equilibrium composition at 600 °C is 90 vol% NO + 10 vol% NO2.48 Therefore, this raises the issue of the feasibility of the selective detection of NO or total NOx if the working temperature is >600 °C. Figure 12.16 examines this issue for a bias under anodic polarisation of approximately +175 mV. Under these conditions, the NiCr2O4 SE was found to be selective to NO over NO2. The semilog plot of the Vs as a function of NO or NO2 concentration is essentially linear for both gases. This linearity suggests that the lower limit of detection limit is ~14–15 ppm NO.29 Thus, it appears that the change in the bias voltage in the three-electrode (oxide SE, Pt CE, and Pt RE) tubular sensor could yield significant improvements in the NO selectivity. Unfortunately, this sensor appears to be insufficiently sensitive to NOx at temperatures >650 °C. Typical characteristics of mixed-potential, YSZ-based, gas sensors using oxide SEs are summarised in Table 12.2. 20 10

NO2

Vs /mV

0 –10 NO –20 –30 –40 10

100 NO or NO2 concentration/ppm

1000

12.16 Dependence of Vs on NO or NO2 concentration for the device using three-electrode structure at 550 °C under the bias of +175 mV (reprinted from Ref. 29 with permission from Elsevier Science).

These results40–45 allow the conclusions that: the sensing mechanism of NOx sensors based on the mixed-potential model when NOx and O2 are presenting simultaneously is complex and the NO2 sensitivity can be determined indirectly by the following factors: • NO2 adsorption-desorption behaviour on the SE. If NO2 shows strong adsorption on the SE, then this may lead to high catalytic activity of the SE for the cathodic reaction (eqn 12.8). • Oxygen adsorption-desorption behaviour on the SE and oxygen sensing


321

Table 12.2 Typical examples of characteristics of the mixed-potential type YSZ-based devices using oxide SE reported by authors Gas

Oxide sensing electrode

Operating temperature (°C)

Measuring concentrations ppm

Year of publication

Reference numbers

CO

CdO

600

20 ~ 4000

1997

12

NOx

CdMn2O4 CdCr2O4 WO3 NiCr2O4 ZnCr2O4 ZnFe2O4

500–600 500–600 500–700 550–650 550–650 550–700

5 ~ 4000 20 ~ 600 5 ~ 200 15 ~ 500 20 ~ 500 20 ~ 500

1996 1997 2000 2001 2001 2002

38 48 39 29 40, 43 41–45

H2

ZnO

400–600

50 ~ 500

1996

24

H 2S

Au - WO3

400

0.6 ~ 50

1998

23

C 3H 6

CdO

600

30 ~ 800

2000

37

performance of the SE. If O2 adsorption on the SE also is strong, the SE behaves as an irreversible oxygen electrode, which means that the catalytic activity of the SE for the anodic reaction (eqn 12.9) is low. • Catalytic activity of the SE for the non-electrochemical gas-phase decomposition reaction (eqn 12.10) of NO2. If the SE has low catalytic activity, this may result in higher NO2 sensitivity of the SE at high temperatures. All of these factors are interlinked with each other in a complicated manner. Investigation of mixed-potential-type NOx sensors using a ZnFe2O4 SE were observed to yield the highest sensitivity to both NO and NO2 in the temperature range 650–700 °C. This sensor was relatively stable even at 700 °C. Thus, although the response rate for the sensor using ZnFe2O4 must be improved,40 it still remains one of the best candidates for the SE for practical hightemperature NOx sensors. An indirect effect, which tends to be neglected in mixed-potential theory, that affects the sensor output signal is the possibility of direct gas-phase reaction between NO and O2. In this case, the surface of the SE can act as a catalyst for such a reaction without contributing to the sensor output emf. Further, the nature of the YSZ/SE/gas TPB and the structure of the SE depend upon the processing methods and heating procedures. Since the fabrication method can have a significant impact on the sensing performance, then it is necessary to obtain a better understanding of the sensing mechanism and the effect that variables, such as the electrode configuration and processing conditions, have on this mechanism.

322


12.2.3 Amperometric gas sensors Amperometric solid-state sensors have been used to detect a wide range of electroactive gases. This type of gas sensor consists of a working electrode (WE) or SE, RE, and an optional CE placed in contact with electrolytic medium into which the measuring gas diffuses. In contrast to potentiometric sensors, amperometric sensors are operated using an externally applied voltage to the WE. The applied voltage induces an electrochemical reaction of the gas, which generates a current proportional in magnitude to the gas concentration. Selectivity to different gases is achieved by using different metals or oxides as the WE and by operating the sensor at different applied potentials. The ability of the WE to provide selectivity to a gas is a result of the different catalytic activities of various electrode materials. For example, CO can be oxidised on a Pt surface but not on Au and some hydrocarbons can be reduced on Ag7. A second method of providing selectivity is to apply different potentials to the WE. In general, an electrochemical reaction will occur only beyond the characteristic equilibrium potential of the reaction. Therefore, a sensor operating at the lowest possible potential will reduce interference from other co-existing gases. Further, the electrode may be biased anodically or cathodically in order to oxidise or reduce gas, respectively.29 Using a specific combination of the WE composition and the sensor operating potential, selectivity to a gas can be obtained with a relatively simple sensor design.5 Most amperometric sensors operate in the diffusion-limited mode. Consequently, each molecule passing through the diffusion barrier reacts at the electrode without delay. A typical example of an amperometric sensor with a channel-type diffusion barrier is shown in Fig. 12.177. The corresponding limiting current is a unique function of the geometric parameters of the diffusion barrier. For a diffusion channel with length L and cross section A, the limiting current Ilimit is given by49

I limit =

4 FDO2 Ptotal A ln(1 – xO2) RTL

12.11

Gas

Diffusion barrier I ZrO2

O2–

U = const

12.17 Experimental set-up of a typical amperometric oxygen sensor with a channel-type diffusion barrier (reprinted from Ref. 7 with permission from Elsevier Science).


323

where DO2 = oxygen diffusion coefficient, Ptotal = total gas pressure, and xO2 = molar fraction of oxygen in the gas. At relatively low O2 concentrations, e.g., below 10 vol%, there is a linear correlation between the Ilimit and oxygen partial pressure PO2 according to: I limit = – 4 FA DO2 PO2 RTL

12.12

Due to this linearity, amperometric O2 sensors are also suitable for operation under conditions of excess oxygen, as in the case of the wide range of oxygen sensors used in lean-burn vehicle engines.50–52 Sensing combustible gases, such as CO, CxHy, and H2, with amperometric sensors requires a priori the presence of oxygen. The operating principle relies on the determination of the remaining oxygen in a restricted volume following combustion of reducing gases on a Pt electrode >700 °C.53 The Ilimit depends linearly on the concentration of the combustible gas. As with oxygen sensing, a double-sensor may be used to achieve superior accuracy. However, such measurements require a known oxygen partial pressure. A recent trend in amperometric YSZ-based sensors is the simultaneous measurement of O2, NOx, and combustibles using a tubular sensor with two WEs.54 However, it is necessary to optimise all of the parameters, including cell design, material comprising the WE, and operational conditions, including temperature and potential at each electrode. A single open-end YSZ tube with an inner RE and outer WE, as shown in Fig. 12.18, can be used as an amperometric gas sensor.55 This sensor has essentially the same design as the YSZ-based potentiometric-type sensors described previously and shown in Fig. 12.3. However, a diametral outer WE consisting of a CdCr2O4 overlayer and/or a Pt underlayer surrounds the YSZ tube. CdCr2O4 layer A

Pt layer

WE

Potentiostat

Pt paste Pt mesh CE (RE)

Air

Pt wire

Sample gas

YSZ WE

12.18 Schematic view of YSZ-based amperometric-type NOx sensor (reprinted from Ref. 55 with permission from Elsevier Science).

The mode of operation for this type of amperometric sensor is different from that previously described. Here, the WE is polarised at a selected

324


potential relative to the RE by means of a potentiostat and the electric current flowing through the YSZ is provides the sensing signal. A YSZ sensor with Pt-only WE exhibited a typical polarisation curve in air, where both anodic and cathodic currents increased almost exponentially on sweeping the potential positively and negatively, respectively. Upon exposure to NO and to NO2, the polarisation curves shifted slightly upwards or downwards in the appropriate positive or negative range. For a YSZ sensor with oxide-only WE, the polarisation curve in air had a shape similar to that obtained for the Pt-only sensor. However, the shifts upon exposure to NO and to NO2 tended to be larger than those of the oxide-only sensor.55 This suggests that the oxide layer plays an important role in promoting the electrochemical reactions of NO and NO2 at the SE. The differences between the two devices were significant in the positive-potential range 50–200 mV. In this range, the oxide-only sensor showed a larger anodic current increment due to NO (the NO current) than that due to NO2 (the NO2 current). In the case of the Ptonly sensor, there was almost no observable trend. On the other hand, NO2 increased the cathodic current more than did NO for both sensors in the negative potential range. This suggests that selective detection of NO and NO2 is possible with the oxide-only sensor if the SE is polarised properly. For example, for selective NO determination, the SE should be set at ~100 mV, which gives the largest ratio of NO current to NO2. Figure 12.19 shows response transients upon exposure to NO and NO2 for the CdCr2O4-only sensor under polarisation of +100 mV at 500 °C. The transients are fairly sharp and stable, viz., the 90% response and 90% recovery times upon exposure to 100 ppm NO were ~20 s and 30 s, respectively. Although the responses to NO and to NO2 are of opposite magnitudes, it is (+ 100 mV) 200 ppm NO

1 µA

100 ppm NO

4 min

100 ppm NO2

4 min

200 ppm NO2

12.19 Response transients to NO and NO2 for the sensor attached with CdCr2O4 under polarization of +100 mV at 500 °C (reprinted from Ref. 48 with permission from Elsevier Science).


325

clear that the current response to NO is considerably larger than that to NO2 for both concentrations. The current response to NO was nearly linear relative to NO concentration in the range 0–200 ppm. Since the cross-sensitivity is a critical characteristic of a sensor, for the oxide-only sensor, the crosssensitivities to 2000 ppm CO2, 200 ppm CO, 200 ppm CH4, 200 ppm H2, and 713 Pa H2O vapour, all in dry synthetic air, were examined. The sensitivities to all these gases were significantly less than that to 200 ppm NO.48 With such unique sensing properties, YSZ-based sensors with oxide-only SEs have considerable promise for the monitoring of NO in combustion exhausts. These results also suggest that this type of sensor can operate reliably if the appropriate applied polarisation can suppress the cross-sensitivities to other gaseous components. This type of amperometric sensor also may be considered to act as a type of oxygen pump.48 Under weakly polarised condition (+100 mV), oxygen tends to be pumped from the CE side to the WE side because the anodic and cathodic reactions of oxygen are favoured at the respective electrodes. The pumping rate in air under this condition is likely to be determined by the anodic process at the WE. Upon exposure to NO in air, the total rate of the anodic processes and, consequently, the pumping current can be increased via another anodic reaction involving O2–: NO + O2– → NO2 + 2e–

12.13

It is probable that, in the case of CdCr2O4 deposited on porous Pt, some of the former comes into contact with the surface of the YSZ, thereby establishing the TPB at which the anodic reaction, given in eqn 12.13, takes place. It also may be noted that the electrochemical conversion of NO to NO2 on the Pt surface takes place in the presence of oxygen at high temperatures. The chemical conversion of NO to NO2, if it occurs over the WE, is the factor that restricts the NO response. Therefore, it is necessary to identify a suitable underlayer material having a lower catalytic activity than Pt in order to reduce or eliminate the chemical conversion of NO to NO2. A practical application for an amperometric YSZ-based sensor is a thickfilm ZrO2 sensor for the measurement of NOx at low concentrations.56 Investigation of the sensing performance of such a sensor located behind a three-way catalyst in the exhaust pipe of a 2.0 l petrol engine revealed that this sensor was capable of measuring NOx concentrations in the range 10– 500 ppm with a precision of ±13 ppm at 100 ppm of NOx. The deviation from accuracy resulted partly from the constantly changing oxygen concentration in the exhaust gas and partly from the working temperature variation (180°–730° C) of the sensor. However, the level of measuring accuracy approximated the requirement for a NOx sensor located immediately behind the catalyst.56 Figure 12.20 shows the dependence of the sensor output signal as a function of NOx concentration for different oxygen

