Materials Aspects in Automotive Catalytic Converters

Material Aspects in Automotive Catalytic Converters Edited by Hans Bode

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

Further Titles of Interest: B. Cornils, W. A. Herrmann, R. Schlägl, C.-H. Wong (Eds.) Catalysis from A-Z ISBN 3-527-29855-X S. Hagen, S. Hawkins Industrial Catalysis ISBN 3-527-29528-3 S. M. Thomas, W. J. Thomas Principles and Practice of Heterogenous Catalysis ISBN 0-471-29239-X G. Ertl, H. Knözinger, S. Weitkamp Handbook of Heterogenous Catalysis ISBN 0-471-29212-8

Material Aspects in Automotive Catalytic Converters

Edited by Hans Bode

Deutsche Gesellschaft für Materialkunde e.V.

Prof. Dr. Ing. Hans Bode Bergische Universität GH Wuppertal FG Werkstofftechnik Gaußstr. 20 D-42097 Wuppertal Germany

International Congress „Material Aspects in Automotive Catalytic Converters“, held from 03–04 October 2001 in Munich, Germany Organizer: DGM · Deutsche Gesellschaft für Materialkunde e.V.

This book was carefully produced. Nevertheless, authors, editor and publisher do not warrant the information contained therein to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate.

Library of Congress Card No. applied for. A catalogue record for this book is available from the British Library Deutsche Bibliothek Cataloguing-in-Publication Data: A catalogue record for this book is available from Die Deutsche Bibliothek ISBN 3-527-30491-6 © WILEY-VCH Verlag GmbH, Weinheim (Federal Republic of Germany), 2002 Printed on acid-free paper All rights reserved (including those of translation in other languages). No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Satz: W.G.V. Verlagsdienstleistungen GmbH, Weinheim Druck: betz-druck GmbH, Darmstadt Bindung: J. Schäffer GmbH + Co. KG, Grünstadt Printed in the Federal Republic of Germany

Preface Based on increased ecological demands, the car and car-supplying industries strive to meet the challenging requirements for higher performance and extended service life of future vehicle generations. Maintaining good performance is mandatory particularly in the view of thinner supports, higher cell densities and higher temperatures. Performance and service life predictions, based on tests or on modelling and simulation techniques, will depend on reliable materials data. Only very close cooperation between researchers and producers will help to meet these requirements. It was the aim of MACC, the second international conference on Materials Aspects in Automotive Catalytic Converters, to foster this cooperation. It refered to papers from both industry and research institutes which concentrate on the high-temperature mechanical and oxidation behaviour of both metal-supported and ceramic supported automotive catalysts. The metalsupported catalyst is based on a ferritic steel with 5–8% aluminum, 17–22% chromium and small additions of reactive elements. More than 11,000,000 units were produced in the year 2000. The ceramic-supported catalytic converter is based on corderite. The production rate is much higher. Both materials have specific advantages and disadvantages which determine the application for a given car model. In addition to these two basic groups of catalytic carriers, coating and canning aspects were also addressed by the conference programme. Especially the influence of coating thickness and composition is becoming more and more important when going to thinner supports and higher cell densities. I am very obliged to the authors for their valuable contribution to a comprehensive programme that covers the whole chain of product development and application, beginning with the melting process and ending with recycling aspects. Munich, October 2001 Prof. Dr.-Ing. Hans Bode Conference Chairman

I Introduction Contribution of Automotive Catalytic Converters R. Searles, Association for Emissions Control by Catalyst, Brussels (B) ...................................3

II Metals Development Status of Metal Substrate Catalysts R. Brück, Emitec GmbH, Lohmar .............................................................................................19 Materials Issues Relevant to the Development of Future Metal Foil Automotive Catalytic Converters J. Nicholls, Cranfield University, Cranfield (GB); W. Quadakkers, Forschungszentrum Juelich (D) ................................................................................................................................31 High Temperature Corrosion of FeCrAlY / Aluchrom YHf in Environments Relevant to Exhaust Gas Systems A. Kolb-Telieps, Krupp VDM GmbH, Altena (D); R. Newton, Cranfield University, SIMS, Cranfield (GB); G. Strehl, TU Clausthal, Institut für Allgemeine Metallurgie Clausthal-Zellerfeld (D); D. Naumenko, W. Quadakkers, Forschungszentrum Jülich, IWV-2, Jülich (D) .....................................................................................................................49 Improved High Temperature Oxidation Resistance of REM Added Fe-20%Cr-5%Al Alloy by Pre-Annealing Treatment K. Fukuda, K. Takao, T. Hoshi, O. Furukimi, Technical Research Laboratories, Kawasaki Steel Corp., Chiba (Japan) ......................................................................................59 Oxidation Induced Length Change of Thin Gauge Fe-Cr-Al Alloys C. Chang, L. Chen, B. Jha, Engineered Materials Solutions, Inc., Attleboro (USA) ................69 Improvement in the Oxidation Resistance of Al-deposited Fe-Cr-Al Foil by Pre-oxidation S. Taniguchi, T. Shibata, Department of Materials Science and Processing, Graduate School of Engineering, Osaka University, Osaka (J); A. Andoh, Steel and Technology Development Laboratories, Nisshin Steel, Osaka (J) ...........................................83 Factors Affecting Oxide Growth Rates and Lifetime of FeCrAl Alloys W. Quadakkers, L. Singheiser, D. Naumenko, Forschungszentrum Jülich (D); J. Nicholls, J. Wilber, Cranfield University, School of Industrial and Manufacturing Science, Cranfield (GB) ...........................................................................................................93 On Deviations from Parabolic Growth Kinetics in High Temperature Oxidation G. Borchardt, G. Strehl, Institut für Metallurgie, TU Clausthal (D) ......................................106

VIII Effect of Reactive Elements and of Increased Aluminum Contents on the Oxide Scale Formation on Fe-Cr-Al Alloys V. Kolarik, M. del Mar Juez-Lorenzo, H. Fietzek, Fraunhofer-Institut für Chemische Technologie, Pfinztal (D); A. Kolb-Telieps, H. Hattendorf, R. Hojda, Krupp VDM GmbH, Werdohl (D) .....................................................................................................117 High Temperature Strength of Metal Foil Materials M. Cedergren, K. Göransson, R&D, AB Sandvik Steel, Sandviken (S) ..................................126 Lifetime Predictions of Uncoated Metal-Supported Catalysts via Modeling and Simulation, based on Reliable Material Data H. Bode, University of Wuppertal (D); C. Guist, BMW AG, Munich (D) ..............................134 Elastic-Plastic Thermal Stress Analysis for Metal Substrates for Catalytic Converters S. Konya, A. Kikuchi, Nippon Steel Corporation, Futtsu (J) ..................................................144 A New Type of Metallic Substrate R. Lylykangas, H. Tuomola, Kemira Metalkat Oy (SF) ..........................................................152

III Ceramics Development Status of Ceramic Supported Catalyst C. Vogt, E. Ohara, NGK Europe GmbH; M. Makino, NGK Insulators Ltd ...........................173 Evaluation of In-Service Properties and Life Time of Automotive Catalyst Support Materials U. Tröger, M. Lang, Zeuna Stärker GmbH & Co. KG, Augsburg (D) ...................................186 Loads, Design and Durability Evaluation of Mount Systems for Ceramic Monoliths G. Wirth, J. Eberspächer GmbH & Co., Esslingen (D) ..........................................................191 High Performance Packaging Materials M. Vermoehlen, D. Merry, S. Schmid, Corning GmbH, Wiesbaden (D) ................................202

IV Catalysts Three-way Catalyst Deactivation Associated With Oil-Derived Poisons J. Kubsh, Engelhard Corporation, Environmental Technologies Group Iselin (USA) ..........217 Catalytic Reduction of NOx in Oxygen-rich Gas Streams, Deactivation of NOx StorageRaduction Catalysts by Sulfur C. Sedlmair, K. Sehan, Technische Universität München, Institut für Technische Chemie II, Garching (D); J. Lercher, A. Jentys, University of Twente, Faculty of Chemical Technology, Enschede (NL) ...................................................................................223

IX Catalytic Reduction of NOx in Oxygen-rich Gas Streams: Progress and Challenges in Catalyst Development W. Grünert, Lehrstuhl Technische Chemie, Ruhr-Universität Bochum (D) ...........................229 Atomic Structure of Low-Index CeO2 Surfaces H. Nörenberg, University of Oxford, Department of Materials, Oxford (GB); J. Harding, University College London, Department of Physics and Astronomy, London (GB); S. Parker, University of Bath, Department of Chemistry, Bath (GB) .....................................237 Nanostructured Ceria-Zirconia as an Oxygen Storage Component in 3-way Catalytic Converters-Thermal Stability B. Djuricic, Austrian Research Centers, Seibersdorf (A), S. Pickering, Institute for Advanced Materials, Petten (NL) ...........................................................................................241

V Recycling Recycling Technology for Metallic Substrates: a Closed Cycle C. Hensel, Demet Deutsche Edelmetall Recycling AG & Co. KG, Alzenau (D) ....................251

VI Miscelleanous Hot-Corrosion of Metal and Ceramic Honeycombs by Alkaline Metals for NOx Adsorption M. Yamanaka, Nippon Steel Technoresearch, Futtsu (J); Y. Okazaki, Nippon Steel, Toukai, (J) ..............................................................................................................................263 The Effect of Trace Amounts of Mg in FeCrAl Alloys on the Microstructure of the Protective Alumina Surface Scales P. Untoro, M. Dani, National Nuclear Energy Agency, Kawasan PUSPIPTEK, Serpong (Ind); H. Klaar, J. Mayer, Gemeinschaftslabor für Elektronenmikroskopie, RWTH Aachen (D); D. Naumenko, J. Kuo, W. Quadakkers, Institut für Werkstoffe und Verfahren der Energietechnik (IWV-2), Forschungszentrum Jülich (D) ................................271

Subject Index* A Adsorption, NOx 263 Al deposition 83 Alkaline metals 263 Alloys 271 Aluchrom YHf 49 Alumina surface 271 Aluminum content 117 Annealing 59 Atomic structure 237 Automotive catalysts 3, 31, 186 C Catalyst 134 - ceramic 173 - three-way 217, 241 - deactivation 217 - development 19, 229 - support materials 186 Catalytic converters 3, 31, 144, 241 Catalytic reduction 223, 229 CeO2 surfaces 237 Ceramic components 241 Ceramic honeycomb 263 Ceramic monoliths 191 Ceramic supported catalyst 173 Ceria-Zirconia 241 Corrosion 49, 263

Elements, reactive 117 Emission limits 152, 202, 251 Environmental protection 251 Exhaust gas systems 49, 152 F Fe-Cr-Al alloy 59, 69, 83, 93, 117, 271 FeCrAlY/Aluchrom YHf 49 Foil 31, 83, 126 G Gas flow 152 Gas streams 223, 229 Gas systems 49 Growth kinetics 106 H High temperature - corrosion 49 - oxidation 59, 106 - strength 126 Honeycomb 263 Hot-corrosion 263 I Increased aluminum content 117 In-service properties 186 K

D

Kinetics 106

Data, reliable 134 Deactivation 217, 223 Development 173 - catalysts 19, 229 - converters 31 Durability evaluation 191

L Length change 69 Lifetime 93, 186, 251 Lifetime predictions 134 Loads 191 Low-index CeO2 surfaces 237

E Elastic thermal stress 144 *

The page numbers refer to the first page of the article

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

281 M Mat, vermiculite 202 Material - data, reliable 134 - issues 31 - packaging 202 - strength 126 Metal foil 31, 126 Metal honeycomb 263 Metallic substrate 152, 251 Metals, alkaline 263 Metal substrate catalysts 19, 144 Metal-supported catalysts 134 Mg 271 Microstructure 271 Mixed gas flow 152 Modeling 134 Monoliths, ceramic 191 Mount systems 191 N Nanostructure 241 NOx - adsorption 263 - reduction 223, 229 - storage 223 O Oil-derived poisons 217 Oxidation 69, 106 Oxidation resistance 59, 83 Oxide growth rates 93 Oxide scale formation 117 Oxygen-rich gas streams 223, 229 Oxygen storage 241 P Packaging materials 202

Parabolic growth kinetics 106 Plastic thermal stress 144 Poisons, oil-derived 217 Pre-annealing treatment 59 Pre-oxidation 83 Protective surface 271 R Reactive elements 117 Recycling technology 251 Reliable material data 134 REM 59 S Separation process 251 Simulation 134 Storage-reduction catalysts 223 Stress analysis 144 Structure, atomic 237 Substrate 19, 144, 152 Sulfur, NOx reduction 223 Surfaces 237, 271 T Thermal stability 241 Thermal stress analysis 144 Thin alloys 69 Three-way catalyst 217, 241 Trace amounts, alkaline 271 U Uncoated metal-supported catalysts 134 V Vermiculite mat 202

I Introduction

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

Contribution of Automotive Catalytic Converters Robert A Searles Association for Emissions Control by Catalyst, Brussels, Belgium

1

Abstract

Catalyst-equipped cars were first introduced in the USA in 1974 but only appeared on European roads from 1985. In 1993 the European Union set new car emission standards that effectively mandated the installation of emission control catalysts on gasoline-fuelled cars. Now more than 300 million of the world’s over 500 million cars and over 85% of all new cars produced worldwide are equipped with autocatalysts. Catalytic converters are also increasingly fitted on heavy-duty vehicles, motorcycles and off-road engines and vehicles. The paper will review the technologies available to meet the exhaust emission regulations for cars, light-duty and heavy-duty vehicles and motorcycles adopted by the European Union for implementation during the new century. This includes low light-off catalysts, more thermally durable catalysts, improved substrate technology, hydrocarbon adsorbers, electrically heated catalysts, DeNOx catalysts and adsorbers, selective catalytic reduction and diesel particulate traps. The challenge is to abate the remaining pollutants emitted while enabling fuel-efficient engine technologies to flourish. This is paramount to the achievement of air quality and greenhouse gas targets given the large increase in the number of vehicles on European roads since 1970 and the projections for further increases in vehicle numbers and greater distances driven each year in future.

2

Introduction

AECC is an international association of European companies making the technologies for automobile exhaust emissions control: autocatalysts, ceramic and metallic substrates, specialty materials incorporated into the catalytic converter and catalyst, adsorber and filter based systems for the control of gaseous and particulate emissions from diesel and other lean burn engines. 2.1

European Emission and Fuel Legislation

The European Union (EU) emission limits for passenger cars set from 1993 have already been lowered from 1996 and again from 2000. For passenger cars and light commercial vehicles the emission standards and fuel composition, including sulfur levels, have been agreed for 2000 and 2005. [1] New test cycles (ESC and ETC) and tougher emission standards for heavy-duty diesel vehicles have been finalized for 2000 and 2005. The limits for Enhanced Environmentally.Friendly Vehicles (EEV) are set and can serve as a basis for fiscal incentives by EU Member States. A

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KgaA ISBN: 3-527-30491-6

4 further reduction in limit values for nitrogen oxides (NOx) in 2008 is subject to a review by the European Commission in 2002 on technical progress. [2] The Working Party on Pollution and Energy (GRPE), an expert group of the World Forum for Harmonization of Vehicle Regulations (WP.29) at UN-ECE in Geneva, is developing a Worldwide Harmonized Heavy Duty Certification procedure and is looking in new measurement protocols in order to ensure that ultra fine particles are controlled by future emission legislation to minimize the health effects of diesel particle emissions. A proposal by the European Commission to set tougher, catalyst-requiring emission limits for motorcycles is being ratified by the European Parliament and Council. Tighter emission limits from 2003 for new types of motorcycles are agreed and correspond to a reduction of 60% for hydrocarbons and carbon monoxide for four-stroke motorcycles, and 70% for hydrocarbons and 30% for carbon monoxide for two-stroke motorcycles. A second stage with new mandatory emission limits for 2006 are expected to be based on the new World Motorcycle Test Cycle (WMTC) which is also being developed by the UN-ECE in Geneva. In the final report of the European Auto Oil II Programme [3], it was concluded that some air quality problems, such as atmospheric levels of particulate matter and ozone, are not yet solved. The challenge is to abate the remaining pollutants emitted while enabling the development of fuelefficient engine technologies. This is paramount to the achievement of air quality and greenhouse gas targets given the large increase in the number of vehicles on European roads and the projections for further increases in vehicle numbers and greater distances driven each year in future. 2.2

Exhaust Emissions from Internal Combustion Engines

Exhaust emissions can be lowered by: · Reducing engine-out emissions by improving the combustion process and fuel management, or by changes to the type of fuel or its composition · Retrofitting catalytic converters and associated engine and fuel management systems if they are not original equipment · Decreasing the time required for the catalytic converter to reach its full efficiency · Increasing the conversion efficiency of catalysts · Storing pollutants during the cold start for release when the catalyst is working · Using catalysts and adsorbers to destroy nitrogen oxides under lean operation · Using particulate filters with efficient regeneration technology · Increasing the operating life of autocatalysts and supporting systems. This paper reviews all the above opportunities, except the first, from the standpoint of material requirements and will also look back into the history of the materials developed for catalytic converters with these requirements in mind.

3

A Brief History of Automotive Catalysts

The first reference to a catalytic converter known to AECC is a patent [4] published to a French chemist, Michel Frenkel, in 1909. The device uses a kaolin (china clay) “honeycomb” with 30 grams of platinum as the active catalytic material. (Figure 1) The patent describes

5 “deodorizing” the exhaust using air blown in by a fan. As far as is known the device was not put into commercial production at the time. No doubt the high loadings of platinum were a deterrent and there was no air pollution concerns in those early, carefree days of motoring.

Figure 1: 1909 Catalytic Converter invented by Michel Frenkel

Figure 2: Eugene Houdry in 1953 with a small prototype catalytic converter

The next report [5] of the concept of catalytic converters was in the 1920s. Another European invention, this time German, was taken to General Motors in the US and was described as a collection of wires and beads, again coated with platinum. The tests were at first a success with the device glowing red, but within seconds the catalyst had failed. This was because tetraethyl lead had been recently introduced in the US as an octane booster, but was at that time unknown in Europe. Lead poisons catalytic converters. The French engineer Eugene Houdry can be considered as the father of the modern catalytic converter. Born in France he moved to the US and invented a revolutionary method for cracking low-grade crude to high-octane gasoline – the “cat cracking” process. After the 1939–1945 war he set up the Oxy-Catalyst company and turned his attention to the health risks from the increasing volumes of automobile and industrial exhausts. In 1962, the year of his death, he patented the first modern catalytic converter (Figure 2). The modern history of the catalytic converter started with the developments that lead to the 1970 US Clean Air Act and the rate of invention has accelerated greatly. Excellent histories of the industry [6, 7, 8] have been published so only a summary will be covered here. The modern catalytic converter, based on platinum group metals deposited on a ceramic honeycomb base or monolith, was first patented in 1965 [9]. However the industrial use of catalysts was then dominated by catalysts deposited on pellet or bead supports. In the first years after 1974 when catalytic converters were used in the US and Japanese markets, both pellet (Figure 3) and monolithic converters (Figure 4) were used. The loss of catalyst material by attrition in pellet converters was largely overcome by reactor design. Early prototypes of ceramic honeycombs were made by two approaches: 1. Dipping paper in ceramic slurries, corrugating them and laying up a unitary structure, firing the composite and shaping 2. Calendaring a plastic material containing ceramic powders between grooved and plain rollers, rolling up into a unitary structure, firing the composite and shaping.

6

Figure 3: Schematic of pelletized catalytic con- Figure 4: “Cutaway” ceramic monolith catalytic converter verter

Both of these developments were ultimately replaced by extruded honeycomb substrates. These are based on cordierite (2MgO .2Al2O3 .5SiO2) and made from natural raw materials and a plastic material that is extruded to form a unitary structure with parallel fine channels and then fired to the final shape. These materials have high thermal shock resistance and high melting and softening points with higher attrition resistance and lower pressure losses than pellet converters, which they ultimately replaced. In the 1970s new ferritic steels became available that could be made into ultra thin foils, corrugated and then laid up to form a honeycomb structure. One such steel was developed at the Atomic Energy Research Establishment in Harwell, UK for “canning” Uranium 235 and was called Fecralloy. This name reflects the components of the alloy - Iron (Fe), Chromium (Cr), Aluminum (Al) and Yttrium (Y). The formation of a self-healing protective “skin” of alumina (Al2O3) allows the ultra-thin steels to withstand the high temperatures and corrosive conditions in auto exhausts. These materials also have high thermal shock resistance and high melting and softening points and facilitated the development of high cell densities with very low pressure losses.

a) b) Figure 5: a: Metallic substrate converter, b: Ceramic substrate converter

7 Further development of metallic and ceramic substrates (Figure 5a & 5b) is described in the next section.

4

Current Catalyst Technology for Emissions Control Autocatalysts

Oxidation catalysts convert carbon monoxide (CO) and hydrocarbons (HC) to carbon dioxide (CO2) and water and decrease the mass of diesel particulate emissions but have little effect on nitrogen oxides (NOx). Three-way catalysts (TWC) operate in a closed loop system including a lambda- or oxygen sensor to regulate the air-fuel ratio. The catalyst can then simultaneously oxidize CO and HC to CO2 and water while reducing NOx to nitrogen. These simultaneous oxidizing and reducing reactions have the highest efficiency in the small air-to-fuel ratio window around the stoichiometric value, when air and fuel are in chemical balance. 4.1

Fast light off catalysts

The catalytic converter needs to work as fast as possible by decreasing the exhaust temperature required for operation so that untreated exhaust is curtailed at the start of the legislated emissions tests and on short journeys in the real world. Changes to the type and composition of the precious metal catalyst (Figure 6) and to the thermal capacity of substrates (figure 7) have together effected big reductions in the required operating temperature and light off times have been reduced from one to two minutes down to less than 20 seconds. [10]

Figure 6: Effect of catalyst technology on light off temperature

Figure 7: Effect of substrate cell density on light off time

The introduction of the new generation platinum/rhodium (Pt/Rh) technology for current and future emission standards is a technically and commercially attractive alternative for current palladium (Pd) based technologies for high demanding applications in close-coupled and under floor positions using different cold start strategies. [11] 4.2

More thermally durable catalysts

Increased stability at high temperature allows the catalytic converter to be mounted closer to the engine and increases the life of the converter, particularly during demanding driving. Pre-

8 cious metal catalysts with stabilized crystallites and washcoat materials that maintain high surface area at temperatures around 1000°C are needed. Improved oxygen storage components stabilize the surface area of the washcoat, maximize the air-fuel “window” for three-way operation and indicate the “health” of the catalytic converter for On Board.Diagnostic (OBD) systems. Figure 8 shows the progress made with mixed cerium and zirconium oxides. [12]

Figure 8: Improvements to thermal stability and oxygen storage capacity (OSC)

4.3

Substrate Technology

The technology of the substrates, on which the active catalyst is supported, has seen great progress. In 1974 ceramic substrates had a density of 200 cells per square inch of cross section (31 cells/square cm.) and a wall thickness of 0.012 inch or 12 mil (0.305 mm). By the end of the 1970’s the cell density had increased through 300 to 400 cpsi and wall thickness had been reduced by 50% to 6 mil. Now 400, 600, 900 and 1200 cpsi substrates are available and wall thickness can be reduced to 2 mil - almost 0.05 mm (Figure 9). [13, 14, 15, 16, 17] In the late 1970's substrates derived from ultra thin foils of corrosion resistant steels came onto the market. In the beginning the foils could be made from material only 0.05 mm thick allowing high cell densities to be achieved. Complex internal structures can be developed and today wall thickness is down to 0.025 mm and cell densities of 800, 1000 and 1200 cpsi are available (Figure 10). [18, 19] This progress in ceramic and metal substrate technology has major benefits. A larger catalyst surface area can be incorporated into a given converter volume and this allows better conversion efficiency and durability. The thin walls reduce thermal capacity and avoid the penalty of increased pressure losses. Alternatively the same performance can be incorporated into a smaller converter volume, making the catalyst easier to fit close to the engine, as cars get more compact. These improvements in substrate technology are now being applied in conjunction with heavy-duty diesel engines with catalysts placed as close as possible to the engine in order to.reduce the time to light off. To improve conversion behavior, catalysts are placed close to

9 the exhaust port before the turbocharger (Figure 11) and close-coupled catalysts using hybrid substrates are fitted (Figure 12). [20]

Figure 9: Progress in ceramic substrate design

Figure 10: Progress in metallic substrate design

Figure 11: Pre-turbo catalyst

Figure 12: Close-coupled hybrid catalyst

4.4

New Technology for Emissions Control Stoichiometric combustion

Conventional three-way catalysts are continually developed to improve high temperature stability and light off performance and to meet the demands of both the most challenging emission legislation in the world and new applications including motorcycles. Their performance can be further extended by the following additional technologies. 4.4.1 Hydrocarbon adsorbers Hydrocarbon adsorber systems incorporate special materials, such as zeolites, into or upstream of the catalyst. Hydrocarbon emissions are collected when exhaust temperatures are too low for effective catalyst operation. The hydrocarbons are then desorbed at higher temperatures when the catalyst has reached its operating temperature and is ready to receive and destroy the hydrocarbons. This technology has the potential to reduce hydrocarbons to less than half the levels emitted from a three-way catalytic converter (Figure 13). [21]

10

Figure 13: Influence of improved three-way catalyst and hydrocarbon adsorber on emissions (European cycle).

