Vous êtes sur la page 1sur 43

The Sintering of Supported Metal Catalysts

SIEGHARD E. WANKE Department of Chemical Engineering University of Alberta, Edmonton. Alberta, Canada PETER C. FLYNN Syncrude Canada, Ltd. Edmonton, Alberta, Canada

I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11. MEASUREMENT O F DISPERSION.

94 95 95 96 97 98 98 99 101 103 109 109 113 115

A. Electron Microscopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. X-Ray Diffraction.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Selective Gas Adsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Comparison of Techniques. . . . . . . . . . . . . . . . . . . . . . . . . .

.. .

. .. .. . . .. .. .. . ..... .. .. .

..

..

. ..

F. Miscellaneous Sintering Results . . . . . . . . . . . . . . . . . . . . . . . . . . G. Empirical Correlation of Sintering Data . . . . . . . . . . . . . . . . . . . . H. General Comments on the Factors Affecting the Rate of 120 Sintering ............................................ N. MECHANISTIC MODELS O F THE SINTERING PROCESS . . . . . . 122 A. Support-Metal Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 B. Crystallite Migration Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 C. Atomic Migration Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 D. Comparison of Crystallite and Atomic Migration Models . . . . . . . 127

111. EXPERIMENTAL DATA AND EMPIRICAL CORRELATIONS . . . A. Variable Time-Variable Temperature Sintering of Supported Pt . . B. Constant Time-Variable Temperature Sintering of Supported Pt C. Constant Temperature-Variable Time Sintering of Supported Pt . D. Thermal Treatment of Supported Pt Resulting in Redispersion . E. Sintering of Supported Rh, Pd, and Ni Catalysts . . . . . . . . . . .

. .

93
Copyright 0 1975 Iiy M.ircel Drkker, Inc. All Rights Hcserved. Neither this worh nor any part may be reproduced or transmitted in any form o r by any means. electronic or niechanical. including photocopying. inicrofilming. and recording. o r hy any information storage and retrieval system. without permission in writing from the publisher.

94
V. CONCLUDINGREMARKS

WANKE AND FLY"

...............................

128

APPENDIX A. CONVERSION OF REPORTED DATA TO DISPERSION .......................................... APPENDIX B. CALCULATION OF APPARENT ACTIVATIONENERGIES ....................................... REFERENCES .........................................

129 130 132

I. INTRODUCTION

Metal catalysts are commonly employed in the form of metal dispersed as small crystallites on high surface area supports. The use of these supported metal catalysts increases the utilization of the metal as a catalyst since a large fraction of the metal atoms a r e at the surface of the small metal crystallites. Another important function of the support is to physically separate the small metal crystallites and thereby hinder the agglomeration of the small metal crystallites into larger crystallites. This agglomeration would decrease the number of surface metal atoms per unit mass of metal, and thereby decrease the utilization of the metal and the activity of the catalyst. Although the use of supported metal catalysts stabilizes the small metal particles, growth of the metal crystallites still occurs, especially if the catalyst is used at elevated temperatures. The process by which this occurs is generally referred to as catalyst sintering or aging. (Note this use of the term sintering is different than the usual use of the term which refers to the agglomeration of particles in physical contact.) In this paper the term sintering will refer to any process which leads to a change in the metal particle size distribution in supported metal catalysts. This change can include increases as well as decreases in the average metal crystallite size. The situation where the term sintering will be used in the narrower sense of fusion of particles in contact will be clearly stated. The aims of this paper are: (1) to review the experimental results of sintering of supported metal catalysts, (2) to present the methods used to correlate the sintering data, (3) to discuss the mechanistic models proposed for the sintering process, and (4) to present some suggestions as to the type of work which would increase our knowledge of the sintering process and thereby lead to the development of catalysts with increased stability. In order to achieve the above objectives it is necessary to describe the techniques used in investigating the sintering phenomenon. The experimental results presented will be restricted to supported Group VIII metals; the majority of thedata being for platinum supported onalumina or silica.

SINTERING OF SUPPORTED METAL CATALYSTS


1 . MEASUREMENT OF DISPERSION 1

95

Thermal treatment of supported metal catalysts results in changes in metal surface area and average metal particle size. Hence it is required that the metal surface area and/or the metal particle size be obtained as a function of treatment conditions. In this paper the metal particle size will be expressed in t e r m s of dispersion. Dispersion is defined a s the ratio of surface to total metal atoms. Various methods and procedures have been reported in the literature for obtaining the dispersion of supported metal catalysts; each technique has advantages and limitations. It is not the aim of this work to review the various methods in detail, a s this has been done elsewhere [l-31. However, a brief description of the various methods used to monitor changes in dispersion with thermal treatment, along with their limitations, a r e required for subsequent discussion of experimental sintering studies.
A. Electron Microscopy

For the characterization of supported metal catalysts, conventional bright-field transmission electron microscopy (TEM) is frequently employed. In TEM a thin specimen of the substance to be examined is placed in an electron beam and the transmitted electrons are focused to form an image. The contrast in the image is due to diffraction contrast for large particles and phase contrast for small particles [4]. While resolution is high, potentially down to single atoms [5], the nature of the image generation for small particles (i.e., phase contrast) leads to major uncertainties as to the size and detection of particles below -2.5 nm in size [4]. Other problems associated with TEM investigations of supported metal catalysts include (1) the interference of the support and (2) representative sampling of the surface. The interference by the support can be eliminated by dissolving the support [6, 71. The dissolution of the support may result in an alteration of metal particle size distribution since some of the metal particles may also be dissolved. The problem of obtaining a representative sample can be partly overcome by measuring a large number of particles, but even measuring in excess of 1000 particles can result in a significant statistical e r r o r [81. Electron microscopy, other than TEM, may also be used to characterize supported metal catalysts. Rather than employing brightfield imaging, dark-field imaging may be employed. This method may be used to obtain information on the crystal structure of the supported metal particles [9]. If the metal particles are large ( > l o 0 nm), scanning electron microscopy (SEM) may be used to

96

WANKE AND FLY

obtain metal particle size distributions. For example, Huang and Li [lo] used SEM to study the growth of platinum particles on various crystal faces of alumina. (For a description of SEM, see Kimota and Russ [ll].) Although the various electron microscopy techniques involve difficulties in interpretation of results and are limited in their applicability, they a r e very useful in the characterization of supported metal catalysts since they are the only direct methods of examining the metal particles. It should be kept in mind that the limitations of TEM for examining highly dispersed supported metal catalyst (average metal particle size < 2 nm) makes this technique qualitative rather than quantitative.
B. X-Ray Diffraction Metal particle sizes can be obtained by two different methods using x-ray techniques: (1)by x-ray diffraction line broadening and (2) by small-angle x-ray scattering (SAXS). The prelqence of small crystalline particles in a sample being examined by x-ray diffraction causes a broadening in the diffraction lines. This broadening can be related to the size of the particles. A detailed description of x-ray diffraction line broadening is given by Klug and Alexander [12], and Dorling [3] has reviewed the application of this method for the characterization of supported metal catalysts. This technique is generally limited to detecting crystalline particles with a size greater than 2 to 4 nm [13, 141, although Adams et al. [15], using a special spectrometer, detected much smaller particles. The difficulty in detecting small particles limits the use of this technique for characterization of well-dispersed catalysts. Small particles in an x-ray beam will also scatter the x-rays due to discontinuities in the electron density between the particles and the surrounding medium (the particles need not be crystalline). SAXS has been used for several decades to determine particle size in the range of 1 to 50 nm [16], but its application to supported metal catalysts is relatively recent. The main reason for not applying SAXS to supported metal catalysts is that the low-angle scattering by the porous supports obscures the scattering by the metal particles. Somorjai [17] eliminated the support interference by compressing the catalyst samples at extremely high pressures to collapse the pores. Recently, pore maskants of electron density similar to the support, such as CH,&, have been used to eliminate support interference [8]. SAXS allows the determination of metal particle size distribution, and it is potentially a very useful tool for characterization of supported metal catalysts.

SINTERING OF SUPPORTED METAL CATALYSTS C. Selective Gas Adsorption

97

By far the most common method for measuring metal dispersions of supported metal catalysts is chemisorption. The method consists of measuring the amount of gas chemisorbed by the metal in the catalyst and converting this quantity to a metal dispersion by assuming an adsorption stoichiometry (adsorption stoichiometry is defined as the ratio of the number of adsorbate atoms o r molecules adsorbed per surface metal atom). Several reviews of the use of chemisorption for determining metal , surface areas for supported metal catalysts exist [ l , 2 18-20]. Therefore, the methods employed will not he discussed here, and only a few general statements of caution when interpreting chemisorption results will be made. As mentioned above, the conversion of chemisorption uptakes to metal dispersion requires an assumption of the adsorption stoichiometry, and in order for the adsorption technique to be an absolute measure of the metal surface area, the correct adsorption stoichiometry has to be used. This is not a serious problem, since as long as the adsorption stoichiometry is constant, i.e., it does not depend on metal particle size, catalyst pretreatment conditions, etc., the obtained surface a r e a s are correct on a relative basis. There is evidence that adsorption stoichiometries may vary considerably with metal particle size [21-241. This can lead to serious difficulties since the changes in adsorption uptake are not directly proportional to the changes in dispersion if the adsorption stoichiometry is a function of the dispersion. For example, the adsorption stoichiometry for oxygen atoms on supported platinum catalysts has been reported to be -0.5 for small Pt crystals ($1.5 nm) and -1.0 for larger P t crystals (>2.0 nm) [21,231. Hence, if oxygen chemisorption were used to monitor the change of dispersion of supported platinum, the situation occurs where growth of metal particles (decrease in dispersion) is not reflected in a proportional decrease in the amount of oxygen adsorbed. The majority of the sintering studies, results of which will be presented in Section ID, employed chemisorption for measuring the metal dispersion. The choice of adsorbates, adsorption and pretreatment conditions, and measuring techniques varied greatly among investigators. In general, each investigator chose conditions which he believed would result in a measure of the metal dispersion. It is beyond the scope of this article to analyze the validity of the assumptions, with regard to the adsorption stoichiometry, that the various investigators used. In Section I11 the adsorption conditions used by the various workers investigating the sintering of supported metal catalysts will be presented. When interpreting these results, variation in adsorption stoichiometry should be kept in mind.