326

Materials for energy conversion devices 2.5 : 1 ~ 4% – O2 : 7 ~ 10% – O2 : 12 ~ 15% – O2

Pumping current, lp2(µA)

2.0

1.5

1.0

0.5

0

0

100

200 300 400 NOx concentration (ppm)

500

12.20 NOx sensor output on diesel engine classified by oxygen concentration (reprinted with permission from SAE paper number 1998-980170 © 1998 Society of Automotive Engineers, Inc.).

concentrations in the engine exhaust. These data make clear that the sensor output signal requires offsetting owing to the residual O2 concentration. In order to minimise this offset dependence, greater control of the oxygen concentration in the first internal cavity must be achieved. Another trend in the development of amperomectric YSZ-based gas sensors is the simultaneous measurement of oxygen, oxygen-containing, and combustible gases, including CO, CxHy, NOx, and H2. Reaching this goal can be achieved through the use of multi-electrode amperometric sensors.7,54 Their main applications for such sensors are the determination of exhaust gases, especially vehicular exhausts, and monitoring of environmentally important pollutants. Multi-electrode amperometric cells contain several electrodes that are separated from the gas phase by a mutual diffusion barrier. The geometric and operational parameters must be adjusted in such a way that only one electrode reaction actually takes place at each electrode. When the sequence of subsequent electrode reactions occurs, the gas phase composition should be changed in a controlled way from one electrode to the next. In this way, good selectivity of the individual electrodes may be achieved.

12.2.4 Impedance-based gas sensors Many potentiometric and amperometric YSZ-based NOx sensors capable of


327

operating at high temperatures have been reported and they have been described previously. Mixed-potential potentiometric sensors appear to be advantageous for on-board NOx sensors owing to their high sensitivities, especially in the lower concentration range of 3 × 10–3 K–1 and ∆T > 300 K. This means that ZT = 1.8 at 600 K is necessary. Thirdly, COP of a commercial refrigerator is 1.2–1.3, which corresponds to ZT = 3–4. Thus much improvement in ZT is needed to replace a freon-gas refrigerator.

13.3

Microscopic theory of thermoelectric phenomena

13.3.1 The Boltzmann theory One-electron states in a periodic potential are exactly solved, and the solution is known as the Bloch function. The Bloch function has a wave number k

Introduction to thermoelectricity

345

30

Efficiency (%)

Z 1 2 3 4

20

× × × ×

10–3 10–3 10–3 10–3

TL = 300 K TH = ∆T + TL

K–1 K–1 K–1 K–1

10

0 0

100

200

300 400 ∆T (K)

500

600

13.3 Energy conversion efficiency plotted as a function of ∆T for various ZTs.

(crystal momentum) as a well-defined quantum number, and its energy ε = ε(k) is written as a function of k (band dispersion relation). To recover a particle picture, we make a wave packet from the Bloch functions. Then the velocity of the particle is equal to the group velocity of the wave packet given by:   v k = 1 ∇ k ε ( k ) = 1  ∂ε , ∂ε , ∂ε  . ∂ k ∂ k ∂ k h h x y z 

13.24

To keep the particle picture, every wave constituting the wave packet should satisfy the relation of h k = mvk with a constant value of m. Then we get m∆vk = h ∆k, and the effective mass (more precisely, the inverse of the effective mass tensor) in a solid is defined by: 1 = 1 ∂v ki = 1 ∂ 2 ε . m ij h ∂k j h 2 ∂k i ∂k j

13.25

Thus the electron in a solid behaves like a charged particle with the charge e, the mass m and the velocity vk. Since electrons are fermions, they obey the Fermi–Dirac distribution f0. Then the electric current density and the thermal current density are written as:

j=

1 4π 3

q=

1 4π 3

∫ ev f d k ∫ (ε ( k ) – µ ) v k

k

3

13.26

k

fk d 3 k

13.27

where fk is the distribution function at an inequilibrium state. fk is given as a solution of the Boltzmann equation written as:

346


∂f v k ⋅ ∇ fk + e E ⋅ ∇ k fk = k ∂t h

,

13.28

scattering

where the right-hand side is the scattering term. In the case of weak perturbation, we can linearize fk as fk = f0 + gk. We further assume the relaxation-time approximation to introduce the relaxation time τ as: ∂f k ∂t

= – 1 gk τ

scattering

13.29

Eventually we find:

ε (k) – µ  ∂f  gk =  – 0  v τ e E + (– ∇ T )  T  ∂ε  ε ( k )= ε k  

13.30

Substituting this into Eqs (13.26) and (13.27), we obtain: j = e2K0E + e K1(–∇T) T

13.31

q = eK1E + 1 K2(–∇T) T where Kn is:

13.32

Kn =

1 4π 3

 ∂f 0 

∫  – ∂ε 

vkvkτ(ε(k) – µ)nd3k.

13.33

ε ( k )= ε

Note that Kn is a second-rank tensor through vkvk, in general. Also note that Eqs (13.31) and (13.32) are identical to Eqs (13.7) and (13.8), and Onsagar’s relation given in Eq (13.3) is readily satisfied. It is reduced to a scalar in the cubic symmetry, and the conductivity and the thermopower are given by:

σ = e2 K0 =

1 4π 3

K S= 1 1 = 1 eT K 0 eT

 ∂f 0 

∫  – ∂ε  ∫

ε ( k )= ε

ν k2 τ d 3 k .

 ∂f 0  ν 2τ (ε ( k ) – µ ) d 3 k –   ∂ε  ε ( k )= ε k

∫

 ∂f 0  ν 2τ d 3 k –   ∂ε  ε ( k )= ε k

13.34

13.35

The thermal conductivity is given as κ′ = K2/T for E = 0, but the electron thermal conductivity is always measured for j = 0. Thus, by substituting E = S∇T (from Eq. (13.9)) into Eq. (13.8), we get: q = S2σ T ∇T + κ ′(–∇T) = κ ′(1 – S2σ /κ ′)(–∇T), and the thermal conductivity observed in real situations κ is:

13.36


S 2σ T   . κ = κ ′ 1 – κ ′  

347

13.37

The second term corresponds to ZT for j ≠ 0. This is usually large in thermoelectric materials, and effectively reduces the real thermal conductivity given in Eq. (13.37).

13.3.2 Asymptotic forms of thermopower Let us discuss thermopower of a metal intuitively. Consider a metal rod subject to a temperature gradient, as shown in Fig. 13.4. Suppose the temperature at one side is T1, and the temperature at the other side is T2 (T1 > T2). Since the average electron velocity is larger at T1, electrons begin to diffuse from the side at T1 to the side at T2. Owing to the charge neutrality, the side at T1 is positively charged, whereas the side at T2 is negatively charged. This implies that the metal rod behaves like a capacitor in the temperature gradient, which is the origin of the thermoelectric voltage Vth. In a steady state:

µ(T1) + eVth(T1) = µ(T2) + eVth(T2)

13.38

is realized, where µ (T ) is the chemical potential at temperature T. In the limit of T1→ T2, the thermopower S (= dVth/dT) reduces to: ∂µ S=1 e ∂T

13.39

This equation means that the thermopower is the entropy per carrier, being compared with Eq. (13.9). Hot T1

Cold T2

13.4 Metal in the temperature gradient.

Equation (13.39) is based on a semi-classical picture, where the electrons can move ‘smoothly’ from edge to edge like a classical particle. A complementary picture is seen in the high-temperature limit, where the transfer energy is much smaller than the thermal energy. From Eq. (13.35), the thermopower is rewritten as:

 ∂f 0 

e ε ν τ –  ∫  ∂ε  1 S= eT  ∂f  e ν τ – ∫  ∂ε  2

2 k

κ

2

2 k

d 3k ε =ε ( k )

0

ε =ε ( k )

d 3k

–

µ eT

13.40

348


The first term of the right-hand side of Eq. (13.40) is of the order of 〈εk〉/eT, and goes to zero as T → ∞. On the contrary, the second term is rewritten with the entropy s as an identity of thermodynamics: –

µ  ∂s  = T  ∂N  E , V

13.41

Thus the thermopower is associated with the entropy per carrier, which is called the Heikes formula written as: S= –

k B ∂ log g e ∂N

13.42

where g is the total number of configurations (Chaikin and Beni, 1976). The third example of the asymptotic expressions is the Mott formula, which is perhaps most frequently used for the thermopower in metals. Equation (13.35) can be associated with Eq. (13.34), when Fermi energy EF is much higher than the thermal energy kBT. By expanding the Fermi–Dirac distribution function in series of kBT/EF, we can show that 2  ∂K  . K1 = π k B2 T 2  0  3  ∂E  E = µ

13.43

On the same assumptions, we may associate the thermopower with σ : 2 k 2 T  ∂ log σ ( E )  B . S=π  3 e  ∂E E=µ

13.44

This is known as the Mott formula. Note that the conductivity-like function σ (E) in Eq. (13.44) is a fictitious conductivity that a metal would show, if its Fermi energy were equal to E. Do not forget that σ (E) cannot be observed in real experiments. Thus the Mott formula should be very carefully applied to analyses of real experiments (Ashcroft and Mermin, 1976).

13.4

Thermoelectric materials

13.4.1 Conventional materials Thermoelectric materials so far used for practical applications are Bi2Te3, PbTe, and Si1–xGex. N-type BiSb is superior at low temperatures, but has no p-type counterpart. Figure 13.5 shows ZT for various thermoelectric materials. Bi2Te3 shows the highest performance near room temperature, and cooling applications such as Peltier coolers are commercially available. PbTe shows the highest performance near 500–600 K, and Si1–xGex is superior above 1000 K. The conventional thermoelectric materials are degenerate semiconductors


349

(a) 1 PbTe

Bi2Te3

ZT

Si1–XGeX 0.5

P-type 0

(b)

1

Si1–XGeX

BiSb PbTe

ZT

Bi2Te3

0.5

0

N-type 0

500 T (K)

1000

13.5 The dimensionless figure of merit ZT for various thermoelectric materials.

of high mobility. Figure 13.6 shows a schematic figure of the conductivity σ, the thermopower S, the thermal conductivity κ and the power factor S2σ as a function of carrier concentration n (Mahan, 1998). Here a simple parabolic band is assumed, and the electron–electron and electron–phonon interactions are neglected. As can be seen in this figure, the thermopower decreases with n, whereas the conductivity increases with n. Then S2σ takes a maximum at an optimal carrier concentration n0, below which the conductivity is too low, and above which the thermopower is too small. Assuming the Boltzmann distribution instead of the Fermi–Dirac distribution, one can evaluate that the optimum concentration is around 1019–1020 cm–3, which is close to n of degenerate semiconductors. Since the conductivity is expressed as σ = neµ, the only way to maximize σ for n = n0 is to maximize the mobility µ. As shown in Fig. 13.6, κ consists of the lattice part κlattice and the electron part κel. Near n = n0, the former part is dominant, and to maximize the figure of merit Z is to minimize κlattice keeping S2σ intact. In the lowest order approximation, κlattice is expressed by (Ashcroft and Mermin, 1976):

κ lattice = 1 C L ν s λ ph , 3

13.45

where CL the lattice specific heat, νs the sound velocity, and λph is the phonon mean free path. Then, a material containing heavy elements (giving

1/σ

S 2/σ

S

Conductivity 1/σ


Thermopower S

350

Thermal conductivity

log n

κel = LT/σ κlattice log n Insulator

Semiconductor

Metal

13.6 Thermoelectric parameters as a function of temperature.

small νs), solid solutions (giving short λph), and many atoms in a unit cell (giving small CL) can be a good candidate. Mahan (1989) has suggested a microscopic parameter for good thermoelectric materials called ‘the B factor’ defined by:  2 mk B T  B=    πh 2 

3/2

µ

κ lattice

∝ m 3/2

µ

13.46

κ lattice

Note that m, µ and κlattice are independent parameters, whereas S, ρ and κ are not. Accordingly, a degenerate semiconductor with heavier effective mass, higher mobility and lower lattice thermal conductivity is extensively searched. Table 13.1 lists the thermoelectric parameters of the conventional thermoelectric materials (Mahan, 1998). The terms σ, S and κ are around 1–2 m Ω cm, 150– 200 µV/K, and 15–25 mW/cmK, respectively. The B factor is around 0.3– 0.4, which is significantly larger than that of other materials.