4.4.2 Electrically heated catalyst systems This uses a small catalyst ahead of the main catalyst. A metallic substrate, onto which the catalyst is deposited, allows an electric current to pass so it will heat up quickly. This brings the catalyst to its full operating temperature in a few seconds. [22] 4.5

Lean Combustion

With the development of lean burn direct injection gasoline engines and increased use of diesel engines, lean combustion is the challenge for automotive catalysis but is essential to reduce fuel consumption and limit CO2 emissions. New diesel technologies with electronic management and direct injection can achieve further fuel consumption improvements. Conventional three-way catalyst technology used on gasoline engines needs a richer environment with lower air-fuel ratios to reduce NOx so a radical new approach is required. DeNOx catalysts and NOx traps hold out the prospect of substantially reduced emissions of oxides of nitrogen. NOx conversion rates depend on exhaust temperature and availability of reducing agents. There are four systems under evaluation by industry: 1. Passive DeNOx Catalysts using reducing agents available in the exhaust stream 2. Active DeNOx Catalysts using added hydrocarbons as reducing agents 3. NOx traps or adsorbers used in conjunction with a three-way catalyst 4. Selective Catalytic Reduction using a selective reductant, such as ammonia from urea. Each of these systems offers different possibilities in the level of NOx control possible and the complexity of the system. Fuel parameters such as sulfur content can affect catalyst performance. 4.5.1 DeNOx (or Lean NOx) Catalysts DeNOx catalysts use advanced structural properties in the catalytic coating to create a rich "microclimate" where hydrocarbons from the exhaust can reduce the nitrogen oxides to nitrogen, while the overall exhaust remains lean. Further developments focus on increasing the operating temperature range and conversion efficiency. 4.5.2 NOx Adsorbers (or Lean NOx Traps) NOx traps are a promising development as results show that NOx adsorber systems are less constrained by operational temperatures than DeNOx catalysts. NOx traps adsorb and store NOx under lean conditions. A typical approach is to speed up the conversion of nitric oxide (NO) to nitrogen dioxide (NO2) using an oxidation or three-way catalyst mounted close to the

11 engine so that NO2 can be rapidly stored as nitrate. The function of the NOx storage element can be fulfilled by materials that are able to form sufficiently stable nitrates within the temperature range determined by lean operating points of a direct injection gasoline engine. Thus especially alkaline, alkaline earths and to a certain extent also rare-earth compounds can be used. When this storage media nears capacity it must be regenerated. This is accomplished in a NOx regeneration step. Unfortunately, alkaline and alkaline earth compounds have a strong affinity for sulfation. As a consequence alkaline and alkaline earth compounds are almost irreversibly poisoned by the sulfur contained in the fuel during the NOx storage operation mode, leading to a decrease in NOx adsorption efficiency during operation. The stored NOx is released by creating a rich atmosphere with injection of a small amount of fuel. The rich running portion is of short duration and can be accomplished in a number of.ways, but usually includes some combination of intake air throttling, exhaust gas recirculation, late ignition timing and post combustion fuel injection. The released NOx is quickly reduced to N2 by reaction with CO (the same reaction that occurs in the three-way catalyst for spark-ignited engines) on a rhodium catalyst site or another precious metal that is also incorporated into this unique single catalyst layer (Figure 14).

Figure 14: NOx adsorber working principle

Under oxygen rich conditions, the thermal dissociation of the alkaline and alkaline earth sulfates would require temperatures above 1000 °C. Such temperatures cannot be achieved under realistic driving conditions. However, it has been demonstrated in various publications [23, 24, 25] that it is in principle possible to decompose the corresponding alkaline earth sulfates under reducing exhaust gas conditions at elevated temperatures. In this way the NOx storage capacity can be restored. The heating of the catalyst, for example by late ignition timing, does however result in a considerable increase in fuel consumption, which is dependent upon the sulfur content. Therefore, reducing the sulfur concentration in the fuel must be regarded as the most effective way of using the full potential of modern direct injection gasoline engines with respect to fuel economy and CO2 reduction. One of the demands for a desulfation strategy must be to avoid any H2S emissions above the odor threshold during desulfation. [26, 27] Developments and optimization of NOx adsorber systems have been and are currently underway for diesel and gasoline engines. These technologies have demonstrated NOx conversion efficiencies ranging from 50 to in excess of 90 percent depending on the operating temperatures and system responsiveness, as well as fuel sulfur content. [28, 29] The system is in production with direct injection gasoline engines.

12 4.5.3 Selective Catalytic Reduction (SCR) SCR is a widespread technology to reduce nitrogen oxide emissions from coal, oil and gas fired power stations, marine vessels and stationary diesel engine applications. SCR technology has been used successfully for more than two decades. SCR technology for heavy-duty diesel vehicles has been developed to the commercialization stage and will be available as an option in the series production of several European truck-manufacturing companies in 2001. SCR technology permits the NOx reduction reaction to take place in an oxidizing atmosphere. It is called “selective” because the catalytic reduction of NOx with ammonia (NH3) as a reductant occurs preferentially to the oxidation of NH3 with oxygen. Several types of catalyst are used, the choice of which is determined by the temperature of the exhaust environment. For mobile source applications the reductant source is usually urea, which can be rapidly hydrolyzed to produce ammonia in the exhaust stream. SCR for heavy-duty vehicles reduces NOx emissions by circa 80%, HC emissions by circa 90% and PM emissions by circa 40% in the EU test cycles, using current diesel fuel ( is a probabilistic term that incorporates the fractional surface area that may spall during thermal cycling and any variance in the rate of oxidation (k) and its effect on the time to breakaway (tB/O). As spallation is rarely observed for foil components the term > takes values equal to, or close to, zero and component life is dominated by the second term on the left hand side (oxide growth) in equation (1). The first term on the left hand side of the equation accounts for oxide spallation, while the right hand side of the equation is a measure of the available aluminium reservoir, locally in the component (if V/A is the local volume/surface area ratio. This behaviour is illustrated in Figure 4, which provides model predictions (assuming parabolic kinetics) compared with breakaway oxidation lifetimes measured over a range of foil/alloy thicknesses from 70mm to 1.8mm (for rectangular geometry test pieces the V/A ratio is approximately half the sample thickness). The open symbols represent samples that have not yet gone into breakaway, while the filled symbols represent samples that show some evidence of breakaway corrosion, usually at one of the corners. Superimposed on this figure are the median predictions assuming a parabolic rate law and various failure criteria. The line marked ‘InCF’, with > = 0 and CB = 0wt%, corresponds to ‘Intrinsic Chemical Failure’. This is the most likely failure mode for foil material of low strength and results in the longest component lives at any foil thickness (V/A ratio). The lines marked ‘MICF’, with > = 0.1 or 1.0 and CB = 1.7 wt%, correspond to ‘Mechanical Induced Chemical Failure’; under these conditions either local spallation or tensile cracking of the alumina scale occurs, usually in areas of constraint, and the repair, cracking and spallation process rapidly depletes the available aluminium reservoir leading to early failure. The term ‘>’ is a measure of the extent of spallation, > = 1

37 being one limiting case whereby the alumina totally spalls at each shut down cycle. In the LEAFA experimental programme this was never observed to happen for the foil samples, but was approached for strong, thick alloy materials on rapid cooling. >=0.1 1.0E+04

MICF >=1.0

Aluchrom Yhf - 1300C Aluchrom Yhf - 1300C (Not.) Kanthal AF - 1300C

2

kp*tB [ (mm) ]

1.0E+03

Kanthal AF - 1300C (Not.) Kanthal APM - 1300C Kanthal APM - 1300C (Not.) PM2000 - 1300C PM2000 - 1300C (Not.)

1.0E+02

1.0E+01 0.01

C0=5.5 wt.% CB=1.7 wt.% (MICF) CB=0.0 wt.% (InCF)

InCF >=0.0 0.10

1.00

10.00

Volume/Surface Area [mm]

Figure 4: Model of breakaway oxidation, based on parabolic rate law oxidation : the lines super-imposed on the figure correspond to median prediction.

It can be seen from Figure 4 that the parabolic based model predictions prove conservative (predict shorter lives) for all foil materials tested, whether weak (Aluchrom YHf, Kanthal AF) or strong (PM2000). This is because a-alumina growth is generally observed to follow subparabolic kinetics [18,19] with an exponent ‘n’ between 2.3 and 3.0; such behaviour would lower the rate of aluminium consumption and therefore lead to extended component lives. It can further be seen from Figure 4, that > (the propensity to spall) is much less than 0.1 for all foil materials in current usage. This means that the dominant foil failure mechanism is ‘Intrinsic Chemical Failure’ and thus CB (the aluminium content at onset of breakaway) can be expected to reduce to essentially zero. 4.2

The Influence of Component Geometry on Catalytic Converter Body Life

It is evident from the foregoing modelling work that specimen thickness is a significant factor in determining the life of the converter matrix. To be exact it is the local volume/surface area ratio that is critical in determining the onset of chemical failure, that is why most rectangular test samples fail at corners. This aspect has particularly been addressed by Strehl et al [22] where it is shown that material thickness, sample shape and local constraints resulting from component geometry may have significant effects on component life. It is also shown that the local oxygen pressure adjacent to the components surface can be reduced by unfavourable geometries, such as crevices, and this can trigger early breakaway corrosion. Probably the most significant contribution of this study is the recognition that is the ‘local’ volume/surface area ratio that controls component life. A simple analysis for plate material, which is obvious once demonstrated, shows that the volume/surface area ratio at a corner will be one third of that for a free surface, while for an edge the volume/surface area ratio is re-

38 duced by a factor of two. For foil samples, these geometry factors are even more critical; when the foil is thin the aluminium concentration within the foil is in equilibrium with two free surfaces (hence the assumption that for an infinite foil of thickness ‘d’ the volume/surface area ratio is d/2), however, at a corner the local volume/surface area ratio approaches d/4, half of that for the bulk of the sheet, while along edges the value approached d/3. Thus the local aluminium reservoir is significantly reduced at edges and corners because of the change in the local volume/surface area ratio in this region and this accounts for the onset of breakaway oxidation usually being noticed at corners first. This has major implications in manufacturing components and emphasises the need for good design particularly at corners, edges and fixings if premature breakaway is to be avoided. However, the local aluminium reservoir is not the sole controlling parameter; for example two small closely spaced holes could be expected to give rise to a geometry where the local volume/surface area ratio results in premature breakaway, however, this does not occur because additional aluminium can diffuse to this region from the bulk of the component. Thus it is the balance between the local aluminium reservoir, its rate of consumption through oxidation and the rate of supply by diffusion from the bulk of the component that determines whether breakaway will occur or not. Geometry also modifies the constraint that the oxidising surface sees and this can alter the oxidation rate. Thin unconstrained foils are free to expand as a result of growth and thermal stress (length increases of 10% have been measured in oxidation measurements of thin foil [21]). Constrained foils, and thick section components, can generate sufficient stress at surface imperfections that scale cracking, can occur locally [22,23] thus oxidation rates in areas of high constraint may differ from those of unconstrained surfaces. This effect of constraint is illustrated in Figure 5, which plots the mass gain per unit area for various geometry components, including a model catalytic converter body, manufactured from a 58mm Aluchrom YHf foil oxidised at 1000 °C for 700h. The lowest mass gain was observed for the rectangular specimen which was free to deform. Both the ring sample and model catalytic converter body showed an increased mass gain, both of similar magnitude. The additional mass gain was associated with cracks observed in the oxide scale on both the ring samples and catalytic converter body. This cracking has the apparent effect of increasing the rate constant by some 60%, during these discontinuous oxidation studies. In practice, this will significantly reduce component life as component failure switches from intrinsic chemical failure (InCF) to mechanically induced chemical failure (MICF) – see Table 2, and Figures 4 and 9. Table 2 provides predictions of the medium component lives for FeCrAl foil material with 5wt% aluminium, based on the LEAFA model, assuming parabolic kinetics, for the cases of intrinsic chemical failure and mechanically induced chemical failure. In calculating the values for mechanical induced chemical failure it is assumed that the catalytic body is constrained (kp is increased by a factor of 1.6) and the alloy concentration at which alumina can no longer reform (CB) is 1.7wt%.

39

Figure 5: Influence of component geometry on the oxidation of Aluchrom YHf: 58mm foil at 1100 °C [22].

One can see from Table 2 that reducing the foil thickness from 50mm to 30mm reduces the life of the component by a factor of 2.77. However, this reduction is not as great as that imposed by the constraint of the cylindrical geometry, which has the effect of reducing the life by a factor of 3.39 over the unconstrained foil. Table 2: Median predicted component lives for FeCrAl-RE foils as a function of foil thickness and degree of constraint

Component life measured as the kp.tB product [mm2] Foil thickness Unconstrained foil Catalytic converter body 20mm 0.90 0.27 30mm 2.03 0.60 50mm 5.63 1.66 70mm 11.03 3.26 4.3 Changes in Operating Temperature, Brought About by Catalytic Converter Positioning

Current underfloor catalytic converter bodies see peak temperatures of 910 °C, while those for close coupled configurations are expected to see 1000 °C. Assuming that a-alumina is the stable oxide formed on the FeCrAl-RE catalytic converter bodies (it is at more elevated temperatures > 1050 °C), then this temperature increase will raise the oxidation parabolic rate constant by a factor of 5.2, from 2.3 × 10–14 g2 cm–4 s–1 at 910 °C to 1.2 × 10–13 g2 cm–4 s—1 at 1000 °C as can be seen from Figure 6 [24]. In terms of numbers suitable to insert into the above mentioned life model the associated scaling rates are 2.16 × 10–2 mm2 h–1 at 910 °C and 1.13 × 10– 1 mm–2h–1 at 1000 °C. Thus assuming a foil thickness of 50 mm, under unconstrained conditions, then the hot exposure time would be reduced from 260 hours to some 50 hours before chemical

40 failure occurs under peak load conditions. Constraint or thinner foil sections would reduce these lives still further. One conclusion, therefore, is that catalytic converters must spend a considerable period of normal operation way below these peak load temperatures, for under such peak load conditions the foil components would not be able to sustain the warranted lives of the vehicle. To achieve the desired warranted life the mean operating temperature of a current, underfloor, catalytic body would have to 780 °C, based on the above FeCrAl-RE life model. These modelling assumptions are based on the rate of growth of a-alumina scales, however, at operating temperatures in the range 780 °C–910 °C it is more likely that transition aluminas will form. This aspect has been extensively studied. One elegant study by Molins et al [25] on a Ugine Savoie alloy examined the oxidation kinetics over the temperature range 850 °C– 1100 °C in flowing synthetic air. The samples were foils 45mm thick. The results obtained are reproduced in Figure 7, as relative mass gain against time (the norm is taken to be 168h at 950 °C).

Figure 6: Arrhenius plot of the parabolic rate constant as determined from TGA studies, for various FeCrAl-RE alloys, over the temperature range 750-1350 °C [24].

41

Figure 7: Oxidation kinetics for a Fe20Cr5.2Al0.01Ce alloy over the temperature range 850 °C–1100 °C [25]

This research shows that two domains exists, at low temperatures (T=0.01 5% 0.1% >=0.1 5% 0.1% >=1.0 5% 0.1%

Aluchrom Yhf - 1300C

1.0E+03

Aluchrom Yhf - 1300C Kanthal AF - 1300C Kanthal AF - 1300C Kanthal AF - 1200C Kanthal APM - 1300C Kanthal APM - 1300C Kanthal APM - 1200C

1.0E+02

PM2000 - 1300C PM2000 - 1300C

1.0E+01 InCF > = 0.0

1.0E+00 0.001

C0=5.5 wt.% CB=1.7 wt.% (MICF) CB=0.0 wt.% (InCF)

0.010

0.100

Volume/Surface Area [mm]

1.000

10.000

PM2000 - 1200C PM2000 - 1200C (Improve) Aluchrom Yhf 1200C (Iso) Kanthal APM 1200C (Iso) PM2000 1200C (Iso)

Figure 9: Stochastic life prediction model for the chemical failure of FeCrAl-RE, alumina forming alloys [4]

46 Thus, chemical failure reflects a balance between the available aluminium reservoir and its rate of consumption due to a complex interplay of oxidation parameters. Key among these is the oxide rate constant, which may follow parabolic or sub-parabolic kinetics, and for thicker sectioned components, or highly constrained geometries, the critical oxide thickness to spall. It has been proposed that the (k × tB) product is a temperature independent parameter, that defines the lifetimes of an alumina forming ferritic steels, whether a foil or sheet materials [4]. Figure 9 presents a plot of (kp × tB) against V/A ratio for a range of alumina forming ferritic steels over the temperature range 1050 °C–1400 °C, confirm the hypothesis that the (kp × tB) product may be used to provide a temperature independent estimate of component life. In Figure 9 alloy lives between a few tens of hours and 20,000 h are plotted, superimposed on the plot are statistical corrosion models, that define the risk of failure [4], two levels of risk are plotted: a 5% chance of failure and a 0.1% chance of failure assuming oxidation follows parabolic kinetics. All foil samples were observed to fail by intrinsic chemical failure (InCF), when unconstrained. A life model, based on intrinsic chemical failure (> = 0.0; CB = 0.0wt%) and parabolic kinetics, provides a conservative estimate of foil component life for a range of FeCrAl-RE materials over the temperature range 1100 °C–1400 °C.

6

References

[1] World Energy Outlook, 2000, International Energy Agency. [2] W. Maus ‘Mobility, Prosperity and Environment Protection – the Catalytic Converter is Indispensable’, in “Metal-Supported Automotive Catalytic Converters” (ed. H. Bode) p3– 13, Werkstoff-Informationsgesellschaft, Frankfurt, Germany (1997). [3] H. Bode (ed) “Metal-Supported Automotive Catalytic Converters” (ed. H. Bode) p3–13, Werkstoff-Informationsgesellschaft, Frankfurt, Germany (1997). [4] J. R. Nicholls, R. Newton, M. J. Bennett, H. E. Evans, H. Al-Badairy, G. Tatlock, D. Naumenko, W. J. Quadakkers, G. Strehl and G. Borchardt, ‘Development of a Life Prediction Model for the Chemical Failure of FeCrAlRE Alloys in Oxidising Environments, “Life Modelling of High Temperature Corrosion Processes’, (eds M. Schutze, W. JH. Quadakkers and J. R. Nicholls) EFC, Publication 28, IoM Communications 2001. [5] H. Bodes ‘Development Status of Materials for Metal Supported Automotive Catalysts’, in “Metal-Supported Automotive Catalytic Converters” (ed. H. Bode) p3–13, WerkstoffInformationsgesellschaft, Frankfurt, Germany (1997). [6] B. H. Engler “Katalysatoren fur den Umweltschutz”, Chem-Ing.Tech. 63, 298–312 (1991); cited in reference 5. [7] T. Nagel and W. Maus “Development of More Exacting Test Conditions for close Coupled Converter Applications”, in Metal-Supported Automotive Catalytic Converters” (ed. H. Bode) p107–126, Werkstoff-Informationsgesellschaft, Frankfurt, Germany (1997). [8] E. Lang (ed) “The Role of Active Elements in the Oxidation Behaviour of High Temperature Materials and Alloys” Elsevier Applied Science, London (1989). [9] M. J. Bennett and G. W. Lorimer (eds), “Microscopy of Oxidation”, Institute of Metals, London (1991). [10] S. B. Newcomb and M. J. Bennett (eds), “Microscopy of Oxidation-2” Institute of Materials, London, 1993.

47 [11] D. Coutsouradis et al (eds) Proc. “Materials for Advanced Power Engineering”, COST 501, Kluwer Academic Publishers, Dordrecht, Netherlands (1994). [12] S. B. Newcomb and J. A. Little (eds) “Microscopy of Oxidation-3”, Institute of Materials, London, 1997. [13] D. A. Shore, R. A. Rapp and P. Y. Hou (eds), Int. Conf. on “Fundamental Aspects of High Temperature Corrosion”, The Electrochemical Soc., USA, (1997). [14] J. Lecomte-Beckers et al (eds), Proc. “Materials for Advanced Power Engineering 1998”, Forschungszentrum, Julich, Germany (1998). [15] M. Schutze and W. J. Quadakkers (eds) “Cyclic Oxidation of High Temperature Materials”, EFC Publication 27, IoM Communications, London, (1999). [16] G. Tatlock and S. Newcomb (eds), Special Issue of Material at High Temperatures on “Microscopy of Oxidation-4” Vol. 17(1) (2000). [17] M. Schutze, W.J. Quadakkers and J. R. Nicholls (eds) “Life-time Modelling of High Temperature Corrosion Processes”, EFC Publication 28, IoM Communications, London (2001). [18] W. J. Quadakkers and K. Bongartz, Werst. U. Korros. 24, 232, (1994). [19] W. J. Quadakkers, K. Bongartz and F. Schutbert in Proc. “Materials for Advanced Power Engineering”, COST 501, (eds D. Coutsouradis et al) part II, p1533, Kluwer Academic Publishers (1994). [20] J. R. Nicholls and M. J. Bennett, “Cyclic Oxidation-guidelines for Test Standardisation aimed at the Assessment of Service Behaviour”, European Federation of Corrosion Publications Vol. 27, pp437–470 (1999). [21] R. Newton, M. J. Bennett, J. P. Wilber, J. R Nicholls, D. Naumenko, W. J. Quadakkers, H. Al-Badiary, G. Tatlock, G. Strehl, G. Borchardt, A. Kolb-Telieps, B. Jonsson, A. Westerlund, V. Guttmann, M. Maier and P. Beaven, “The Oxidation Lifetime of Commercial FeCrAlRE Alloys” in ‘Life Modelling of High Temperature Corrosion Processes’, (eds. M. Schutze, W. J. Quadakkers and J. R. Nicholls) EFC Publication 28, IoM Communications 2001. [22] G. Strehl, V. Guttmann, D. Naumenko, A. Kolb-Telieps, G. Borchardt, J. Klower, P. Beaven, J. R.Nicholls, “The Influences of Sample Geometry on the Oxidation and Chemical Failure of FeCrAl(RE) Alloys”, Life Modelling of High Temperature Corrosion Processes’, (eds M. Schutze, W. J. Quadakkers and J. R. Nicholls), EFC Publication 28, IoM Communications, 2001. [23] H. Al-Badairy, G. J. Tatlock and J. LeCoze, “An Auger Study of Thermally Spalled Oxides on Fe-20Cr-5Al Based Alloys” in “Microscopy of Oxidation-3” (eds S. Newcomb and J. A. Little) p105, Institute of Materials, London (1997). [24] M. Gobel, J. Schimmelpfennig, A. Glazkow, G. Borchardt, “Growth of a-alumina Scales on Fe-Cr-Al Alloys” in Metal-Supported Automotive Catalytic Converters” (ed. H. Bode) p191, Werkstoff-Informationsgesellschaft, Frankfurt, Germany (1997). [25] R. Molins, A. Germidis and E. Andrieu “Oxidation of Thin FeCrAl Strips: Kinetic and Microstructural Studies in “Microscopy of Oxidation-3”, (ed. S. B. Newcomb and J. A. Little, p3, Institute of Materials, London (1997). [26] A. Kolb-Telieps, U. Miller, H. Al-Badairy, G. Borchardt, G. Tatlock, D. Naumenko, W. J. Quadakkers, G. Strehl, R. Newton, J. R. Nicholls and V. Guttmann, “The Role of Bioxidant Corrodants on the Life Time Behaviour of FeCrAlRE Alloys” ‘Life Modelling of

48 High Temperature Corrosion Processes’, (eds M. Schutze, W. J. Quadakkers and J. R. Nicholls), EFC Publication 28, IoM Communications, 2001. [27] G. J. Tatlock and H. Al-Badairy “The Oxidation of Thin Foils of FeCrAl-RE Alloys in Moist Air”, accepted for publication in Materials at High Temperatures (2001). [28] M. J. Bennett, R. Perkins, J. B. Price and F. Starr in Proc. “Materials for Advanced Power Engineering”, COST 501 (eds D. Coutsouradis et al), p1553, Kluwer Academic Publishers (1994). [29] J. Klower and A. Kolb-Telieps, “Effect of Aluminium and Reactive Elements on the Oxidation Behaviour of Thin FeCrAl Foils in “Metal-Supported Automotive Catalytic Converters” (eds H. Bode), p33, Werkstoff-Informationsgesellschaft, Frankfurt, Germany, (1997). [30] J. Klower, Materials and Corrosion, 49, 758 (1998). [31] A. B. Smith, A. Kempster and J. Smith, “Characterisation of Aluminide Coatings formed on Nickel Based Superalloys by Vapour Aluminising” in “High Temperature Surface Engineering” (eds J. R. Nicholls and D. S. Rickerby) p13, IoM Communications, London (2000). [32] L. Vandenbulcke, G. Leprince and B. Nciri, “Low Pressure Gas-Phase Pack Cementation Coating of Complex-Shaped Alloy Surfaces”, Materials Science and Engineering, A121, 379 ff, (1989). [33] W. J. Quadakkers, T. Malkow, H. Nickel and a. Czyrska-Filemonowics in Proc. 2nd Int. Conf on “Heat Resisting Materials”, Gatlinburg USA, p91, ASM International, Ohio, USA (1999). [34] J. G. Smeggil, “Some Contents on the Role of Y in Protective Oxide Scale Adherence”, Materials Science and Eng. 87, 261 (1987). [35] J. L. Smialek, “Sulphur Impurities and the Microstructure of Alumina Scales” in “Microscopy of Oxidation-3” (eds S. B. Newcomb and J. A. Little), p127, Institute of Materials, London, (1997). [36] B. Pint, Oxid. Met. 45, 1 (1996). [37] J. Klower and J. G. Li, Materials and Corrosion 47, 545 (1996). [38] P. A. Van Manen, E. W. A. Young, D. Schlakoord, C. J. an der Wekken and J. H. W. de Wit, Surface and Interface Analysis 12, 391 (1988). [39] W. J. Quadakkers and L. Singheiser, “Practical Aspects of the Reactive Element Effect”, in “High Temperature Corrosion”, Les Embiez, France, (May 2000). [40] D. Naumenko, W. J. Quadakkers, P. Beaven, H. Al-Badairy, G. Tatlock, R. Newton, J. R. Nicholls, G. Strehl, G. Borchardt, J. Le Coze, B. Jonsson, A. Westerlund, “Critical Role of Minor Elemental Constituents on the Lifetime Oxidation Behaviour of FeCrAl-RE Alloys”, “Life Modelling of High Temperature Corrosion Processes”, (Eds. M. Schutze, W. J. Quadakkers and J. R. Nicholls) EFC Publication 28, IoM Communications 2001. [41] H. Al-Badairy, G. Tatlock, H. E. Evans, G. Strehl, G. Borchardt, R. Newton, M. J. Bennett, J. R. Nicholls, D. Naumenko and W. J. Quadakkers, “Mechanistic Understanding of Chemical Failure for FeCrAl-RE Alloys in Oxidising Environments” in “Lifetime Modelling of High Temperature Corrosion Processes”, (eds M. Schutze, W. J. Quadakkers and J. R. Nicholls) EFC publication 28, IoM Communications, London (2001).