98

WANKE AND FLY"


D. Comparison of Techniques

The above sections have emphasized the disadvantages and the limitations of the various techniques used to measure metal dispersions. The picture presented above is overly pessimistic since numerous studies (e.g., Refs. 8, 14, 21, 25) have compared metal particle sizes determined by various techniques. The results are often in excellent agreement. Poor agreement among the techniques results if the metal particle size distribution is broad o r bimodal [8, 22, 26, 271. This is of concern in sintering experiments since sintering frequently results in a broadening of metal particle size distributions.

111. EXPERIMENTAL DATA AND EMPIRICAL CORRELATIONS

There are few studies reported in the literature in which time and temperature were systematically varied to determine the influence of thermal treatment on the metal surface area of supported metal catalysts. In many studies (e.g., Refs. 13, 28-31) thermal treatments were employed to change the metal crystallite size in order to determine the effect of metal crystallite size on adsorption uptakes and/or rates of reaction. In these studies, however, the rate of metal particle growth a s a function of treatment conditions was not of interest, and hence the conditions used a r e often not described in detail. Even in studies where treatment conditions are presented in detail, the data a r e difficult to interpret because
1. the problems associated with the measurement of metal dispersion (Thisproblem has been discussed in the previous sec-

tion.);
2. the support material may undergo changes, such a s collapse of the pore structure resulting in the trapping of metal within the support (This problem is particularly severe if silica is used

as the support.);
3. the state of the metal (elemental, oxide, or salt) during treat-

ment, or at least for part of the treatment, is not known. In spite of these limitations, general conclusions of the effect of various treatment conditions on metal sintering will be made. In order to facilitate comparisons among the various studies, all the reported data on metal surface area, particle size, etc. are converted to dispersion. The methods used for these conversions are described in detail in Appendix A.

SINTERING OF SUPPORTED METAL CATALYSTS

99

In the following section (III-A to III-F) experimental results on the sintering of supported catalysts is presented. The analysis of these data are presented in Section III-G (empirical correlation of the data) and III-H (general conclusions regarding rate of sintering).
A. Variable Time-Variable Temperature Sintering of Supported Pt

The main factors affecting the rate of sintering of a specific catalyst are the temperature, time, and atmosphere. Herrmann et al. [32]were among the first to report a detailed study of the changes in P t surface area for Pt/AhO, ca&lysts treated a t various temperatures and periods of time in a nitrogen atmosphere. Their results for the two extensively studied catalysts a r e summarized in Table 1. Hydrogen chemisorption at 200C and 9 T o r r was used to measure the Pt area. The measured hydrogen uptakes were such that a dispersion greater than unity resulted for fresh catalysts if an adsorption stoichiometry of one hydrogen atom p e r surface Pt atom was used. This was probably due to the short reduction time (20 sec at 500C) and the long evacuation time (overnight) which could have resulted in surface contamination with oxygen. For this reason, the dispersions reported in Table 1 are normalized with respect to the dispersion of the fresh catalysts, i.e., D/D, is the ratio of the dispersion of the sintered to the fresh catalyst. Somorjai [17] reports the effect of treatment in oxidizing and reducing atmospheres at 600 and 700C for periods of up to 96 h r on the average metal particle size for a 5% Pt/AbO, catalyst. SAXS was used to determine the average Pt crystallite size. The results of this study a r e summarized in Table 2 (The values reported in . Table 2 were obtained from Figs. 5 and 6 in Ref. 17.) Recently, Bett et al. [33] investigated the sintering of Pt/carbon catalysts in nitrogen and hydrogen at temperatures of 600, 700, and 800C. They measured the Pt a r e a by an electrochemical technique [34]. The results of this study a r e summarized in Table 3 (values obtained from Figs. 1 and 5 in Ref. 33). The methods of catalyst preparation a r e described in Table 4. Hughes et al. [35] used CO chemisorption to measure the changes in Pt dispersion of a 0.4% Pt/AhO, catalyst caused by treatment in hydrogen a t 900 and 1000F for treatment times of up to 1000 hr. They found that the CO adsorption uptake, U, as a function of treatment time, t, could be correlated by a function of the form (see Fig. 9, Ref. 35)
U = atb (1) where a and b a r e constants at a fixed temperature. Assuming that

100
TABLE 1

W A W AND F L Y "

Effect of Thermal Treatment in Nitrogen on Dispersion of Pt/Al,03 Catalysts [ 321

__________

Treatment conditions Catalyst description


-

Temperature, "C Fresh

Hydrogen adsorption uptake, cc (m)/g of Time, hr catalyst D/Do

0.7748P t / ~ A I , 0 3 Support area = 225 mz/g


Prepared by impregnation with H PtC1, solution ,

0.68 44 70.5 167 353 24 48 93 4 8 18 40 0.431 0.372 0.292 0.116 0.326 0.215 0.135 0.355 0.141 0.123 0.080 0.300 0.136 0.057 0.045 0.051 0.030 0.117 0.077 0.043 0.088 0.035 0.030 0.023

1.00 0.63 0.55 0.43 0.17 0.48 0.32 0.20 0.52 0.21 0.18 0.12 1.00 0.45 0.19'" 01' .5 0.17 0.10 0.39 0.26 0.14 0.29 0.12 0.10 0.08

564

594

625

0.375% Ptiy-Al, 0, Support area = 176 m2lg


Commercial reforming catalyst

Fresh

564

594

625

______

44 47.5 70.5 167 353 24 48 93 4 8 18 40

two data points do not fit the general trend, i.e., 48 hr treatment at 594C 47.5 hr treatment at 564"C,and 167 hr treatment at 564 causes a smaller loss in dispersion than 70.5 hr treatment at 564.Two other data points have been omitted because the original workers assumed them to be incorrect. "These
causes a smaller loss in dispersion than

CO adsorption occurs as one CO molecule per Pt surface atom, a questionable assumption [36], the results in Fig. 9 (of Ref. 35) can be expressed as dispersion, D, as function of time. This yields D
:

0.73t-O*l3

a t 900F

SINTERING OF SUPPORTED METAL CATALYSTS


TABLE 2 Effect of Temperature, Time, and Atmosphere on Dispersion of 5% Pt/r)--Al,O, [ 171 (area of support not given) Treatment conditions Average Pt particle diameter, nm

101

Atmosphere Air

Temperature, C
~

Time hr

Dispersion

600

700

1 6 24 48 1 3 6 24 48 1 3 6 24 48 96 1 3 6 24 48 96

12.4 20.0 23.6 25.4 22.0 27.0 29.6 32.9 34.3 10.8 17.0 18.4 20.4 21.4 22.1 13.6 18.6 20.0 22.4 23.4 24.4

0.082 0.051 0.043 0.040 0.046 0.038 0.034 0.031 0.030 0.094 0.060 0.055 0.050 0.048 0.046 0.075 0.055 0.051 0.045 0.044 0.042

H, (or H,/CO)

600

700

and
D = 0.67t-0.4

a t 1000F

(3)

with t in hours. The results expressed by Eqs. (2) and (3) are valid for 2 t 6 1000 hr.
B. Constant Time-Variable TemDerature Sinterina of Sumorted Pt

A common method of obtaining catalysts with varying metal dispersion is to treat a fresh catalyst sample a t different temperatures for a fixed period of time. In Table 5 the results of a number of such

102
TABLE 3

WANKE AND FLY"

Effect of Thermal Treatment on Dispersion of Pt/Carbon Catalysts [ 331 Catalyst description' Composition Method of preparation Ih
N1

Treatment conditions Atmosphere Temperature, "C Time, hr Dispersioh

5% Pt/carbon

Fresh

03 .1
600

0.5
1 6 16 30 65 90

0.30 0.30 0.29 0.27 0.23 0.24 0.21 0.24 0.21 02 .1 0.19 0.23 0.21 0.18 0.27 0.26 0.26 0.24 0.19 0.19 0.19 0.26

700

1 2 6 16 03 . 1 2

800

12% PUcarbon

VII

Fresh

N*

600

3.5 8 16 23 46 96

20% Pt/carbon

Ih

Fresh

N2

600

05 . 15 . 2 6 16 65

0.16 0.16 0.16 0.16 0.14 0.13 0.16 01 .5 0.16 0.13

Hi

600

05 .
2

6 16

aCarbon support was graphitized carbon (Vulcan XC-72;Cabot Corp.) with a surface area of 80 ml/g.