Table 13.1 Thermoelectric parameters of conventional thermoelectric materials

Bi2Te3 PbTe Si1–xGex

Temperature for maximum ZT (K)

Effective mass

Mobility (m2/Vs)

Lattice thermal conductivity (W/mK)

ZT

300 650 1100

0.2 0.05 1.06

0.12 0.17 0.01

1.5 1.8 4.0

1.3 1.1 1.3


351

13.4.2 Filled skutterudite compound Since the discovery of Bi2Te3 in the mid 1950s, thermoelectric materials were extensively searched in binary systems. In fact, many promising materials were found through the research, but ZT did not exceed unity. Filled skutterudite CexFe3CoSb12 is the first unambiguous example whose ZT exceeds unity, and is going to be used for the next generation of thermoelectric power generation (Sales et al., 1997). Figure 13.7(a) shows the crystal structure of the skutterudite CoSb3. The unit cell of cubic symmetry consists of the eight subcells whose corners are occupied by Co atoms. Six subcells out of the eight are filled with Sb plackets, forming the valence band. According to the band calculation, CoSb3 is a narrow gap semiconductor with an indirect gap of 0.5 eV, which is favourable for a thermoelectric material. In fact, the hole mobility of CoSb3 exceeds 2000 cm2/Vs at 300 K, which is much higher than that for Bi2Te3 (Caillat, 1996). Co, Fe

Co

Ce

Sb

Sb

(a)

(b)

13.7 Crystal structures of (a) the skutterudite and (b) the filled skutterudite.

Figure 13.7(b) shows the crystal structure of the filled skutterudite CeFe3CoSb12. In the two vacant subcells of the skutterudite, two Ce ions are filled. In order to compensate the charge valance, six Fe atoms are substituted for the eight Co sites, because Ce usually exists as trivalent. The most remarkable feature of this compound is that ‘filled’ Ce ions reduce the lattice thermal conductivity several times lower than that for an unfilled skutterudite CoSb3. Ce ions are weakly bound in an oversized atomic cage so that they will vibrate independently from the other atoms to cause large local vibrations. This vibration and the atom in the cage are named ‘rattling’ and ‘rattler’, respectively. As a result, the phonon mean free path can be as short as the lattice parameters. In particular this compound has a poor thermal conduction like a glass and a good electric conduction like a crystal, called ‘an electron crystal and a phonon glass’ by Slack (1995).

352


Figure 13.8 shows how the rattlers reduce the lattice thermal conductivity (Sales et al., 1997). The κ of CoSb3 is one order of magnitude higher than the κ of Bi2Te3, which means that Z of CoSb3 is much smaller. In the filled skutterudite, however, κ is drastically reduced, and the lattice thermal conductivity has nearly the same value as SiO2 glass. This has been a piece of evidence for phonon glass, but in the writer’s opinion, it should be examined carefully whether or not the reduction of κ comes only from rattling. The filled Ce ions induce the high carrier density of the order of 1021 cm–3, which seriously suppresses the phonon mean free path through the electron–phonon interaction. Also, the lowest κ is realized in a Ce deficient sample, and thus disorder also significantly affects the reduction of κ. In fact, κ is also dramatically reduced upon solid solutions in CoSb3 (Anno and Matsubara, 2000). Nevertheless, the concepts of rattling and phonon glass have been a strong driving force in thermoelectric material search in recent years. Accordingly, many promising materials, such as Sr6Ga16Ge30 (Nolas et al., 1998) and CsBi4Te6 (Chung et al., 2000), have been synthesized. 0.35

κlattice(w/cm-K)

0.3 CoSb3

0.25 0.2 0.15 0.1

CeFe4Sb12

0.05 0 0

50

100

150 T (K)

200

250

300

13.8 Effect of rattling (adapted from Sales, 1997).

13.5

Oxide thermoelectrics

13.5.1 Layered Co oxides As mentioned in the previous section, the state-of-the-art thermoelectric materials are Bi2Te3, PbTe and Si1–xGex, all of which are degenerate semiconductors of high mobility. Since Te is scarce, toxic and volatile at high temperature, the application of Bi2Te3 and PbTe has been limited. By contrast, oxide is chemically stable at high temperature in air, and thus use of oxide thermoelectrics is expected in much wider areas. However, most oxide semiconductors show very low mobility, and have thus been mostly dismissed. Since we discovered the large thermopower and the low resistivity in a


353

NaCo2O4 single crystal (Terasaki et al., 1997), we have proposed that some kinds of oxides can be a thermoelectric material (Koumoto et al., 2002). Fujita et al. (2001) have succeeded in measuring the thermal conductivity of a NaCo2O4 single crystal, and found that ZT exceeds unity at 800 K. These results strongly suggest that NaCo2 O 4 is a promising candidate for thermoelectric oxides. Another fascination of NaCo2O4 is the existence of various related oxides. Following NaCo2O4, Ca3Co4O9 (Funahashi et al., 2000; Shikano and Funahashi, 2003), (Bi,Pb)2Sr2Co2O8 (Funahashi and Mastubara, 2001), TlSr2Co2Oy (Hébert et al., 2001), and (Hg,Pb)Sr2Co2Oy (Maignan et al., 2002) have been found to show good thermoelectric performance. Some single crystals show ZT > 1 at 1000 K. As shown in Fig. 13.9, the CdI2-type hexagonal CoO2 layer is common to these cobalt oxides, which reminds us of the CuO2 plane in high-Tc superconductors (Tokura and Arima, 1990). Thus the hexagonal CoO 2 layer should be a key ingredient for the unusually high thermoelectric performance of the layered Co oxides.

Co SrO

Co

BiO

CaO

Na0.5

Co NaCo2O4

CoO

BiO

CaO

SrO

Co

Co Ca3Co4O9

Bi2Sr2Co2Oy

13.9 Crystal structures of the layered cobalt oxides.

Not all the transition metal oxides can be a good thermoelectric material. Figure 13.10 shows the resistivity and the thermopower of various layered transition metal oxides. The layered Co oxide NaCo2O4 shows as low resistivity as the layered Cu oxide Bi2Sr2CaCu2O8 (one of high-Tc superconductors), whereas the layered Ni and Mn oxides show hopelessly high resistivity. For thermopower, the difference between the Co oxide and the other oxides is more remarkable. NaCo2O4 shows 100 µV/K at room temperature, while the layered Cu, Ni, and Mn oxides show very small thermopower of the order of 1–10 µV/K. Thus the most peculiar feature of the layered Co oxide is the unusually high thermopower.

354


(a)

Resistivity (Ωcm)

100

La1.3Sr1.7Mn2O7

10–1 10–2 10

La2NiO4

–3

NaCo2O4

10–4

Bi2Sr2CaCu2O8

10–5 (b)

Thermopower (µV/K)

100

NaCo2O4 Bi2Sr2CaCu2O8 50 La2NiO4

0

La1.3Sr1.7Mn2O7 0

100 200 Temperature (K)

300

13.10 (a) Resistivities and (b) thermopowers of layered transitionmetal oxides.

13.5.2 Physics of the layered Co oxides As an origin of the large thermopower, Koshibae, et al. (2000) proposed an extended Heikes formula for transition metal oxides given by:

S=

g p kB log A C gB 1 – p

13.47

where gA and gB are the degeneracy of the electron configuration of A and B ions, C is the charge difference between A and B ions, and p is the atomic g p is equal to the entropy per content of the A ion. Since k B log A gB 1 – p carrier, Eq. (13.47) is a special case of Eq. (13.9). Let us apply the above formula to NaCo2O4. Assuming that Na and O exist as Na+ and O2– in NaCo2O4, we expect that Co ions exists as Co3+ and Co4+ with a ratio of Co3+:Co4+ = 1:1. Then p for NaCo2O4 is equal to 0.5, and g k S for p = 0.5 is simply reduced to S = B log A . Magnetic measurements C gB reveal that the Co4+ and Co3+ ions are in the low spin state in NaCo2O4. As shown in the upper part of Fig. 13.11, the configuration of the low spin state Co3+ is (t2g)6, whose entropy is zero. On the other hand, the low spin state


355

Co4+ has a hole in the t2g states, which is six-fold degenerate (two from spin and three from t2g orbitals) to carry large entropy of kBlog6. Suppose electric conduction occurs by exchanging Co3+ and Co4+, as is shown in the lower part of Fig. 13.11. Then a hole on Co4+ can carry a charge of +e with entropy of kBlog6, which causes a large thermopower of kBlog6/e (~150 µV/K). This is very close to the high-temperature value of the thermopower. Note that carriers in degenerate semiconductors have no internal degrees of freedom: they can only carry entropy due to their kinetic energy. In this sense, a hole in NaCo2O4 can carry much larger entropy than degenerate semiconductors, which leads us a new design for thermoelectric materials. Co3+

Co4+

eg

eg

t2g

t2g

Co3+

Co3+

Co4+

Co3+

Co3+

Co3+

13.11 Electronic states and electric conduction in the layered Co oxides.

Although Koshibae’s theory has successfully explained the high-temperature limit thermopower of NaCo2O4, the remaining problem is not so simple. The thermopower of NaCo2O4 is 100 µV/K at 300 K, which is about 2/3 of kBlog6, which means that the large amount of entropy of kBlog6 in the hightemperature limit (~104 K) survives down to 102 K. We think it important that NaCo2O4 shows no structural, electric and magnetic transitions from 2 to 1000 K. Usually various phase transitions occur in order to release an excess entropy per sites in the strongly correlated systems. Then, if all the phase transition were blocked, the large entropy would inevitably point to the conducting carriers (Terasaki et al., 2002).

13.6

Summary and future trends

In this chapter, we have briefly reviewed thermoelectric phenomena and thermoelectrics. Since thermoelectrics is a direct energy conversion between heat and electric power, it has various advantages. It can get some electric energy back from waste heat, and can cool materials without an exchange

356


media like a freon gas. Thus this technology has attracted renewed interest from the viewpoint of increasing needs for environment-friendly energy sources. In the last decade, new thermoelectric materials have been researched extensively, some of which have better thermoelectric properties than the conventional thermoelectric materials. From a viewpoint of basic science, the thermoelectric power is an entropy (or heat) carried by an electron. This is more or less controversial terminology, because entropy and heat are concepts in the macroscopic world, whereas the electron is a concept in the microscopic world. Thus a new thermoelectric effect is lying near the boundary between microscopic and macroscopic worlds, which will give a new insight or direction to condensed matter science.