High Temperature Corrosion of FeCrAlY/Aluchrom YHf in Environments Relevant to Exhaust Gas Systems Angelika Kolb-Telieps1), Gernot Strehl2) , Dmitry Naumenko3), Willem. J. Quadakkers3) , Rachel Newton4) 1)

Krupp VDM GmbH, Kleffstr. 23, 58762 Altena, Germany TU Clausthal, Institut f³r Allgemeine Metallurgie, Robert-Koch-Str. 42, 38678 Clausthal-Zellerfeld, Germany 3) Forschungszentrum Juelich, IWV-2, 52425 J³lich, Germany 4) Cranfield University, SIMS, Cranfield, Bedfordshire, MK43 0AL, UK 2)

1

Abstract

Fe-20Cr-5.5Al-Y/Aluchrom YHf foils were exposed in air, N2 + 5000vppm NO, a simulated fuel-rich exhaust gas, air + SO2 and air + 50vppm HCl at 1200°C. N2 + NO and the exhaust gas act as shielding gases, a behaviour which probably can be explained by the higher oxygen partial pressure in air compared to that of the mixed gas. Air + 0.3% SO2 leads to earlier breakaway, which is supposed to be induced by internal sulphidation. Air + 50vppm HCl seems to result in the formation of volatile species and active oxidation.

2

Introduction

Assessments of different drive concepts with regard to emissions and weight/cost ratios show that the spark-ignition engine will remain the preferred drive concept during the next years [1]. The necessity of further improving the cold-start efficiency, which is highly dependent on the cell density of the catalyst substrate, leads to the objective to provide the largest possible catalytic surface with the lowest possible heat capacity, as can be seen in Figure 1.

Figure 1: Dependence of cold-start factor (catalytic surface/heat capacity) on cell density and foil thickness [1]

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

50 Increasing the cell density also means to reduce the thickness of the foil of the catalyst substrate, for which iron chromium aluminium alloys with additions of reactive elements proved to be an excellent solution. The chemical composition of the foil has been optimised with respect to its oxidation resistance, especially with respect to thickness [2]. However, for a thickness of 50 Ám or less environmental influences have to be considered in more detail than for the thicker foils. Therefore this paper will compare results gained in air to data obtained in atmospheres relevant to exhaust gas systems.

3

Experimental

The tests were performed on Aluchrom YHf, an FeCrAl alloy with additions of reactive elements. The chemical composition is given in Table 1. 0.05mm × 20mm × 10mm coupons were cut from the foil and exposed to multicomponent corrodants at 1200 °C. Table 1: Chemical compositions (mass%) element Cr (wt.-%) Al (Wt.-%) Y (ppm) C (ppm) S (ppm) O (ppm) N (ppm) P (ppm) Zr (ppm) V (ppm) Ti (ppm) Cu (ppm) Ca (ppm) Hf (ppm) Mn (ppm) Si (ppm) Nb (ppm) Mg (ppm) Mo (ppm)

Aluchrom YHf 19.7 5.5 460 2100 1.3 < 10 40 130 540 860 98 110 12 310 1800 2900 < 50 78 100

The atmospheres (in vol.% res. vppm) were the following: 1. N2 + 5000 vppm NO, which simulated the major NOx components in exhaust gas. The only important NOx compound NO was chosen, since engine temperatures reach values from 900 °C to 1300 °C, where the equilibrium partial pressure of NO is most relevant at levels between 500 ppm – 3000 ppm, as can be seen in Figure 2. The cycle time was 20 hours. 2. A simulated fuel-rich exhaust gas (N2 + 12% CO2 + 2% CO + 10% H2O). The oxygen partial pressure at this temperature was calculated to be about 10 Pa–15 Pa. The cycle time was again 20 hours.

51 3. Foils were exposed in air + 0.3 vol.% SO2. Although a lower SO2 concentration, e.g. of 10 vppm, is more relevant in automotive exhaust systems, air + 0.3 % SO2 was chosen as worst case condition. Figure 3 shows the equilibrium partial pressures for both concentrations. From these calculations an influence of the SO2 enrichment at 1200 °C should not be visible for the 10 vppm. The cycle time was 20 hours. 4. 0.05mm and 1 mm thick samples were discontinuously exposed in air + 50 vppm HCl with 100 hours cycles.

Figure 2: Equilibrium composition of air with 5000 ppm NO in the temperature range 0 °C to 1400 °C, calculated with ChemSage [3]

a) Air + 10 vppm SO2

b) Air + 0.3 vol.% SO2

Figure 3: Equilibrium composition of air with (a) 10 vppm resp. (b) 0.3 vol.% SO2 in the temperature range 0 °C to 1400 °C at 1.013 bar, calculated with FactSage

After every cycle the furnace was cooled down to room temperature and the mass changes of the species were measured. No differences between net and gross mass changes were found in

52 N2 + 5000 vppm NO, N2 + 12% CO2 + 2% CO + 10% H2O and air + SO2. In air + 50 vppm HCl the gross mass change could not be determined due to volatile species. The surface coloration was evaluated and a grey colour was associated with (-alumina, a green colour with chromia and red and black with iron oxides. The microstructure has been characterised by optical and scanning electron microscopy. The test results were compared to those gained in parallel tests performed in laboratory air.

4

Results

4.1

N2+5000 vppm NO

As shown in Figure 4, the oxidation rate in the N2-NO mixture is always smaller than in air but shows a similar development. This is also true for the coloration changing from metallic to grey and then to green before breakaway occurs. No evidence of internal nitridation has emerged.

Figure 4: Comparison of the net mass gain of 50 mm foils after cyclic oxidation in N2 with 5000 vppm NO and in laboratory air at 1200 °C [3]

4.2

Fuel-rich exhaust gas (N2 + 12% CO2 + 2% CO + 10% H2O)

Specimen mass gains are shown in figure 5. The mass increase of the foil was slower in the fuel-rich exhaust gas (N2 + 12% CO2 + 2% CO + 10% H2O) than in air.

53

Figure 5: Specimen mass change in artificial exhaust gas at 1200 °C [3]

4.3

Air + 0.3 vol.% SO2

Figure 6 indicates little effect on growth kinetics when adding 0.3 vol.% SO2 to the air environment. But the onset of breakaway at 1200 °C occurred after 150 hours in air in comparison to 120 hours in the sulphur dioxide containing environment. Breakaway occurred with the formation of chromia underlying the alumina scale, and later non-protective iron oxide formation, which resulted in a rapid increase of mass gain.

Figure 6: Comparison between the gross mass gain in air + 0.3% SO2 with that in laboratory air at 1200 °C [3]

4.4

Air + 50 vppm HCl

As can be seen from Figure 7a, the net mass gain at 1200 °C is lower in the HCl containing gas than in air. In contrast to the tests performed in laboratory air, HCl induces spallation in all samples (see Figure 7b). Small red and black spots can be recognised on the surfaces after relatively short exposure times in the HCl containing gas.

54

a)

b)

Figure 7: a) Net mass gain in air and air+50 vppm HCl at 1200 °C [3]; b) cross section of 1 mm thick Aluchrom YHf after 100 hours

5

Discussion

The nitrogen-oxygen-bioxidant and the synthetic exhaust gas retard the breakaway in comparison to exposures in air. In these cases the alumina scale, which could grow due to sufficiently high oxygen levels, protects the alloys against nitridation [4]. So these environments actually can be used as shielding gases for FeCrAlRE alloys. Similar results in exhaust gas have been found for thicker species by Sigler [5] and have been attributed to different oxide morphologies. The oxidation mechanisms are the same in these environments and in air. The formation and almost parabolic growth of alumina is followed by that of chromia and subsequently iron oxides, which then leads to chemical breakaway failure. The experiments show that the growth rates of the aluminium oxide in the two low- pO2 test gases (N2/NOx and exhaust gas) are substantially slower than in air. Two-stage oxidation studies using 18O-Tracer [7] have shown that gas tight alumina scales on FeCrAl alloys containing reactive elements grow by oxygen grain boundary diffusion. For scales exhibiting this growth mechanism, the relation between scale thickness x and oxidation time t is given by:

x2 = k p × t

(1)

The parabolic rate constant kp has been calculated in /8,9/ to be

kp =

4@ DB æ DmO2 ö ç ÷ r è RT ø

(2)

55 with DB the oxygen grain boundary diffusion coefficient along grain boundaries with a width @and a grain diameter of r , DmO2 the chemical potential gradient across the oxide scale, R the gas constant and T the temperature. Previous experiments [7,8] have shown that in an alumina scale with optimum protective properties the grains develop a columnar type structure whereby the grain size tends to increase in growth direction. Thus, the grain size at the scale-metal interface increases with increasing scale thickness, i.e. in the above equation r = r(x). Based on theoretical considerations and experimental observations it was previously shown [7,8] that this increase in grain size can with reasonable accuracy be described by the relation r = r0 × x p

(3)

whereby p is close to one and r0 is the initial grain size, when diffusion along grain boundaries becomes the rate determining process. The scale thickness can easily be transformed into æ Dm ö mass gain ç ÷ using a factor f, because this is the common way to measure oxidation kiè A ø netics. x = 0.53374

mm Dm Dm = f g / m2 A A

(4)

The equations (1) to (4) can now be combined to achieve a comprehensive understanding of the oxidation process.

x2 = k p × t x2 =

4d DB æ DmO2 ç r è RT

Û ö æ DmO2 ÷×t = ç ø è RT

ö 4d DB ÷ ×t × p ø r0 × x

æ 4d DB ö x2+ p = ç ÷ DmO2 t è r0 RT ø

Û

1/ 2 + p

x= f

Dm æ 4d DB ö =ç ÷ A è r0 RT ø

Using the definitions

Dm 1/ n =  kct  . A

Û

 Dm t 

1/ 2 + p

O2

DmO2 4d DB 1 1 = the growth law simplifies to: and kc = f n r0 RT n 2+ p (5)

For the tested foils p equals one. This reveals a cubic time dependence for the scale thickening and thus a scale thickness dependence on the oxygen partial pressure

x : DmO1/2 2 + p = DmO1/23

(6)

The chemical potential gradient across the scale is proportional to the difference between the logarithm of the oxygen activities at the scale/alloy and the scale/oxide interface, respectively.

56 DmO2 = RT  ln aO2 |surface - ln aO2 |int erface 

(7)

The activity of oxygen at the surface is given by the oxygen partial pressure in the surrounding atmosphere. The oxygen activity at the scale/alloy interface has been calculated from thermodynamic data [10] and activity coefficients measured in the LEAFA [3] project to be 2.305 × 10–25 at 1200 °C. Table 2 gives the oxygen partial pressures for the tested atmospheres and the expected retardation in the oxide growth. Table 2: Oxygen partial pressures for the tested atmospheres and the expected retardation in the oxide growth

atmosphere air N2-NO exhaust gas

aO2 0.21 2.5 × 10–3 1 × 10–10

,mO2 / kJmol–1 675.5 621.6 412.8

retardation 1 0.9 0.6

The retardation factor is calculated assuming that breakaway oxidation in air and in the reduced oxygen partial pressure atmosphere occurs for the same mass gain. 1/ n Dm air air 1/ n  kc tB  =  kcatmospheretBatmosphere  Û kcairtBair = kcatmospheretBatmosphere A

retardation =

t Bair t Batmosphere

(8)

atmosphere

=

kcatmosphere DmO2 = kcair DmO2

air

(9)

In Table 3 the calculated data are compared to the measured data. They are the same for the exhaust gas but differ for N2+NO. However, this difference might be explained by the cycle time. The data gained in air were measured with 20 hours cycles whereas those measured in the bioxidant refer to 100 hours cycles. Table 3: Comparison of calculated and measured retardation factors after 100 and 150 hours exposure

atmosphere N2-NO exhaust gas

calculated retardation 0.9 0.6

measured retardation after 100 and 150 hours 0.8 0.6

Results gained for PM 2000, a Plansee FeCrAl alloy, support this idea. Isothermal thermogravimetric tests were performed in Ar + 20% O2 and in Ar + 4% H2 + 7% H2O at 1200 °C. Figure 8 shows the retarded gross mass gain in Ar + 4% H2 + 7% H2O, the atmosphere with the lower oxygen partial pressure. The exponential fits of these data resulted in k values of 0.31 mg2cm4/h for Ar + O2 and 0.19 mg2cm4/h for Ar + H2 + H2O. This method which allows to predict results from retardation factors is not applicable for atmosphere where volatile species or inclusions occur, like in air + HCl or air + SO2. A similar oxidation mechanism is found in air and in air + SO2, but 0.3% SO2 induce earlier breakaway. The SO2 bearing environment affects the oxidation behaviour dramatically by internal sulphidation. The sulphidation can also be explained by the oxygen partial pressures of

57 SO2 and SO2 at 1200 °C, which are shown in Figure 3. Since SO2 contents more realistic in automotive exhaust systems are in the range of 10 vppm, the conclusion is that SO2 has not to be considered for oxidation processes in vehicles with spark-ignition engines.

Figure 8: Gross mass gain of PM 2000 in atmospheres with different oxgen partial pressures gained at 1200 °C

In the above mentioned environments the species fail due to chemical breakaway when most of the aluminium is consumed and other oxides types form. Therefore for the species with 5.5 mass% Al, the foil thickness is critical. Spallation was only noticed on thicker coupons. However, the growth of alumina in the HCl containing atmosphere is not parabolic. Even for the foils the alumina scale spalled. The mechanistic understanding for the behaviour of species in HCl containing atmospheres is based on the observation, that an influence is visible before the first cracking or spalling of the alumina scale occurs. Furthermore very fine spall was found early on and also the scale adherence in some species seems to be weaker. Red and black spots on top of the alumina indicate that iron is involved in the process, but not chromium, because green spots are missing. Together with the oxygen HCl can react at 1200 °C to H2O and Cl2. Similar reactions have already been found at 600 °C [6]. This might promote the formation of volatile chlorides. These compounds deteriorate the alumina scale, which in turn allows access of the gaseous species O2, HCl, H2O and Cl2 to the metal oxide interface. Thermodynamical calculations revealed that the chlorides with the highest fugacity at 1200 °C are aluminium chlorides. These ideas are confirmed by the observation that the processes accelerate as soon as the first cracks appear in the alumina.

6

Conclusions

For wrought Aluchrom YHf N2 + 5000 vppm NO and N2 + 12% CO2 + 2% CO + 10% H2O act as shielding gases at 1200 °C. An explanation for this behaviour is thought to be the higher oxygen partial pressure in air compared to that of the mixed gas.

58 Air + 0.3 vol.% SO2 leads to earlier breakaway due to sulphidation. Air + 50 vppm HCl seems to result in the formation of volatile aluminium chlorides and active oxidation at 1200 °C.

7

Acknowledgements

We are grateful to the European Commission for financial support under the LEAFA project no. BRPR-CT97-0562 and to our partners for the supply of the alloys tested, for the chemical analysis of alloys and for their contribution to scientific input in discussing these results. We also want to thank Dr. K. Hack, Gesellschaft für Technische Thermochemie und -physik, Herzogenrath, for the calculations of figure 3.

8 [1] [2]

References

W. Maus, R. Brück, G. Holy, Int. Congress in Graz, Sept. 2–3, 1999 J. Klüwer, A. Kolb-Telieps, M. Brede, Int. Conf. MACC ’97 in Wuppertal, Oct. 27–28, 1997 [3] BRITE-EURAM project, Contract no. BRPR-CT97-0562 [4] M.J. Bennett, R. Newton, J.R. Nicholls, Eurocorr 2000, available as CD [5] D.R. Sigler, Oxidation of Metals, Vol. 40, No.5/6, 1991, p. 555–583 [6] A. Zahs, M. Spiegel, H.J. Grabke, Materials and Corrosion 50, 1999, p. 561–578 [7] W.J. Quadakkers, H. Holzbrecher, K.G. Briefs, H. Beske, Oxidation of Metals 32, (1/2) (1989), 67–88 [8] K. Bongartz, W.J. Quadakkers, J.P. Pfeifer, J.S. Becker, Surface Science 292 (1993) 196–208 [9] S.N. Basu, J.W. Halloran, Oxidation of Metals 27 (1987) 143 [10] Ihsan Barin, Thermochemical Data of Pure Substances, Weinheim, Basel, Cambridge, New York, VCH, 1989

Improved High Temperature Oxidation Resistance of REM Added Fe-20%Cr-5%Al Alloy by Pre-Annealing Treatment K. Fukuda, K. Takao, T. Hoshi and O. Furukimi Technical Research Laboratories, Kawasaki Steel Corp., Chiba (Japan)

1

Introduction

Fe-Cr-Al alloys exhibit outstanding oxidation resistance at high temperatures because a protective α-Al2O3 scale forms on their surface. As a practical application of this good oxidation resistance, Fe-20mass%Cr-5mass%Al alloy foils have recently been used as a catalytic converter substrate for automobiles, in which the material is exposed to high temperature exhaust gas. (1)-(3) The thickness of these foils is usually limited to 30 µm–50 µm so as not to increase the back pressure in the exhaust system. Since the surface-to-volume ratio of foils is higher than that of thicker alloy sheets, the Al in the foils is consumed as Al2O3 in a shorter time than in sheets. Therefore, reducing the growth rate of Al2O3 scale is important for this application. It is widely known that the addition of reactive elements such as rare earth metals (REM), Ti, Zr, and Hf to these alloys improves their oxidation resistance by preventing spalling of the Al2O3 scale.(4)-(7) Considered these studies, Fe-20Cr-5Al-La-Zr alloy foils have been widely used for catalytic converter substrates. Foils for this application are usually supplied as-cold rolled or after pre-annealing in a reducing atmosphere. However, the effect of pre-annealing conditions on the growth rate of Al2O3 scale formed on Fe-20Cr-5Al with the addition of reactive elements such as rare earth metals (REM), Zr, and Hf alloys had not been clarified. Therefore, in this study, the effect of pre-annealing in a hydrogen atmosphere on improvement of high temperature oxidation resistance in Fe-20mass%Cr-5mass%Al alloy foils with a small content of La-Zr or La-Hf was investigated.

2

Experimental Procedure

2.1

Specimen Preparation

The chemical compositions of the alloys used in this study are shown in Table 1. The basic composition was Fe-20mass%Cr-5mass%Al with a small addition of La, La-Zr, or La-Hf. All the alloys were melted in a vacuum induction furnace and cast as 10kg ingots, and were then hot-rolled to 3mm thick plate. These plates were annealing and cold-rolled to foils with a thickness of 50 mm or 300 mm. Some of the foils were annealed at 1223K and polished with #600 SiC paper, and some of the polished foils were also pre-annealed in hydrogen gas or air at 1223K for 60 seconds. These foil specimens were cut into coupons with a size of 20–30mm, which were degreased in acetone and in alcohol before oxidation. Alloy specimens with a

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

60 thickness of 300 mm were used for observation by scanning electron microscope (SEM) and transmission electron microscope (TEM). Table 1: Chemical compositions of specimens Alloy A B C D E F 2.2

Cr (mass%) 19.8 20.0 20.1 20.1 20.1 20.6

Al (mass%) 5.5 5.7 5.6 5.6 5.7 5.6

La (mass%) 0.024 0.045 0.082 0.098 0.088 0.072

Zr (mass%)

Hf (mass%)

0.029 0.032

Oxidation tests

Oxidation tests were carried out in air at 1373 K or 1423 K using an electric furnace. The mass change was measured by weighing the specimens at certain time intervals after cooling to room temperature. 2.3

Analysis method

The surfaces of the foils after polishing, pre-annealing in hydrogen, and oxidation were investigated by secondary ion mass spectroscopy (SIMS). Mass filtered O2+ primary ions (acceleration voltage: 15 kV) were rastered over areas of 100 × 100, 150 × 150, and 200 × 200 µm on the targets. Specimens of the 300 µm thick sheets of the alloys after continuous oxidation at 1373 K for 172.8 ks were cracked by immersion in liquid nitrogen to observe the cross-section of the scale by SEM. The Al2O3 scale which formed on the specimen foils after continuous oxidation at 1373 K for 172.8 ks was examined with a field emission type TEM equipped with an energy-dispersive X-ray spectrometer (EDX). Specimens of the foils were prepared by mechanical grinding of the alloy substrate and subsequent ion-milling with an Ar gun, aiming at the mid-thickness region of the scale. Specimens of the 300 mm thick sheets of the alloys were prepared by a focused ion beam (FIB) system using Ga-ions to observe the cross-section of the scale.

3

Results and Discussion

3.1

Effect of pre-annealing on oxidation rate

Figure 1 shows the mass change curves of 300 µm thick foil samples of alloy C during oxidation at 1423 K. No spalling of the scale was observed with any of the specimens, and a protective Al2O3 layer formed on all the specimens after the oxidation test. The mass gain results were virtually the same with the as-cold rolled, polished, and pre-annealed in air specimens. The mass gain of the specimen which was pre-annealed in hydrogen gas was lower than that of the other specimens. Figure 2 shows the oxidation gain after oxidation in air at 1423K for 86.4ks with 300 µm thick foil samples of alloys A, B, C, and D with several La contents in the as-cold rolled con-

61 dition and with pre-annealing in hydrogen. The oxidation gain of the as-cold rolled specimens decreased as the La content increased up to about 0.04 wt%, and remained approximately constant 5.0 (g/m2) above about 0.04 wt%. In contrast, the oxidation gain of the samples which were pre-annealed in hydrogen continued to decrease as the La content increased even at values of more than 0.04 wt%. This demonstrates that pre-annealing treatment in hydrogen gas is more effective for decreasing the oxidation rate as the La content increases.

Mass gain (g/m2)

20.0

15.0

as rolled annealed in air polished after annealing annealed in hydrogen

10.0

5.0

0 0

200

400

600

800

Oxidation time (ks) Figure 1: Effect of annealing treatment on oxidation behavior of 300 µm thick sheets of alloy C at 1423K in air

Mass gain (g/m2)

10 as rolled annealed in hydrogen

8 6 4 2 0 0.00

0.02

0.04

0.06

0.08

0.10

La content (mass%) Figure 2: Effect of La content and hydrogen annealing on mass gain after oxidation in air at 1423K for 86.4ks

Figure 3 shows the mass change curves during oxidation at 1373K with 50 µm thick foil samples of alloys C, E, and F in the as-cold rolled condition and with pre-annealing in hydrogen. The oxidation gain results of the as-cold rolled alloys E (La + Zr) and F (La + Hf) were smaller than that of alloy C, which contained only La. The oxidation gain of these La + Zr and La + Hf co-added alloys was also smaller when the samples were given pre-annealing treatment in hydrogen.