SINTERING OF SUPPORTED METAL CATALYSTS

103

TABLE 4
Methods of Catalyst Preparation Method designation Description of preparation method Impregnation with aqueous solutions containing the following dissolved salts H2PtCl, Pt(NH3 )2(N02 )z WNH, ),(OH), RhCl, H2PdC14 Ni(N0, h H2PtCI, followed by treatment with H2S Pt(NH3)z (NO,), Ion exchange with Pt(NH,); ;ref. 38 Colloidal PtS deposited on Al(OH), in aqueous suspension Pt vacuum deposited onto y-Al, 0,microcrystals a b VI Commercial catalysts Cyanamid Ketjen Katalysator, CK306 Engelhard, Lot 18-381 Cogelling of alumina sol and H,PtCI, solution by addition of aqueous ammonia Deposition of colloidal Pt

I
a b
C

d e
f

g h

I1
I11

IV
V

VII

studies a r e summarized. The methods of catalyst preparation (column 3) and the methods of measuring the Pt dispersion (column8) are described in Tables 4 and 6, respectively. A more detailed description of the treatment methods presented in Table 5 is necessary since this will influence subsequent conclusions. Additional comments, generally relating to the state of the metal during thermal treatment, a r e tabulated in Table 7. C. Constant Temperature-Variable Time Sintering of Supported P t Another method to vary the metal dispersion of supported catal y s t s is to treat catalysts a t an elevated temperature for various periods of time. Results of such studies are summarized in Table 8. Gruber [42] carried out studies of this type with two Pt/q-AbO,

catalysts. One catalyst, containing 0.7% Pt, was prepared by impregnation with Pt(NH,),(OH)2; the other one, containing 0.6% Pt, was prepared by impregnation with H,PtC!&. The area of the support

TABLE 5
I -

Effect of Treatment Temperature on Dispersion of Supported Pt Catalysts Treatment conditions Time, hr Atmosphere Fresh vacuum Dispersion Temperature, "C Method for measuring dispersion

Catalyst description

Composition

support area, m2/g

Method of preparation

Ref. 8

3.7%Pt/l)-AIZO3

131 600 700 800 250 400 500 600 700 0.39 0.35 0.27 0.15 0.11 400 500 700 700 100 350 500 600 650 500 700 800

Ia

0.38 0.38 0.42 0.33 0.34 0.44 026 0.23 0.31 0 2 6 022 0.19

la.5,6

2.1% Pt/Si02
5

285

Ia

Air

lb

37

2.1% Pt/Si02

285

I1

Ar i

lb

37

16 5

2.1.0 2.1.0 0.56 0.47 0.063 0.056 0.048 0.041 0.039 0.66 0.51 0.40

2.1% Pt/SiO,

285

Ib

Air

lb

37

1.6%Pt/SiO, 2

370

Ia

lc

38

1 6 Pt/Si02 .% 2
H 2

370

I1

500 700 800 500 600 650 700 800 550 600 650 700 500 600 800 460 500 556 770 0.48 0.48 0.04 0.01 0.57 0.54 0.53 0.39 0.23 4

1.oa 0.87 0.69


lc

38

0.5% Pt/A1203 2

Ig

Air

39

0.4%PtIAI, 0,

I11

4
Air

39

2.8% Pt/Si02
1
H2

250 4
'L1
H2

Ia
Air

0.52 0.28 0.003 0.97 0.81 0.88 0.28

lb

40

3.6% Pt/Si02

250

I1

lb

40

aDispersion normalized with respect to 500C treated catalyst.

106
TABLE 6

WANKE AND FLY"

Description of Methods Used for Measuring Dispersions Method designation Description of method for measuring dispersions Hydrogen chemisorption using the following conditions: Flow system at 200C with addition of pulses containing 5.7% H2 in N2 Static system at 25C using extrapolated, zero-pressure uptake as monolayer coverage Flow system at 0C with continuous addition of 0.13% H2 in argon stream Static system at 250C using uptake at 100 Torr as monolayer coverage Flow system at room temperature with addition of pure H2 pulses to nitrogen carrier gas Static system at 70C using the uptake at 1 0Torr as mono. layer coverage Static system at -78C using the uptake at 250 Torr as monolayer coverage Static system at 200C using uptake at 9 Torr as monolayer coverage (only 10 sec reduction of catalyst at 500C and 10 Torr before adsorption measurement) Oxygen chemisorption in a flow system at room temperature with addition of pure O2 pulse into helium carrier gas

1
a
b
C

d e
f

3
a
b
4

Carbon monoxide adsorption using the following conditions:


Static system at room temperature with uptake after evacuation with Toepler pump taken as monolayer coverage Flow system at room temperature with continuous addition of 0.5% CO in helium. Hydrogen titration of adsorbed oxygen at 25C in a flow system Small-angle x-ray scattering Transmission electron microscopy X-ray line broadening

was 200 m2/g. The thermal treatment was carried out in hydrogen at 500C for 1 to 82 days. The Pt dispersion was measured by CO adsorption at 25C using pulse addition of CO to a helium carrier stream. The results were correlated by Eq. (1). The results, expressed as Pt dispersion as a function of treatment time, are (from

SINTERING OF SUPPORTED METAL CATALYSTS


TABLE 7 Comments on Results Presented in Table 5 Ref. Probable state metal during thermal treatment Elemental Pt Pt salt during early stages of treatment and oxidized Pt during later stages Comments

107

Samples were reduced in H2 at 500C for 3 hr before thermal treatment Samples were only mildly reduced in hydrogen (2OO0C, length of time not given) before thermal treatment in air. The results indicate that catalysts prepared by Method l b were probably not reduced, and the low dispersion resulted from the agglomeration of the salt Samples were reduced in H2 at 500C for 1hr and subsequent treatment was in H2. was It not clear whether the treatments at 500, 700,and 800C were carried out sequentially on the same sample, or whether a different sample was used for each treatment temperature. A decrease in the silica support surface area of 20% occurred during the 800C treatment Samples were oxidized prior to sintering in air Samples were reduced in H2 at >400C before thermal treatment. The length of thermal treatment is only approximate

37

38

Elemental Pt

39
40

Oxidized Pt Elemental Pt for catalysts treated in H2;oxidized Pt for sample treated in air

Fig. 4, Ref, 42): for the 0.6% Pt/q-A&O, Catalyst


D = 0.38t-'"'
(4)

and for the 0.7% Pt/q-A&O, Catalyst


D = 0.465t-0.073

(5)

with t in hours. The results in Eqs. (4) and (5) a r e for 24 C t C 2000 hr. Huang and Li [lo] studied the sintering of large Pt crystals (>150 nm) on various crystal faces of aluminum oxide (sapphire) at 900C in air at 1/2 and 1 atm. They used SEM to measure Pt particle size changes as a function of treatment time. Treatment times of up to

TABLE 8

Effect of Treatment Time at Elevated Temperature on Dispersion of Supported Pt Catalysts neatment conditions Temperature, "C Atmosphere Dispersion
Air

Catalyst description

Composition

support area, m2/g

Method of preparation

Time, hr
1 3 16 70 1 3 16 70

Method of measuring dispersion

Ref.

Ptlr-Alz 0,

Iv
700

0.062 0.051 0.035 0.023 0.064 0.058 0.050 0.044

41

Pt/r-A1203

Iv
700 2% 0 in Nz 2

41

0 6 Pt/r-AI,O, .%
780
Air

179

Va

0 2

lh

4
10 17 72 500

0.85 0.73 02 .6 0.12 0.055 0.018

1.1% Pt/9-AI2O3

210

Ic

H Z

0 72 200 1200

0.82 0.52 0.43 0.32

Id

36

SINTERING OF SUPPORTED METAL CATALYSTS

109

4 days were studied. The results were well correlated by a powerlaw function of the form

- :f

= kt

(6)

where i-is the average Pt particle diameter a t time t, and f,, is the average Pt particle diameter a t t = 0. The rate constant k varied by more than a factor of 10, depending on which crystal face of the alumina was used as the support. The rate constant decreased by as much a s a factor of 3 when the sintering atmosphere was charged from a i r a t 1 atm to a i r at 1/2 atm pressure. D. Thermal Treatment of Supported Pt Resulting in Redispersion Generally, treatment of supported metal catalysts to elevated temperature (>500C) results in a decrease in metal surface area. Under certain treatment conditions it appears to be possible to cause a redispersion of sintered metal catalysts. This is very desirable for the regeneration of deactivated catalysts. The patent literature contains various claims for the regeneration of deactivated reforming catalysts; for example, deactivated supported Pt catalyst was regenerated by treatment at 370 to 550C in an inert gas stream containing 0 5 to 2% 0 [43]. The regenerated catalyst had a higher . , activity than the original fresh catalyst. There a r e several studies in the literature that indicate that metal redispersion has occurred after thermal treatment. The results of these studies a r e summarized in Table 9. It is worth noting that the atmosphere for all the cases resulting in an increase in dispersion contained oxygen. For a more detailed description of the treatment indicated in Table 9, the original references should be consulted. E. Sintering of Supported Rh, Pd, and Ni Catalysts Most of the sintering studies reported in the literature a r e for supported platinum. Considerably less information is available on the sintering of other supported noble metal catalysts. In Table 10 some results of the effect of thermal treatment on the metal dispersion of supported Rh, Pd, and Ni catalysts a r e presented. Sintering studies of catalysts in which the metal made up more than 40% of the total catalyst mass (e.g., Refs. 26, 53, 54)are not included in Table 10, since these catalysts cannot be considered supported metal catalysts. The Rh in the 5% Rh/q-A&O, sample [22, 481 was present as Rh metal prior to the treatment shown in Table 10 since the samples

TABLE 9 Experiments Showing Redispersion of Pt During Thermal Treatment

Catalyst description

Composition Fresh After used under reforming conditions

support area, mz/g Dispersion

Method of preparation

Treatment conditions Time, Temperature, Atmosphere hr O C Method of measuring dispersion

Ref.

0.4% Pt/A1203

Ia

0.53 0.37

39

Air
Not given
H 2

0.55% Pt/q-AI,O,
Followed by air
H 2

160
Followed by air

Ia

44

0.57% Pt/7-Alz03
Followed by air
H 2

160
500

Ia

H Z 600
600
Followed by air Fresh
0 2

4 4 4 4 4 4 4 4 16

480 500 500 600 600 500

44

0.5% Pt/A12O 3

Vb

le

45

450 600 700 800

2 0 Pt/Alon .%
0 2

100
Followed by O2
0 2
0 2

Ia

Fresh

16 16 16 16

575 610 630 700

0.72 0.54 0.76 0.53 0.60 05 .4 0.65 0.63 0.37 0.23 03 .2 0.26 0.10 0.01 0.25 0.26 0.17 0.07 0.03

45

Ia
0 2

Fresh Followed by 0,
0 2

45

16 16 16 575 610 630

0.20 0.22 0.09 0.006

2.0%PtIAlon (made by mixing fresh and 700C treated catalysts from above sample) 0.6% Pt/Al,O,
Used
Air (1atm) Air (1 atm)

V
3 3 20 3
440 480 505 505

3a

46

0.58%PtIAI, 0,

210

VI
H 0 z
5 2 5 2 5 2 705 620 705 620 750 620 5 2 5 2 705 620 750 620

Followed by air (10 atm) Air (1 atm) Fresh 35%H,O in air Followed by 0,(5 atm)

lh

47

0.3%PtlAl, 0,

210

VI

lh

47

0.3% PtIAl, 0,

250

Ia

Followed by 0,(5 atm) Fresh 35%H 2 0 in air 0, (5 atm) Fresh 35%H,0 in air 0, (5 atm) H, 0 0,(5 atrn)

0.20 0.29 0.32 0.48 0.35 1.008 0.12 0.58 0.15 0.19 1 .ooa 0.13 0.83 1 .ooa 0.45 0.47 0.05 0.07

lh

47

aDispersions normalized with respect to fresh catalyst.