13.7

Acknowledgments

I would like to thank K. Matsubara, H. Anno, T. Caillat and C. Uher for fruitful discussion on thermal properties of skutterudites. Our work cited in this manuscript was partially supported by PRESTO and CREST projects of Japan Science and Technology Agency.

13.8

References

Anno, H. and Matsubara, K. (2000), Recent Res. Devel. Applied Phys., 3, 47–61. Ashcroft, N.W. and Mermin, N.D. (1976), Solid State Physics, Philadelphia, Saunders. Caillat, T., Borshchevsky, A. and Fleurial, J-P. (1996), ‘Properties of single crystalline semiconducting CoSb3’, J. Appl. Phys. 80, 4442–9. Callen, H.B. (1985), Thermodynamics and an introduction to thermostatistics, 2nd edn, Chapter 14, New York, John Wiley & Sons. Chaikin, P.M. and Beni, G. (1976), ‘Thermopower in the correlated hopping regime’, Phys. Rev., B13, 647–51. Chung, D.Y., Hogan, T., Brazis, P., Rocci-Lane, M., Kannewurf, C., Bastea, M., Uher, C. and Kanatzidis, M.G. (2000), ‘CsBi4Te6: A high-performance thermoelectric material for low-temperature applications’, Science, 287, 1024–7. Fujita, K., Mochida, T. and Nakamura, K. (2001), ‘High-temperature thermoelectric properties of NaxCoO2–δ single crystals’, Jpn. J. Appl. Phys., 40, 4644–7. Funahashi, R. and Matsubara, I. (2001), ‘Thermoelectric properties of Pb- and Ca-doped (Bi2Sr2O4)xCoO2 whiskers’, Appl. Phys. Lett., 79, 362–4. Funahashi, R., Matsubara, I., Ikuta, H., Takeuchi, T., Mizutani, U. and Sodeoka, S. (2000), ‘An oxide single crystal with high thermoelectric performance in air’, Jpn. J. Appl. Phys., 39, L1127–29. Hébert, S., Lambert, S., Pelloquin, D. and Maignan, A. (2001), ‘Large thermopower in a metallic cobaltite: The layered Tl-Sr-Co-O misfit’, Phys. Rev., B64, 172101-1–4. Koshibae, W., Tsutsui, K. and Maekawa, S. (2000), ‘Thermopower in cobalt oxides’, Phys. Rev., B62, 6869–72. Koumoto, K., Terasaki, I. and Murayama, N. (eds.) (2002), Oxide Thermoelectrics, Trivandrum, Research Signpost.


357

Mahan, G.D. (1989), ‘Figure of merit for thermoelectrics’, J. Appl. Phys., 65, 1578–83. Mahan, G.D. (1998), ‘Good thermoelectrics’, Solid State Phys., 51, 81–157. Maignan, A., Hebert, S., Pelloquin, D., Michel, C. and Hejtmanek, J. (2002), ‘Thermopower enhancement in misfit cobaltites’, J. Appl. Phys., 92, 1964–7. Nolas, G.S., Cohn, J.L., Slack, G.A. and Schujman, S.B. (1998), ‘Semiconducting Ge clathrates: Promising candidates for thermoelectric applications’, Appl. Phys. Lett., 73, 178–80. Sales, B.C., Mandrus, D., Chakoumakos, B.C., Keppens, V. and Thompson, V.R. (1997), ‘Filled skutterudite antimonides: Electron crystals and phonon glasses’, Phys. Rev., B56, 15081–9. Shikano, M. and Funahashi, R. (2003), ‘Electrical and thermal properties of singlecrystalline (Ca2CoO3)0.7CoO2 with a Ca3Co4O9 structure’, Appl. Phys. Lett., 82, 1851– 3. Slack, G.A. (1995), in CRC Handbook of Thermoelectrics, Rowe, D.M. (ed.) Boca Raton FL, CRC Press, Chap. 34. Terasaki, I., Sasago, Y. and Uchinokura, K. (1997), ‘Large thermoelectric power in NaCo2O4 in single crystals’, Phys. Rev., B56, R12685–7. Terasaki, I., Tsukada, I. and Iguchi, Y. (2002), ‘Impurity-induced transition and impurityenhanced thermopower in the thermoelectric oxide NaCo2–xCuxO4’, Phys. Rev., B65, 195106-1–7. Tokura, Y. and Arima, T. (1990), ‘New classification method for layered copper oxide compounds and its application to design of new high Tc superconductors’, Jpn. J. Appl. Phys., 29, 2388–402.

14 The measurement of thermoelectricity S S U G I H A R A, Shonan Institute of Technology, Japan

14.1

Introduction

After a brief boom in the 1960s, thermoelectric (TE) materials have attracted renewed interest. This interest was exemplified by the first international conference on thermoelectricity held in 1993 in Yokohama, Japan. Since then the conference has been held regularly, rotating between Europe, USA and Asia in turn. During the late 1990s TE materials started to be used commercially, for example in the cooling system for an optical fibre relay station in Japan. New TE materials have started to emerge, including oxide, skutterudite delforssite, clathrate, alongside more established materials such as bismuth telluride and silicon germanium. TE devices are, however, still at an early stage of development. In particular, their use in power generation has suffered because of lower efficiency compared to solar energy. In addition, there had not been a standardized system of measurement until a Japanese Industrial Standard (JIS) was established. The Japanese Industrial Engineering Bureau established a committee to develop the measurement of TE materials in 1998. The author chaired the committee of seven drawn from both industrial and academic fields. In its first year, the committee conducted research among companies, national institutes and universities on thermoelectric energy conversion, commercial applications and future trends. Finally, the committee analysed the data it had gathered, and producing three JIS volumes on TE in 2002, covering the Seebeck coefficient, electrical resistivity and thermal conductivity (including diffusivity and heat capacity).1 This chapter discusses the measurements of thermoelectricity (Seebeck coefficient), electrical resistivity and thermal conductivity as set out in the JIS document.

14.2

Seebeck coefficient

At the junction between two different conductors, the Peltier effect produces the generation or absorption of heat (depending on the direction of the current). This is shown as Q in the following equation when a current I flows through the conductors: 358

The measurement of thermoelectricity

Q=Π·I

359

14.1

The Peltier coefficient (Π) closely relates to the Seebeck effect and to the Seebeck coefficient α: Π=α·T

14.2

A thermoelectric voltage V is produced in an open circuit consisting of two different conductors when there is a temperature difference dT between their ends: dV = α dT

14.3

Generally, any change in voltage is measured in the time that a current of 10 mA flows between the conductors. A specimen is placed between the higher temperature and lower temperature conductors, keeping a constant value of 10 K. The size of a specimen is usually 3–5 mm wide and 13–15 mm in length, or 3–5 mm in diameter in the case of a cylindrical specimen. There is also a third aspect of thermoelectric physics called the Thomson effect. If there is a temperature drop Th–Tc along a conductor with an electrical current, Thomson heat (QT) is generated or absorbed besides Joule heat. Heat flux is shown in the following equation: QT = r (Th – Tc)/L

14.4

where T is temperature, r is Thomson coefficient, L is length, and r = T (d α /dT)

14.5

Figure 14.1 shows the so-called four probe method for measuring these effects. The fixed current (ex. 10 mA) flows from both ends and two probes contact the side of the specimen to measure the voltage drop. The measurement precision of a digital voltmeter is required to be 0.5 µV. Thermocouple

Current Specimen Probe for voltage measurement Current Thermocouple

14.1 Four probe method for measurement.

Gotoh describes the AC method (see Fig. 14.2).2 Thermopower (∆E) is measured at both ends (at high and low temperature) and the Seebeck coefficient

360

Materials for energy conversion devices Thermometer Aluminum block Thermocouples Module Switch

DC power

Heat sink

14.2 Block diagram for evaluation of module.

(S) is calculated by the relationship S = ∆E/∆T. The whole specimen is heated by an infrared lamp up to 1500 K and the temperature of specimen is measured by Pt-PtRh13% (0.025 mm φ) under vacuum (approx. 0.1 Pa) or Ar gas (67 kPa). This method has the advantages of precision compared with a laser flash method and a shorter measuring time. The current direction is changed by switching repeatedly, and cooling and heating are also repeated in a certain time interval.

14.3

Electrical resistivity

Industrial standards have already been published on measuring metal and bulk electrical resistivity. There are two methods for electrical resistivity measurement: the DC and AC methods. TE material possesses generally very low resistivity (usually 10–6 ~ 10–4 Ω · m). A larger current density is required to obtain a higher voltage. Larger Peltier heat, however, is generated at the interfaces of both electrodes and specimen when a larger current flows in a TE material, resulting in an error in measuring resistivity since the drop voltage is added to the measured thermoelectric power. This error can be avoided by the AC method. To reduce an error due to Peltier heat, the voltage take-out points should be placed at the shortest side and kept as far as possible from the electrode. As a rough rule of thumb: width/voltage takeout distance ≥ 4. In addition, the current should flow in as short a time as possible, preferably less than a second. The four probe method is more popular for measuring resisitivity than the two probe method. Usually resisitivity is measured at the same time as the Seebeck coefficient, as shown in Fig. 14.1. The standard method also describes the size of specimen which should be 3–5 mm wide and 13–15 mm in length, and 3–5 mm in diameter in the case of a cylindrical specimen. A caliper and micrometer are used for size measurement. A thermocouple of less than 0.3 mm in diameter is used for measuring the temperature at both ends. The AC method is facilitated if the two probes are set at the ends of both electrodes and the voltage take-out points. In particular, both ends should use a heat sink to absorb Peltier heat. Electrical resistivity is also popularly


361

measured with Hall effect measurement equipment, where the current I (A) flows in longitudinal direction and voltage drop is measured in parallel to the current flow. In the following equation, the voltage represents VH (V), L (m) stands for the distance of the voltage take-out points, width w (m) and thickness d (m), the resistivity (ρ):

ρ = VHwd/IL (Ωm)

14.6

Temperature gradient in the longitudinal direction due to Peltier heat is added to the voltage drop resulting in electrical resistivity and in the Nernst effect. The Hall coefficient measurement has not been covered in the JIS. Briefly the measurement is introduced here. The Hall coefficient RH is obtained by inducing the voltage vertically to both magnetic field B(T) and electric current I(A) and is evaluated by the following equation: RH = VH d/IB (m3/C)

14.7

The Hall coefficient also relates to carrier concentration. Carrier concentration n is described as follows: n = γH/eRH (1/m3)

14.8

γH depends on scattering factor γ, which is 0 for an atomic lattice, 1 for an ionic lattice and 2 for scattering by impurity ions. The Hall coefficient of a Bi2Te3 system, which is widely used as a TE material, is in the order of 10–7 m3/C and 10–9 m3/C for n-FeSi2.3 As shown in Fig. 14.3, the Hall coefficient by van der Pauw method is described in the following equation: RH = t ∆RBD,AC/B

14.9

∆ RBD,AC is the change of Hall resistivity induced between ends of A and C when magnetic field (B) is applied in a perpendicular direction. RAB,CD is defined as the ratio of voltage drop of VCD at the ends and current IAB, VCD /IAB when current IAB flows at the ends of A and B. The Hall coefficient measurement is difficult to assess in materials like oxides with a low carrier density. A

A

B

C

14.3 Sample shape of the van der Pauw measurement.

362


It is noticeable that TE materials may produce a large voltage through a small temperature difference. The induced voltage, due to a number of thermomagnetic effects, may add to the steady state VH when a magnetic field is applied perpendicularly to the electrical current. In the Ettinghausen effect, there is a temperature difference in the measuring direction of VH which is perpendicular to both electrical current and magnetic field. Thermoelectric power VE is added to VH. VE due to a temperature difference cannot be cancelled by reversing the electrical current and magnetic field since VE changes direction at the same time as VH. In the Nernst effect, when there is a heat flow in parallel to electrical current, the voltage of VN is generated in the measuring direction of VH. In a TE material, VN coincides with VH since heat flow is caused by a temperature difference due to Peltier heat. Therefore, it is not possible to cancel VN even by a reversal process between electrical current and magnetic field. In the Righi-Leduc effect, a temperature difference appears in the measuring direction of Hall voltage, VH when a heat flow exits in the direction of electrical current, so that thermoelectric power, VRL overlaps VH.