62

5

Alloy C as rolled Alloy E as rolled Alloy F as rolled Alloy C annealed in hydrogen Alloy E annealed in hydrogen Alloy F annealed in hydrogen

100 (b) Mass gain (g/m2)

10

(a)

2

Mass gain (g/m2)

15

Alloy C as rolled Alloy E as rolled Alloy F as rolled Alloy C annealed in hydrogen Alloy E annealed in hydrogen Alloy F annealed in hydrogen

10

0.4 0 0

200

400 Time(ks)

600

800

1 10

100 Time (ks)

1000

Figure 3: Effect of hydrogen pre-annealing annealing on oxidation behavior of 50 µm thick foils of alloy C, E, and F at 1373K in air.

The mass change curves of the 50 mm thick alloys during oxidation at 1373 K are expressed in a log-log plot in Figure 3(b) in order to understand the kinetics. The slope of each sample is about 0.4. This means that the oxidation curve basically conforms to the parabolic rate law. 3.2

SIMS analysis

Figure 4 shows the depth profiles of 50 µm thick foils of alloy E in the as-cold rolled condition and with pre-annealing in hydrogen as obtained by SIMS. No apparent peak was detected near the surface of the as-cold rolled foil. However, with the pre-annealed foil, the intensities of Al, La, and Zr secondary ions near the surface of the foil were stronger than that inside the substrate. This indicates that a thin layer of Al2O3 which contained La and Zr had formed on the surface as a result of the pre-annealing treatment. Figure 5 shows the depth profiles after oxidation at 1373 K for 10.8 ks for the 50 µm thick foils of alloy E in the as-cold rolled condition and with pre-annealing in hydrogen as obtained by SIMS. With the as-cold rolled sample, the intensities of Cr and Fe were high at the outer side of the oxide layer. This indicates that the Al2O3 which formed on the as-cold rolled foil contained Fe and Cr oxides. However, with the foils which were pre-annealed in hydrogen, no apparent peaks of Fe or Cr were detected. This means that the Al2O3 which formed on the alloy pre-annealed in hydrogen contained little Fe or Cr.

63 109

(b) annealed in hydrogen

(a) as cold rolled

Secondary ion intensity (counts)

108

27Al

27Al

53Cr

53Cr

107 106 105 104

57Fe

57Fe

103

16O

102

139La

16O 139La

101 100

90Zr 0

500

1000 Time(s)

1500

2000 0

500

1000

1500

90Zr 2000

Time(s)

Figure 4: Depth profiles of 50 µm thick foils of alloy E as obtained by SIMS. (a) As-cold rolled (b) annealed in hydrogen at 1223 K for 60 s in hydrogen; contined litle Fe or CR

109

(a) as cold rolled

27Al

27Al

108 Se co nd ary ion int ens ity (co unt s)

(b) annealed in hydrogen

107 106 105 104

53Cr

53Cr

57Fe

16O

16O 139La

103 102

57Fe 139La

90Zr

90Zr

101 100 0

2000

4000 Time(s)

6000

8000 0

2000

4000 Time(s)

6000

8000

Figure 5: Depth profiles of 50 µm thick foils of alloy E after oxidation at 1373 K for 10.8 ks as obtained by SIMS. (a) As-cold rolled (b) annealed in hydrogen at 1223 K for 60 s before oxidizing

64 3.3

SEM observation

Figure 6 shows cross-sectional SEM images of the Al2O3 scale which formed on the specimens of alloys E and F in the as-cold rolled condition and with pre-annealing in hydrogen after oxidation at 1373 K for 172.8 ks in air. The alloy/scale interface was smooth with no voids at the interface. The Al2O3 scale which formed on the cold rolled foil consisted of two layers. The outer layer was about 0.5 µm in thickness and showed a morphology characterized by equiaxed grains. The inner layer was about 2.0 µm in thickness and had a columnar grain morphology, which is the same morphology as that reported by Golightly et al. (8) In as-cold rolled foils, Al2O3 scale forms in equiaxed grains during the initial oxidation period and then grows by forming columnar grains. On the other hand, the Al2O3 scale which formed on foil that had been pre-annealed in hydrogen consisted of only one layer, which was approximately 1.5um in thickness and had a columnar grain morphology. (a)

(b)

2.0um (c)

2.0um (d)

2.0um

2.0um

Figure 6: Cross-sectional SEM images of Al2O3 scale formed on alloy E and F after oxidation at 1373 K for 86.4 ks in air; (a) and (b) as-cold rolled , (c) and (d) annealed in hydrogen at 1223 K for 60 s before oxidizing

3.4

TEM observation

Figure 7 shows the results of TEM observation of the Al2O3 scale which formed on alloys E and F in the as-cold rolled condition and with pre-annealing in hydrogen after the specimens were oxidized at 1373K for 86.4ks in air. No secondary phases were apparent at the grain boundaries, as shown in Figure 7. As can be seen in this figure, the Al2O3 scale which formed on the as-cold rolled foil consisted of two layers, an outer layer of equiaxed grains and an inner layer of columnar grains. In contrast, the Al2O3 scale which formed on the foil that was preannealed in hydrogen consisted of only one layer and had a columnar grain morphology. Moreover, the size of the columnar grains of Al2O3 which formed on the pre-annealed foil was slightly larger than that of the Al2O3 grains which formed on the cold rolled foil.

65

(a)

(b)

0.5um (c)

0.5um (d)

0.5um

0.5um

Intensity (counts)

Figure 7: Cross-sectional TEM images of Al2O3 scale formed on alloy E and F after oxidation at 1373 K for 86.4 ks in air; (a) and (b) as-cold rolled, (c) and (d) annealed in hydrogen at 1223 K for 60 s before oxidizing

X-ray intensity ratio La-L/Al-K, Zr-L/Al-K

(a)

0.06 0.05

(c)

0.04 0.03

50 (b) 40 Al-K 30 Fe-K 20 O-K Cr-K 10 Zr-L La-L 0 0 2 4 6 8 10 X-ray energy, E (keV)

La/Al Zr/Al

0.02 0.01 0

40 30 20 10 0

10 20 30 40

Distance from grain boundary, d / nm Figure 8: (a) TEM image of grain boundary in parallel section of Al2O3 scale formed on alloy E after oxidation at 1373 K for 86.4 ks in air; (b) EDX spectrum from grain boundary in parallel section of Al2O3 scale formed on alloy E and (c) X-ray intensity ratios of La-L/Al-K and Zr-L/Al-K in EDX spectrum across grain boundary

66 One of the authors has reported previously that the segregation of La, Zr, and Hf at grain boundaries in A l 2 O 3 scale suppresses oxygen diffusion along the Al2O3 grain boundaries, resulting in a decrease in the growth rate of the Al2O3 scale. (9) In the present study, a distinct La-L intensity and Zr-L peak were also detected at each grain boundary by EDX analysis, as shown in Figure8 (b). In Figure 8(c), the X-ray intensity ratios of La-L to Al-K and Zr-L to AlK were taken from the points indicated by the dots in Figure 8(a) and plotted against the distance from one grain boundary. The results revealed that La and Zr had segregated to the grain boundary. Figure 9(a) and 9(b) show TEM images of the grain boundary in a parallel section of the Al2O3 scale formed on alloy F in the as-cold rolled condition and with pre-annealing in hydrogen after the specimens were oxidized at 1373 K for 86.4 ks. In the EDX spectrum from the columnar grain boundary of the Al2O3 scale on the as-cold rolled specimen of alloy F, a distinct Fe-L intensity and Cr-L intensity peak were detected, as shown in Figure 9(c). Moreover, from the equiaxed grain boundary of the Al2O3 scale on alloy F with pre-annealing in hydrogen, LaL intensity and an Hf-L intensity peak were detected, but an Fe-L intensity and Cr-L intensity peak were not detected, as shown in Figure 9(d). These results were in good agreement with the results obtained by SIMS. (b)

(a)

0.5um Intensity (counts)

Intensity (counts)

0.5um 50 (c) Al-K 40 Cr-K 30 O-K Fe-K 20 10 0 0 2 4 6 8 10 X-ray energy, E/keV

50 (d) Al-K 40 30 Cr-K 20 O-K La-L Fe-K Hf-L 10 00 2 4 6 8 10 X-ray energy, E/keV

Figure 9: (a) TEM image of grain boundary in parallel section of Al2O3 scale formed on as-cold rolled alloy F; (b) TEM image of grain boundary in parallel section of Al2O3 scale formed on alloy F annealed in hydrogen; (c) EDX spectrum from grain boundary of columnar grain in parallel section of Al2O3 scale formed on as-cold rolled alloy F; (d) EDX spectrum from grain boundary of equiaxed grain in parallel section of Al2O3 scale formed on alloy F annealed in hydrogen

Figure 10 is a plot of the X-ray intensity ratios of La-L, Zr-L, and Hf-L to Al-K at the Al2O3 grain boundary, against the distance from the alloy/scale interface. These results show that the degree of La, Zr, and Hf segregation at the Al2O3 grain boundaries was higher at the columnar grain boundaries than at the equiaxed grain boundaries. In addition, the intensity ratios of La-

67 L, Zr-L, and Hf-L at the Al2O3 c o l u mn a r grain boundaries were stronger with the preannealed alloys than with the as-cold rolled alloys. Based on the results of SEM observation, Golightly has reported that the Al2O3 scale on FeCr-Al alloys with small contents of rare earth metals forms at the alloy/scale interface mainly by inward diffusion of oxygen through the scale. (7) Similarly, based on experiments using an O18 tracer, Reddy has reported that the Al2O3 scale on an Fe-Cr-Al alloy grew by inward diffusion of oxygen through the grain boundary of the Al2O3. (10) In this study, in Fe-Cr-Al alloys with small contents of La-Zr or La-Hf, the alloy/scale interface was smooth with no voids at the interface, as shown in Figure 5. This means that outward diffusion of Al was suppressed, and the Al2O3 scale grew by inward diffusion of oxygen through the Al2O3 scale. Considering the fact that the oxidation curves of the alloys which were pre-annealed in hydrogen basically conformed to the parabolic rate law, it was inferred that the pre-annealing treatment did not change the mechanism of oxidation, but rather suppressed the inward diffusion of oxygen through the Al2O3 scale. (c) X-ray intensity ratio , La-L/Al-K

as cold rolled annealed in hydrogen

0.01

00.0 0.5 1.0 1.5 2.0 Distance from interface (um) (b) 0.02 as cold rolled annealed in hydrogen 0.01

0 0.0

0.5

1.0

1.5

2.0

Distance from interface (um)

X-ray intensity ratio , Hf-L/Al-K

X-ray intensity ratio , Zr-L/Al-K

X-ray intensity ratio , La-L/Al-K

(a) 0.02

0.02

as cold rolled annealed in hydrogen

0.01

0

0.0 0.5 1.0 1.5 2.0 Distance from interface (um) (d)

0.02

as cold rolled annealed in hydrogen

0.01

0

0.0

0.5

1.0

1.5

2.0

Distance from interface (um)

Figure 10: X-ray intensity ratios, (a) La-L/Al-K, (b) Zr-L/Al-K, (c) La-L/Al-K, (d) Hf-L/Al-K in EDX spectra from grain boundaries in columnar grain and equiaxed grain oxide layers of Al2O3 scale on alloys E and F

As with the cold rolled alloy specimens, Fe and Cr were oxidized together with Al during the initial oxidation period because the partial pressure of oxygen at the surface was high. The large amount of Fe and Cr oxides in the Al2O3 prevented the Al2O3 from growing by forming columnar grains. Instead, Al2O3 grew by forming small equiaxed grains during the initial oxidation period. The growth of these equiaxed grains then reduced the partial pressure of oxygen at the alloy/scale interface, and as a result, only Al was oxidized. After this point, Al2O3 grew by forming columnar grains at the alloy/scale interface, resulting in a two layer structure, and La, Zr, and/or Hf segregated at the grain boundary.

68 However, when pre-annealing in hydrogen was performed, Fe and Cr were not oxidized in this reducing atmosphere. The Al2O3 oxide, including a small amount of La, Zr, and/or Hf, grew by forming large columnar grains beginning in the initial oxidation period. It may be inferred that the low density of grain boundaries in the Al2O3 and the high segregation of La, Zr, and/or Hf at the Al2O3 grain boundaries suppressed oxygen diffusion along these grain boundaries. For this reason, the pre-annealing treatment in hydrogen reduced the growth rate of the Al2O3 scale.

4

Conclusion

The effect of pre-annealing on the oxidation behavior of Fe-20mass%Cr-5mass%Al alloy foils containing a small amount of La-Zr or La-Hf was examined in a cyclic oxidation test at 1373 K in air. 1. The oxidation rate of these alloys was reduced by pre-annealing in hydrogen. 2. In the Al2O3 scale which formed on the pre-annealed alloys, the outer equiaxed grain layer was thinner and the grain size of the inner columnar grain layer was larger than in the scale which formed on as-cold rolled alloys. 3. The segregation of La, Zr, and/or Hf at the columnar grain boundaries of the Al2O3 scale was higher with the pre-annealed alloys than with as-cold rolled alloys. It may be inferred that the low density of grain boundaries in the Al2O3 and high segregation of La, Zr, and/or Hf in the Al2O3 grain boundaries suppressed oxygen diffusion along these grain boundaries, reducing the growth rate of the Al2O3 scale.

5 [1] [2] [3] [4] [5] [6]

References

S. Isobe: Denki-seikoh, 58 (1987), 104. D. R. Sigler: Oxid. Met, 32 (1989), 337. D. R. Sigler: Oxid. Met, 40 (1993), 555. . A. Golightly, F. H. Stott and G. C. Wood: Oxid. Met, 10 (1976), 163. K. Ishii and T. Kawasaki: J. Japan Inst. Metals, 56(1992), 854. H. Hindam and D. P. Whittle: Proc. 3rd JIM Int. Symp. on High Temperature Corrosion of Metals and Alloys, The Japan Institute of Metals, Supplement to Trans. JIM, 24 (1983), 261. [7] F. A. Golightly, F. H. Stott and G. C. Wood: J. Electrochem. Soc., 126 (1979), 1035. [8] T. A. Ramanarayanan, M. Raghavan and R. Petkovic-Luton: J. Electrochem. Soc., 131(1984), 923. [9] K. Fukuda, K. Ishii, M. Kohno and S. Satoh: Proc. Int. Symp. on High-Temperature Corrosion and Protection 2000, (2000) Hokkaido ISIJ, p. 309. [10] K. P. R. Reddy, J. L. Smialek and A. R. Cooper: Oxid. Metals, 17(1982), 429.

Oxidation Induced Length Change of Thin Gauge Fe-Cr-Al Alloys C. Steve Chang, Leigh Chen, and Bijendra Jha Engineered Materials Solutions, Inc., Attleboro, MA 02703 USA

1

Abstract

Alloys of Fe-20Cr-5Al have been used extensively as the material of choice in metallic catalytic converter substrates. The alloy chemistry has been developed through the last decade to provide the oxidation weight gain resistance that was thought to be adequate. Mainly by the addition of rare earth and active elements, the cyclic oxidation behavior has been improved to the point to meet the regulation requirements on the durability of catalytic converters. The major uncertainty on the understanding of oxidation behaviors of thin gauge Fe-Cr-Al foil is the cause of foil length change. It has been suggested that this length change is one of the causes of buckled honeycomb observed in the fractured substrate. Analysis on the stress and strain of thin gauge Fe-Cr-Al foil under the oxidation condition has been published to account for the length changes due to the oxidation. However, the potential metallurgical factors that control the length change has not been rationalized yet. In this paper we will present (1) the oxidation test results on Fe-Cr-Al foils with different gauge and chemistry, (2) the microstructure evolution and (3) oxide scale development during the oxidation test. The oxidation failure mechanisms will be demonstrated. A phenomenological model to incorporate the alloy chemistry, microstructure and oxide scale will be described to account for the oxidation behaviors of these Fe-Cr-Al alloys.

2.

Introduction

Catalytic converters have become the universal solution for automobiles to meet the emission regulations. The catalytic converters with metallic type substrates have seen ever-wider acceptance because of several advantages over conventional ceramic-based converters[1]. One of the advantages of metallic substrate is the thinner substrate wall (30 to 50 microns) which provides lower backpressure and smaller package. However, the desirable thin gauge of alloy foils limit the usable life of the substrates. This limit is due to the fact that the metallic alloys require the formation of stable, protective scales such as aluminum and chromium oxides to slow down the oxidation of the alloy. The scale acts as a barrier to slow the diffusion of oxidizing agents such as oxygen to reach the alloy. However, the continuous oxidation of Al or Cr to form scale is like a reservoir being continuously drained and eventually the protective elements will be exhausted. At this point, oxygen will be able to react with the rest of the alloys to form the non-protective oxides, which causes the alloy to gain weight rapidly. The rapid and catastrophic oxidation weight gain is usually referred to as breakaway oxidation [2].

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

70 The alloy of choice for the metallic converter substrate has been the ferritic stainless steel with a nominal composition of 20wt% Cr, 5wt% Al and the balance of Fe. The addition of 5wt% Al provides the stable scale for the alloy to be used above 1100°C while the 20wt% Cr provides the oxidation and corrosion resistance from the ambient to where the Al oxidation becomes significant. The cyclic oxidation resistance is improved by the addition of rare earth elements such as Y, La and Ce. The addition of active elements such as Zr, Hf provides further improvement on the oxidation resistance [3]. The oxidation resistance is commonly measured by the amount of sample weight gain due to the scale formation. The oxidation weight gain is important, as it is an indication of the effectiveness of protective scale. The oxidation weight gain for the Fe-20Cr-5Al alloy has seen significant improvement and known to be one of the most oxidation resistant materials. Nonetheless, Fe-Cr-Al foils for the catalytic converter applications have to be dimensionally stable to avoid rupture of the substrate during the service [4]. The source of the dimensional instability has been attributed to the stress between scale and ally, which causes the creep deformation of substrate. Heats of Fe-Cr-Al alloys having identical oxidation weight gain behaviors have shown drastically different dimension changes. This paper summarizes the effort to rationalize the oxidation length change mechanism and attempts to draw a guideline to prevent dimensional instability. On the practical aspect of producing Fe-Cr-Al foils, it has been known that the conversion of Fe-20Cr-5Al alloy to thin gauge has been difficult and contributed to the high cost of the materials. A commercially feasible process [5,6] to produce thin gauge Fe-Cr-Al foils has been developed to address the issue. The process starts out with roll bonding the Fe-Cr alloys (AISI 4xx type ferritic stainless steels) to proper amount of Al alloys to form a three layer composites. The composite coil, which was roll bonded with proper attention so it can be cold rolled to an intermediate thickness. The intermediate thickness was selected to allow a homogenization heat treatment to be conducted at the temperature and time, which are commercially acceptable. After the heat treatment, the strip is cold rolled to provide the desirable temper and finish. Obviously, the advantage of alloying the Al to Fe-Cr by the roll bonding process is to circumvent the limits of Al content and the conversion difficulties[7]. In this study, oxidation tests on the oxidation weight gain and dimension stability were conducted on materials taken from roll bonding produced Fe-Cr-Al alloys. The effects of chemistry and heat treatment on the oxidation behaviors, in particular the dimensional stability is rationalized with a phenomenological model developed from examining the length change behaviors of hundreds of samples. This oxidation length change model will attempt to show that from the synergetic effects of physical (e.g. density and thermal expansion) and metallurgical (e.g. scale adhesion) changes, the various length change behaviors can be accounted for.

3

Experimental Procedure

3.1

Materials

The Fe-Cr-Al alloy foils were produced via the roll bonding process. In brief, Fe-Cr alloy (stainless steel) strips were clad with Al strips on both sides by feeding the strips into a fourhigh rolling mill. The cladding process was developed to apply sufficient reduction to form a well adhere three layer (Al/SS/Al) composite. The roll-bonded composites were cold rolled to

71 an intermediate thickness followed by heat treatment to homogenize the Al layers with the FeCr alloy. A finial cold rolling was applied to reduce the alloyed strip to the finish foil thickness. Fe-Cr-Al alloys with Al content range between 5 to 8 wt% are readily produced by this roll bonding process. Rare earth addition in the manner of La+Ce was accomplished by casting the Fe-Cr alloy with misch-metals. Typical alloy chemistry is shown in Table 1 in weight %. Table 1: Nominal composition of Fe-Cr-Al alloys in wt% C Mn P S Si Cr Ni Al N O La Ce 0.02 0.2 0.02 1.1 Mpa +/– 1.6 mm

Closed-coupled catalyst 1050 °C ~ 30 g up to 600 °C > 0.7 Mpa +/– 1.6 mm

This data was generated with todays systems, which are mainly built as modular converters with a ceramic substrat in a case covered by expanding paper or ceramic fiber mat and a metal can. An air gap insulated cone is welded on both ends of this middle part to make the connec-

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

192 tion to inlet and outlet tubes with lower diameters. Depending on the space and shape of monolith the cross sectional area of the monolith is always much larger than the tube diameter to maintain a low backpressure and a low space velocity for the exhaust gases to get good catalytic effects. For european driving conditions the high loads and temperatures are common for long time use in cars and systems have to be tested for the integrity of the system for more than 100 000 miles (more than 160 000 km).

2

Requirements for the Canning of Ceramic Monoliths

The change in the environment of the converter lead to the search for new mount materials for the ceramic monoliths and for a new design of the mount system, as it was soon found out, that the standard expanding paper mounting could not work under all of these conditions. The design had to fullfil all the following requirements for the canning of the bricks: · Maintain sufficient holding forces under all loads between ceramic monolith and can · Compensation of all tolerances between monolith, can and mat and all influences from the production process to prevent monolith crushing in mass production and under load · High temperature stability and aging resistance in hot, acidic and humid atmosphere for the cars live · Stability against erosion from the exhaust gas pulsation, even under can deflection and humid atmosphere (enlarged gaps) · Insulation properties to reduce can temperature for minimum gap changes through thermal expansion and for low heat transfer into the engine compartment · Gas sealing properties to prevent a bypass without catalytic treatment Besides these technical requirements there is always the demand for a reasonable price of the system and as a general target a minimum space for the mount system around the monoliths and a reliable mass production process. In practice the mounting is done by wrapping an elastic ceramic fiber mat or an expanding paper around the substrat and compressing this material with a metal sheet case on to the monolith. The holding force is given by the friction between the monolith, can and mat, the contact surface and the compression wich is given by the mat (Figure 1). It has to be higher than the actual load from the pressure drop over the monolith and the inertia force of the monolith under vibration of the system. In service the whole system heats up from inside to outside. As the substrates are specially made with a very low coefficient of thermal expansion (good thermal shock durability), the system mainly changes dimension at the outer shell, even if this is at lower temperature than the monolith. This leads to an enlargement of the mounting gap of 7 % to 20 % under high load and temperature. A further gap enlargement of up to 20 % is from the can deflection under the mat pressure on limited areas of the system at flat parts of the can (Oval or Racetrack cross section).

193

Figure 1: Monolith wrapping

To limit the heat expansion of the can it is common to use ferritic steels for the can with a relatively low coefficient of thermal expansion. The effects on the gap are also reduced, if enough cooling air comes to the surface of the part, the absolute gap height is enlarged (more space and mount material required) or a material with better insulation properties is used (Figure 2).

Figure 2: Gap enlargement

The actual loads on the system are measured at the engine in the car or on a test bench. Temperatures are taken with thermocouples and infrared scanners as Figure 3 shows:

Figure 3: Infrared thermography of a closed coupled catalyst with ceramic monoliths under full load

194 The mechanical loads in the system are taken by water cooled accelerometers. Piezo–accelerometers are mounted on different measuring points of the whole exhaust system in three directions (multiaxial) and the system is stimulated by the engine at all revs. and loads. The maximum acceleration in the flow direction at the mount system is calculated out of the directional values. To withstand the mechanical loads under thermal changes in the system the mat has to hold a high pressure on the monoliths surface. A limitation for the maximum value is given by the strength of the monolith. Ultra-Thin-Wall Substrates show remarkaby less strength and larger spread of strength in serial production than common products. As an example the following strength data shows almost comparable average strength for new products in relation to todays products (6/400) with 6 mil wall thickness and 400 cells per square inch (acceptable for 4/600, 4/400 and 3/600).

Figure 4: Isostatic strength of thin wall substrates, relative to standard substrates 6/400

A point of concern is the lower strength of a significant number of parts as the lower wall thicknesses lead to a broader strength distribution in the production process. Therefore the usable maximum pressure on the mat is limited to lower values to avoid monolith crushes in mass production. With lower wall thicknesses the actual shapes of the parts show bigger deviations than standard substrates. The contour tolerance, describing a mixture of dimensional and shape deviation is now merely used for shape deviations, which are hard to overcome with monolith measurements and calibrated canning systems. As a result higher mechanical and thermal loads on todays and future catalytic converter systems in close coupled position allow lower maximum mounting pressures on the mat systems. The mounting has to show more constant pressures under all loads and tolerances of the parts. This can be done by larger gaps and thicker mount mats, lowering the relative influences on the mat or with better mount materials.