TABLE 10 Effect of Thermal Treatment on Dispersion of Supported Rh, Pd, and Ni Catalysts Treatment conditions Time, Temperature, Atmosphere hr O C Dispersion
lb

Catalyst description Method of Composition preparation Method of measuring dispersion

Ref.

5% Rh/q-AlI 0
0, Followed by O2
0 2

Id

Fresh

22,48

600 700 800


lb

5% Rh/Si02
Air Air

Id

49
3b
If

538 800 900

0.3% Rh/A1203
N2
H2

35 50
If

5% PdIAI2 0,
H2

Ie

3 3 8 Fresh 4 4 Fresh 72 Deactivated 20 20

2% Pd/Si02

Ie

400 600 800 400 600 800 900

50

3% Ni/AI2 0
N2
H2

If
Fresh

3b

35 900 370 450 500 580 700


1g

10% Ni/A1203-Si02

If

72 1

51

6.7% Ni/Si02
Air

If

7 450

Fresh 16

0.79 0.59 0.37 01 .2 0.53 0.26 0.086 0.77 0.25 0.01 0.41 0.15 0.06 0.40 0.30 0.13 00 .8 0.021 0.002 0.112 0.066 0.051 0.039 0.028 0.144 0.092

52

SMTERING OF SUPPORTED METAL CATALYSTS

113

were reduced in hydrogen at 500C for several hours prior to the sintering in oxygen. The Rh in the 5% Rh/SiO, sample [49] was present as RhC1, prior to the a i r treatment shown in Table 10 since the dried impregnated catalyst was not reduced prior to the sintering. The state of the Rh and the N i in the 0.3% Rh/AbO, and 3% Ni/AbO, prior to the treatment in nitrogen is not known. The two Pd catalysts [50] were calcined at 500C before the treatment in hydrogen. Since the treatment temperatures were 2400"C, the oxidized Pd was probably converted rapidly to elemental Pd. The N i in the 10% Ni/ AbO,-SiO, [51] and 6.7% Ni/SiO, [52] catalysts was present as the nitrate at the beginning of the treatment. The treatment in hydrogen at 370"C, according to the authors [51], resulted in complete reduction.
F. Miscellaneous Sintering Results

Numerous miscellaneous observations of changes in catalytic activity due to thermal treatment of supported metal catalysts can be found in the literature. Armstrong et al. [29] studied the effect of thermal treatment at 1000 to 1200C in steam on the ignition tem0,-3% H2-95% He mixtures by supported Pt, Ir, Pdperature of 1% Ru, Pd-Pt, Pd-Ir, Ir-Pt, Pt-Ru, and Pt-Rh. It is difficult to interpret the results in terms of loss of dispersion since significant alterations in the supports occurred at the elevated treatment temperatures. It is interesting to note that for some of the catalysts (Pt-Rh and Pt-Ru) steaming at 1000 to 1200C resulted in an increase in activity. Conflicting results a r e reported on sintering in vacuum of reduced supported Pt. Boudart et al, [55] found that treating a 1%Pt/ carbon catalyst at 900C for 16 h r in vacuum did not result in a loss of dispersion for one sample, but decreased the dispersion by a factor of -3 in another case. They attributed this difference in behavior to a poor vacuum in the later case, and concluded that sintering in vacuo at temperatures of up to 900C does not cause a loss of dispersion. This is in agreement with the results of Spindler [56] who states that treatment in high vacuum at 800C results in a metal dispersion which is essentially the same a s that obtained by only reduction in H, a t 500C. These observations do not agree with the results of Renouprez e t al. [8], which a r e reported in Table 5, since they observed a loss of dispersion for sintering in vacuo at temperatures a s low a s 600C. There is general agreement that treatment in hydrogen for prolonged periods (>15 hr) at temperatures >500"C results in a loss of dispersion for supported Pt catalysts [25, 551 (also see data in Tables 5 and 8).

114

WANKE AND FLY"

The effect of thermal treatment of impregnated, unreduced catalysts is not well defined. Wilson and Hall [21] found that calcination of a dried unreduced 0.75% Pt/AbO,, prepared by impregnation with H , P Q solution, in pure nitrogen at 600C for 4 h r did not reduce the Pt dispersion. Calcination of unreduced, supported Rh in air at 538C for 4 hr caused the Rh dispersion to decrease by a factor of 2 [49I. Unexpected growth rates of Pd particles supported on charcoal have been observed a t temperatures below 50C. Pope et al. [57] observed a 30% loss of Pd surface due to reduction of a 10% Pd/charcoal catalyst at 25C for 2 hr. Brownlie et al. [58] employed a Pd/ charcoal catalyst, prepared by vapor deposition of Pd, for the hydroisomerization of 1-butene. They observed, by TEM, a change in average Pd crystallite size from 14 nm for the fresh catalyst to 130 nm for the catalyst used for reaction. The reaction temperature was 43"C, the feed consisted of a 1:l 1-butene:hydrogen mixture, and the total initial pressure was 100 Torr. Little particle growth was observed for another Pd/charcoal sample which was prepared by impregnation with palladous chloride and exposed to the same reaction condition. Various investigators report on regeneration techniques without specifying the conditions. Spindler [56] states that complete redispersion of Pt catalysts on clay supports can be obtained by treatment in air o r other media. Blume et al. [59] states that the removal of coke from reforming catalysts can be accompanied by a redispersion of the Pt if proper conditions are chosen. No specific detail of the regeneration conditions is presented. Emelianova and Hassan [30] report that Pt redispersion occurs if thermal treatment at 400 to 650C was followed by rapid cooling to room temperature. If the cooling was carried out slowly, no redispersion was observed. Plank et al. [60] sintered commercial Pt/A&O, catalysts in K, hydrocarbons, and oxygen. They never observed a redispersion of Pt. They found that hydrocarbon and hydrogen atmospheres result in a slow change in metal dispersion while in oxygen the loss of dispersion was very much faster, e.g., treatment in H, at 870C for 2 h r changed the average Pt crystallite size from 2.0 to 2.8 nm, while treatment in O2 at 760C changed the Pt crystallite size from 2.0 nm to a bimodal particle size distribution with 65% of the particles having an average diameter of 3.0 nm and 35% of the particles having an average particle diameter of 42.5 nm. Similar bimodal Pt particle size distributions were observed for commercially aged catalysts. The Pt crystallite sizes were determined by x-ray diffraction line broadening. The use of supported noble metals in catalytic automobile mufflers

SINTERING OF SUPPORTED METAL CATALYSTS

115

is a new use for these catalysts. The operating conditions encountered in these catalytic converters can be very severe, e.g., intermittent temperature in the excess of 1000C are not uncommon. These severe operating conditions, together with catalyst poisons in the exhaust, limit the life of the catalyst. Very little information on the loss of metal dispersion with use is available in the literature. Most of the reported life studies of exhaust catalysts deal with the catalyst stability toward poisons, and the thermal deactivation has been largely ignored. The information available [61-641 is generally qualitative in nature. Substantially more information on the sintering behavior of these catalysts should become available in the near future.

G. Empirical Correlation of Sintering Data In the previous sections a large amount of experimental data on the sintering of supported metals has been presented. Most of the investigators discuss their data qualitatively. In this section the majority of the data presented in Tables 1-3,5,8, and 10 willbe correlated by a power-law rate function of the form -dD/dt
=

kDn

(7)

where the rate constant k will be assumed to obey the Arrhenius law, i.e.,

Wherever possible, both the power-law order, n, and the activation energy, E, will be evaluated. In order to obtain values of n from experimental data, it is necessary to have D as a function of time at constant temperature. At constant temperature k is assumed to be constant, and Eq. (7)can be integrated to yield
k =+In@

for n = 1

(9)

where Do is the initial metal dispersion (i.e., a t t dispersion at time t.

0) and D is the

116

WANKE AND FLY"

The constant temperature-variable time sintering data can then be fitted by Eqs. 9 and 10, and the value of n which results in relatively constant values of k will be the power-law order. For the two cases where the dispersion a s a function of time was given by

D = atb
the order is determined by the following method. If it is assumed that n > 5 and D,/D > 1.4, then Eq. (10) can be approximated by

(1)

By comparing Eqs. (11)and (1) it can be seen that n = 1-3 The power-law orders obtained by the methods described above a r e presented in Table 11. For many cases a range of n is presented, since the values of k were not constant for a specific value of n. Increasing the value of n increases the value of k at large t in comparison to the value at small t. The range of n shown in Table 11 is such that at the low value of n the value of k at short sintering times (the experimental times) is larger than the value of k at the long sintering times, while at the high value of n the value of k at short sintering times is less than the value at long sintering times. The results in Table 11 are arranged according to the atmosphere in which the sintering was carried out. The general trend is that the order for reducing atmospheres is larger than for oxidizing atmospheres. The orders in nitrogen atmospheres are approximately equal to those for air. Another trend in the values of n is the effect of initial dispersion; higher initial dispersions generally result in higher values of n. Mechanistic interpretations of the n values will be presented in Section IV. Apparent activation energies can be obtained from variable timevariable temperature sintering data. Combining Eqs. (8) and (lo), one can obtain
1

where At, is the time required to change the dispersion from D, to D at a temperature T,, and At, is the time required for the same , change in dispersion (i.e., D, to D) at temperature T,. (For a de,

TABLE 1 1 Power-Law Orders from Sintering Rates

Treatment conditions Temperature, "C Data shown in table Initial dispersion Power-law order (range)a
Ref.