14.4

Thermal conductivity

The JIS for TE materials includes measurements of heat diffusivity, specific heat capacity and thermal conductivity. JIS R 1611 discusses thermal conductivity for fine ceramic material. The standard (JIS R1650-3) suggests the use of a laser flash method for specimens with porosity less than 10%. This technique is the most popular way to measure the thermal conductivity of TE material at ranges from room temperature up to a maximum of 1000 K.1 In the laser flash method, heat diffusivity α and specific heat capacity C are measured at room temperature, Thermal conductivity (κ) is then calculated using the following equation:

κ = α · C · ρ/(1 + ν)3

14.10

where κ is thermal conductivity (W/(m · K)), α is thermal diffusivity (m2/s), C is heat capacity (J/Kg · K), ν is thermal expansion, and ρ is bulk density of a specimen at room temperature (Kg/m3). The range of thermal diffusivity measured by this method is 10–7~10–4 m2/s, and TE material will be mostly around the lower level of this range. Heat diffusivity is calculated by the half-time method. This measures how long it takes for the furthest side of the sample to reach half of the maximum temperature of the laser:

α = 1.37 L2/π2t1/2

14.11

Furthermore, heat capacity is calculated as follows: C = QL /D d Tmax

14.12


363

where QL is heat absorbed by the sample surface (J/m2), ρ is the same as above, d is thickness of sample, and Tmax is maximum temperature on the furthest side of the sample. The thickness of a specimen is required to be 4 mm at most and the thermocouple should be less than 0.1 mm in diameter to avoid heat loss. Furthermore, a glassy carbon should be coated on the specimen. A detailed explanation is provided in JIS R 1650-3.3 In this standard, the definitions are precisely presented for terms such as effective pulse width, half-time method, temperature history curve, and so on. In the static method, thermal conductivity at room temperature is measured by the same equipment as the α calculation method (Fig 14.4). The specimen is pinched at both ends by copper plates. The thermal conductivity of the grease used on the copper plates is typically 0.502 W/m · K. A standard specimen is transparent silica glass (κ = 1.36 W/m · K at room temperature). The thermopower difference between the two thermocouples, correlated with the electric power given to the heater, is measured to within 1 K, and thermal conductivity is obtained by the gradient of thermopower/electric power. Pressure Water Thermostat Cu plate

Cu plate is put by thermal conductive grease to avoid bubbles No. 1 thermocouples (75 µ m φ )

Specimen

Cu plate Cu plate Cu plate

Heater (Pt 50 µ m φ) Specimen

No. 2 thermocouples (75 µ m φ )

Thermostat Pressure

14.4 Static method for measuring thermal conductivity.

14.5

Simple evaluation of Z for module

The Harman method describes an approach to measuring thermoelectric properties and the figure of merit of the Peltier module. A small current I (such as 10 mA) passes through the system generating a slight temperature differential along the module. By measuring the Joule and Seebeck voltage drops, we can measure thermoelectric properties:

α · IT0 – (1/2)I2R – κ ∆T = (a0/N)(Ta – T0), α · IT1 + (1/2) I2R – κ ∆T = (a1/N)(T1 – Ta)

14.13

where T0 is cold side temperature; Ta is ambient temperature; T1 is hot side temperature; ∆T = T1 – T0; κ is specimen (device) thermal conductivity; R is

364


electrical resistivity; N is the number of device; a0 is cold side heat flux and a1 is hot side heat flux. We can obtain the following relationship: a0 = a 1 = a

14.14

The figure of merit (Z) can be obtained by the equation: ∆Tmax = (1/2) Z (Tej)2

14.15

Here ∆Tmax represents the maximum of temperature difference and Tcj describes the lowest temperature at the cold junction. One needs 30 min to one day duration for one temperature point.

14.6

Future trends

Thermoelectric energy conversion in such areas as cooling devices is a major area of interest. In the energy sector, thermoelectric methods remain less attractive than other techniques. Their most likely application in this sector will be as a back-up system to fuel cell and solar power. On the other hand, small-scale applications are feasible. A typical example is a wristwatch using micro-thermoelectric modules developed by Seiko Instruments in Japan.4 The material used is a Bi-Te compound, while the size of the TE module is 80 × 80 × 600 µm and contains 104 elements resulting in Qmax = 0.16 W. In the industry, cooling systems have been used for relay stations in the optical fibre business. There will also be an increasing need for the use of TE materials in cooling for high speed integrated circuits, for example for charge coupled device cameras, and precise temperature control of semiconductor processing equipment. New types of TE material have also been developed recently. The development of oxide materials for thermoelectricity have become popular in Japan in recent years. The most popular is Na2CoO4, followed by CaMnO3, (ZnO)(In2O3), ZnO and CuAlO2. Another trend is development of film type thermoelectric materials. Film or superlattice TE materials have become popular in micro devices.

14.7

References

1. Sugihara, S. (ed.), Testing method for fine ceramics thermoelectric materials, Part 1; Thermoelectric power, Part 2; Resistivity, Part 3; Thermal diffusivity, specific heat capacity, and thermal conductivity, JIS R1650-1, JIS R1650-2 and JIS R16503, respectively. 2. Gotoh, T., ‘Measurement of Seebeck coefficient thermoelectric material by AC method.’ Private communication. 3. Uemura, K. and Nishida, I., Netsuden Handoutai to sono Ouyou (Thermoelectric Semiconductor and its Application), Nikkan Kogyo Shinbunsha, 1989 (in Japanese). 4 Kishi, M. et al., 18th International Conference on Thermoelectrics, 301–7, Baltimore, USA, 1999.

15 Environmentally friendly hydrogen generation by nuclear energy M Y A M A W A K I, Tokai University, Japan, T N I S H I H A R A, Y I N A G A K I, K M I N A T O, H O I G A W A, K O N U K I, R H I N O and M O G A W A, Japan Atomic Energy Research Institute, Japan

15.1

Introduction

It is universally admitted that hydrogen is one of the best energy media and its demand will increase greatly in the near future, because it can be used as clean fuel in a variety of energy end-use sectors including conversion to electricity without CO2 emission, and also can be stored and transported over long distances with lower loss compared to electricity. If hydrogen is produced with nuclear energy, it could greatly contribute to the solution of the global warming issue. A high temperature gas-cooled reactor (HTGR), which provides hightemperature heat at above 900 °C, can generate hydrogen economically without CO2 emissions. In Japan an HTGR called the high temperature engineering test reactor (HTTR), was constructed at the Oarai Establishment of Japan Atomic Energy Research Institute (JAERI), with the coolant output temperature of 950 °C being achieved in April, 2004. Since hydrogen generation from water is considered as an ideal method for hydrogen generation using the HTGR due to no CO2 emissions being expected from the system, JAERI has been conducting R&D on the thermochemical iodine-sulfur (IS) process for the hydrogen generation process by water splitting. The IS process utilizes plural chemical reactions and works like a chemical engine to generate hydrogen by absorbing high temperature heat from the HTGR. Continuous hydrogen generation by the IS process was successfully achieved for the first time in the world, using a bench-scaled apparatus in August, 2003. As for long-lived radioactive waste generated by nuclear reactors, it has been proposed and eagerly studied to transmute minor actinides (MA) and long-lived fission products (LLFP) into stable or short-lived nuclides by means of accelerator-driven systems (ADS). By applying new technologies, it is expected to significantly reduce the load on the human environment caused by long-lived radioactive nuclear waste.

365

366

15.2


Activities on hydrogen generation in Japan

Hydrogen demand (billion Nm3/year)

Current society depends on fossil energy, and raises the issues of global warming, acid rain, etc. To mitigate the issues, the Japanese government has been conducting R&D on hydrogen energy. For example, the WE-NET (World Energy NETwork) project was carried out to develop technologies on hydrogen generation, hydrogen use such as a fuel cell and a hydrogen combustion turbine, transportation and storage, a hydrogen station, etc. Now, the JHFC (Japan Hydrogen & Fuel Cell Demonstration) project is under way. The project consists of the fuel cell demonstration program, included in the support project for ‘empirical and other research on solid high-polymer fuel cell systems’ under the auspices of the Ministry of Economy, Trade and Industry, and the demonstration study of hydrogen fueling facilities for fuel cell vehicles. Figure 15.1 shows the prediction of hydrogen demand for fuel cells in Japan. The target for the introduction of fuel cell vehicles is 50,000 by 2010, 5 million by 2020 and 15 million by 2030. The target for stationary fuel cells for residential use is 2.1 GW by 2010, 10 GW by 2020 and 12.5 GW by 2030. Each hydrogen demand is predicted to be 6 billion Nm3 in 2010, 28.3 billion Nm3 in 2020 and 46 billion Nm3 in 2030, respectively. Hydrogen is generally generated from carbonized hydrogen and oxidized hydrogen, namely fossil fuel and water because little hydrogen exists naturally. Accordingly, hydrogen is decomposed from fossil fuel or water by providing much energy such as heat and electricity. Then, how can we generate a large amount of hydrogen economically and reduce CO2 emissions simultaneously? Hydrogen generation with nuclear energy is one of the solutions to this question. In Japan, the Basic Plan for Energy Supply and Demand based on the Basic Law on Energy Policy Making was decided by the Cabinet on October 6, 50

46 (total)

Fuel cell vehicle Stationary fuel cell

17 (15 m vehicles)

40

28.3 (total)

30

6.5 (5 m vehicles) 20

10

0

0.4 (0.05 m vehicles) 6 (total)

2010

21.8 10GW

29 (12.5 GW)

5.6 (2.1 GW) 2020 Year

2030

15.1 Hydrogen demand for fuel cells in Japan.

Environmentally friendly hydrogen generation by nuclear energy

367

2003. The plan prescribes that commercialization of hydrogen generation systems using nuclear, solar and biomass, not fossil fuels, is desired.

15.3

Hydrogen generation by nuclear energy

Nearly 90% of hydrogen generated in the world today is produced with a steam reforming process industrialized mainly based on methane, where combustion heat of fossil fuel is supplied for the chemical reaction of steam reforming. The steam reforming process exhausts approximately 0.9 kg-CO2 to generate 1 Nm3-H2. The electrolysis of the water exhausts more than 1.6 kg-CO2 for 1Nm3-H2 when electricity is generated with fossil fuel. Therefore, the raw material and the energy source containing little carbon should be selected to realize hydrogen generation without CO2 emission. The methods of hydrogen generation through nuclear energy are roughly classified into two groups, that is, electrolysis and thermal decomposition. Further, research on a hybrid method using both electricity and heat is under way. The characteristics of each method are described below.

15.3.1 Electrolysis Technical development is not necessary for the electrolysis of water using electricity supplied from a light water reactor (LWR), which is currently popular as a nuclear power reactor throughout the world. In this case, however, the efficiency of hydrogen generation is lower than 30%, because the efficiencies of electricity generation and electrolysis are approximately 35% and 80%, respectively. If power generation by means of a HTGR helium gas turbine system is put to practical use, the efficiency of hydrogen generation would be improved to 40%, because the efficiency of electricity generation in this case will be approximately 50%. Moreover, the research on the hybrid method, namely high temperature electrolysis, is on-going to improve efficiency. By electrolysis of steam at approximately 900 °C, about 20% electricity can be saved by using high temperature heat supplied from a HTGR. However, development of an electrolysis cell to be used in a high temperature environment is a key technical issue to realize the system.