195

3

Mounting Systems

To fix the substrates the mat system has to exert higher forces than the load. This is done by: · The radial pressure in the system [p] · The contact area between mounting mat and monolith/can [U] · The coefficient of friction between the parts [m] F=p×U×m Pressure and friction change their values in service with time, temperature, humidity and corrosion. Practical values are: p = 10 ... 1000 kPa, m = 0.2 ... 0.6 The main factor to influence the holding force in canning is the radial pressure on the mounting mat (the gap). These forces are applied primarily in the canning of the monoliths. There are mainly three methods to do the mounting (Figure 5).

Figure 5: Half shell canning and tourniquet technique (wrapping)

Figure 6: Tube stuffing

All these methods compress the mat to a predetermined height around the monolith.They are all in serial production with their special pro’s and con’s. While the half shell canning has fixed dimensions of the shells coming out of the tools, the tourniquet technique and the tube

196 stuffing (Figure 6) have some possibilities to match the dimension of the case to the actual dimension of the monolith in the production process. Depending on the facilities it is possible to calibrate the tourniquet can to different dimensions or to calibrate the tube for stuffing in different sizes, matching to the actual sizes. But the main part of tolerance compensation, especially if the faults in shape are larger than the size-deviations, has to be done by the mounting mat.

4

Mount Materials-Tests and Results

Today ceramic fiber mats with vermiculite as expanding papers or alumina fibermats are the common mount material to fix the ceramic monolith in the metal can. These mats fill the gap between the parts and hold the ceramic part to prevent it from moving and hitting the metallic can. The composition of these mats is seen under the microscope (Figure 7).

Figure 7: Mat composition, left: expanding paper with vermiculite, right: alumina fibermat

The expanding paper (XP) consists of vitreous alumina-silica-fibers and grainy vermiculite, a mineral that expands under high temperatures. The alumina-silica-fibers are made by blowing a molten ceramic beam into fibers and cooling it with the driving air. This leads to a broad range of fiber diameters and length and to impurities in the form of glassy beads and drops in the material. For years XP was the common material for underfloor catalysts. The fibers show sufficient elasticity to hold the monolith under low loads. With increasing temperatures at higher loads the vermiculite expands and braces the parts with high pressure against each other, which is explained in the following thermo mechanical analysis (Figure 8). As the temperature activity of the vermiculite diminished over the years, some systems showed erosion in the mount mat especially at underfloor catalysts with large gaps as a result of can deflection and low temperatures. This problem was overcome with expanding papers with less vermiculite and higher content of washed fibers. These high fiber content expanding papers showed much better resiliency after canning and worked well, even at exhaust gas tem-

197 peratures below 400 °C. The elastic effect is shown in the following compression diagram. In the second cycle the high fiber content XP shows much better resilience (Figure 9).

Figure 8: Thermomechanical analysis of XP

Figure 9: Compression of expanding papers in first and second cycle

But for new systems in close coupled position to the engine the expanding papers still showed problems at high gas temperatures above 900 °C. The whole mat loses tension as the crystal water leaves the vermiculite and the alumina-silica fibers sinter together. Furthermore the expanding papers still are very progressive in the compression graph against the gap. With common tolerances at the gap the highest pressures at narrow gaps are higher than the guaranteed isostatic strength of the Ultra-Thin-Wall monoliths. Tests on different mount systems showed, that crystalline alumina fibermats for high temperature insulations worked well under these conditions and are now under improvement. This

198 material is good for long time use at more than 1000 °C without becoming weak or brittle. These mats have lower pressure progression at narrow gaps compared to XP’s (Figure 10).

Figure 10: Pressure progression in mats

In the graph the upper curves show the pressure against the gap height in the first compression. In the car the gap opens with every drive cycle mainly because of thermal expansion at the can. With mechanical cycling in a compression tester these gap changes are simulated for 1000 cycles from an estimated gap to a maximum value as it can be seen in the actual mount and load situations. The results are the „aged“ graphs at the bottom. Without thermal expansion from the vermiculite the preheated expanding paper shows almost no resilient pressure at the large gap (high load-hot can) while the alumina fiber mat still shows reasonable values. The improvement in this system goes to lower mass of the expansive fibers (narrow gaps with worse insulation effects) and to oriented fibers in the mats to use the strength of all fibers – leading to the use of less material with the same elasticity of the mat. Also there are developments with other materials to have similar temperature stability with blown fibers or to have good elasticity at lower temperatures (diesel applications).

5

Practical Tests at the Whole System

As the mat tests only show an estimated behaviour in the laboratory, practical tests were developed and ran at the whole system. The next step is the hot shake test of the whole system in the flow direction to give the monoliths inertia load to the mount system at practical temperatures for the monolith and can. This is done by a servohydraulic actuator, while the system sees hot gas from an oil burner. With gas temperature, gas flow and an axial acceleration at a given frequency, all calibrated to the actual load in the car, the system is aged. With increased levels of load over long times the limit to failure is proved. If the system does not fail in the test, a push out test shows the left load bearing capability and is normalized as shear stress at slipping

199 of the monolith. As an example the following graph will show the maximum shear stress in the mount system for different expanding papers, gaps and temperatures (Figure 11).

Figure 11: Shear stress at XP after different preconditioning

While these tests reflect the maximum mechanical stabiltiy of the system quite well, some other effects in the car need to be simulated on the test bench with real engines. As an example remarkable erosion has been detected in systems with large gap enlargement at new expanding papers in car durability runs after very short tests (15 000 km’s). With a standard thermal shock test for 250 hours one can’t see any of these effects on the test bench. Optimised tests, related to the data of the test runs in the cars, resulted to a simulation of internal cooling and condensate. With such thermal shock tests similar effects over 50 to 80 hours at the test bench can be observed as at the car (Figure 12).

Figure 12: Car bench test, left: 15000 km car test, right: 80 h test bench

200 These tests lead to a good separation between the standard expanding papers and the different expanding papers with high fiber content. Compared to very good old expanding papers we foundacceptable materials for standard applications in underfloor condition (Figure 13).

Figure 13: Durability runs to failure at the engine test bench with different expanding papers

Alumina fibermats did not show any erosion after 500 hours in these tests, but the high material costs did not lead to the use in this underfloor position. To test close coupled systems, thermal shock tests are mainly used. As an example the engine runs for 10 Minutes with almost maximum load and speed (adjusted to the desired exhaust gas temperature) and 10 minutes at idle. Under these conditions with exhaust gas inlet temperatures up to 1050 °C in the system expanding papers failed. Sintered inlet edges and erosion in the paper led to significant damage after 20 to 50 hours. With alumina fiber mats we found no changes in the mat after durability runs under these conditions of more than 250 hours. Checking the holding forces left (shear stresses in the mount system) with push out tests after the durability runs showed remarkably high values at room temperature. Even with worst case samples with large gaps and estimating the thermal expansion of the system, the values showed good security against practical loads.

6

Conclusion

The loads on catalytic converters and how to measure them was explained and the methods to test mount materials for practical use and the general design of the system was shown. Changes in the properties of expanding papers and limits of their use in future systems require

201 new mount materials. Due to the high price of the usable alumina-fibermats they can only be recommended for critical applications. So the special properties of the mounting materials lead to the following fields of use: Expanding paper with high fiber content for underfloor catalysts Alumina-fibermat for systems with gas temperatures above 900 °C and or in combination with Ultra-Thin-Wall substrates For diesel applications XP with high fiber content is used, only in special applications with high accelerations, low temperatures or weak monoliths alumina-fibermat is recommended.

High Performance Packaging Materials M. Vermoehlen, D. Merry 3M Deutschland GmbH, Neuss,

Steffen Schmid Corning GmbH, Wiesbaden

1

Abstract

Catalytic converters have been undergoing major changes in recent years in order to meet tighter emission regulations. Catalytic converters are being moved closer to the engine, sometimes even attached directly to the exhaust manifold, in order to achieve quicker light off to reduce emissions during cold start. At the same time, there has been a trend to thinner wall monoliths often accompanied by higher cell densities that also improve emission performance. An important component of a catalytic converter is the mounting system. It has traditionally been a heat-expanding mat that acts to hold the ceramic monolith securely in place despite the significant difference in thermal expansion of the ceramic monolith and metal housing. Additionally, the mat provides a seal to prevent exhaust gas from bypassing the monolith and thermal insulation to keep to keep the housing cooler and to reduce thermal gradients in the monolith. Recently, mounting mats have been changing to reflect the more stringent requirements of close-coupled converters and/or thinner wall monoliths. This requires mats with higher temperature capability and mats with reduced holding force for the thinner wall monoliths. This paper will examine recent developments in this area of high performance mounting materials for the new requirements and compare the properties of these new mats with traditional mats.

2

Introduction

Mounting materials for catalytic converters have always had to adapt to changing requirements. When catalytic converters were first introduced in the US in the middle of the 1970’s, the predominant mounting system was wire mesh. As emission laws tightened, exhaust gas bypass through the wire mesh could no longer be tolerated and gas seals had to be added to the wire mesh to prevent gas bypass. This lead to the introduction of the first intumescent mounting mats which were invented by 3M. These mats consisted of ceramic fibers, an organic binder and a heat expandable material, vermiculite, which generates pressure as it is heated. The latter is needed to compensate for the thermal expansion difference between monolith.and.shell. As catalytic converters were introduced in Europe in the middle 1980’s, the thickness of the intumescent mats increased somewhat to compensate for the higher temperatures of the high speed German driving, but the general composition of the mats remained essentially the same.

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

203 In the middle 1990’s, certain OEM’s, particularly Japanese OEM’s began to use thin wall 400/4 (i.e. 400 cell per square inch/6 mil wall thickness) and 600/4 monoliths. These monoliths generally were considered to have lower compression strength due to their thinner walls. 3M introduced the first lower vermiculite mats to mount these somewhat weaker monoliths. Wall thickness of monoliths has continued to decrease and now non-intumescent ceramic mats are also available as a possible solution for the ultra thin-wall monoliths (600/3, 900/2, 1200/2). In addition, the temperatures in today’s close coupled and manifold mounted converters can sometimes exceed the limits of standard intumescent mats. For this reason mats with higher temperature capability have been introduced that can handle the most severe temperatures.encountered.to.date. Tests have been developed that can measure and compare various properties of mounting mats and help select the proper mat for a given application. Test descriptions and results for three new mats will be given in the next section and compared with standard intumescent mat.

3

Description of mats investigated

This paper will examine and compare the properties of four different mat types and assess their suitability as high temperature, thinwall and close coupled catalytic converter mounting materials. The four mat types are standard intumescent mat, reduced vermiculite intumescent mat, non-intumescent mat and hybrid mat, see description below: · Standard intumescent mat This type of mat consists of three basic raw materials (Table 1). Table 1: Basic raw materials of mats Ceramic fibers (alumina silicate) Unexpanded vermiculite

Organic binder

provide high temperature resiliency and contain the vermiculite provides the intumescent property i.e. creates pressure when the mat is heated and the mounting gap is opening due to the different expansion coefficient of cordierite and steel shell. provides strength to the product during manufacture and assembly

These mats are widely used in millions of converters all over the world to mount ceramic monoliths into converters. The most commonly used basis weight is 4070 g/m² mat which is appropriate for a 4 mm mounting gap. By increasing the basis weight, for example to 6200 g/m² (6 mm mounting gap), this mat can be suitable for a properly designed close coupled converter system. · Intumescent mat with reduced vermiculite level This mat consists of the same basic raw materials as standard mat but employs a lower vermiculite level and cleaner fibers. This result in a mat offering lower peak pressures at temperature and enhanced erosion resistance. · Non-intumescent mat Non-intumescent mats contain no vermiculite and use a special high alumina, polycrystalline fiber. These fibers maintain their resiliency at temperatures over 1000 °C. These fibers

204 contain essentially no shot (nonfiberous particles), resulting in excellent resiliency. Mats made with these fibers offer low, uniform holding pressure over the converter temperature range and enhanced erosion resistance. · Hybrid mat consisting of two layers The hybrid mat is a coformed mat consisting of two distinct layers - non-intumescent layer, which is positioned against the monolith - intumescent layer, which is facing the shell. This results in a mat capable of withstanding very high temperature and offering lower compression pressures as well as reduced peak operating pressures.

4

Property Comparison

Each of the four mats were evaluated for canning pressure simulation (compression test), operating pressure simulation (real condition fixture test), shear stress (resistive thermal push test) and high temperature resistance (engine dyno test). 4.1

Canning Pressure Simulation

The compression characteristics of intumescent and non-intumescent mats are important because they determine the amount of pressure exerted on the monolith during canning operations. If the compressive force is too high, monolith breakage can occur. For each mat there is specific mount density (i.e. density of the mat after canning) recommended to assure proper mat performance. Compression force is measured at target gap i.e. the gap that corresponds to the recommended mount density. To account for tolerances of monolith, can and mat, the compression force is also measured at gaps larger and smaller than target gap. F @ Speed 25.4 mm/inch

Uncompressed Thickness

Figure 1: Compression test

Compressed Thickness

205 The compression force of the mat is tested using equipment shown in figure 1. A sample diameter of 25.4 mm is used and the closing speed of the crosshead is 25.4 mm/min. Each mat is compressed to its recommended target gap and also compressed to gaps that are plus and minus 1 mm of target to simulate the tolerance stack up of the assembly. A target gap of 6.0 mm is used for all mats except the non-intumescent mat. Because of the lower thermal conductivity and higher temperature resistance of the non-intumescent mat, it can generally be used in a smaller gap than intumescent mats. Therefore, a target gap of 4.0 mm was used for the nonintumescent mat. Both the instantaneous or peak pressure value and the relaxed compression value are recorded. The relaxed pressure value is the pressure recorded 15 seconds after closure. The results of this experiment can be seen in figure 2 and 3. Figure 2 shows the peak compression values and Figure 3 shows the relaxed compression values for all four mats. The highest peak pressure recorded is for the standard intumescent mat, followed by the intumescent mat with reduced vermiculite level, and the Hybrid mat. The non-intumescent mat has the lowest peak compression pressure. Actual canning results are dependent on canning technique, know-how and closing speed, but in general the following can be concluded from this compression test: · Non-intumescent mats are suitable for all wall thicknesses. · Hybrid mats are suitable for all wall thicknesses. · Low vermiculite mats are suitable for thin-wall monoliths and possibly ultra thin-wall monoliths. · Standard intumescent mats are suitable for normal wall thickness monoliths. Peak Compression Pressures (Closing Rate 30.5 cm/min) 2500

Standard Intumescent Mat 2000

Pressure [kPa]

Reduced Vermiculite Mat 1500

Hybrid Mat

1000

Non Intumescent Mat 500

0 Target Gap +1 mm

Target Gap

Figure 2: Peak compression force at target gap and +/– 1 mm

Target Gap -1 mm

206 Relaxed Compression Pressures (15sec) (Closing Rate 30.5 cm/min) 700 Reduced Vermiculite Mat 600 Standard Intumescent Mat

Pressure [kPa]

500

Hybrid Mat

400

Non Intumescent Mat 300

200

100

0 Target Gap +1 mm

Target Gap

Target Gap -1 mm

Figure 3: Relaxed compression force at target gap +/– 1 mm

Looking now at the relaxed compression values seen in figure 3, it is clear these values are considerably lower for all gaps than the peak values. This is due to the viscoelastic nature of the mat. It is interesting to note that the relaxed compression values of the low vermiculite intumescent mat are higher than the relaxed compression values for the standard intumescent mat. This is the result of the lower vermiculite mat employing cleaner fibers, which gives the mat greater resiliency and results in less relaxation. 4.2

Operating Pressure Simulation

A real condition fixture test (RCFT) is used (see figure 4) to measure mat pressure under simulated catalytic converter conditions. The fixture consists of two heated plates and thermocouples for temperature control. The upper plate is used to simulate the hot side or monolithmat interface temperature and the lower plate is used to simulate the mat-can interface temperature. A measuring device measures the position of the plates or test gap at any given time. The heated plates, thermocouples, and gap measuring device are connected to a computer. A thermal cycle is chosen based on the actual temperatures of monolith and can that occur in a catalytic converter during operation. From these temperatures the gap change during operation is calculated. The temperature model and gap change are programmed into the computer. The computer controls the temperatures of both plates and the gap desired for each temperature condition within the cycle precisely and repeatably to these programmed conditions. The pressure exerted by the mat is recorded during the cycle. The RCFT is very useful in converter and mounting system development to understand the mat pressure generating characteristics as they relate to the converter design parameters and operating conditions.

207 Load Cell Water-cooled Heatshield

Gap Measurement Device

Heated Top Plate Heated Bottom Plate

Thermocouples Insulation Test sample

Figure 4: Real condition fixture test equipment

Operating Pressure Simulation @ Target Mount Density 1400

First Cycle Second Cycle -----------1200

Standard Intumescent Mat

1000

Pressure (kPa)

15 min. soak 800

Reduced Vermiculite Matt

600

Hybrid Mat 400

200

Non Intumescent Mat

50 50

100 60

150 95

200 130

250 155

300 180

350 215

400 250

450 275

500 300

550 329

600 358

650 387

700 416

750 445

800 474

850 502

900 530

900 530

900 480

900 430

850 325

800 220

750 185

700 150

650 125

550 85

600 100

500 70

450 60

400 50

350 45

300 40

250 38

200 35

150 30

25 25

100 25

0

skin monolith

Temperature (C)

Figure 5: RCFT measurement on four different mats

RCFT pressure curves for the four mats tested are shown in figure 5. Both first and second thermal cycles are shown for each product. There is not much change during subsequent cycles, so only the first two cycles are shown.

208 For all products there is a considerable difference in operating pressure of the first cycle as compared to the second cycle. This is because the organic binder portion of the mat burns out during the first cycle causing the pressure to decrease as it is heated. For the intumescent products, the pressure starts to increase once the expansion temperature of the vermiculite is reached. The increase in operating pressure of the hybrid mat starts later than either the standard or low vermiculite mat because the ceramic fiber layer acts as insulation. Since the ceramic fiber layer is against the hot side of the RCFT, it prevents the vermiculite from heating up as quickly. All the intumescent mats build their maximum pressure during the first cycle. Subsequent cycles show significantly lower peak pressure, but always have their maximum pressure at high temperature and lowest pressure at room temperature. The standard intumescent mat has the highest operating pressure followed by the low vermiculite mat and hybrid mat, respectively. The non-intumescent mat shows the lowest pressure at temperature. The second cycles of the intumescent mats are quite different from the first cycles. There is no pressure decrease because the binder has already been burned out. Also the pressure starts to build very quickly since the vermiculite has already been activated. Peak pressure still reaches a maximum at high temperature and minimum at room temperature. The same order of pressure is maintained i.e. the standard intumescent mat exerts the highest pressure, followed by the low vermiculite mat and hybrid mat, respectively. The operation pressure of the non-intumescent mat differs from the intumescent mats in several ways. First, the pressure does not vary as much over the entire temperature range. Second, the maximum holding force is at room temperature, and the minimum at high temperature, which is the opposite of intumescent mats. This is because it contains no intumescent material, so the pressure decreases as the gap opens up during heating. Finally, the non-intumescent mat has the lowest peak operating pressure of any of the mats making it very compatible with ultrathinwall monoliths. The hybrid mat also shows a peak operating pressure that is compatible with ultra-thinwall monoliths, but has the advantage of increased pressure at high temperature where normally the maximum vibration level occurs. 4.3

Temperature Resistance in Resistive Exposure Testing (RTE)

The temperature characteristics of the intumescent and non-intumescent materials are tested in a heated push test. The axial push test is conducted while the converter is heated. A useful way to do this employs a resistive thermal exposure (RTE) technique developed by Corning, Inc1,2. Resistive wires are inserted into the outermost cells of the ceramic monolith close to the outer skin as shown in figure 6. The ends of the wires are spot welded together to produce a continuous circuit resistance heater. The heater is attached to a temperature controller and the substrate can be heated to the desired temperature. In this way, the substrate can be aged at a predetermined temperature while having a temperature gradient across the mat. This temperature gradient is normally monitored with two thermocouples, one located on the monolith skin and the other at the can surface. The intumescent, non-intumescent (not 3M) and hybrid mats were aged using the RTE test. To increase the severity of this test and to simulate a close coupled situation the RTE converter was aged inside a furnace (see figure 6, first step).

209 Heating Wires @ 980°C

Furnace @ 550°C

Load Cell

Heating Wires temperature off

Furnace @ 550°C

Mat Mount Sample 1st step 100h aging

2nd step push test at 550°C

Figure 6: RTE test equipment

Three samples of each mat type were heated/aged to 980 °C as shown in the first step of figure 6. After this aging the mats were pushed at 550 °C (second step, figure 6). Three additional samples were pushed at the same temperature (at 550 °C) but without prior heat aging. The results can be seen in figure 7. 800

unaged 700

Residual Shear Strength @ 550°C [kPa]

600

500

400

unaged aged @ 980°C

300

aged @ 980°C 200

unaged

100

aged @ 980°C

0

Intumescent Mat

Figure 7: Results of RTE testing

Non Intumescent Mat

Hybrid Mat

210 Before the shear strength of the mat is reached slippage occurs. The residual strength therefore refers to the force needed to cause slippage divided by the area of the mat. In the RTE test the intumescent mat loses approximately 73 % of its shear strength after aging (from 725 kPa unaged to 197 kPa aged). Intumescent mats have an upper temperature limit for the mat – substrate interface of approximately 950 °C and an average mat temperature limit over the entire mat thickness of 700 °C. When temperature exceeds this limit the mat in contact with the monolith sees the high temperature first. The fibers in this area close to the substrate can become rigid and prevent the vermiculite in the outer mat layers from transmitting pressure to the monolith. At the same time some of the vermiculite is losing its effectiveness because of the high temperature. This combined effect causes the holding pressure to decrease. This effect is the same for all intumescent mats independent of their vermiculite level and explains the large reduction in shear strength. The residual shear strength of the non-intumescent mat decreased approximately 23 % (from 85 kPa unaged to 65 kPa.aged). Because of the different type of fibers used there is no rigidizing effect as described above. However the pressure level of this mat as reflected in shear strength is much lower when compared to vermiculite containing intumescent and hybrid mats. The best results are obtained with the hybrid system showing a minimal decrease of approximately 11% (from 312 kPa unaged to 276 kPa aged) and overall higher values than for non-intumescent mat. This mat construction combines the best features of both mats (i.e. the intumescent pressure building property of vermiculite containing mat with the high temperature resistance of the non-intumescent fiber mat.) The above test results of the hybrid mat encouraged us to test this construction in high temperature engine tests. 4.4

High Temperature Resistance in Engine Tests

Final testing included only the hybrid mat. Engine dynamometer testing was conducted at mat-monolith interface temperatures up to 1100 °C. The conditions were as follows: Test conditions: · 7.5 Liter V8 engine · 3000 rpm/299 Nm torque · 1 converter per exhaust bank (close coupled) · Secondary air added at exhaust manifold · 1050 °C inlet gas temperature · 1100 °C Monolith-Mat interface temperature · 100 hours test duration · · · ·

Converter design Monolith Ø 85.5 mm, 80 mm long 350/5,5 cpsi endcones with and without insulation hybrid mat

211

Figure 8: Shows the substrate wrapped with the hybrid mat

At high temperature the whole converter system is under severe stress. It is therefore important that all components are designed in the most robust way. For this reason a round substrate with insulated double wall end cones was selected. Figure 9 and 10 show the dramatic difference in converters with and without insulated double wall endcones, respectively.

Figure 9: Showing a converter system with insulated endcones

212

Figure 10: Converter design without insulated inlet and outlet cones

The converter with the insulated endcones successfully completed the 100 hour test duration. No failure and no erosion observed after disassembly. Based on the good test results Aston Martin decided to use a hybrid mat system for their new 450 horsepower 6.0 litre V12 Vanquish model. The catalyst system consists of two manifold mounted 400/4 and one close coupled 400/6 substrate. The actual converter system is shown in figure 11.

Figure 11: Aston Martin converter system

213 The Aston Martin will be launched autumn 2001. The whole car was developed together with Ford RVT (Research & Vehicle Technology). According to Autotechnology3 Aston Martin ran a test program with 50 prototypes covering more than a million miles.