Catalyst

Atmosphere

5%PtIAl, 0,

1.1% Ptlq-All 0, 0.7%Ptlv-Al, 0 3 0.6% PtIq-Al2 0, 0.4% PtIAl, 0,

2 2 8

17 17 36 42 42 35

35

0.774%Pt/y-A120,

0.375%PtlY-Al, 0 3

5%Ptlcarbon

1 1 1 1 1 1 3 3 3 3 8 8

12%Pt/carbon Ptly-All O3 PtIr-Al2 0 3 PtlAl203 5%Pt/A12 o3 2


8

0.094 0.075 0.82 x.3 x.3 x.7 x.7 x.7 x.7 x.7 XI .7 x.7 x.7 0.31 0.31 0.31 0.27 0.064 0.062 <o .01 0.082 0.85

0.6% PtlY-Al, 0

600 700 500 500 500 482 538 564 594 625 564 594 625 600 700 800 600 700 700 900 600 700 780

12 to 16 14 to 16 6to 7 15 1 1 9 8 2 to 3(2) 2 (2) 2 to 5 (2) 2 (2) 2 (2) 2 to 5 (2) 5 to 12 (5) 5 to 14 5to 8 4 to 7 (5) 12 to 13 (13) 5 (5) 9 to 10 8 to 14 2

32 32 32 32 32 32 33 33 33 33 41 41 10 17 17 7

aThe numbers in parentheses refer to the power-law orders reported by the original investigators.

118

WANKE AND F L Y "

tailed development of Eq. 13, see Appendix B ) Apparent activation . energies estimated by this procedure for the data shown in Tables 1

to 3 are given in Table 12.


TABLE 12 Apparent Activation Energies from Variable Time-Variable Temperature Studies Treatment Conditions Catalyst Atmosphere Temperature range, "C Apparent activation energy: kcdmole

Ref.

0.4% Pt/A1203 5% Pt/Al,O, 5% Pt/carbon 0.774% Pt/TAlz 0 3 0.375% PtlyAl, O3 5% PtlAl, o3

H Z H Z
N1
Nl

N1

Air

482538 600-700 600800 564-625 564825 600-700

15 25 (20) 40 (42) 90 (70) 85 60 (52)

35 17 33 32 32 17

aValues in parentheses refer to activation energies reported by original investigators.

An apparent activation energy can be estimated from constant time-variable temperature sintering data if the order n is assumed. Using Eq. (lo), one can obtain

where Do is the initial dispersion, D, is the dispersion after treatment at temperature T, for a time t', and D, is the dispersion after treatment at temperature T, for the same period of time, t'. (For a detailed development of Eq. 14, see Appendix B.) Using the data presented in Table 5, values of E as a function of n were calculated by Eq. (14). The results of these calculations are summarized in Table 13. If the sintering process is described by the power-law rate equation, then the activation energies calculated by Eq. (14)should be constant for the "correct" value of n. Unfortunately, most of the sintering data are only for two o r three sintering temperatures and it is difficult to determine the value of n which results in relatively constant E values. The results presented in Table 13 show E for various values of n. Only in two cases, the 1% Pt/SiO, samples [37], was it possible to determine the order a s being -2 by examining the variation of E with n. The other reported values are be-

SINTERING O F SUPPORTED METAL CATALYSTS


TABLE 13
Apparent Activation Energy from Constant Time-VariableTemperature Studies Treatment conditions
~

119

Catalyst

Atmosphere
3

Temperature range, "C

Apparent activation energies, kcdmole, power-law order

6 30 30

10 40 50

14 50

Ref.

3.7% Pt/TAIZ 0 1.6% Pt/Si02

Vacuum
H2

600800 700800

8 38

(prepared by Method Ia) 1.6% Pt/Si02 (prepared by Method 11) 1% Pt/Si02 1%Pt/Si02 0 5 Pt/A12O 3 .% 0.4% Pt/AI, O 3 5% Rh/Si02 5% Pd/A120, 2% Pd/Si02 10% NUAl, 03-Si02

H Z
Air Air Air Air Air

700800

35

50

38

H Z
H2 H2

400-700 350-650 600800 650-700 538800 600800 600-900 500-700

13 6 35 60 10 10 17 7

45 170 35 40 30 20

75 40 60

37 37 39 39 49 50 50 51

lieved to reflect the range of E for the various catalysts and sintering conditions. It should be pointed out that higher values of n always result in higher values of E. Examining the results in Tables 12 and 13, it can be concluded that the apparent activation energies for sintering in an oxidizing atmosphere a r e higher than those in reducing atmospheres. The notable exceptions to this generalization are the low values reported for the 1% Pt/SiO, samples (E = 13 and 6 kcal/mole) and the 5% Rh/SiO, sample (E = 10 to 35 kcal/mole). None of these three catalysts were thoroughly reduced prior to the sintering treatment. Hence the low activation energies a r e possibly due to migration of the metal salt causing large metal salt crystals which upon subsequent reduction result in low metal dispersions. On the basis of the limited data available it is difficult to compare the activation energies and order of sintering for various metals, but it appears that N i has a lower , activation energy than the other metals, and that Pd in an H atmosphere has a lower power-law order than Pt. The above discussion shows that the attempts to correlate sintering data by the simple power-law rate function (Eq. 7) result in only

120

WANKE AND FLY"

limited success. The failure of this model has implications in regard to the mechanism of sintering since various mechanisms predict a power-law type behavior. This aspect will be discussed in Section IV.
H. General Comments on the Factors

Affecting the Rate of Sinterinn Based on the data presented in Section 111-A to G, it is possible to draw some general conclusions in regard to factors that influence the sintering of supported metal catalysts. The most important factors affecting the sintering phenomenon are temperature and atmosphere; the nature of the support and the degree of metal loading appear to be of secondary importance. In general, the rate of loss of dispersion increases with increasing temperature. No exceptions to this rule have been found for nonoxygen-containing sintering atmospheres. In oxygen-containing atmospheres, on the other hand, many investigators have observed an increase in metal dispersion at certain temperatures (see Table 9). Redispersion in oxygen-containing atmospheres appears to be restricted to temperatures in the range of 400 to 620C. At temperatures above this range, or even at lower temperatures for prolonged periods of treatment, a loss of dispersion occurs. The rate of loss of dispersion, except for the case where redispersion occurs, is larger in oxygen atmospheres than in hydrogen atmospheres [17]. The rate of loss in dispersion decreases with decreasing oxygen partial pressure [lo, 411. The rate of sintering in nitrogen atmospheres is comparable to the rate in hydrogen atmospheres (cf. data in Tables 1 and 5. The rate of sintering in ) vacuum is lower than if a gas phase is present, but care has to be taken to avoid contamination [55]. The effect of the support on the rate of sintering is difficult to establish. The work of Huang and Li [lo] showed a strong dependence on the support, but the applicability of these results to typical supported metal catalysts is questionable since the size of metal crystallites in this work was much larger than those encountered in supported metal catalysts. Although sintering data for A&OS,SiO,, and carbon-supported catalysts is included in Tables 1-3, 5, and 8 , it is not possible to determine the effect of the support on the sintering rate since sintering conditions, initial metal dispersions, and metal loadings are different for all these samples. Bett et al. [33] report that the rate of sintering for Pt/carbon catalysts depends on the metal loading, i.e., the amount of metal per unit area of support. Gruber [36, 421 reports sintering data at 500C

SINTERING OF SUPPORTED METAL CATALYSTS

121

in H for 0.6, 0 7 and 1.1% Pt/v-AbO, catalysts. The area of the , ., supports was 200 mz/g for the 0.6 and 0.7% catalysts, and 210 m2/g for the 1.1% catalyst. Hence the metal loading has the same sequence a s the percent Pt. The initial dispersions for the 0.6 and 0.7% catalysts a r e not known (see Eqs. 5 and 6), but the data can be normalized with respect to the dispersion at 72 hr. This normalized dispersion is plotted against time in Fig. 1. As can be seen in this figure, the rate of sintering is not a monotonic function of metal loading. The 1.1% P t catalyst has the highest rate of loss, but the 0.6% Pt catalyst has a greater relative loss in dispersion than the 0.7% Pt catalyst. This discrepancy may be due to the different initial metal crystallite size distributions since Flynn and Wanke [45] found that catalysts with bimodal (wide) size distributions sinter more rapidly than catalysts with narrow size distributions.
1.0
0.9

0
v) OL Y

=
Y P

'"
. a

0.8

z
r

0.7

0.6

0.5 250
500

750

1000

1200

TIME.

HOURS

FIG.1. Normalized dispersion (see text) as a function of treatment time at 500C in hydrogen for 0.6,0.7, and 1.1%Pt/l)-AI,O,. (Data for 1.1%catalyst in Table 8; dispersion as a function of time for 0.6 and 0.7% Pt catalysts given by Eqs. 4 and 5.)
A comparison of the resistance of the various metals toward sintering is difficult due to the nonuniform conditions employed by the various investigators, but it appears that in non-oxidizing atmospheres the stability toward sintering increases with increasing melting temperature. That is, the sequence of stability toward sintering in non-oxidizing atmospheres is N i < Pd < Pt < Rh. In oxidizing atmospheres and with catalysts containing unreduced metal salts, the sequence of resistance toward sintering cannot be determined from the available data. The general observations presented above will be discussed in terms of sintering models in the following section.