15.3.2 Thermal decomposition Hydrogen generation by direct thermal decomposition of water requires hightemperature heat of about 4000 °C. However, by combining high-temperature endothermic chemical reactions and low-temperature exothermic chemical reactions, in which the net chemical change resulting from the sequence of component chemical reactions is the water decomposition, it is possible, in

368


principle, to decompose water with heat of about 900 °C. The IS process, which is a kind of thermochemical method for hydrogen production by water splitting, was proposed and studied by the General Atomic Co. [1] and has been studied also in Germany [2], Canada [3], and Japan [4]. The IS process produces hydrogen by absorbing high-temperature heat at 800–900 °C supplied from a HTGR, and is composed of the three chemical reactions shown in Fig. 15.2. Summation of eqns 15.1 to 15.3 results in the water splitting reaction of eqn 15.4: H2SO4 → H2O + SO2 + (1/2)O2

(oxygen generation)

15.1

2HI → H2 + I2

(hydrogen generation)

15.2

2H2O + SO2 + I2 → H2SO4 +2HI

(Bunsen reaction)

15.3

H2O → H2 + (1/2)O2

15.4

67 J 67 J Hydrogen

Oxygen Nuclear heat

H2

400C 2HI I2

H2SO4

2HI

I2

I2

O2

900 C 33J Rejected heat 100 G

H2

I (lodine) circulation

1 2

76J

24

H2O

SO2+H2O

H2SO4

1 2

O2

SO2+H2O

S (sulfur) circulation

H2O

SO2 + H2O

Water

15.2 Chemical reactions of the IS process.

The process works like a chemical engine to produce hydrogen by absorbing high-temperature heat in the endothermic decomposition of eqn 15.2 and discharging low-temperature heat in the exothermic reaction of eqn 15.4. The IS process has attractive features in that all the process chemicals are used in its fluid phase and the endothermic sulfuric acid decomposition reaction proceeds stoichiometrically with large entropy change. First, H2SO4 decomposes spontaneously into SO3 and H2O in the temperature range of 350–500 °C. By further heating up to over 800 °C, SO3 decomposes into SO2 and O2 in the presence of a solid catalyst. Both reactions are strongly endothermic and the temperature range of the reactions is well matched with


369

that of HTGR. The current status of the IS process is described in section 15.5. Research on the hybrid method using high-temperature heat and electricity, the Westinghouse process [5], is under way: 2H2O + SO2 → H2SO4 + H2

(hydrogen generation)

15.5

H2SO4 → H2O + SO2 + (1/2)O2

(oxygen generation)

15.6

Hydrogen generation of eqn 15.5 is carried out by means of electrolysis, and the method of oxygen generation is the same as the IS process. The system of this process can be simplified because the iodine process is not used. However, there is a possibility that hydrogen generation cost becomes expensive compared with the IS process because of the use of electricity, and development of an electrolysis cell is a key technical issue to realize the system.

15.3.3 Hydrogen generation cost The cost of hydrogen production was roughly evaluated, taking into account CO2 fixation costs of 21 Yen/kg-CO2. Here, hydrogen cost is defined as Japanese Yen per unit energy. The cost of hydrogen production consists of (i) energy cost, (ii) raw material cost, (iii) capital cost (including operational and maintenance cost). Costs of nuclear production of hydrogen are calculated under the assumptions tabulated in Table 15.1. Figure 15.3 shows costs of hydrogen produced by four methods; steam reforming process using combustion heat of fossil fuel (SR/FF), electrolysis of water with a light water reactor (LWR), electrolysis of water with renewable energy and IS process with the HTGR. The costs in Fig. 15.3 are normalized by the cost of SR/FF. The hydrogen cost of the steam reforming process is 1.0 with CO2 fixation and 0.55 without CO2 fixation. On the other hand, the cost of the IS process is 0.8 because the cost of CO2 fixation is not necessary.

15.4

Features of HTGR

The LWRs use water as coolant/moderator and metal as in-core materials. The outlet temperature of the coolant is approximately 300 °C, and the nuclear heat application is limited to electricity generation. On the other hand, HTGR can generate high-temperature heat above 900 °C using graphite as in-core materials and helium gas as coolant. The high-temperature heat can be used for hydrogen generation as well as electricity generation as shown in Fig. 15.4. Furthermore, HTGR has excellent inherent safety. The fuel particle coated with graphite and silicon carbide has high thermal integrity and a high FPs retention capability, and the core made of graphite has no possibility of meltdown. Chemical reaction between the coolant and core

370


Table 15.1 Assumptions for calculation of hydrogen cost Hydrogen production rate Service life

200,000 Nm3/h 40 years

Steam reforming process Capital cost Natural gas cost Operation rate Maintenance cost CO2 fixation cost

27 billion yen 1.8 yen/1,000 kcal 90% 21.3% of capital cost 21 yen/kg-CO2

IS process with HTGR Nuclear cost Thermal output Operation rate Thermal efficiency

38 billion yen 600 MW 90% 55%

Electrolysis of water Capital cost Electricity generation cost with LWR Electricity generation cost by wind

CRIEPI Report, Y91005 (1991) Design of GTHTR-300

5.8 yen/Nm3-H2 5.9 yen/kWh 6 yen/kWh

0.2 Methane steam reforming with fossil fuel (existing plant)

WE-NET FY 1997 Report

0.05

0.3

NEDO-GET-005 FY2000 Report

0.1

0.35

1.0

1.3 Electrolysis of water with LWR

CO2 fixation Cost (energy) CO2 fixation cost (raw material) Energy cost Raw material cost Capital cost

Electrolysis of water with renewable energy

IS process with HTGR

2.0

0.8

1 Ratio of hydrogen cost

2

15.3 Ratio of hydrogen cost in the case of CO2 fixation.

components does not occur in a high-temperature environment because helium gas is inert. Therefore, it can be said that there is no danger of severe accidents causing large-scale fuel failure or core meltdown.

Environmentally friendly hydrogen generation by nuclear energy 0

200

371

Temperature (°C) 400 600 800 1000 1200 Power generation with GT Hydrogen production from water Hydrogen production from natural gas Petroleum refining

Region heating, sea desalination HTGR LWR

FBR

15.4 Temperature regions on heat application.

Figure 15.5 shows the cutaway view of the HTTR reactor. The HTTR can supply high-temperature heat of 950 °C at the reactor outlet with the thermal power of 30 MW, using helium gas as coolant and graphite as materials of core and reactor internals such as fuel elements, replaceable and permanent reflector blocks and core support structures [6]. The reactor consists of a

Stand pipe

Permanent reflector block Replaceable reflector block Core restraint mechanism Fuel element Hot plenum block Support post Lower plenum block Carbon block Bottom block Support plate Core support grid

Auxiliary coolant outlet pipe Main coolant outlet pipe

15.5 Cutaway view of the HTTR reactor.

372


reactor pressure vessel, fuel elements, replaceable and permanent reflector blocks, core support structures, control rods, etc. Thirty columns of fuel elements and seven columns of control rod guide blocks form the reactor core called the fuel region, which is surrounded by replaceable reflector blocks and large-scale permanent reflector blocks. The fuel element of the HTTR is a so-called pin-in-block type. The reactivity of the HTTR is controlled with sixteen pairs of control rods in the fuel and replaceable reflector regions of the core. Figure 15.6 shows the concept of the HTGR hydrogen generation system. The heat generated in the core is exchanged from the primary helium gas to the secondary one with the IHX, and the secondary one is transported to the hydrogen production facility passing through the hot gas duct. The transported heat is used in the facility for the endothermic reaction of hydrogen production.

Reactor

Hydrogen production plant

Chemical reactor High-temperature isolation valve

Nuclear reactor Hot gas duct

Intermediate heat exchanger (IHX)

15.6 Concept of the HTGR hydrogen production system.

15.5

R&D activities in hydrogen generation

The HTTR attained the first criticality in November, 1998. The rise to power test of HTTR started in September 1999 and then HTTR reached the full power of 30 MW with the reactor outlet coolant temperature being 850 °C in December, 2001. Then, on March 6, 2002, JAERI received a certificate of the pre-operation test from the Government, that is, the operation permit of HTTR at the rated operation mode (operation at the reactor outlet coolant temperature of 850 °C), completing the rise to power test at rated operation mode. The HTTR accomplished the reactor outlet coolant temperature of 950 °C at high-temperature test operation mode in April, 2004. This temperature


373

is enough for hydrogen production by the thermochemical water splitting method. The safety demonstration test using the HTTR is under way to demonstrate the inherent safety features of HTGR. For developing the IS process the (i) close cycle test, (ii) high efficiency component test and (iii) material test are under way. The close cycle test aiming at establishment of reaction control has been carried out with the bench scale test apparatus as shown in Fig. 15.7. Demonstration of the first continuous hydrogen production in the world, at the hydrogen production rate of approximately 0.03m3/h, was successfully achieved for 6.5 h in August, 2003 and 20 h in December, 2003 as shown in Fig. 15.8 [7]. Furthermore, hydrogen was continuously produced for one week in June, 2004. As a result, it can be said that the reaction control technology has been established. Oxygen generation unit (sulfuric acid decomposer)

Bunsen Reactor

Hydrogen generation unit

15.7 Bench-scaled test apparatus of the IS process.

Production rate [Nm3/h]

0.04 0.03 0.02 0.01 0

H2 O2 0

5

10 15 Time [h]

20

15.8 Continuous hydrogen production for 20 hours.

In the high efficiency component test, a liquid phase separator to be placed in the reaction of eqn 15.3 at elevated temperature condition (0 °C → 95 °C) is being developed to achieve better separation of HI and H2SO4, and introduction of advanced separation technologies, which includes method of

374


electrodialysis, etc., for efficient HI decomposition, is being studied to improve the process scheme. In parallel with these process studies, materials for the pilot scale facility having a hydrogen production rate of 30 Nm3/h are being developed to meet the corrosive process conditions such as boiling sulfuric acid and SO2-SO3-H2O-O2 gaseous mixture at about 800 °C. As for the boiling sulfuric acid condition, iron-silicon alloys and silicon impregnated SiC are promising candidates from the viewpoint of corrosion resistance [8, 9]. JAERI is carrying out R&D on hydrogen production with a HTGR as shown in Fig. 15.9. As for the reactor technology, the HTGR operational experience is accumulated and tests on safety demonstration and up-grade will be carried out. As for the IS process, the pilot test is scheduled from 2005. In the test, the test plant will be developed, which is made of industrial materials and can be operated under prototypical high-pressure condition. High-temperature helium gas heated by a simulated electric heater will be used to drive the process. Operation of the test plant will demonstrate the technical feasibility of the IS process, and the test data will be used to verify the analytical codes to be developed. After completion of the pilot test of the iodine-sulfur cycle, it is planned to proceed to the demonstration test of nuclear hydrogen production using a HTTR based on the technology development mentioned above.