5

Conclusions

The four tests that were used to determine the suitability of mats for high temperature and thin wall mounting work quite well and can be used to help select the proper mounting mat. This is especially important in areas where there is not as much long-term field experience such as in high temperature, ultra thin-wall converters. Of course, the actual design of the catalytic converter can also have a very strong positive or negative influence on the performance of a mat4 (e.g. insulated double wall end cones can have a very positive influence, whereas insulated heat shields can have very negative influence). For any given converter design selection of the right mat is essential. Based on results of the four mats tested, the following conclusions can be made. · For standard applications where temperature requirements are within recommended guidelines standard mats are an economically and technically sound solution. Many of today’s applications are using this mat in close-coupled situations. It is important that the overall design is robust e.g. uses insulated cones etc. A recent SAE paper5 describes the durability of a thin-wall, converter system in a close coupled application in a high speed road test · Low vermiculite mats are better suited for thin wall monolith mounting than standard intumescent mats but have the same temperature limitations. · Non-intumescent mats are well suited for ultra-thin-wall monoliths and also for converters exceeding temperature guidelines for standard intumescent mats. Their disadvantage is reduced pressure performance caused by increasing gap at temperature. Maximum holding force is realized at low temperature. · Hybrid two layer constructions can be used for all described applications. They offer high temperature capability, lower peak pressures suitable for ultra thin-wall applications and increased holding pressure with temperature.

6

References

[1]

K.P Reddy, J.D. Helfinistine, and S.T. Gulati, Corning Inc “New test for characterizing the durability of a ceramic catalytic converter package”, SAE Paper 960559 R.J. Locker and C.B. Sawyer, and M.J. Schad, Corning Inc. “Quantification of ceramic preconverter hot vibration durability”, SAE Paper 960563 Auto Technology Volume No 1 August 2001 P.D. Stroom, R.P. Merry, 3M Company and S. T. Gulati, Corning Inc. “System approach to packaging design for automotive catalytic converters”, SAE Paper 900500 J. Kallenbach, and P. Floerchinger, Corning GmbH Wiesbaden “Durability of a thinwall converter system in close coupled application“, SAE 982897E, SAE Brazil 98

[2] [3] [4] [5]

IV Catalysts

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

Three-Way Catalyst Deactivation Associated With Oil-Derived Poisons Joseph Kubsh Engelhard Corporation, Environmental Technologies Group Iselin, New Jersey, U.S.A.

1

Introduction

During much of the 1990s the development efforts associated with three-way catalysts for automotive applications focused on improving the thermal durability of these catalysts (1–3). This stemmed largely from the general trend in the industry of moving these catalysts closer to the engine in order to accelerate their activation during the early stages of vehicle operation. These thermal durability development efforts resulted in the commercialization of new generations of three-way catalysts capable of long-term operation in exhaust environments at temperatures in excess of 1000 °C. The success of developing catalysts with high temperature durability has more recently focused more attention on the deactivation mechanisms for threeway catalysts associated with the deposition of poisons present in the exhaust environment. More specifically, deactivation of three-way catalysts by poisons derived from the consumption of engine lubricating oils such as phosphorus and zinc has been revisited (4–6). These poisoning mechanisms have been found to be of importance in advanced emission systems targeting ultra-low emission performance such as the recently adopted California Low Emission Vehicle II Program (LEV 2) requirements for ULEV2 (ultra-low emission vehicle 2) and SULEV (super ultra-low emission vehicle) classifications. In these near-zero tailpipe emission applications catalyst light-off often dominates a vehicle’s emission performance. Poison deposition on the inlet section of a close-coupled converter could be sufficient to depress the catalytic light-off characteristics of the three-way catalyst outside of these very low emission performance limits. The aim of this work was to revisit oil-derived poisoning issues on three-way catalysts. In particular an investigation was made into the development of an engine-based, catalyst aging protocol that could be used to more closely impart the impacts of oil-derived poisons found under real world driving conditions.

2

Experimental Details

Previous studies aimed at understanding the impacts of oil-derived poisons on three-way catalysts have used a variety of accelerated engine-based aging protocols. In most cases these aging protocols have attempted to accelerate catalyst deactivation by using high levels of oil additives that contain catalyst poisoning agents such as phosphorus or heavy metals such as zinc. An example of one such lubricant additive is zinc dialkyldithiophosphate (ZDDP). An additive such as ZDDP can be incorporated into an aging protocol in several manners to deac-

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

218 tivate a catalytic converter. These options include: 1) doping gasoline with the oil additive of interest, 2) doping the engine oil with higher levels of the additive, or 3) injection of doped oil containing the additive into the exhaust manifold upstream of the catalytic converter. In all cases the intent is to expose the catalyst to high levels of poison precursors that deposit on the catalyst surface and deactivate catalyst performance. In this study, the option of oil injection was selected for investigation. Initial investigations indicated that the effectiveness of this oil injection procedure in deactivating the catalyst was a strong function of the exhaust temperature present in the catalytic converter, as well as the time of catalyst exposure under potential catalyst poisoning situations. For example, using a 75 h catalyst aging protocol in which the bed temperature of the catalyst never reached temperatures below 850 °C, catalyst performance remained similar with and without ZDDP-doped oil injection into the engine exhaust manifold upstream of the converter. In contrast, dropping the exhaust temperature to 450 °C and exposing the catalyst to an additional 24 h of aging with oil injection resulted in a significant degree of catalyst deactivation as measured by the hydrocarbon light-off characteristics of the catalyst. The amount of ZDDP present in the oil injected in the manifold was also expected to have a strong impact on the level of catalyst deactivation observed in such an accelerated engine aging protocol. As a result, a design of experiments approach based on Box-Benkhen methods was used to investigate the impacts of these three key parameters (aging temperature, aging time, level of ZDDP present in the oil used for injection into the exhaust manifold) on three-way catalyst performance. 2.1

Design of Experiment Aging Variable Definition

A 75 hour engine aging protocol was used as the basis for this investigation into catalyst poisoning. The aging protocol contained two primary modes of operation: a high temperature, exothermic mode with a catalyst inlet temperature of 850 °C (maximum catalyst bed temperature of 1000 °C), and a low temperature, mode with a catalyst inlet temperature of either 600 °C or 400 °C. At this low temperature mode the aging engine is run at an idle condition. When oil-containing ZDDP was injected into the exhaust manifold of the engine upstream of the catalyst, this injection occurred during both primary modes of engine operation. The oil injection rate was 0.95 liter of oil over 24 hours of engine operation. For the injected oil, the ZDDP level was doped to be 1.5 wt.%. In some aging experiments no manifold oil injection was used, in which case the catalyst exposure to potential oil-derived poisons stemmed only from the oil consumption rate of the engine itself. This oil consumption rate was also approximately 0.95 liter of oil over 24 hours of engine operation. The concentration of ZDDP in the oil used for normal engine lubrication was 0.1 wt.%. Table 1 provides details on the levels selected for each of the three key design variables: low temperature aging temperature, time at the low temperature aging mode, and wt.% ZDDP in the oil exposed to the catalyst. In aging runs where the low temperature mode was set to be 50 % of the total aging time, the engine was cycled between high temperature and low temperature modes every 20 minutes. For runs in which the low temperature mode was only 20 % of the total aging time, a 20 minute low temperature mode was followed by 80 minutes of time at the high temperature mode.

219 Table 1: Aging cycle parameters for catalyst poisoning studies Aging run 1 2 3 4 5 6 7 8

Low temperature Injection (°C) 600 600 600 600 400 400 400 400

Time at low temperature mode (%) 50 50 20 20 50 50 20 20

ZDDP in oil (wt.%) 0.1 1.5 0.1 1.5 0.1 1.5 0.1 1.5

The catalyst used in all experimental aging runs was identical. A Pd-only, three-way catalyst was used with a total Pd content of 1.90 g Pd/liter of monolith. This catalyst was coated on a ceramic honeycomb monolith with 62 cells/cm2 (0.165 mm wall thickness). The total volume of catalyst present in each experimental converter was 0.69 liters.

3

Aging Results

After aging the converters with the aging cycle options discussed above, catalyst light-off characteristics were used as a measure of catalyst activity. Light-off characteristics of the aged converters were measured on an engine using a standard procedure with a stoichiometric exhaust gas composition. This light-off test procedure used a temperature ramp rate of approximately 20 K/min with a catalyst space velocity of approximately 80,000/hr (at STP). The light-off curves (catalyst conversion vs. inlet exhaust temperature) generated in this fashion were then used to determine the appropriate T50 light-off temperatures (inlet exhaust temperature at which 50% conversion efficiency is observed) for each converter with respect to both hydrocarbons and NOx emissions. These light-off characteristics became the observed outputs for the design of experiment analysis. Analysis of the data was completed using Design-Expert (version 5) software. This software package generated the typical contour plots that stem from the statistical analysis of a Box-Benkhen design. From these contour plots the relationships between the design variables and catalyst performance (T50 light-off temperatures) can be observed. Representative interaction results that stem from the statistical analysis are shown in Figure 1 for catalyst hydrocarbon and and NOx performance. These results are interpolated from the data set for an aging cycle that uses 35 % of the total time at the low temperature mode. With respect to both hydrocarbons (HC) and NOx, catalyst light-off performance degrades when oil is injected into the engine exhaust manifold upstream of the converter (i.e., exposure to higher ZDDP concentrations results in more severe catalyst deactivation). The degree of performance degradation is also more severe when the low temperature aging mode is lowered from 600 °C to 400 °C.

Temperature for 50% Efficiency (C)

220 500 450 400

HC

350

NOx

300 250 600 C, no oil injection

600 C, oil injection

400 C, no oil injection

400 C, oil injection

Aging condition (35% duration for low temp. mode) Figure 1: Impact of aging conditions on catalyst light-off for hydrocarbons (HC) and NOx

4

Aged Catalyst Characterization

Catalysts aged with the various two-mode aging cycles described previously were also characterized for the levels of P and Zn accumulated on the catalyst during the aging protocol. These characterizations included chemical analyzes for total P and Zn levels on the catalyst and electron microprobe investigations to determine the distribution of P and Zn within the catalytic coating. For chemical analyses of P and Zn, aged catalysts were cut in half so information regarding the axial distribution of these poisons could be determined. In general these chemical analyses showed the expected higher distributions of poisons at the inlet side of the converter with lower poison concentrations at the outlet end of the converter. Converters aged with oil injection into the exhaust manifold, and its higher ZDDP levels, showed more than an order of magnitude higher P and Zn levels on the catalyst compared to converters aged without oil injection. For example, with oil injection P levels on the front half of the aged converters were in the 4-5 wt.% range, compared to 0.1-0.25 wt.% for converters aged without oil injection. Poison concentrations in the rear half of the aged catalysts were in general no more than 50% of poison concentrations found in the inlet half of the catalyst. P and Zn levels measured on the converters aged with oil injection were slightly higher for the aging runs completed with the longer duration low temperature mode at 400 °C compared to the 600 °C case. Aging completed with oil injection and the shorter duration low temperature mode showed similar levels of P and Zn for both the 400 °C and 600 °C cases. Based on the throughputs of P and Zn associated with the oil consumption of each aging run and the chemical analyses done on the aged catalysts, the capture rate of P and Zn for each aged catalysts was calculated. Figure 2 summarizes these capture efficiencies for aging runs done with oil injection. With oil injection into the exhaust manifold and the high levels of ZDDP used in the injected oil, capture efficiencies for P and Zn in these aging runs were always greater than 20 % and 10 %, respectively. Consistent with the lower levels of P and Zn measured on the catalysts aged without oil injection, capture rates for P and Zn on aging runs completed without oil injection were significantly lower as shown in Figure 3. P capture efficiencies without oil injection were always less than 9 % and less than 3 % for Zn.

Capture Efficiency (%)

221

40 30 20 10 0

P Zn 600 C, 50% low T 600 C, 20% low T 400 C, 50% low T 400 C, 20% low T Aging condition (all with oil injection)

Capture Efficiency (%)

Figure 2: Capture efficiency of P and Zn by the catalyst during aging with oil injection

10 8 6 4 2 0

P Zn

600 C, 50% low 600 C, 20% low400 C, 50% low 400 C, 20% low T T T T Aging condition (all without oil injection)

Figure 3: Capture efficiency of P and Zn by the catalyst during aging without oil injection

Microprobe analyses of catalysts aged with oil injection in the exhaust manifold showed significant penetration of P and Zn into the washcoat materials coated on the ceramic substrate. The amount of P and Zn, as well as the degree of penetration, was most severe at the inlet end of the catalyst. Consistent with the elemental analyses, the poison levels observed by microprobe decreased in moving toward the rear, outlet end of the catalyst.

5

Discussion and Summary

The results presented here indicate that oil-derived catalyst poisons such as P and heavy metals like Zn are most significantly deposited on three-way catalysts during low temperature engine operating modes. The design of accelerated catalyst aging protocols require modes with low temperature operation with high levels of potential poisons in order to impart the kinds of catalyst deactivation associated with poison accumulation on the catalyst surface. Combining low temperature modes that facilitate poison deposition on the catalyst surface with higher temperature modes appears to drive the poisons deeper into catalyst structure. The aging cycle developed in this study makes use of oil injection doped with high levels of ZDDP upstream of the converter in combination with both low temperature and high temperature operating modes. Adjustable aging parameters such as the level of ZDDP present in the injected oil, the relative duration of the low temperature mode, and the temperature associated with the low temperature mode can all be used to try and create poison profiles and catalyst performance observed after real world vehicle operation. The penetration of P and Zn into the

222 washcoat of three-way catalysts observed by microprobe on the catalysts aged here, has been observed on real vehicle fleets that are known to have large relative consumption levels of engine oil and considerable operating time under low speed, low temperature exhaust conditions. Significantly higher catalyst light-off temperatures for regulated emissions such as hydrocarbons and NOx characterize the catalyst deactivation observed with this oil injection aging protocol and under real world conditions that favor high P and Zn levels on three-way catalysts. As discussed by Darr et al. (6), even modest amounts of poison deposition on a close-coupled catalyst’s inlet face that lead to modest deactivation of a catalyst’s light-off characteristics could be the difference between meeting a SULEV emission requirement or failing to meet these near-zero emission levels. In order to minimize these negative impacts on catalyst performance stemming from oil-derived poisons, it is critical that engine lubricant consumption be maintained at low levels throughout the catalyst regulated lifetime, or alternative lubricant additive packages that minimize the levels of potential poisons such as P and Zn be developed for emission critical applications.

6

References

[1] [2] [3] [4] [5] [6]

Z. Hu and R. Heck, SAE Paper No. 950254 (1995). P. Andersen and J. Rieck, SAE Paper No. 970739 (1997). P. Andersen and T. Ballinger, SAE Paper No. 1999-01-0308 (1999). J. Thoss, J. Rieck, and C. Bennett, SAE Paper No. 970737 (1997). D. Ball, A. Mohammed, and W. Schmidt, SAE Paper No. 972846 (1997). S. Darr, R. Choksi, C. Hubbard, M. Johnson, and R. McCabe, SAE Paper No. 2000-011881 (2000).

223

Catalytic Reduction of NOx in Oxygen-Rich Gas Streams, Deactivation of NOx Storage-Reduction Catalysts by Sulfur Ch. Sedlmaira,b, K. Sehanb, A. Jentysa and J. A. Lerchera a b

Technische Universität München, Institut für Technische Chemie II, Garching, Germany University of Twente, Faculty of Chemical Technology, Enschede, The Netherlands

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

224

225

226

227

228

Catalytic Reduction of NOx in Oxygen-Rich Gas Streams: Progress and Challenges in Catalyst Development Wolfgang Grünert Lehrstuhl Technische Chemie, Ruhr-Universität Bochum

1

Introduction

The forthcoming introduction of tighter emission regulations for car exhaust has created much research effort aiming at new NOx abatement technologies for Diesel and lean-burn engines. This task presents the challenge to reduce NO and NO2 to nitrogen in the presence of excess oxygen, which tends to compete for the reductant employed. Such “selective catalytic reduction (SCR) is well known from flue-gas treatment in power plants where the reductant is ammonia. Since the storage of ammonia in cars is impractical, different solutions have been sought, most of them involving heterogeneous catalysts. The catalytic approaches may be divided into storage-reduction and SCR processes, the latter being differentiated according to the reductant used – hydrocarbons, urea (as an ammonia source), or soot. In the NOx storage-reduction (NSR) approach [1], NOx is stored on a catalyst that contains a storage medium (mostly barium compounds) together with a noble-metal component. When the storage capacity becomes exhausted, the stored nitrate is catalytically reduced during an excursion into rich regime (fuel injection). With real exhaust, the sulfur present in the fuel blocks the storage component rapidly and has to be removed by a high-temperature treatment. It is not yet clear if the forthcoming reductions of fuel sulfur content will be sufficient to pave the way for commercialization of NSR. Selective catalytic reduction of NOx by hydrocarbons (HC) was discovered during research about other NOx abatement approaches – SCR with urea [2] and NO decomposition [3]. The latter has not gained practical interest since then. SCR with hydrocarbons would be the most practical approach utilizing the fuel as an on-board reductant. However, the selectivity of hydrocarbons toward NO (as opposed to O2) is much less than that of ammonia, hence the technique involves a severe fuel penalty. Moreover, SCR with hydrocarbons is much more demanding toward the catalyst than SCR with ammonia (urea), hence progress in catalyst development was unexpectedly slow. On the other side, SCR with urea involves the problem of urea distribution while the catalytic processes (including a urea decomposition stage) appear to be well controlled. There is, however, concern about N2O that may be formed under certain reaction conditions. In addition, the catalysts contain vanadium, which should not be spread into the environment. SCR with soot describes the attempt to use NOx for the combustion of soot held in traps with walls that contain catalytically active materials. This should be differentiated from the Continuously Regenerating Trap (CRT), where soot is burned non-catalytically by NO2 previously produced from the NO present on a noble-metal catalyst. In the CRT, NO2 is reduced only to NO while N2 is the desired product of SCR with soot. Among these techniques, SCR with urea is already commercialized [4] while NSR and SCR with hydrocarbons have real chances for a breakthrough to practical application [5]. While

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

230 NSR is subject of another paper in this book, the present contribution summarizes recent progress in catalyst development for SCR with hydrocarbons and ammonia.

2

Selective Catalytic Reduction with Hydrocarbons

2.1

Basic Problems

SCR with hydrocarbons was originally found with ZSM-5 catalysts highly loaded with copper, which were considered most promising for quite a long time. Meanwhile, many catalyts have been discovered [6–8], most of them consisting of redox components introduced into zeolite matrices. Analogous catalysts based on refractory-oxide supports, e.g. Al2O3, were long considered insufficiently active. A typical feature of most SCR catalysts is a relatively narrow temperature window, in which NO is reduced to N2. Figure 1 shows this for several catalysts. It reflects mostly the situation found in studies with small molecules (propane, propene, isobutane), which are preferred in academic studies. The use of higher hydrocarbons sometimes leads to a significant broadening of this temperature window.

100

a

NO conversion, %

80

c 60

d 40

b

20

0

500

600

700

800

Figure 1: Temperature dependence of NO conversion found with typical SCR catalysts. a) Cu-ZSM-5, reductant propene, 32 000 h–1, from [9], b) Fe-ZSM-5, reductant isobutane, 30 000 h–1 [10], c) Fe-La-ZSM-5, reductant isobutane, 42 000 h–1, from [11], d) Pt/Al2O3, reductant propene, ca. 100 000 h–1, from [12]

231 The SCR reaction appears to be initiated by an oxidative attack of adsorbed NO2 or nitrate intermediate on the HC reductant, which subsequently reduces the N species. This constitutes a high reductant excess (according to stoichiometry, one CH4 molecule reduces 4 NO molecules), and the non-used parts of the reductant molecule are apparently prone to unselective attack by the oxygen present. This leads to a rather modest hydrocarbon utilization, which generally decreases with increasing reaction temperature. It is in the order of 8 %–10 % for good catalysts and small HC molecules at the temperature of peak NO conversion and tends to decrease with increasing chain length of the reductant. Only for methane, which is however difficult to activate, HC utilization may be in the order of 40 % at peak NO conversion. Although Cu-ZSM-5 is considered particularly bad in this respect (which depends, however, on the reductant considered), the catalysts found subsequently offer only gradual improvements of hydrocarbon utilization. Selectivity to N2 is another important issue in SCR with hydrocarbons because N2O may be formed as an undesired side product. Indeed, N2O selectivities between 70 % and 20 % [12] are the major drawback of supported Pt catalysts, which are promising in many other respects (high activity at low temperatures – though with narrow temperature window, poisoning resistance, durability). This disadvantage has not been overcome by now and rules out Pt-based SCR catalysts for practical application. Formation of N2O (even of NO2, at total HC consumption [13]) may be also a problem with oxide catalysts while N2 is almost exclusively formed over redox zeolites. The major obstacles to the practical application of the HC-SCR approach have been, however, the poisoning of many catalysts by H2O and SO2 and the durability problem. While Cu-ZSM-5 is only moderately poisoned by water and SO2 [3], it proved unstable in real exhaust, at least at temperatures above 500 °C [14]. Major goals of catalyst development have been, therefore, to avoid N2O formation with Pt-based systems, to stabilize redox zeolites without compromizing activity properties, and to increase the activities of the more rugged oxide-based systems. It appears that the best success has been achieved so far with the latter catalyst type. 2.2

Examples of new catalyst developments

The effort to counteract poisoning and to stabilize redox zeolite catalysts for the use in real media has lead to several interesting results. Thus, it was found that In-ZSM-5, which belongs to the few systems able to activate methane for the SCR reaction, can be effectively promoted by noble metals [15] and by CeO2 [16]. Examples of the latter case are given in Figure 2, where the pronounced effect of physical CeO2 admixture is demonstrated. In dry medium, NO conversions of >70 % are achieved even at 100 000 h–1 with the methane reductant. Unfortunately, at lower temperatures, which are more relevant to application, Ce-In-ZSM-5 suffers strongly from water poisoning. Figure 2 shows also activity data for a catalyst prepared via a different route (RSSIE, reductive solid-state ion exchange, cf. [16]), where an increased In content resulted in improved durability at a moderate loss of activity. Meanwhile it was found that the low-temperature activity of this system is improved when other hydrocarbon reductants are used. Notably, no loss of activity was found in SCR with higher hydrocarbons after 162 h treatment of the catalyst at 773 K in moist air [17].

232

100

Mixture CeOx/In-ZSM-5

X(NO), %

-1

80

30 000 h -1 100 000 h

60

Ce-In-ZSM-5 (RSSIE) -1 24 000 h 7 % H 2O

40

20

In-ZSM-5 30 000 h

0

600

700

800 T, K

-1

900

Figure 2: NO conversions in the SCR with methane over Ce-promoted In-ZSM-5. In-ZSM-5 with and without CeO2 physically admixed; Ce-In-ZSM-5 catalysts prepared by RSSIE of In and precipiation of Ce [16]. 1000 ppm NO, 1000 ppm methane, 2 or 10 % O2, water content and space velocity indicated in the panel.

Cobalt-exchanged ZSM-5 was the first catalyst found to activate methane for the SCR reaction [18]. It is also useful with other reductants [19], but its stability in feed containing water and SO2 was insufficient although these poisons alone did not deactivate the catalyst [20]. Remarkable progress was made by introducing Co ions into zeolite Beta [21, 22]. This catalyst, which exhibits remarkable stability (cf. Figure 3) is being commercialized for applications with stationary sources. 100

Co-Beta

X(NO, CH4)

80 60

CH4

CH4

40

NO

Co-ZSM-5

20

NO 0

0

1000

2000 3000 Time-on-stream, h

4000

Figure 3: Stability test with Co-ZSM-5 and Co-Beta. Feed: 150 ppm NO, 500 ppm C3H8, 1000 ppm CH4, 10 % O2, 9 % H2O, 0.3 ppm SO2, 500 ppm CO, 250 ppm H2; T = 673 K, 15 000 h–1. From [21].

233 Recently, much attention has been paid to “overexchanged” Fe-ZSM-5, in which iron is introduced into ZSM-5 to a Fe/Al atomic ratio of 1. This interest was generated by work of Feng and Hall [23] describing excellent activity and stability for overexchanged Fe-ZSM-5 obtained by a preparation, which has never been reproduced despite considerable effort. Later, Chen and Sachtler [24] proposed a preparation based on chemical vapor deposition of FeCl3, which produces catalysts that are less active but still have high resistance to poisoning and encouraging stability properties. Figure 1 gives some activity data for this type of catalyst. The best results were obtained with La-promoted materials [11]. Figure 4 illustrates the stability of the Fe-La-ZSM-5 catalyst at the temperature of peak NO conversion [11]. The slow activity decay is due to coke deposits, hence, catalyst regeneration is straightforward.

Conversion, Yield, %

100

NO 80

isobutane 60

CO

40

CO2

20 0

10 % O2/He 2 h, at 773 K 0

20

40 60 80 100 Time-on-stream, h

120

Figure 4: Stability test with Fe-La-ZSM-5. 2000 ppm NO, 2000 ppm isobutane, 3 % O2, 10 % H2O, T = 623 K, 42 000 h–1. From ref. [11].