122

WANKE AND F L Y "

IV. MECHANISTIC MODELS OF THE SINTERING PROCESS


Two mathematical models based on postulated mechanisms for the sintering of supported metal catalysts have recently appeared in the literature, The model developed by Ruckenstein and Pulvermacher [65, 661 envisages the sintering of supported metal catalysts to occur by migration of metal crystallites over the surface of the support, and the resulting collision and fusion of metal particles causes the loss in dispersion. The model will be referred to subsequently a s the crystallite migration model. The second model, proposed by Flynn and Wanke [67, 681, considers the sintering to occur by dissociation of atomic o r molecular species from the metal crystallites. These atomic (or molecular) species migrate over the support surface and become incorporated into metal crystallites upon collision with the stationary metal crystallites. This model will be referred to a s the atomic migration model. The two proposed mechanisms are not new. The crystallite migration model is similar to the mechanism proposed by Smoluchowski [69] for the coagulation of colloidal suspension by Brownian motion. The atomic migration model is similar to the model proposed by Ostwald (Ostwald ripening) [70]for the change in mercury oxide particle sizes in solution. The migration of atoms across a support surface was advanced by Shekhter [71] to account for the decrease in number and increase in size of globules of metal catalysts on filament supports. The transport of metal (or metal compounds) through the vapor phase is a third possible mechanism for the sintering of supported metal catalysts. The transport of metal by this mechanism in non-oxidizing atmospheres is negligible for Group VIII metals at temperatures below 1000C; in oxidizing atmosphere the transport via metal oxides may be appreciable at lower temperatures [72, 731. Substantial loss of Ru by volatilization of ruthenium oxides occurs when supported Ru catalysts a r e used in catalytic mufflers [74]. Little loss of Pt was observed for supported Pt catalysts under similar conditions "751. The vapor phase transport as a mechanism of sintering will not be discussed in detail since the normal temperatures encountered in the sintering of supported metal catalysts a r e too low for appreciable vapor phase transport (Ru catalysts are an exception to this general statement). In the subsequent sections the two models presented above will be discussed in some detail. Since the interaction of the metal (both in atomic and crystalline states) with the support is important in determining which of the mechanisms predominates, a section on metalsupport interaction is included. It should be mentioned that an excellent review of the growth of supported metal crystallites has been prepared by Wynblatt and Gjostein [76]. The discussion of the models

SINTERING OF SUPPORTED METAL CATALYSTS

123

in this review will be quantitative, and the above-mentioned review should be consulted for a theoretical presentation of the supportmetal interaction and metal crystallite growth.
A. Support-Metal Interactions

In considering the agglomoration of metal which rests on a substrate, interactions between the metal crystallites and the support material are significant. Geus [77] has reviewed data on the interaction of metal atoms and crystallites with various supports from a large number of film growth studies. Both mechanical methods and contact angle measurements have been used to determine adhesion between metal crystallites o r films and ionic o r oxide surfaces. From a number of studies cited by Geus 78-851, the following picture emerges: for alkali halide, oxide, and glass substrates, when ultrahigh purity conditions and thorough reduction are employed, interaction between metal and substrate is due to van der Waals forces. In these cases, typical adhesion energies of 20 kcal/g atom o r less are observed. However, the presence of oxygen, either gaseous o r chemically available a t the interface, can change the interaction from physical adsorption to chemisorption, with much higher energies of adhesion. These effects a r e particularly significant on glass and oxide substrates, For example, Pask and Fulrath [81] found significant decreases in contact angle (indicating greater adhesion) between molten glass and Au, Fe, and Pt metal surfaces when oxygen was present, a s compared to results in vacuo o r reducing atmospheres. They postulated incorporation of metal atoms into the substrate structure in an oxygen atmosphere. While such compound formation between metal and substrate has been observed in some cases (e.g., nickel aluminates), Darling et al. [86] showed that platinum does not form compounds with refractory oxides at temperatures below 1200C. Benjamin and Weaver [82, 831 found that aging in air of metal films such a s Fe, Cr, Ni, and T i on glass increased adhesion. Moore and Thornton [84] and Mattox [85] also observed that the presence of oxygen increased adhesion of gold to substrates. Atmosphere and surface contamination also affect the adhesion of individual metal atoms to substrates. This effect has been investigated in film growth studies, primarily on alkali halide surfaces; Geus [77] reviews the results of numerous workers. In general, cleavage of salt crystals in a i r gives faces with higher sticking probabilities for metal atoms than crystals cleaved in vacuo. Similarily, hydrocarbon deposits on the surface increase the adhesion of metal atoms to a surface [87]. In two studies, metal atoms adsorbed on substrates by vapor deposition were found to alter the

124

WANKE AND FLY"

electrical conductivity to a different extent than metal crystals, indicating electronic interaction. The effect of the metal atoms was found to decay gradually, presumably due to the migration and incorporation of the atoms into metal crystals [88, 891. The atmosphere to which a substrate is exposed will also affect the degree of epitaxy that metal crystals show toward the support [77]. Adhesion of metal atoms during condensation onto substrates is also sensitive to the heterogeneity of the substrate surface. The well-known decorating effect (e.g., Ref. 77), where nuclei concentrate around surface ledges o r dislocations in film growth studies, is attributed to an increased metal atom interaction with these areas of the surface, and hence a higher metal adatom population. Thus the interaction of metal atoms and crystallites with typical substrates is extremely complex. In ultrahigh vacuum on homogeneous surfaces, binding energies are low enough to be attributed to van der Waals interactions. However, in the presence of atmospheres, particularily oxygen, stronger interactions a r i s e which are not necessarily removed by reduction or evacuation. Surface contamination and heterogeneity also increase the energy of interaction. B. Crystallite Migration Model The crystallite migration model postulates that metal crystallites migrate a s entities along the surface of the support. Rapid diffusion of metal atoms on the surface of a metal crystallite will cause metal atoms to accumulate on one side of the crystallite by random fluctuations. This rapid, random surface diffusion will cause Browniantype motion of the particles on the support [76]. Direct evidence of this crystallite migration is difficult to obtain. The usually cited references in support of crystallite migration [go-931 use microscopy to investigate this phenomenon, and the results show either very slight crystallite motion, such as small rotations, o r they are anomolous in that large particles appeared to move faster than small particles. If appreciable particle migration occurs, it is restricted to metal crystallites smaller than 5 nm in diameter [76]. The model presented by Ruckenstein and Pulvermacher [65, 551 is hence restricted to the early stages of sintering when crystallites are < 5 nm in size. Two limiting cases of the crystallite migration model were developed: (1) surface diffusion controlled (i.e., the rate of migration of crystallites is the rate-determining process), and (2) sintering controlled (i.e., the merging of two metal crystallites coming into physical contact by collision is the rate-determining process). For both of these cases the authors were able to reduce the kinetic equations of the sintering process to power-law functions (Eq. 7) where

SINTERING OF SUPPORTED METAL CATALYSTS

125

the value of the power-law order, n, depends on the rate-determining step. For the sintering controlled case (i.e., merging of particles), n is equal to 2 o r 3, and for the surface diffusion controlled case . n 2 4 The magnitude of n for the surface diffusion controlled case is related to the dependence of the surface diffusion coefficient on the average metal crystallite size. Recently, Pulvermacher and Ruckenstein [94]described mathematical techniques for differentiating between sintering and diffusion controlled cases. The application of these techniques to sintering data obtained by TEM, SAXS, x-ray line broadening, and chemisorption was presented. In light of the limitations of the various methods for measuring dispersion (Section 11), it is doubtful whether the characterization of the catalysts is sufficiently precise to use the developed mathematical methods. Furthermore, Wynblatt and Gjostein [76]showed on the basis of theoretical predictions that the merging of metal crystallites cannot be the rate-determining process at temperatures above 500"C, i.e., the sintering controlled case is generally not realizable under normal sintering conditions. This is supported by observations that metal blacks sinter appreciably (fusion of particles in physical contact) a t temperatures below 200C [95, 961. Ruckenstein and Pulvermacher [65, 661 also concluded that for the diffusion controlled case the particle size distributions as a function of time can be represented by a unique dimensionless distribution (see Fig. 9, Ref. 66) after short sintering times. This type of unique, dimensionless distribution function has been used by Swift and Friedlander [97]in describing coagulation processes. The experimentally observed bimodal particle size distributions obtained by sintering supported metal catalysts [48, 601 do not agree with this conclusion. During the initial stages of sintering some crystallite migration may occur, but it is a highly unlikely mechanism for metal agglomeration for larger metal particles. Platinum agglomeration has been observed to continue when the size of the metal crystallites exceeds the size of the alumina support particles [45]. W e conclude that during the latter stages of sintering a mechanism other than crystallite migration causes the observed growth of metal crystallites.

C. Atomic Migration Model


The model presented by Flynn and Wanke [67, 681 envisages sintering to occur a s a three-step process: (1) escape of metal atoms (or molecules such as metal oxides in an oxygen atmosphere) from the metal crystallite to the support surface, (2) migration of these atoms along the support surface, and (3) capture of these migrating atoms by metal crystallites upon collision of these migrating atoms with stationary metal crystallites. This process results in the growth

126

WANKE AND FLY"

of the large metal crystallites and in a decrease in size of the small metal crystaflites since the rate of loss of atoms is smaller than the rate of capture for large crystallites, while for small crystallites the rate of loss is greater than the rate of capture. This occurs because large crystallites are in equilibrium with a lower concentration of migrating atoms than small crystallites (this is the two-dimensional analog of Ostwald ripening). Two cases of this model were analyzed by Flynn and Wanke: (1) the rate of capture of migrating surface atoms is large (this results in negligibly small concentrations of migrating surface atoms), and (2) the rate of capture is small (this can result in an appreciable concentration of migrating surface atoms). For both cases the rate of surface migration was assumed to be large, i.e., there are no concentration gradients of migrating atoms on the support surface. The rate of loss of atoms from the metal crystallites was taken to be independent of the crystallite size; this assumption approximates the rate of loss predicted by the Kelvin equation for metal crystallites with sizes of 2 to 50 nm [67]. (A more complex size-dependent rate of loss can be incorporated into the model.) The temperature dependence of the rate of loss of atoms from the crystallites was assumed to follow the Arrhenius law. Both cases of the atomic migration model predict a strong dependence of the rate of sintering on the metal particle size distribution (PSD). Catalysts with broad or bimodal PSD a r e predicted to sinter more rapidly than catalysts with narrow PSD. Initially, unisized PSD on homogeneous supports would not sinter at all according to the model, and statistical fluctuations in particle sizes to cause deviations from the unisized PSD would be required to initiate sintering. For short periods of time the dispersion as a function of time predicted by the atomic migration model can be approximated by a power-law rate equation. This approximation is not valid for long sintering times since the kinetic equations of the atomic migration model, both for slow and rapid atom capture, have a PSD dependence, and hence these equations cannot generally be approximated by a power-law rate function which contains only a dependence on the average particle size (dispersion). Bett et al. [33] state that the rate of sintering for Ostwald-ripening-type mechanisms is independent of metal loading. This conclusion is correct if the buildup of species in the transporting medium (in the case of supported metal catalyst sintering, the metal migrating on the surface) is small. However, when the atomic migration model predicts a significant buildup of transported metal, the surface loading becomes important; the rate of collision of migrating atoms with stationary metal crystallites depends on the surface concentration of migrating atoms, which in turn depends on the concen-