Program item 2000 HTTR operation and test

2005

2010

2015

2020

Accumulation of HTGR operational experience Safety demonstration test

Up-grade of HTTR fuel Demonstration of IS process

Hydrogen production

System integration technology

(IS process) Bench-scale test

Pilot test

HTTR test

15.9 R&D schedule on hydrogen production with HTGR at JAERI.

15.6

An innovative option for radioactive waste management

When the nuclear reactors are used, the management of the high-level radioactive waste (HLW) is one of the most important issues to be solved. Although most countries with nuclear reactors have their own plans to dispose of HLW into a deep geological repository, it does not seem easy to determine


375

Radio-toxicity (ALI ingestion hazard index of HLW per one metric ton of fresh fuel)

appropriate sites and to get public acceptance. One of the reasons for this difficulty is due to the fact that the HLW contains long-lived hazardous nuclides such as minor actinides (MA: Np, Am, Cm) and long-lived fission products (LLFP: Tc-99, I-129) whose radio-toxicity lasts for millions of years [10]. JAERI has proposed and developed the technology ‘Partitioning and Transmutation (P&T) of MA and LLFP’ as an innovative option for radioactive waste management within the framework of the OMEGA program launched by the Atomic Energy Commission of Japan in 1988. OMEGA is the acronym derived from Options Making Extra Gains from Actinides and fission products. The objective of the P&T technology is to reduce the long-lived nuclides in HLW. Figure 15.10 shows the radio-toxicity of HLW as a function of the time after reprocessing. The radio-toxicity is defined as the amount of a nuclide in HLW per ton of fresh fuel divided by the ALI (annual limit of intake) of the nuclide. The results of the analyses are also shown in Fig. 15.10, which indicate that radio-toxicity can be reduced by two orders and the time period 1010

109 Without transmutation 90% transmutation for MA and LLFP

108

107 Natural uranium (5 ton) 106

105 100% transmutation only for MA

104 99.5% transmutation for MA and LLFP 103

102 100

100% transmutation for MA and LLFP 101

102 103 104 105 106 Time after reprocessing (year)

107

15.10 Reduction of radio-toxicity of HLW by transmutation.

376


to reduce the radio-toxicity below the level of natural uranium used as raw material can be shortened from 50,000 years to 500 years if 99.5% transmutation of MA and LLFP is achieved. The reduction of amounts of MA and LLFP in HLW would make radioactive waste management much easier. To achieve the transmutation efficiently, JAERI has proposed the concept of the double-strata fuel cycle [11, 12]. The first stratum is the power-reactor fuel cycle and the second stratum the dedicated transmutation fuel cycle, as shown in Fig. 15.11. The minor actinides partitioned from HLW of the first stratum are fed into the second stratum, where they are transmuted to fission products through fission reactions by fast neutrons (~1 MeV). Tc-99 and I129 are transmuted to stable nuclides of Ru-100 and Xe-130 through neutron capture reactions. In this concept, the power-reactor fuel cycle and the dedicated transmutation fuel cycle may be optimized independently for the safe and economical use of plutonium and the efficient reduction of long-lived hazards, respectively. Power reactors: 3000 MWth × 10 units Spent fuel

Reprocessing

Fuel fabrication

Actinide burner 800 MWth MA fuel

Spent fuel

U, Pu

HLLW

P-T complex HLLW Partitioning

Pyroprocess Short-lived FPs

Disposal

15.11 Double-strata fuel cycle.

For the transmutation fuel cycle, nuclear fuel mainly consisting of MA without uranium is used to enhance the transmutation efficiency; uranium would be converted into MA through neutron captures and decays. Using such MA fuel, the critical reactor would encounter many difficulties in its safety and controllability aspects. The accelerator-driven subcritical system (ADS) has potential advantages in comparison with critical reactors: (i) various fuel compositions are flexibly acceptable since the Doppler effect


377

does not seriously affect system safety, and (ii) a small value of delayed neutron fraction is also acceptable since the margin to the prompt critical state can be kept by the subcriticality. The ADS is, therefore, considered suitable to transmute the MA in the second stratum [10]. The reference ADS design proposed by JAERI is the 800 MWth (MW of thermal output) fast subcritical core fueled with MA nitride, cooled by PbBi eutectic and driven by the spallation neutron source using a Pb-Bi target and a proton accelerator [13]. The nitride fuel has been chosen as a candidate of the transmutation fuel because of the possible mutual solubility of the actinide mononitrides and the excellent thermal properties. In addition, it supports a hard neutron spectrum needed for fissions of MA. In this ADS, 250 kg of MA can be burned per year by fission reactions, which corresponds to the amount of MA produced in ten units of LWRs per year. To realize the ADS, JAERI has conducted R&D in the fields of the proton accelerator [14], the Pb-Bi technology [15], and the MA nitride fuel [16]. Furthermore, a new experimental facility, the Transmutation Experimental Facility (TEF), is planned to demonstrate the feasibility of the ADS from the viewpoints of the reactor physics and the target engineering [17]. The roadmap to realize the ADS is illustrated in Fig. 15.12. After the basic R&D and the experiments in the TEF, an experimental ADS with thermal power of about 80 MW is considered to be necessary in late 2010s to demonstrate the feasibility of the ADS from engineering aspects. The demonstration of the MA transmutation will be completed by 2030, including the reprocessing of the nitride fuel irradiated in the experimental ADS. After that, the transmutation plants with 800 MW will be built.

Power

Transmutation Plant 20 MW-beam, 800 MWth Transmutes MA from 10 LWRs Experimental ADS 2 MW-beam, 80 MWth Demonstrates ADS performance Transmutation experimental facility 200 kW-beam, Pb-Bi Target Technology ADS physics experiment Year

Loop experiment 2000

2010

2020

15.12 Future plan for ADS development.

2030

378

15.7


Conclusion

Hydrogen generation from water using nuclear energy is one of the promising solutions for reducing CO2 emission from the viewpoint of the global warming issue. Especially, HTGR has a possibility to generate hydrogen economically compared with other types of nuclear reactor. The HTTR project of JAERI is a very important milestone to commercialize hydrogen generation with a HTGR in terms of the demonstration of coupling of reactor and the hydrogen generation plant, and hydrogen generation directly from thermal energy supplied from a nuclear reactor. JAERI is also contributing to the solution of the highlevel radioactive waste management by developing the ADS related technologies, where MA and LLFP will be transmuted to less hazardous nuclides with much shorter half lives. Thus, the application of nuclear energy will provide an environmentally friendly and economical method to generate hydrogen in the near future.

15.8

References

1. Norman, J.H., et al. Thermochemical water-splitting cycle, bench-scale investigations, and process engineering. GA-A16713, 1982. 2. Roth Mand, M. and Knoche, K.F., Thermochemical water splitting through direct HI-decomposition from HI/I2/H2O solutions. Int J Hydrogen Energy 1989; 14: 545– 549. 3. Oeztuerk, I.T., et al. A new process for oxygen generation step for the hydrogen producing sulfur-iodine thermochemical cycle. Trans IChemE 1994; 72 ( Part A): 241–250. 4. Onuki, K., et al. R&D program on thermochemical water-splitting iodine-sulfur process at JAERI. GENES4/ANP2003, Sep. 15–19, 2003, Kyoto, Japan, p. 1072. 5. Goossen, J.E., Improvements in Westinghouse Process for Hydrogen Production, Global 2003, Nov. 16–20, New Orleans, LA, USA, p. 1509. 6. Saito, S., et al. Design of High Temperature Engineering Test Reactor (HTTR). JAERI-1332, 1994. 7. Kubo, S., et al. A demonstration study on a closed-cycle hydrogen production by the thermochemical water-splitting iodine-sulfur process. Nucl Eng Des 2004; 233: 347– 354. 8. Futakawa, M., et al. Viscosity of amorphous oxide scales on SiSiC at elevated temperature. J Am Ceram Soc 1998; 81: 1819–1823. 9. Ioka, I., et al. The characterization of passive films on Fe-Si alloy in boiling sulfuric acid. J Materials Science Letters 1999; 18: 1497–1499. 10. Oigawa, H., et al. Research and development on accelerator-driven system for transmutation of long-lived nuclear waste at JAERI. Proc the 13th Pacific Basin Nuclear Conference (PBNC 2002), Oct. 21–25, 2002, Shenzhen, China. 11. Takano, H., et al. Transmutation of long-lived radioactive waste based on doublestrata concept. Progress in Nuclear Energy 2000; 37: 371. 12. Mukaiyama, T., et al. Review of research and development of accelerator-driven system in Japan for transmutation of long-lived nuclides. Progress in Nuclear Energy 2001; 38: 107.


379

13. Tsujimoto, K., et al. Neutronics design for lead-bismuth cooled accelerator-driven system for transmutation of minor actinide. J Nucl Sci Technol 2004; 41: 21. 14. Ouchi, N.. R&D status of superconducting proton LINAC and KEK/JAERI High Intensity Proton Accelerator Project. Proc the 3rd International Workshop on the Utilization and Reliability of High Power Proton Accelerators (HPPA 2002), May 12–16, 2002, Santa Fe, USA. 15. Kikuchi, K., et al. Lead-bismuth eutectic compatibility with materials in the concept of spallation target for ADS, JSME International Series B 2004; 47(2): 1–8. 16. Minato, K., et al. Fabrication of nitride fuels for transmutation of minor actinides. J Nucl Mater 2003; 320: 18. 17. Oigawa, H., et al. Conceptual design of transmutation experimental facility. Proc International Conference on Back-End of the Fuel Cycle, GLOBAL 2001, Sep. 10– 13, 2001, Paris, France.

16 Immobilisation of high-level radioactive waste from nuclear reactor fuel E R V A N C E and B D B E G G, ANSTO, Australia

16.1

Summary

Spent nuclear power plant fuel and the waste fission products and actinides from the reprocessing of nuclear fuels for commercial power or weapons production are classed as high-level waste (HLW). A brief history of the technical development of immobilisation strategies for high-level wastes and the desirable performance characteristics of the waste-immobilising solids (waste forms) are given. The pros and cons of different types of waste forms – borosilicate, phosphate, and other glasses; silicate, aluminate, phosphate, and titanate ceramics; glass-ceramics; cements and geopolymers; as well as spent fuel itself – are outlined, together with common production methods – melting, sintering, and hot uniaxial or isostatic pressing. Some of the issues in the fundamental science of waste form behaviour in a repository are presented. Geological disposal scenarios, together with some of the political and ethical issues inherent in waste disposal, are discussed. Likely future developments in waste form science and technology, including the impact of waste form research on nuclear fuel improvements, are put forward. At present, this factor is having some limited influence on the disposition of excess plutonium in inert matrix fuels. The use of transmutation via fast reactors or accelerators is not seen as an inexpensive, short-, or long-term solution to the disposition of radioactive waste. The future of HLW disposition around the world is discussed. The material presented is focused on activities taking place in the US but this focus is driven largely by the facts that: (i) the US has had the world’s longest nuclear programs in terms of both power production and military applications and (ii) the US therefore has a large fraction of the worldwide HLW inventory.

16.2

Generation of high-level waste from nuclear fuel

The first nuclear reactor was built by Enrico Fermi’s team at Chicago University in 1942. Nuclear power was first utilised for weapons production in the US 380

Immobilisation of high-level radioactive waste

381

during World War II and stockpiling nuclear weapons, principally by the US and Russia, has continued in parallel with the progressive development of peaceful applications of nuclear power for commercial electricity production in many countries after the war. Nuclear power derives from the neutron-induced fission of 235U or certain (reactor-produced) transuranic nuclides, such as 239Pu. Each fission event produces nearly 200 MeV of energy, manifested as kinetic energy of two fission product nuclei of unequal atomic numbers and masses plus several fast neutrons and gamma rays. If the neutrons are suitably moderated by slowing them down without their being captured by other nuclei, a critical mass and concentration of fissionable elements will give rise to a controlled chain reaction and produce controlled power, as distinct from an atomic bomb. Many of the fission products are highly radioactive. Figure 16.1 shows the relative distributions of the fission product abundances. The buildup of actinides (transuranic elements) derived from successive neutron-capture reactions depends non-linearly on the total burn-up of the fuel. Table 16.1 indicates the halflives of some of the key radioactive fission products and actinides. Only a few percent of the fissionable nuclei in nuclear fuel actually are fissioned during its useful life. This is because some of the fission products

103 Fission product total activity

102 10

Actinides total activity

1 10–1 10–2 10–3 10–4 10–5 10–6

1

10 102 103 104 105 106 107 Time (years)

16.1 Time dependence of radioactivity of reprocessing waste from commercial power plant nuclear fuel.

382

Materials for energy conversion devices Table 16.1 Half-lives of key fission product radionuclides Isotope

Half-life (y)

Isotope

Half-life (y)

Fission products

90

Sr Nb 129 I

30 2.4 × 104 1.6 × 107

93

Zr Tc 137 Cs

1.5 × 106 2 × 105 30

Actinides

235

7 × 108 87 6.5 × 103 3 × 106

238

U Pu 241 Pu 244 Cm

4 × 109 2.4 × 104 14 18

94

U Pu 240 Pu 237 Np 238

99

239

are strong absorbers of neutrons and inhibit the chain reaction, with the result that the fuel is no longer capable of producing significant amounts of nuclear energy. Thus, the general idea in the early days of nuclear power was to reprocess chemically the used (or spent) fuel to separate out the waste fission products so that the uranium (plus the Pu produced) could be recycled to make more fuel. In the mid-1970s, the future of reprocessing for nuclear power plant fuel was thrown into doubt when inexpensive uranium became widely available. Indeed, today the US, unlike France, for instance, does not reprocess commercial reactor fuel.