Fe-ZSM-5 remains a challenge for academic research since recent, still unpublished results of several groups suggest that the original claims by Feng and Hall [23] were real, but that different ways have to be found to obtain the corresponding catalyst structure. The present versions of overexchanged Fe-ZSM-5 are considerably less active with other hydrocarbons (propane, propene). Severe durability tests have not been reported by now. However, there is considerable potential for improvement even with Fe-ZSM-5 prepared by CVD of FeCl3. Alumina-based SCR catalysts were developed by several groups, typically with Ag or Co promoters. While their stability was attractive, their activity was insufficient. Significant progress has been made recently by application of a new mesoporous alumina developed in the Institute of Applied Chemistry Berlin-Adlershof. Figure 5 shows activity data of a material containing 0.6 % Ag and 0.4 % Co on this alumina [26]. High conversions are obtained in a broad temperature window with the reductant n-decane. The N2 selectivity, which is sometimes a problem with alumina-based SCR catalysts, was >90 % at all temperatures. A completely different, interesting concept was proposed by Iwamoto [26]. Based on the fact that NO2 is an intermediate in many catalytic systems, the NO oxidation (mostly over Pt) and the remaining steps are performed in separate catalyst layers at different temperatures, with intermediate reductant addition. Significant improvements of activity and HC utilization were obtained.

234

4

Selective Catalytic Reduction with Ammonia (Urea)

Although there are many potential catalysts for the SCR with NH3, the use of promoted V2O5/ TiO2 catalysts, in particular V2O5/WO3/TiO2 is predominant in practical applications. These catalysts are very active and robust and have, therefore, been employed also in NOx converters for diesel engines. Similar to SCR with hydrocarbons, the reductant ammonia becomes oxidized at high temperatures. Over SCR catalysts based on the V2O5/TiO2 system, this oxidation is not selective for the harmless product N2, but NO and N2O are also formed. The former limits the NO conversion, while the latter will be emitted.

100

N2 yield, %

80

60

60

40

40

20

20

0 500

Decane conversion, %

100

80

0

600

T, K

700

800

Figure 5: Catalytic behavior of a Co-Ag/Al2O3 catalyst. 1000 ppm NO, 550 ppm decane, 6 % O2, 10 % H2O, 12 % CO2, at ca. 100 000 h–1 [25]

Recently, highly active catalysts for the SCR of ammonia have been developed on the basis of redox zeolites. Among the new catalysts are Ce-MOR [27], Cu-zeolites (Y, MOR, ZSM-5, [28]), and overexchanged Fe-ZSM-5 [29, 30]. Figure 6 shows the example of Fe-ZSM-5 pre-prepared by CVD of FeCl3. With a 1:1 NO/NH3 feed, NO conversions of 80 % are obtained at 300 000 h–1 over a wide temperature range, and the reaction is even promoted by moisture in the feed. With slight NH3 excess, the conversions are even higher, and the excess NH3 is oxidized to N2 over the whole temperature range [29]. In a 50 h durability test at 848 K (16 h each in dry feed, moist feed (2.5 % H2O) and SO2-containing feed (+300 ppm SO2)), no evidence of deactivation was found. An interesting stabilization effect has been observed with Cu-ZSM-5 catalysts, the activity of which is similar to that of overexchanged Fe-ZSM-5 [31, 32]. Steaming at 873 K (17 h, 3 % H2O/Ar) led to a significant loss of NO conversion at all reaction temperatures studied. However, this effect was absent when the zeolite crystals were supported on a Ni surface (Raney Ni), instead, the NO conversion at high temperatures increased after steaming. It was proposed that Ni ions migrate into the zeolite where they become catalytically active and exert a stabilizing influence on the zeolite matrix[32].

X(NO), X(NH3), %

235

100 + 2.5 % H O 2 NH3 NO 80

dry

60 40

CVD

conventional

20 0 500

600

700

800

900

Figure 6: SCR of NO with ammonia over Fe-ZSM-5 prepared by CVD of FeCl3, comparison with conventional preparation. 1000 ppm NO, 1000 ppm NH3, 2 % O2, 300 000 h–1. From [29].

5

Conclusions

Since the early days of SCR with hydrocarbons, new materials have been found that give rise to the expectation of a breakthrough of this approach to commercial use including mobilesource applications. New zeolite-based catalysts for SCR with ammonia also offer an interesting application potential.

6

References

[1] [2] [3] [4] [5]

N. Takahashi, H. Shinjoh, T. Ijima, T. Suzuki, et al., Catal. Today 1996, 27, 63. W. Held, A. König, T. Richter, L. Puppe, SAE-Paper 1990, No. 900 496. M. Iwamoto, H. Yahiro, S. Shundo, Y. Yu-u, N. Mizuno, Shokubai 1990, 32, 430. G. Fränkle, W. Held, W. Hosp, W. Knecht et al., VDI-Ber. Ser. 12, 1997, 365. J. Armor, “Opportunities, Strategies and Innovation for Catalysis in the New Millenium” (Plenary lecture), Europacat-V, Limerick (Ireland), September 2001. A. Fritz, V. Pitchon, Appl. Catal. B, 1997, 13, 1. V. I. Parvulescu, P. Grange, B. Delmon, Catal. Today 1998, 46, 233. Y. Traa, B. Burger, J. Weitkamp, Microporous Mesopor. Mat., 1999, 30, 3. T. Liese, W. Grünert, J. Catal. 1998, 172, 34. F. Heinrich, E. Löffler, W. Grünert, Stud. Surf. Sci. Catal. 2001, 135, 30–P-27. H.-Y. Chen, W. M. H. Sachtler, Catal. Lett. 1998, 50, 125. R. Burch, P. Millingham, Catal. Today 1995, 26, 185. F. C. Meunier, J. P. Breen, V. Zuzaniuk, et al., J. Catal. 1999, 187, 493. D. R. Monroe, C. L. Di Maggio, D. D. Beck et al., SAE-Paper 1993, Sp-968, 195. E. Kikuchi, M. Ogura, N. Aratani, Y. Sugiura, et al., Catal. Today 1996, 27, 35.

[6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

236 [16] F.-W. Schütze, H. Berndt, M. Richter, B. Lücke, C. Schmidt, T. Sowade, W. Grünert, Stud. Surf. Sci. Catal., 2001, 135, 10-O-01. [17] F.-W. Schütze, H. Berndt, to be published. [18] Y. Li, J. N. Armor, Appl. Catal. B 1992, 1, L31. [19] S. Sato, Y. Yu-u, H. Yahiro, N. Mizuno, M. Iwamoto, Appl. Catal. 1991, 70, L1. [20] Y. Li, J. N. Armor, Appl. Catal. B 1993, 3, L257. [21] T. Tabata, M. Kokitsu, H. Ohtsuki, O. Okada, et al., Catal. Today 1996, 27, 91. [22] T. Tabata, H. Ohtsuki, et al., Microporous Mesopor. Mat. 1998, 21, 517. [23] X. Feng, W. K. Hall, Catal. Lett. 1996, 41, 45. [24] H.-Y. Chen, W. M. H. Sachtler, Catal. Today 1998, 42, 73. [25] M. Richter, M. Langpape, S. Kolf, R. Fricke, lecture at Europacat-V, Limerick (Ireland), September 2001. [26] M. Iwamoto, Stud. Surf. Sci. Catal. 2000, 130, 23. [27] E. Ito, R. J. Hultermans, M. H. W. Burgers et al., Appl. Catal. B 1994, 4, 95. [28] S. Kieger, G. Delahay, B. Coq, B. Neven, J. Catal. 1999, 183, 267. [29] A. Z. Ma, W. Grünert, J. Chem. Soc., Chem. Commun. 1999, 71. [30] Q. Sun, Z. X. Gao, H. Y. Chen, W. M. H. Sachtler, J. Catal. 2001, 201, 89. [31] A.-Z. Ma, M. Muhler, W. Grünert. Chem. Eng. Technol. 2000, 23, 3. [32] A.-Z. Ma, M. Muhler, W. Grünert, Appl. Catal. B 2000, 27, 37.

Atomic Structure of Low-Index CeO2 Surfaces Holger Nörenberga, J. H. Hardingb and S. C. Parkerc a

University of Oxford, Department of Materials University College London, Department of Physics and Astronomy c University of Bath, Department of Chemistry b

1

Introduction

Polycrystalline cerium oxide is used in catalytic converters for cars as an additive to the washcoat [1]. The cerium oxide acts as an oxygen buffer for reduction and oxidation reactions because it can easily transform between CeO2 and Ce2O3. Polycrystalline materials contain crystallites, which mainly exhibit low-index surfaces because they are usually the low-energy surfaces. We have chosen single crystals of low index (111) and (001) orientation to study the atomic structure of CeO2 because polycrystalline material is unsuitable for detailed investigation by Scanning Tunnelling Microscopy (STM). The geometric and electronic structure of sufficiently conductive surfaces can be studied by STM with atomic resolution.

2

Experiments and Modelling

Commercial Crystal Labs Inc. supplied single crystals of CeO2, cut and polished to obtain lowindex surfaces. The crystals were placed in the UHV-chamber and then annealed to remove surface contamination [2]. STM experiments were carried out at elevated temperatures with a JEOL JSTM 4500-XT at a base pressure in the low 10–10 mbar range at temperatures up to 400 ºC. Auger-electron spectroscopy and X-ray photoelectron spectroscopy were used to check for the cleanness of the surface. No contamination was detectable. We have simulated the surface structures using the MIDAS and CHAOS [3] programs. The first program considers the crystal as a stack of planes defined by the interface normal. Atoms in the planes close to the interface are relaxed to positions of zero force. Atoms in the outer planes are held rigid, but the stack is allowed to move to permit the interface to dilate. The second program takes the relaxed surface structure from the first program and inserts the point defect. The crystal is then divided into two regions. In the inner region the ions are relaxed to positions of zero force. In the outer region, the response of the crystal to the defect charge is calculated using a dielectric continuum approximation. We use a central-force pair potential model for ceria (including electronic polarization by the shell model). The electrostatic terms are handled by the standard two-dimensional summations. A number of model potentials are available for CeO2; we use that of ref [4]. A more detailed discussion of methods and the model can be found in ref. [5]

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

238

3

Results and Discussion

3.1

CeO2(111)

Figure 1a shows an STM image of the CeO2(111) surface after preparation. The surface is (1×1) terminated and shows no reconstruction. This is in agreement with previous calculations, which found the (111) surface to be lowest in energy [6]. Oxygen is located in the top layer and the top layer is imaged by STM [2]. After annealing to about 1000 °C defects appear on the surface. Figure 1b shows a number of triangular defects. It has already been shown that arrangements of oxygen vacancies in pairs lead to a lower energy than two single oxygen vacancies on the surface. In order to study defect agglomeration in more detail we have calculated the energy of a number of defect arrangements consisting of three oxygen vacancies. Preliminary results suggest that indeed the triangle appears to be the energetically most favourable vacancy arrangement. The cluster consists of the triangle of vacancies together with the polarons (Ce3+ ions) on the layer below to balance the charge. This cluster is stabilised by the attraction of the polarons to the vacancies and also by the polarisation of the lattice arising from the defect dipole.

Figure 1: STM images of the CeO2(111) surface at different stages of annealing under UHV conditions, image size 6 nm × 4 nm, a) showing a (1×1) termination after annealing to 950°C, TSTM = 300°C, VBIAS = –2.5 V, IT = 0.2 nA, b) triangular defects after annealing to 1000 °C, TSTM = RT, VBIAS = –2.5 V, IT = 0.01 nA c) line defects after annealing to 1030 °C, TSTM = 500 °C, VBIAS = –2.5 V, IT = 0.08 nA, d) “black holes” after repeated annealing to 1030 °C, TSTM = 500 °C, VBIAS = –2 V, IT = 0.05 nA

239 Figure 1c shows line defects consisting of oxygen vacancies introduced after further annealing. They are running along the [1¯10], [0¯¯ 11] and [10¯1] directions. At this stage there is not enough space left on the surface to form additional triangular defects. In the vicinity of the triangular defects the stoichiometry is Ce2O3 rather than CeO2. Figure 1d shows the surface structure after repeatedly annealing the same crystal to 1030 °C. At this stage the atomic ordering of the surface is destroyed and a hexagonal pattern of “black holes” has formed. It is noteworthy that the distance between these “black holes” is about 1.6 nm, which is the minimum possible distance between two triangular defects on the (111) surface. The rough surface structure (apparent height of 0.25 nm for the “black holes”) indicates the involvement of deeper layers in the formation of this structure. The temperature above 1000 °C (at which the crystal was annealed) is the temperature range where catalytic converters irreversibly lose their efficiency. 3.2

CeO2(001)

Reconstruction is expected to occur on this surface because an unreconstructed surface would have a macroscopic dipole and therefore infinite surface energy. This is due to the stacking sequence of alternating planes with charge densities of equal magnitude but opposite sign in the [001] direction. Figure 2 shows an STM image of the CeO2(001) surface. The surface reconstructs as Ö2/2(3×2)R45°. The bright, vertical rows in figure 2a are separated by 1.15 nm which is equal to three times the lattice constant in the [110] direction. On these bright lines, atomically resolved features 0.77 nm apart (twice the lattice constant in the [¯ 110] direction) are visible. Because an oxygen terminated CeO2(001) surface is much lower in energy than a Ce-terminated one we conclude that these bright features in fig. 2a are related to oxygen.

[¯110] [110] 1 nm a) b) Figure 2: a) STM image of the CeO2(001) surface, TSTM = 300°C, VBIAS = –2.5 V, IT = 0.2 nA, the inset shows an area which the model in b) refers to; b) model of a Ö2/2(3×2)R45° reconstructed surface with an oxygen coverage of 33 % (top view and side view)

We have used the experimental observation as a starting point for atomistic modelling [5]. We have calculated the surface energy for the CeO2(001) surfaces. We varied the symmetry of the surface and the amount of oxygen in the top layer. The result of the calculations is that indeed a Ö2/2(3×2)R45° reconstruction is lowest in energy. This type (3×2) symmetry can be maintained over a rather wide range of oxygen coverage, namely between 50 % and 25 %.

240 Figure 2b shows a Ö2/2(3×2)R45° reconstructed surface with an oxygen coverage of 33 % which has a 0.7 eV (per oxygen atom removed) lower energy compared to reduction in the bulk. In this structure two thirds of the cerium ions in the second layer have been reduced from Ce4+ to Ce3+ in this structure. Based on the apparent height in the STM images and our calculations we conclude that the bright features shown in fig. 2a are oxygen atoms in the top layer of the surface. 3.3

CeO2(110)

The (110) surface of CeO2 has a higher surface energy than the (111) surface. After annealing to about 940 °C we found a (2×1) surface reconstruction [7]. After further annealing to about 1030 °C the surface forms (111) facets to reduce the surface energy. Ridges (>130 nm long) running along the [1¯¯ 10] direction are formed.

4

Conclusions

We have shown the structure of low-index surfaces of CeO2 with atomic resolution. STM at elevated temperatures and/or very low tunnelling currents are crucial to obtain this resolution. The reduction of CeO2 as it may happen in a catalytic converter was illustrated on the (111) surface of a single crystal of CeO2. Starting from a (1x1) terminated surface, agglomerations of oxygen vacancies in triangular arrangement could be observed after annealing the crystal under extremely oxygen lean conditions (UHV). Upon further annealing more oxygen vacancies are introduced forming lines in different, equivalent crystallographic directions. The CeO2 (001) surface shows a strong tendency for reconstruction. Calculations showed that it is able to maintain the same surface symmetry for oxygen coverage between 0.25 and 0.5.

5

References

[1]

G. Ertl, E. S. J. Lox, B. H. Engler, Environmental catalysis – mobile sources, in: G. Ertl, H. Knözinger, J. Weitkamp (Eds.) Handbook of Heterogeneous Catalysis, Vol. 4, VCH, Weinheim, 1997, p. 1559. H. Nörenberg and G. A. D. Briggs, Phys. Rev. Lett. 1997, 79, 4222–4225, Surf. Sci. 1999, 424, L352–L355. P.W. Tasker, Philos. Mag. A 39, 1979 119–136; D.M. Duffy and P.W. Tasker, A guide to CHAOS; a program for the calculation of point defect energies near interfaces in polar crystals 1983 Harwell Report AERE-R 11059. G.V. Lewis and C.R.A. Catlow, J. Phys. C 1985, 18, 1149–1161 H. Nörenberg and J. H. Harding, Surf. Sci. 2001, 477, 17–24. J. C. Conesa, Surf. Sci. 1995, 339, 337–352. H. Nörenberg and G. A. D. Briggs, Surf. Sci. 1999, 433–435, 127–130.

[2] [3]

[4] [5] [6] [7]

Nanostructured Ceria-Zirconia as an Oxygen Storage Component in Three-Way Catalytic Converters-Thermal Stability Boro Djuričić1 and Stephen Pickering2 1 2

Austrian Research Centers, A-2444 Seibersdorf, Austria Institute for Advanced Materials, 1755 ZG Petten, The Netherlands

1

Abstract

Nanostructured CeO2 and CeO2-ZrO2 solid solutions with Ce/Zr ratios of 0.75/0.25 and 0.50/0.50 were produced from aqueous solutions by precipitation/coprecipitation and hydrothermal crystallization in an autoclave at 180 °C for times up to 10 h. The thermal stability of these systems was studied by X-ray diffraction (XRD), differential thermal analysis (DTA) thermogravimetric analysis (TG) and transmission electron microscopy (TEM). Initial crystallite size was about 3nm for all powder compositions as measured by XRD peak broadening. Powders were calcined in air at 300 °C, 500 °C, 800 °C and 1000 °C for periods of up to 600 minutes. The crystallite sizes in pure Ceria , Ce75/Zr25 and Ce50/Zr50 powders were 137 nm, 23 nm and 37 nm respectively after 600 minutes at 1000 °C indicating a significantly higher thermal stability, i.e. greater resistance to coarsening, in nanostructured ceria-zirconia solid solutions. TEM images of the pure ceria powder showed clear evidence of sintering after calcination at 1000 °C. In contrast, TEM images of the ceria-zirconia solid solution powder calcined under similar conditions showed it to be still loosely agglomerated with significantly smaller crystallite size. This finding suggests that ceria-zirconia solid solution have higher effectiveness in service compared with pure ceria due to higher thermal stability.

2

Introduction

The three-way catalytic converter has probably made a greater contribution to reducing emissions from automobile exhausts than any other technology. Nevertheless, there is a continuing need to improve the performance of catalytic converters, particularly with respect to their ability to withstand high temperatures in order to allow them to be mounted closer to the engine and to extend their useful operating life. Cerium oxide plays an important role in catalytic converters as an oxygen storage medium. To maximise its effectiveness in this role, the cerium oxide should be distributed uniformly throughout the washcoat layer as very fine particles to ensure that a high specific surface area is exposed to the exhaust gas stream. Unfortunately, nanostructured cerium oxide powders suffer from pronounced coarsening on heating to temperatures typically encountered in exhaust systems. This paper demonstrates that crystal growth in ceria can be significantly reduced by means of a suitable additive.

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

242 Zirconia is a suitable additive because it can also contribute to oxygen storage through its ability to change stoichiometry. About 3–5 times higher oxygen storage capacity in ceriazirconia system than in ceria have been reported [1] which is attributed mainly to the enhancement of the bulk O2– ion mobility via vacancy diffusion in ceria-zirconia system. The goal was to produce material that could maintain a high specific surface area and good oxygen storage characteristics at high temperature. Compositions of ceria-zirconia mixed oxide with Ce/Zr ratios of 0.75/0.25 and 0.50/0.50 were produced for comparison with pure ceria.

3

Experimental

3.1

Preparation of powders

The powders were produced from aqueous solution by a novel two-stage precipitation process using hydrogen peroxide and ammonium hydroxide solutions. A 0.1 mol/l solution was made by dissolving zirconyl oxynitrate solution in de-ionised water. This solution remained slightly white and cloudy which may be attributed to the tendency of many Zr(IV) species to polymerize [2]. Addition of hydrogen peroxide resulted in a fully transparent solution with a pale yellow-green color. These changes in color and transparency indicate that a more soluble zirconium species was present after addition of hydrogen peroxide. The addition of cerium(III) nitrate solution to hydrogen peroxide is believed to result in the oxidation of cerium ions to cerium(IV) which is more easily hydrated and readily forms complexes to initiate the precipitation process. Subsequent addition of ammonium hydroxide completed the precipitation with the formation of an orange-brown gel accompanied by an intensive evolution of gas. The color of the precipitate changed on washing to orange-brown and finally to yellow-orange. The powder was autoclave treated at 180 °C to improve dispersability by removing hard agglomerates. For ceria this two-stage precipitation process [3] was earlier found to yield smaller crystallites (3nm) than by precipitation with ammonia (15nm). Samples of the powders were calcined in air at 300 °C, 500 °C , 800 °C and 1000 °C for periods of up to 600 minutes. 3.2

Analysis of powders

The ceria and ceria-zirconia powders were analysed by various techniques to determine phase composition, particle size and particle morphology. The precipitated powders were characterized by differential thermal analysis and thermogravimetry (DTA/TG, Netsch STA 409) in a dry air atmosphere using a heating rate of 5 °C/min to a temperature of 1400 °C. Samples of precipitated and calcined powders were prepared for examination by transmission electron microscopy (TEM, Philips EM 400) by ultrasonically dispersing the powder in ethanol and dipping a carbon-coated copper grid into the dispersion. After removing the grid from the suspension, the alcohol was allowed to evaporate. The phase composition was determined by x-ray powder diffraction (XRD, Philips PW 3710). A Voigt fitting routine was used to measure the broadening of the (220) and (311) peaks and the apparent crystallite size was calculated using the Scherrer equation: Dapp = Kl/Bcosq where K = 0.9, B = peak width, and l = 1.5406 Å for CuK= radiation. The line broadening due to the limits of instrument resolution was determined using a LaB6 standard.

243

4

Results

4.1

Differential thermal analysis/ thermal gravimetry

The precipitated pure ceria powder precursor showed a weight loss of 17 %–18% on heating to 500 °C and a well-defined exothermic peak at 260 °C–280 °C which probably corresponds to crystallization of CeO2. These data indicate that the precipitated species was CeO2.2H2O which has a theoretical weight loss of 17.3 %. 0.15

100

95

CeO 2

0.05

90

0.00

85

95

Ce0.75Zr0.25O 2

0.05

90 0.00

TG, % initial weight

0.10

0.10 DTA, mV/mg

100

85 -0.05 0

200

400

600

80 800 1000 1200 1400

-0.05 0

200

400

600

800 1000 1200 1400

Temperature, oC

Temperature, o C

Figure 1: DTA and TG analysis of ceria and ceria-zirconia powders after drying at 85 °C

The DTA and the TG data for the ceria-zirconia powders were broadly similar to that for pure ceria powders. In particular there was no sign of the ‘glow exotherm’ at about 400 °C typical of pure zirconia powders. The data therefore indicate that the two cation species precipitated as a solid solution rather than as separate phases. 4.2

X-ray diffraction

The initial crystallite size was about 3nm for all powder compositions as measured by XRD peak broadening The X-ray spectra for pure cerium oxide samples showed that cubic cerium oxide was the only phase present. The amount of peak broadening was the same for samples that were air-dried at 85 °C, autoclave treated at 180 °C for 4 h, or calcined in air at 300 °C, and corresponded to a calculated size of 3 nm. Calcination at higher temperatures resulted in narrower peaks corresponding to significantly larger crystallite sizes as shown in Table 1. After calcination at 1000 °C the powder was seen to be more coarsely agglomerated. Table 1: Average crystallite size [nm] after calcination in air 300 °C 60 min Ce100 3 Ce75Zr25 3 Ce50Zr50 3

500 °C 60 min 6 4 4

6 min 19 7 6

800 °C 60 min 25 8 8

600 min 40 10 10

6 min 72 18 13

1000 °C 60 min 79 25 14

600 min 137 37 23

244 The X-ray spectra of the ceria-zirconia powders were all generally similar to those for pure ceria, showing that cubic solid solutions of zirconia in ceria were present. The spectra for the Ce75Zr25 samples corresponded most closely to JPDS card 28-0271 for Ce0.75Zr0.25O2. The spectra for the Ce50Zr50 samples, however, corresponded slightly better to JPDS card 38-1439 for Ce0.6Zr0.4O2 (cubic) than to card 38-1439 for Ce0.5Zr0.5O2 (tetragonal). This shift is indicative of a change in lattice parameters of the solid solution. The valence change from Ce4+ with ionic radius 1.18 Å [4] to Ce3+with ionic radius 0.88 Å [5] causes phase separation [6]. The reduction of Ce4+to Ce3+, means that the system ZrO2-CeO2 transforms into the system ZrO2-CeO1.5 (Fig 2) with an entirely different phase composition; in fact the ternary system ZrO2-CeO2-CeO1.5 needs be considered. The presence of many defective tetragonal and cubic structures at temperatures below 1000 °C were reported for 50 mol% CeO2 in zirconia [10] and 65 mol% CeO2 in zirconia [11]. The cubic compound Ce2Zr2O7 with the pyrochlore (P) structure is stable in the composition area from about 44 mol to about 57 mol CeO1.5 [7,8,9]. Ce2Zr3O10 (B-phase) has been reported over a wide range of compositions in the CeO2-ZrO2 system at temperatures below about 800 °C [12,13]. The change in crystal symmetry is strongly effected by oxygen stoichiometry (i.e. by the atmosphere to which the material is exposed). Despite of the slow rate at which phases separate to the equilibrium state, these phase separations might be capable of playing a crucial role in the behaviour of this catalyst component, both with respect to the oxygen storage and catalytic activity. T °C

T°C

3000

3000

liquid

liquid

C2 + A

C2

A

C1

C

2000

C11 + P

T + C1

2000

T T

P

t+p T+P

1000

P + C2

1000

T+C

M M+P M

20

ZrO2

40 6060

Mol% Mol% Mol%

8080

CeO2

20

ZrO2

40

60

Mol%

80

CeO1.5

Figure 2: Phase diagram: ZrO2-CeO2 [7] and ZrO2-CeO1.5 [8,9]

If the reversible phase separation were to occur rapidly then, in principle, recrystallisation should lead to a “grain-refining” effect in which crystallites which have grown large at high temperatures would nucleate small crystallites of different phase composition, thereby maintaining a high specific surface area. This phenomenon could make it possible to design materials with a “self-repairing” oxygen storage capability.