SINTERING OF SUPPORTED METAL CATALYSTS

127

tration of metal crystallites (i.e., metal loading). For high metal loadings, the rate of capture is larger than for low metal loadings, and hence the buildup of migrating metal is lower. For the case of rapid capture, the metal dispersion decreases monotonically with time, while for low rates of capture the metal dispersion can increase initially and then decrease. This increase in dispersion is due to the migrating surface atoms which have a dispersion of unity. Redispersion will occur i f these migrating atoms a r e trapped at high energy sites on the support surface o r the nucleation of small metal particles from these migrating species is brought about by rapid cooling. (It is interesting to note that the results of Emelianova and Hassan [30]show an effect of the rate of cooling following thermal treatment on the observed dispersion. Rapid cooling results in higher dispersions.) The rate of loss, which partly governs the concentration of migrating atoms, is greatly influenced by the atmosphere due to the influence of atmosphere on the metal-support interaction (Section IV-A). Large metal-support interactions increase the ease with which atoms o r molecules can escape from the crystallite to the support. The atomic migration model can be used to explain many of the observations of sintering studies, but from a mechanistic point of view it has one marked deficiency. In order for appreciable rates of sintering to be predicted by this model at temperatures of 500 to 800C,the activation energy for the escape of metal atoms (or molecules) from the crystallites has to be 680 kcal/g atom. The heat of sublimation of Pt, for example, is 135 kcal/g atom [98], and hence the metal support-interaction has to be of the order of 55 kcal/g atom o r higher. The qualitative discussion of the metal-support interactions presented in Section IV-A showed that metal-support interactions stronger than van der W a a l s interactions can occur, but to require these interactions to be 255 kcal/g atom in order for sintering to occur does not appear to be justifiable on the basis of the discussion presented. A problem with the practical application of the atomic migration model is that the initial PSD has to be known. Reliable data of this type a r e difficult to obtain.
D. Comparison of Crystallite and Atomic Migration Models

Experimental attempts to determine which of the two mechanisms is predominantly responsible for the sintering of supported metal catalysts have not been conclusive. In large part this inconclusiveness arises from the complexity of the sintering process. If support surfaces were energetically homogeneous, a number of tests could distinguish between crystallite and atomic migration. For example, the crystallite migration model predicts unisized o r narrow PSD

128

WANKE AND F L Y "

catalysts will sinter readily, while the atomic migration model predicts such catalysts would sinter extremely slowly. However, real surfaces are, of course, energetically heterogeneous; if one includes strong trapping sites and preferred adsorption areas, either model can account for a wide variety of phenomena. Model discrimination will be extremely difficult for this case. Bett et al. [33]concluded, on the basis of increasing sintering rates with increasing metal loading, that sintering occurs via the particle migration mechanism. However, the results of Gruber [36, 421, shown in Fig. 1, do not show the same dependence of sintering rates on metal loading, and furthermore, the atomic migration model also predicts a sintering rate which is dependent on metal loading. Flynn and Wanke [45]attempted to differentiate between the two mechanisms by examining identical areas of a catalyst surface before and after thermal treatment by means of TEM, but their results were also inconclusive. Some data useful in discriminating between the two mechanisms are: (1) the observed redispersion (Table 9), (2) the formation of a bimodal PSD during sintering [60], (3)the absence of appreciable sintering in vacuum at 800 and 900C [55, 561, (4)the increase in the sintering rate due to an initial bimodal PSD [45],and (5) the observation of continued metal crystallite growth in regions where crystallite migration is implausible [45]. All of these observations a r e consistent with the atomic migration model. With the exception of redispersion and growth of very large crystallites, the observations can also be explained by crystallite migration on a heterogeneous support. The experimental identification of the sinter ing mechanism(s) is a problem that awaits solution.
V. CONCLUDING REMARKS
The previous sections have illustrated that the processes involved in the sintering of supported metal catalysts are, at best, only partially understood. Empirically, it has been established that oxidizing atmospheres cause more rapid sintering than reducing and inert atmospheres. The rate of sintering is very temperature dependent, but the Arrhenius law applied to a power-law rate function does not correlate sintering data adequately. The identification of the sintering mechanism is of prime importance for determining the factors which affect the sintering rate. The crystallite migration model predicts a different dependence on PSD and metal-support-atmosphere interaction than the atomic migration model. The crystallite migration model predicts that the rate of sintering is relatively unaffected by the initial PSD, while the atomic migration model predicts a marked dependence on the

SINTERING OF SUPPORTED METAL CATALYSTS

129

initial PSD. The crystallite migration model predicts that for large metal-support interactions sintering would be slow due to the small mobility of the metal particles, while the atomic migration model predicts that large metal-support interactions would result in a high sintering rate due to the large atom migration along the support. Once the sintering mechanism has been established, the results can then be used as a guideline for the development of supported metal catalysts with greater thermal stability. For example, if atomic migration is the sintering mechanism, then a catalyst with narrow PSD and the use of alloys which increase the metal-metal interaction and decrease the metal-support interaction would result in more stable catalysts; on the other hand, if crystallite migration is the sintering mechanism, then catalysts with strong metal-support interactions would be stable catalysts. W e have not been successful in determining the processes which occur during sintering of supported metal catalysts, but if we have correctly identified areas in which further investigation will yield conclusive evidence of the sintering mechanism then the main objective of this paper will have been met.

APPENDIX A. CONVERSION OF REPORTED DATA TO METAL DISPERSION In order to convert metal surface areas, average metal particle size, o r gas adsorption uptakes to metal dispersion values for the adsorption stoichiometries, surface area per metal atoms, etc. have to be known o r assumed. The values used in this work are tabulated in Table A-1. Using the values and assumptions listed in Table A-1, the following relationships can be obtained:
1. Dispersion a s a function of metal surface area, A, and weight fraction metal, x, yields
D = 3.63 x 10-3 A/X D = 1.50 x 10-3 A/X

for Pt for N i

where A is the metal surface area per gram of catalyst. If the area is reported per gram of metal, x is set to u i y nt. 2. Dispersion a s a function of average crystallite size yields D = 0.509/r D = 0.505/r for Pt for Ni

130
TABLE A-1

WANKE AND FLY"

Values and Assumptions Used for CalculatingMetal Dispersion Metal Area per surface metal atom (nm2) Density (g/cm3) Pt Ni Rh Pd

0.089 21.45

0.065 8.90

CO adsorption stoichiometry

One CO molecule per surface metal atom for all metals One hydrogen or one oxygen atom per surface metal atom for all metals Spherical crystallites for all metals

O2 and H2 adsorption
stoichiometry Crystallite shape

where r is the crystallite radius in nm. The above relationships are valid for r 2 1 nm. 3. Dispersion as a function of gas adsorption uptake, U, and weight fraction metal yields

where M is the atomic weight of the metal, U is in moles of gas adsorbed pe r gram of catalyst, and I = 2 for 0, and H, , and I = 1 for CO. APPENDIX B: CALCULATION OF APPARENT ACTIVATION ENERGIES All the calculated apparent activation energies are based on the simple power-law rate function -dD/dt
=

A exp(-E/RT)Dn

(B-1)

Integrating Eq. (B-1) from t = t, to temperature T, yields (for n # 1)

4 and D = D, to D, at constant

Integrating Eq. (B-1) over the same limits of D, i.e., D, to Dz, and the corresponding range of times t,' to 4' at a temperature T yields ,

SINTERING OF SUPPORTED METAL CATALYSTS

131

Letting (t, - t,) = Atl, and (6 Eq. (B-3) yields

- t,) = Ah, and dividing Eq. (B-2) by

or
03-51

Equation (B-5) can be used to evaluate E i f sintering data are available at two o r more temperatures as a function of time and if the range of dispersions at the different temperatures overlaps, i.e., it is necessary to have the same values of D, and D, at different temperatures in order to obtain At, and A h . Much of the variable temperature sintering data in the literature reports the change in dispersion a s a function of temperature for a constant period of treatment. Equation (B-5) cannot be used to determine a n apparent activation energy from these data. A value of E as a function of n can be obtained for this case. Integrating Eq. (B-1) at temperature T, and T, from t = 0 to t, with the corresponding changes in dispersion from Do to D, for T = T, and Do to D, at temperature T, yields

and

Dividing Eq. (B-6) by Eq. (B-7) yields

132

WANKE AND FLY"

03-91
If dispersion data as a function of time and temperature are correlated by

D = atb

(B-10)

the value of a at two temperatures can be used to estimate E since for large values of n, D = [(n - 1 A e - E / R T ) (see Eq. 1 ) Hence 1. (B-12)
Using an average value of 1/(1

Il / l - n t l / l - n t

(B-11)

- n) for T, and T, gives


(B-13)

REFERENCES

T. E. Whyte, Jr., Catal. Rev.-Sci. Eng., 8,117(1973). T. A. Dorling, Warren Spring Laboratory, Report LR 144 (CA), Department of Trade and Industry (London), 1971: T. A. Dorling, Warren Spring Laboratory, Report LR 145 (CA), Department of Trade and Industry (London), 1970. P. C. Flynn, S. E. Wanke, and P. S. Turner, J. Catal., 33,233(1974). E.B. Prestridge and D. J. C. Yates, Nature (London), 234,345(1971). T. Paryjczak and B. Zarzycka, Rocz. Chem., 48,1731(1974). H. J. Maat and L. Moscou, Proceedings o f the International Congress on Catalysis, 3rd, North Holland, Amsterdam, 1965,p. 1277. A. Renouprez, C. HoangVan, and P. A. Compagnon, J. Catal., 34,411 (1974). N.R.Avery and J. V. Sanders, Ibid., 18,129 (1970). F.H. Huang and C. Li, Scr. Metdl., 7,1239(1973). S. Kimoto and J. C. Russ, Amer. Sci., 57,112 (1969). H. P. Klug and L. E. Alexander,X-ray DiffractionProcedures, Wiley, New York, 1954,Chap. 9.