16.2.1 Types of radioactive waste There are various classifications of radioactive waste: (i) short-lived and long-lived low-level wastes (LLWs), (ii) intermediate-level waste (ILW), and (iii) high-level waste (HLW). The most dangerous waste from largescale nuclear origin is high-level waste arising from the reprocessing of spent fuel or the spent fuel itself. The definition of HLW relates to its method of production and so it cannot be converted to low-level waste by dilution. Thus, it is forbidden to dispose of it in the oceans for instance. Activities of this kind of waste are typically 1000 Ci/L. (Ci = Curie, the activity of 1 g of 226 Ra; 3.7 × 1013 disintegrations/second (Becquerels, Bq)). In addition to spent fuel and reprocessed waste from power plant fuel, HLW exists in other forms. In military applications, i.e., Pu production, burn-ups are relatively small (otherwise, the Pu is converted to higher actinides) and the wastes often are diluted strongly by process chemicals used in the chemical extraction of the Pu. Over the years, several methods have been used to separate out Pu for military use, which has increased the general variability of HLW from a chemical point of view. Typically, the radioactivity per unit volume of military wastes is only ~0.1–1.0% of that of reprocessed nuclear power plant fuel. Other types of waste are low-level waste from reactor operations, decontamination of radioactive samples, hospital wastes, spent sources, mining operations, etc. Intermediate-level wastes also are recognised in some countries.


383

The International Atomic Energy Agency (IAEA) is a good source of formal definitions of radioactive wastes around the world.1 Nuclear waste becomes less radioactive the longer it is stored owing to radioactive decay but some of its radioactivity persists for millions of years. Figure 16.2 shows the approximate time dependence of the activity for commercial reprocessing waste. Military wastes follow approximately the same time dependence but, as mentioned above, the activities are considerably less than those derived from commercial nuclear fuel. Note that these wastes are chemically and radiologically very diverse in nature. Since the activity of the waste falls with increasing time, it is technically advantageous to store it as long as possible. However, the method of storage is critical. For instance, a strong initial driver of HLW clean-up in the US derives from the history of the Hanford reservation in Washington state, where the stainless steel tanks containing the old military wastes leaked into the surrounding environment. Moreover, tanks in which the water largely had evaporated over the years of storage owing to radiogenic heating evolved gas in the form of periodic large hydrogen bubbles. This had safety implications associated with: (i) potential radionuclide removal from the tank into the atmosphere and ground area adjacent to the tanks, and (ii) potential ignition and explosion. All countries generating nuclear power and/or nuclear weapons produce HLW but the

10

14 MeV

Fission yield (%)

1

0.1

0.01 Thermal 0.001

0.0001 70

90

110 130 Mass number

150

16.2 Relative atomic abundances of fission products and transuranic elements in commercial power plant nuclear fuel.

384


urgency of particular countries to dispose of the HLW is a strong function of the maturity of their programs.

16.3

Historical waste form development for reprocessing HLW

It soon was realised that the highly radioactive spent power reactor fuels, whether reprocessed or not, would need careful management to prevent the spread of the unwanted radioactivity from fission products and transuranic elements into the public domain – the biosphere. The reference disposal scenario is a deep geological repository and this strategy will be discussed subsequently. The waste form is the primary immobilisation barrier that isolates the waste from the biosphere, so it is a key plank in the whole immobilisation process. The waste form is subject to laboratory investigation, so its performance can be optimised and validated. Borosilicate glasses for the immobilisation of HLW from nuclear fuel reprocessing were developed by the US Atomic Energy Commission (AEC) in the 1950s and were scaled up in the late 1960s to the full size dictated by the standard US disposal canisters, which are 3 m high × 0.61 m outer diameter. The scientific basis for their use generally was not articulated in detail during this era but the concept was that: (i) fission products generally were soluble in such glass, (ii) the glass could be produced easily in large quantities by melting at modest temperatures (~1100 °C) in a Joule-heated melter, and (iii) self-radiation damage from the decay of the incorporated radionuclides had little effect on the major properties of the glasses. The glasses could accommodate ~20 wt% fission products and actinides. However, Pennsylvania State University (PSU) workers in the mid-1970s noted that glasses were unstable fundamentally from a thermodynamic point of view. Consequently, they devised ceramic waste forms for the reprocessing of HLW based on the known natural longevity of crystalline silicate, phosphate, and molybdate minerals. These so-called ‘supercalcine ceramics’2 were sintered in air at ~1100 °C and had very high loadings of fission products – typically 70 wt% fission product oxides – and the chemistry of the different phases was driven by the fission products as majority components. Typical phases were pollucite (CsAlSiO4), powellite (CaMoO4), and rare earth apatites and phosphates (e.g., monazite (REPO4), where RE = trivalent rare earth). All of these had mineral analogues that were known to be very durable in the hot and wet conditions likely to characterise a deep geological repository for the waste. Following work at Sandia Laboratories in the US on phase assemblages occurring upon the heating of sol-gel titania particles on which simulated HLW fission products and actinides were sorbed, Ringwood and his coworkers3 in Australia in the late 1970s devised multi-phase titanate ceramics,


385

termed ‘synroc’, in which nearly all the fission products and actinides in HLW from nuclear fuel reprocessing were incorporated substitutionally in the various mineral analogue phases. Typical waste loadings were ~20 wt% HLW oxides and the production technology was: (i) slurry mixing of the waste and precursor oxides, (ii) calcination of the waste/precursor mixture in a reducing atmosphere, and (iii) hot uniaxial pressing at ~1100 °C to make a dense ceramic. These ceramics will be discussed in more detail subsequently. At around the same time, there was a worldwide surge in interest in this topic. However, in the US, a key decision was made during 1981–82 to use borosilicate glass to immobilise HLW at the Savannah River Defense Waste Processing Facility (DWPF), South Carolina site, accompanied by a substantial decrease in US funding for HLW waste form research from then on. Nevertheless, a variety of alternative waste form development work was taking place around the world; the book by Ewing and Lutze4 gives an excellent survey up to nearly the end of the 1980s. Candidate materials included glasses, ceramics, glass-ceramics, cermets, coated materials, and cements. However as time progressed, it has been agreed widely, but not universally, that the only real remaining candidate types of material for HLW immobilisation are glasses, ceramics, and glass-ceramics. In principle, these can be produced by either Joule or cold-crucible melters, sintering, hot uniaxial pressing, and hot isostatic pressing. Waste form development is still continuing in some shape or form in different countries with nuclear programs, although Japan chose borosilicate glass in the mid-1990s and thereafter ceased work on alternatives except in some niche areas, such as immobilisation of 129I. France instituted the ‘Law of 1991’, which placed a moratorium on waste disposal until 2006, giving them 15 years of research to make a decision on the best choices of waste forms for their particular wastes. Spent nuclear fuel itself also has been studied in the waste form context since the late 1970s. At this stage, it is appropriate to reiterate the diverse nature of HLW, depending in part on whether it derives from commercial Purex reprocessing or military Pu production. Generally speaking, the wastes consist of a concentrated solution of salts plus a sludge of hydroxides, a mixture that is very inhomogeneous and largely uncharacterised, even in single tanks. Hence, there is a potential need to separate individual wastes into solution and sludge fractions and there is a definite need to design waste forms that can cope with diversity and compositional uncertainty. Moreover, it must be recognised that different wastes require different technical solutions, for both the chemical design of the waste form and its mode of processing.

386


16.3.1 Desirable performance characteristics of high-level nuclear waste forms It is imperative that waste forms have very high chemical durabilities in terms of resistance to leaching by water. The durability of the waste form can be subjected to laboratory study and then optimised. The performance of a waste form is based primarily on its resistance to aqueous dissolution. The effect of self-irradiation (see subsequently) must be taken into account in these determinations. The higher the proportion of waste that can be incorporated per unit volume of the waste form, the less repository space will be needed and so the costs will be minimised. The waste form needs to be processed easily and reliably in a remote environment and minimisation of secondary wastes, such as off-gases during the production of the waste form, is important. Having established a given waste form for a given HLW chemical composition, it is important that the waste form properties are flexible and not unduly compromised due to: (i) mismatches of waste/additives ratios and (ii) variations of waste form chemistry, noting that HLWs are almost always inhomogeneous mixtures of solutions and sludges, calcines, etc. Flexibility derives from the use of multiple phases and chemical buffering via the presence of a phase that does not include radionuclides. In this case, variations in chemical composition result merely in a change in the proportions of the phases present, not the identities of the phases themselves. The US is the leader in waste performance acceptance criteria and, beginning in the late 1970s, a battery of aqueous tests designed for glass waste forms was undertaken at the Materials Characterization Center, which is associated with the Pacific Northwest National Laboratory (PNNL), WA. The most popular of these is the so-called MCC-1 test.5 Here, a polished cylindrical or cuboid sample of ~2 cm2 in surface area is immersed in ~ 20 mL of deionised water in a closed container and leached without agitation for a given time. In this test, a satisfactory candidate glass will yield a normalised leach rate of

Materials for energy conversion devices

Materials for energy conversion devices

Fluorinated Materials for Energy Conversion

Nanostructured Materials for Solar Energy Conversion

Nanostructured Materials for Solar Energy Conversion

Molecular Catalysts for Energy Conversion

Molecular Catalysts for Energy Conversion

Energy Conversion (Mechanical Engineering)

Data Conversion Handbook (Analog Devices)

Fundamentals of Thermophotovoltaic Energy Conversion

Energy conversion: electromechanical, direct, nuclear

Dielectric Materials and Devices

Materials for High-Temperature Semiconductor Devices

Handbook of Materials for Medical Devices

High Energy Density Materials

High Energy Materials

Smart Materials for Energy, Communications and Security

Spintronics: Materials, Applications, and Devices

Materials for Electrochemical Energy Conversion and Storage (Ceramic Transactions, Vol. 127) (Ceramic Transactions Series)

Biohydrogen III Renewable Energy System - Biological Solar Energy Conversion

Third Generation Photovoltaics: Advanced Solar Energy Conversion

Third Generation Photovoltaics Advanced solar energy conversion

Transport and Energy Conversion in the Heliosphere

Ferroelectric Devices (Materials Engineering, 16)

Advances in Wind Energy Conversion Technology

Renewable Energy Conversion, Transmission, and Storage

Energy: Production, Conversion, Storage, Conservation, and Coupling

Conversion

Conversion

Nanotechnology and Nanoelectronics: Materials, Devices, Measurement Techniques

Nanotechnology and Nanoelectronics: Materials, Devices, Measurement Techniques

Materials for energy conversion devices