245 The peak broadening of samples treated at up to 300 °C was the same in the ceria-zirconia samples as in the pure ceria samples, corresponding to a crystallite size of 3nm. After calcination at 800 °C and 1000 °C both the Ce75Zr25 and Ce50Zr50 ceria-zirconia samples showed a small but systematic shift in peak positions to smaller angles. The pure ceria samples showed no such shift. The peaks widths of the ceria-zirconia samples remained broader than for the pure ceria samples calcined under the same conditions, indicating significantly less crystallite growth than in pure ceria as shown in Table 1. The crystallite sizes in pure ceria, Ce75/Zr25 and Ce50/Zr50 powders were calculated to be 137 nm, 23 nm and 37 nm respectively after 600 mins at 1000 °C indicating a significant beneficial effect of the zirconia additive. 4.3

Electron microscopy

Figure 3 shows the morphology of weak agglomerates in ceria and ceria-zirconia powders after hydrothermal crystallization.

A

10nm

B B

10nm

Figure 3: Powders after hydrothermal crystallisation; (A) Pure ceria and (B) Ce50/Zr50

Figures 4 and Figure 5 show the morphology of calcined ceria and ceria-zirconia powders. The crystallite sizes visible in TEM images always appeared to be considerably smaller than those calculated from XRD peak broadening. To some extent this is to be expected as peak broadening values are very sensitive to the presence of a few large particles. Peak broadening values relate more to the mass median diameter whereas TEM image appearance relates more closely to the count median diameter so that considerable differences can be expected between the two values for the type of log-normal particle size distributions typically found in ceramic powders. Nevertheless, the TEM images of the pure ceria powder after calcination systematically showed a significantly larger crystallite size than in the Ce75/Zr25 and Ce50/Zr50 powders, thus confirming the same general trend shown by the calculated XRD values. There was clear evidence of sintering in the pure ceria powder after calcination at 1000 °C for 10h. In contrast, TEM images of the Ce75/Zr25 and Ce50/Zr50 powders calcined under similar conditions showed them to be still loosely agglomerated and the powder could be easily dispersed in alcohol to form a stable suspension.

246

A

B

10nm

C

10nm

10nm

Figure 4: Powders after calcination in air at 800 °C for 1 h; (A) Pure ceria, (B) Ce75/Zr25 and (C) Ce50/Zr50

50nm

A

B

50nm

Figure 5: Powders after calcination in air at 1000 °C for 10 h; (A) Pure ceria and (B) Ce50/Zr50

5

Conclusions

Ceria and ceria-zirconia mixed oxide powders were produced by a two stage precipitation process using hydrogen peroxide and ammonia. This precipitation process produced a significantly smaller crystallite size than is obtained by conventional precipitation by ammonium hydroxide alone. The ceria-zirconia compositions showed considerably less crystallite coarsening than pure ceria on calcination in air at temperatures up to 1000 °C. The ceria-zirconia compositions therefore have potential advantages over pure ceria for application as the oxygen storage component of 3-way catalytic converters.

247

6

References

[1] [2] [3] [4] [5] [6] [7] [8]

C. E. Hori et al, Applied Catalysis B: Environmental,1998, 16, 105–17. A.Clearfield, Rev. Pure and Appl. Chem. 1964, Vol 14, 91–108. B.Djuričić and S.Pickering, J. Eur. Ceram. Soc. 1999, Vol 19, 1925–1934. W. M. Goldschmidt, Berichte Deutsche Keram. Ges. 1963, 60, No 5. W. H. Zachariasen, Phys. Rev. 1948, 73, No 9. C. Leach and N. Khan, J. Mater. Sci. 1991, 26, [8], 2026–30. P. Duwez and F. Odell, J. Amer. Ceram. Soc.1950, 33, [9], 274–83. A. I. Leonov et all, Izvestiya Akademii Nauk USSR, Neorganicheskie Materialy, 1966, Vol. 2, No 1, 137–44. A. I. Leonov et al, Ogneupory,1966, No 3, 42 – 48. M. Yashima et al, J. Amer. Ceram Soc.1993, 76, [7], 1745–50. M. Yoshima et al, J. Amer. Ceram. Soc.1993, 76, [11], 2865–68. E. Tani et al, J. Amer. Ceram. Soc. 1993, 66, [7], 506–510. P. Duran et al, J. Mater. Sci. 1990, 25, 5001–06.

[9] [10] [11] [12] [13]

V Recycling

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

Recycling Technology for Metallic Substrates: a Closed Cycle Clemens Hensel Demet Deutsche Edelmetall Recycling AG & Co. KG, Alzenau (Germany)

1

Abstract

Emissions during the useage phase of vehicles are of increasing interest in environmental protection, and consequently, there is considerable interest in exhaust systems. The automotive exhaust system including the catalytic converter is, because of the precious metals in the catalyst, of particular interest for the recycling of automotive parts. The paper will describe the recycling technology of ceramic and metal catalyst substrates. The process will be analyzed in detail with the example of metal supports. As a result the complete life cycle and the recycling efficiency are presented.

2

Introduction

As early as the sixties, the first attempts were made to reduce the harmful substances in car exhaust gases by means of catalytic converters. As a result, converters were first employed in the USA in the seventies and were introduced to Europe in the eighties. While the emphasis at first was on the oxidation of carbon monoxide (CO) and hydrocarbons (HC), nitrogen oxides (NOx) are also reduced in today's three-way converters. The chemical reaction is achieved in a stoichiometric equilibrium with precious metal catalysts, platinum, rhodium and palladium. In the reduction processes, this chemical equilibrium ensures that precisely enough oxygen is released for oxidation. As its name indicates, the catalyst is not a collector (filter) but a converter of exhaust gases. A large reactive surface of the catalyst is necessary to convert the exhaust gases which are led through the exhaust train at high speed. This is ensured by the structure of the substrate and by the surface-enlarging wash coat. The surface of a conventional substrate is approx. 3 m² at 1 liter of volume and is increased to several thousand m2 by the wash coat. Efforts to increase catalyst efficiency tends to bring about further increases in specific geometrical surface area. The first catalysts for motor vehicles were manufactured from coated loose material (pellets). However, the relative movements of the parts led to wear and tear due to friction. Ceramic monoliths therefore displaced them almost completely. Increasingly stricter exhaust gas legislation made it necessary for the efficiency of converters to be continually increased. The metal converters – developed at the end of the eighties which are manufactured from thin brazed metal foil – offer some advantages in this respect. One advantage is the increased specific geometric surface area compared to that of the ceramic substrate. As efficiency is dependent first and foremost on the area of the geometric surface (where the parameters are otherwise

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

252 identical), the metal catalyst can be reduced in volume while remaining equally effective. This means that the quantity of precious metals required is likewise reduced. In addition, the employment of especially thin metal foil reduces the thermal mass of the substrate so that it reaches its working temperature far more quickly. Metal substrates are therefore employed to an increasing degree. The recycling procedure used for ceramic substrates is only partly suitable for metal substrates. For this reason, a new process, described in the following, was developed.

2

Differences in the Structure of Catalysts

2.1

Structure of a ceramic monolith catalyst

The ceramic substrate has a honeycomb structure. The exhaust gas flows through the coated channels and contacts the precious metals. The monolith is surrounded by an insulating element (ceramic fibre mat) whose purpose is not just to insulate the catalyst thermally but also to secure it mechanically. As with a silencer, the steel case is the outer “packaging”.

Casing Ceramic fibre mat Ceramic substrate

Figure 1: Ceramic catalyst incl. ceramic fibre mat and casing [1]

2.2

Structure of a Catalytic Converter with Metal Substrate

The metal substrate is made of flat and corrugated metal foils. These are stacked in layers and are wound into a cylindrical or elliptical body. This body is then pressed into a metal casing and is joined by a high-temperature brazing process. The alternating flat and corrugated layers produce a channeled structure which can then be coated in order to increase its surface area – just like the ceramic substrate. Due to the integrated casing, no further measures are required to

253 secure the substrate. If need be, additional heat insulating sheets can be applied for purposes of thermal shielding. These two types of monoliths differ in form, size, weight and precious metal loadings as a result of different designs for the exploitation of specific advantages of the two techniques.These differences in structure bring about different requirements with regard to the recycling of these components.

Figure 2: Structure of a catalyst with metal substrate

3

Recycling Procedures

3.1

Requirements for Recycling

As the precious metals are applied only to certain parts of the converter (in the wash coat), the first step to be taken in the technical process of recycling is a relatively simple one: mechanically seperate the washcoat and precious metals from the substrate. The next step is then to homogenize the precious metal fraction. This is crucial for determining the value of recycling batches, i.e. statistically representative samples to acquire an analytical sample which is analyzed for its precious metal content. The results of this analysis form the basis for the settlement of accounts with the supplier. As described earlier, the ceramic monolith is only layed into the casing, surrounded by the ceramic fibre mat. Since the mechanical properties of the monolith are extremely different

254 from those of the outer steel mantle, the monolith can be seperated relatively easily. The ceramic material obtained by breaking down the catalyst can be easily ground to a homogeneous powder which is highly suitable for sampling. Initially, the first recycling procedures used for metal substrates had to do without this sort of separation. Instead, the entire catalyst was processed in a melting procedure. Apart from the high costs of melting, this has a further disadvantage in that instead of concentrating the precious metals, the precious metals are diluted throughout the whole melt (including the case). Such a process is less effective; in addition, it is difficult to obtain a representative sample from the homogeneous material. In the procedure described below, these disadvantages can be eliminated. 3.2

Newly Developed Separation Process for the Recycling of Metal Catalysts

3.2.1 General Considerations, Patent Special emphasis was placed in development to find a useful sampling procedure for metal catalysts. Melting a few complete components is costly and not sufficiently representative. An obvious approach was to process all of the components into a “homogeneous mass” – and to do this by mechanical rather than pyrometallurgical means. According to this approach, a sample could then be taken from this “homogeneous mass”. However, the first attempts ended with an alarming result: the crushing produced a fine dust rich in precious metals, and this dust was in danger of being lost unless a costly dust collection process was employed. The wash coat had become partly seperated from the substrate! This counterproductive effect would have meant the end of these attempts had it not been for a further attempt to deliberately turn the negative effect to good use. Thus the idea came about to no longer produce a uniform “homogeneous mass” but to remove the entire wash coat with its precious metal content from the remaining material by means of an explicit separation procedure. This was subsequently demonstrated in testing. This procedure – since patented throughout the world (US Patent 005279464A) – enables not just the wash coat with its precious metals to be recovered, but also the steel fraction. 3.2.2 Stages of the Process This procedure is based on the empirically acquired experience that by observing certain process parameters, the wash coat can be seperated from the substrate by mechanical means. If this produces a mixture of particles differing from each other in form, size and weight, then these can be seperated from each other. The end result is that different but homogeneous fractions of high purity are recovered. a)

Mechanical Reduction

In the first stage, the catalysts are reduced to a defined particle size by means of a shredder. Through the high degree of deformation of the substrate foil, caused by the applied mechanical energy, a large part of the wash coat already becomes detached and can be collected by means of air separation.

255 b)

Magnetic Separation

In most cases (depending on the materials used), the remaining metallic fraction can be magnetically seperated. The casing material – often austenitic – is not processed any further. On the other hand, a residue of wash coat still adheres to the substrate foil. c)

Mechanical Aftertreatment

In a further processing stage, the residue of precious metals on the substrate foil is seperated from the steel parts by means of a mechanical separation unit. A hammer mill is employed to “beat” the remaining particles of the wash coat away from the foil; here too, the foil and ceramic particles can be seperated from each other by means of air separation. Metal Substrate Converters

Shredder Washcoat

Metal

Magnetic Separation

Substrate Foil

Stainless Steel

Mechanical Separation Washcoat

Blending & Sampling

Figure 3: Separation process

256 All fractions of the wash coat are brought together for homogenization and sampling. The substrate foil and casing material can be reutilized in steel smelting. Fractions after Separation After separation, the fractions of the wash coat containing the precious metals, the substrate foil, and the casing material are available individually for further processing as shown in Figure 4.

Mantle

Foil Washcoat Figure 4: Fractions after separation

3.2.2 Sampling All subsequent steps for the recovery of the precious metals now focus exclusively on the seperated wash coat. As already mentioned, sampling is of major importance in the recycling of precious metals so that the content of the material can accurately be determined by means of an analysis. For ceramic materials, multiple-stage homogenization and separation processes are generally employed. The first step here is a synchronous milling and sieving procedure. Steel residue can thus be seperated and larger parts of the wash coat can be reduced to a uniform maximum size. The sample produced is also subjected to a drying process to fix its moisture content. As a ceramic material, the wash coat always contains a slight amount of moisture; this can amount to over 3 % of the weight of the catalysts if these are stored under unfavorable conditions.

257 Washcoat

Blending

SampleSplitting

Moisture Determination

Fine Milling

Screening

Basis for settlement of account

Wash-Coat for Refining zum Refining

Fines

Coarse

SampleSplitting

SampleSplitting

Sample

Sample

Figure 5: Sampling

4

Comparison of Catalyst Types in Recycling

4.1

Advantages and Disadvantages in Summary

The decanning of ceramic catalysts can be effected with far simpler means when compared to the shredding process employed for metal catalysts. The result in both cases is a ceramic material, which means that no costly technical procedure is required for the sampling. The only constituent requiring further processing in the case

258 of metal catalysts is the wash coat with precious metals, with ceramic catalysts the substrate is also involved. This has a substantial effect on costs. The subsequent recovery of the precious metals is therefore less costly because only a far smaller quantity of material needs to be refined. In addition, the substrate fraction of the metal catalyst can be used for steel smelting while the ceramic substrate is reduced to slag. The greater expenditure on pretreatment is rewarded by lower costs in further processing. Neither of the procedures can be credited with advantages of economy. From the ecological point of view, however, it is advantageous when only a small fraction of the material which cannot be returned to the materials cycle - this is the case with metal catalysts. To determine the financial expense involved in the recycling of metal and ceramic catalysts, two different systems employed in a modern 2 liter engine are examined. The two systems are equally effective and have similar geometric surface areas. The characteristics of the metal substrate are such that the metal catalyst system has less volume. Comparison is made by setting the values to 100 % for ceramic catalysts as the basis. The differences then are shown by calculating the ratio, giving the percentage of Metalits. Table 1: Geometrical data of catalysts Size Cell density and foil thickness Catalyst volume Geometric surface area

Metalit™ Ø118×174 mm 400 cpsi / 50 µm 1.9 l 6.1 m²

Ceramic substrate Ø143×152 mm 400 cpsi / 6.5 mil 2.44 l 6.2 m²

With the same volume-specific quantity of wash coat, this likewise means a smaller quantity of precious metals. Table 2: Data of catalytic coating Washcoat mass Precious metal loading Value of precious metals

Metalit™ 150 g/l – 332.5 g 50 g/ft³, Pt/Rh 5:1 – 3.39 g 77.9 %

Ceramic substrate 150 g/l – 427.0 g 50 g/ft³, Pt/Rh 5:1 – 4.357 g 100 %

There are considerable differences in the methods employed in the recycling process, especially in the way the materials are broken down. With the ceramic catalyst, the substrate including the coating is ground into a homogeneous powder. With the metal catalyst, the metallic substances and the wash coat powder have been seperated in the mechanical shredding and separation processes. It is somewhat more expensive to break down the metal catalyst with the separation of the coating powder but it results in a higher concentration of precious metals. Table 3: Downsizing of substrates Costs of shredder process Costs for dismantling and milling

Metalit™ 568 %

Ceramic substrate 100 %

259 This means less powder is subsequently processed in order to recover the precious metals, which leads to substantial cost savings in this stage of the process. Table 4: Precious metal recycling costs Value of precious metal content Metal fees Processing costs powder Value of recycled PM Overall loss

Metalit™ 77.9 % 77.9 % 21.3 %

Ceramic substrate 100 % 100 % 100 %

77.9 % 78.9 %

100 % 100 %

A further advantage over processes formerly employed is the possibility of recovering the materials of the converter seperated into fractions. Table 5: Value of steel recycled Metalit™ Mantle + matrix 292 % Mantle

Ceramic substrate 100 %

As a result of these cost savings, the absolute costs for the recycling of catalyst systems are somewhat less for the metal catalyst and the value of the material recovered in relation to the costs of recycling is comparable for both types of catalyst substrates. Table 6: Overall recycling costs Costs / substrate Recovered value / costs

5

Metalit™ 79.5 % 2.58 DM/DM

Ceramic substrate 100 % 2.58 DM/DM

Summary

It has been demonstrated that compared to the traditional melt process, a more efficient alternative is mechanical pretreatment. · Sampling The original development objective of producing a material capable of being sampled was attained despite the change in research plan. The wash coat powder recovered is ideal for sampling as it is a conventional ceramic material. · Recycling of steel In contrast to the fusion procedure, mechanical separation allows the steel fractions to also be reutilized. This is advantageous not only economically but also – and especially – ecologically. · Costs From today's point of view, the recycling of both designs is to be categorized as equally expensive even if the respective cost-intensive factors are in different stages of the processing.

260 In conclusion, it can be stated that with the mechanical pretreatment of catalytic converters having metal substrates, a further step has been taken that contributes both economically and ecologically towards the improved recycling of automobiles.

VI Miscellaneous

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

Hot-Corrosion of Metal and Ceramic Honeycombs by Alkaline Metals for NOx Adsorption Mikio Yamanaka Nippon Steel Technoresearch, Futtsu, Chiba

Yuichi Okazaki Nippon Steel, Toukai, Aichi, Japan

1

Introduction

Alkali metals and alkaline-earth elements are used for the NOx adsorption catalyst of lean burn engines. The NOx storage efficiency is the higher with the higher basicity of these metals (1)(2). But highly basic metals cause a hot-corrosion especially on ceramic honeycombs. In the previous paper, the authors dealt with the hot-corrosion caused by Li, K and Cs sulfates and it was turned out that a serious hot-corrosion takes place on cordierite honeycombs with g-Al2O3 coating containing K2SO4 above 1173 K (900 °C) (3). These corroded honeycombs lose their mechanical strength. In addition, reactive g-Al2O3 is easily converted to a type in the condition where the hot-corrosion takes place. On the other hand, no damage was observed for 20Cr-5Al metal foils under the same condition, but their weight increase was linear with the oxidation time. In case of real catalyst, alkali metals and alkaline-earth elements are added to g-Al2O3 as carbonate compounds for NOx storage. Thus in this paper, the effect of K2CO3 is presented in comparison with those by alkali metal sulfates, and their mechanism will be discussed focusing on the difference of hot corrosion behavior by the kind of alkaline elements.

2

Experimental Details

Test conditions were nearly the same with those of the previous experiment, except that K2CO3 was used as the additive to g-Al2O3 and that the slurry was applied only on the outer surface of honeycombs, not inner surface of caves this time. 0.2 mol of K2CO3 was added to 100g of g-Al2O3 with distilled water. The slurry of their mixture was applied on the surface of cordierite honeycombs and 20Cr-5Al foils. The size of each specimens was 20 × 20 × 2.8 mm for cordierite honeycombs and 20 × 20 mm × 50 mm for metal foils. After dried, each test piece was heat treated in wet air (DP = 323 K) at 873 K, 1023 K and 1173 K (600 °C, 750 °C and 900 °C) for 360 ks (100 hrs). These samples were weighed and subjected to folding test for metal foils and to 3 points bending test for ceramic honeycombs as shown in figure 1. The mechanical strength of these ceramic samples s was calculated by the equation as follows:

Material Aspects in Automotive Catalytic Converters, Hans Bode Copyright © 2002 Wiley-VCH Verlag GmbH &Co. KGaA ISBN: 3-527-30491-6

264 s = 3PL/2WT2 where P is the maximum load before failure, L the distance between two bars beneath the test piece, W and T the width and thickness of the sample, respectively. Both the ceramic and metal structure were observed by an optical microscopy and were analyzed by EPMA .

Figure 1: Schematic illustration of 3 points bending test for ceramic samples

3

Experimental Results

No nodule or abnormal oxidation was observed for metal foils heat treated even at 1173 K (900 °C). Their weight increases are listed in Table 1 with data given by the previous experiment using alkali metal sulfates. No cracking took place in the folding test of any metal foil after heat treatments, but the surface film of a foil heat treated at 1173 K (900 °C) was not uniform, as shown in figure 2. EPMA mapping of the metal foil is shown in figure 3. A few percent of K (potassium) is detected in the surface film but not in the metal matrix.

Figure 2: Cross section of metal foil folded face to face after heat treatment at 1173K

265 Table 1: Weight increase(×10–2kg/m2) of metal foil by heat treatment for 3.6 × 103 Ks (100 h) Additive to g-Al2O3 Li2SO4 Cs2SO4 K2SO4 K2CO3

873 K ~0 ~0 ~0 ~0

1023 K 0.01 0.02 0.03 0.03

1173 K 0.09 0.15 0.18 0.56

Figure 3: EPMA mapping of metal foil after heat treatment at 1173 K with the coating containing K2CO3

266 On the other hand, ceramic honeycombs heat treated above 1023 K (750 °C) with the coating containing K2CO3 changed their color from yellow to a little reddish, and some of them showed cracking, as shown in figure 4. Mechanical strengths of ceramic honeycombs with no crack were plotted in figure 5 with data given by the previous experiment. Ceramic honeycombs heat treated above 1023 K (750 °C) lost their mechanical strength. The ceramic structure of a honeycomb heat treated at 1173 K is shown in figure 6. A distortion and microcracks can be seen in the honeycomb lattice. Figure 7 shows the EPMA mapping of a ceramic honeycomb after heat treatment at 1173 K (900 °C). K (potassium) was observed to invade from the outer surface of the honeycomb sample to the whole wall thickness

Figure 4: Ceramic honeycombs heat treated at 1023 K with the coating containing K2CO3

Figure 5: Mechanical strength of ceramic honeycombs before and after heat treatment

267

Figure 6: Structure of ceramic honeycomb after heat treatment at 1173 K with the coating containing K2CO3

4

Discussion

In the previous paper, the authors reported that 20Cr-5Al metal foils for metal supports have a hot-corrosion resistance to K sulfate at least up to 1173 K (900 °C), but that cordierite honeycombs are attacked by K2O of K2SO4 and lose its mechanical strength at 1173 K. But in case of real catalyst, alkali metals and alkaline-earth elements are added to g-Al2O3 as carbonate compounds for NOx storage. Thus in this experiment, effects of K2CO3 both on metal foils and on ceramic honeycombs were examined. Even to K2CO3, metal foils exhibited a hot-corrosion resistance up to 1173 K (900 °C) as mentioned above. But the weight increase at 1173 K was three times of that by K2SO4 and five times of that with no coating at the same temperature given by another experiment. The value at 1173 K (900 °C) is about 50 % of the weight increase when the whole Al in the foil of 50 m is consumed for the surface oxide film. But this time, it contains the weight increase by K2O which attacked the Al2O3 film, because a few percent of K was detected in the surface film of the metal foil as shown in figure 3. Its surface film is not uniform and thick compared with one given by the mere oxidation in air or exhaust gas from an engine without coating. Nevertheless, the surface film prevented K from entering into the metal matrix. In case of cordierite honeycombs, K2CO3 attacked the ceramic matrix at the lower temperature of 1023 K (750 °C) than K2SO4, though the coating was applied only on one side of the honeycomb wall in this case. Cordierite 2(MgO)2(Al2O3)5(SiO2) is based on silica network and is presumed to be somewhat acidic compound. K2CO3 is more basic than K2SO4 , thus cordierite is more sensitive to K2CO3 than to K2SO4. In the previous experiment, cordierite honeycombs heat treated with the coating containing Li2SO4 or Cs2SO4 did not lose their mechanical strength, but Cs was detected to invade into the surface layer of ceramic matrix by EPMA. Thus the order of alkali attack activity to cordierite can be regarded as follows: Li2SO4