SINTERING OF SUPPORTED METAL CATALYSTS

133

H. Spindler and M. Kraft, 2.Anorg. Allg. Chem., 391,155(1972). R.L.Moss, Platinum Met. Rev., 11,141(1967). C. R. Adams, H. A. Benesi, R. M. Curtis, and R. G. Meisenheimer, J. Catal., 1,

336 (1962). A. Guinier, Phys. Today, 22,25 (1969).


G. A. Somorjai, in X-ray and Electron Methods of Analysis (H. van Olphen and W. Parrish, eds.), Plenum, New York, 1968,Chap. 6. J. MUller, Reu. Pure Appl. Chem., 19,151(1969). E.G. Schlosser, Chem.-Ing.-Tech., 39,409(1967). H. L.Gruber and A. Hausen, Kolloid-2. 2. Polym., 214,66 (1966). G. R. Wilson and W. K. Hall, J. Catal., 17,190(1970). S . E. Wanke and N. A. Dougharty, Ibid., 24,367(1972). R. A. Dalla Betta and M. Boudart, Proc. Int. Congr. Catal., 5th, 2,96-1329

(1972).

E. Kikuchi, P. C. Flynn, and S. E. Wanke, J. Catal., 34,132(1974). J. Freel, Ibid., 25,149 (1972). F. E. Shephard, Zbid., 14,148(1969).
T. A. Dorling and R. L. Moss, Ibid., 7 , 378 (1967). G. A. Mills, S. Weller, and E. B. Cornelius, Proceedings of the 2nd Znternationol Congress on Catalysis, North Holland, Amsterdam, 1961,p. 2221. W. E. Armstrong, T. J. Jennings, and H. H. Voge, J. Catal., 24,502(1972). G. I. Emelianova and S A. Hassan, Proceedings of the International Congress on . Catalysis, 4th, Rice Univ. Printing, Houston, 1969,p. 1329. L. Spenadel and M. Boudart, J. Phys. Chem., 64,204(1960). R. A. Herrmann, S. F. Adler, M. S. Goldstein, and R. M. DeBaun, Ibid., 65, 2189 (1961). J. A. Bett, R. Kinoshita, and P. Stonehart, J. Catal., 35,307 (1974). J. Bett, K.Kinoshita, K. Routsis, and P. Stonehart, Zbid., 29,160(1973). T . R.Hughes, R. J. Houston, and R. P. Sieg, Ind. Eng. Chem., Process Des. Deuelop., 1,96 (1962). H. L.Gruber, J. Phy. Chem., 66,48(1962). N. H. Sagert and R. M. L. Pouteau, Can.J. Chem., 49,3411(1971). H. A. Benesi, R. M. Curtis, and H. P. Studer, J. Catal., 10,328(1968). Z . JaworskaGalas and J. Wrzyszcs,Znt. Chem. Eng., 6,604(1966). G. R.Wilson and W. K. Hall, J. Catol., 24,306(1972). P.Wynblatt and N. A. Gjostein, Scr: Metall., 7,969(1973). H. L.Gruber,Anal. Chem., 34,1828(1962). Chem. Abstr., 68,31814b (1968) (patent, Netherlands Appl. 6,614,074). M. h a f t and H. Spindler, Proceedings of the Internationol Congress on Catalysis, 4th, Rice Univ. Printing, Houston, 1969,p. 1252. P. C. Flynn and S. E. Wanke, J. Catal., 37,432 (1975). M. F.L.Johnson and C. D. Keith, J. Phys. Chem., 67,200(1963). S . F. Adler and J. J. Keavney, Ibid., 64,208 (1960). S.E.Wanke, Ph.D. Thesis, University of California, Davis, 1969. D. J. C. Yates and J. H. Sinfelt, J. Catal., 8,348 (1967). P. C. Aben, Zbid., 10,224(1968). J. L.Carter, J. A. Cusumano, and J. H. Sinfelt, J. Phys. Chem., 70,2257(1968). R.van Hardeveld and A. van Montfoort, Surface Sci., 4,396 (1966).

134

WANKE AND FLY


A. Williams, G. A. Butler, and J. Hammonds, J. Catal., 24,352(1972). P. M. Selwood, S. Adler, and T. R. Phillips, J. Amer. Chem. Soc., 77,1462

(1955).
M. Boudart, A. W. Aldag, L. D. Ptak, and J. E. Benson,J. Catal., 11,35(1968). H. Spindler, Int. Chem. Eng., 14,725 (1974); 2.Chem., 15,1(1973). also D. Pope, W. L. Smith, M. J. Eastlake, and R. L. Moss,J. Catal., 22,72 (1971). I. C. Brownlie, J. R. Fryer, and G. Webb, Ibid., 14,263(1969). H. Blume, C. Szkibik, F. Pfeiffer, H. Klotzsche, E. R. Strich, K. Becker, and G. Weidenback, Chem. Tech. (Leipzig), 18,449 (1966). C. J. Plank, G. T. Kokotailo, and L. C. Drake, 140th Meeting of the American Chemical Society, Division of Colloid Chemistry, Chicago, September 1961, Paper 45. R. A. Haslett, SAE Meeting, Detroit, February 1974,SAE Paper 740246. B. D. Lockhart and S. L. Genslak, SAE Meeting, Detroit, February 1974,SAE Paper 740245. M. F. L. Johnson, J. Mooi, H. Erickson, W. E. Kreger, and E. W. Breder, SAE Meeting, Toronto, October 1974,SAE Paper 741079. A. Koenig, H. Hellbach, and G. Doering, SAE Meeting, Toronto, October 1974, SAE Paper 741059. E. Ruckenstein and B. Pulvermacher, AIChE J., 19,356 (1973). E. Ruckenstein and B. Pulvermacher,J. Catal., 29,224 (1973). P. C. Flynn and S. E. Wanke, Bid., 34,390 (1974). P. C. Flynn and S. E. Wanke, Ibid., 34,400(1974). M. v. Smoluchowski, 2.Phys. Chem., 92,129(1917). W. Ostwald, Zbid., 34,495(1900). A. B. Shekhter, A. I. Echeistova, and I. I. Tretyakov, Izv. Akad. Nauk SSSR Otd. Khim Nauk 1950,465.(This reference taken from Ref. 28). J. C. Chaston, Platinum Met. Rev., 8,50(1964). H. Schafer, Chemical Transport Reactions, Academic, New York, 1964,p. 39. T. P. Kobylinski, B. W. Taylor, and J. A. Young, SAE Meeting, Detroit, February 1974,SAE Paper 740250. W. R. Pierson, R. H. Hammerle, and J. T. Kummer, SAE Meeting, Detroit, February 1974,SAE Paper 740287. P. Wynblatt and N. A. Gjostein, Progr. Solid State Chem., 9,In Press. J. W. Geus, in Chemisorption and Reactions on Metallic Films (J. R. Anderson, ed.), Academic, London, 1971,Chap. 3 . M. Humenik and W. D. Kingery, J. Amer. Ceram. SOC.,37,18 (1954). W. D. Kingery, Ibid., 37,42(1954). R. M. Pillar and J. Nutting, Phil. Mag., 16,181 (1967). J. A.Pa& and R. M. Fulrath, J. Amer. Ceram. SOC..45,592(1962). P. Benjamin and C. Weaver, Proc. Roy. SOC., A , 254,163(1960); Ser. 261,516

(1961); 274,267(1963).
C. Weaver, Chem. Ind., 91,370 (1965). D. C. Moore and H. R. Thornton,J. Res. Nat. Bur. Stand., 62,127(1959). D. M. Mattox, J. Appl. Phys., 37,3613(1966). A. S. Darling, G. L. Seman, and R. Rushforth, Platinum Met. Rev., 14,54

(1970); 14,95(1970); 14,124(1970);15,13(1971).

SINTEFUNG O F SUPPORTED METAL CATALYSTS

135

M. Krohn and A. Barna, in Proceedings of the Second Colloquium on Thin Films, Budapest, 1967 (E. Hahn, ed.), Van den Hoeck and Ruprecht, Gottingen, 1968,
p.

45.

D. G. Tabatadze, I. A. Myansikov, and L. A. Evstigneeva, Russ. J. Phys. Chem.,

46, 1488 (1972).


V. H. Gentsch, V. Hartel. and M. Kopp, Ber. Bunsenges, Phys. Chem., 75,1086

(1971).
D. M. Pashley, M. J. Stowell, H. M. Jacobs, and T. J. Law,Phil. Mag., 10,127

(1964).
W. B. Phillips, E. A. Deslodge, and J. G. Skofronick, J. Appl. Phys., 39,3210

(1968).
M. J. Thomas and P. L. Walker, Jr., J. Chem. Phys., 41,587(1964). G. W. Sears and J. B. Hudson, Zbid., 39,2380 (1963). B. Pulvermacher and E. Ruckenstein, J. Catal, 35,115 (1974). D. W. McKee,J. Phys. Chem., 67,841(1963). S . A. Khassan, S . G. Fedorkina, G. I. Emelyanova, and V. P. Lebedev, Russ.J. Phys. Chem., 42,1324(1968). D. L. Swift and J. Friedlander, Colloid Sci., 19,621(1964). E. R. Plante, A. B. Sessoms, and K. R. Fitch, J. Res. Nut. Bur. Stand., A, 74,

647 (1970).

Vous aimerez peut-être aussi