Carbon supported noble metal catalysts in the oxidation of glucose 1-phosphate and related alcohols

(1)

Carbon supported noble metal catalysts

in the oxidation of glucose 1-phosphate

and related alcohols

(gasphase)

/ ^ ^

(bulk)

O

TR diss

(2)

(3)

Carbon supported noble metal catalysts

in the oxidation of glucose 1-phosphate

and related alcohols

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Technische Universiteit Delft

op gezag van de Rector Magnificus, prof.drs. P.A. Schenck in het openbaar te verdedigen ten overstaan van

een commissie aangewezen door het College van Dekanen op dinsdag 28 februari 1989 te 16.00 uur

door

Hans Erwin van Dam

geboren te Delft scheikundig ingenieur

(4)

Dit proefschrift is goedgekeurd door de promotor prof.dr.ir. H. van Bekkum.

(5)

(6)

The investigations described in this thesis have been supported financially by the Netherlands Organisation for Scientific Research (NWO) and the Innovation Oriented Program for Biotechnology (IOP-B).

The financial support for the printing of this thesis by Norit N.V. is also gratefully acknowledged.

(7)

1 Introduction

This thesis deals with different aspects of the catalytic oxidation of polyols in the liquid phase, combined with the preparation of the supported noble metal catalysts used. These topics fall into the field of heterogeneous fine chemistry catalysis. Before entering the detailed discussions in the following chapters, the opportunity is taken to present a broader view on fine chemistry catalysis in general and heterogeneous fine chemistry oxidation catalysis in particular.

Traditionally, the term 'catalysis' has been more often associated with bulk chemistry than with fine chemistry. Bulk chemistry is largely based on heterogeneous catalysis. In table 1, heterogeneous catalysis in bulk chemistry and in fine chemistry are compared. Reviewing both fields, it appears that in bulk chemistry (as compared to fine chemistry) more money is spent on catalysis research, and less (selectivity) problems are encountered.

Still, in terms of added 'value, fine chemistry catalysis is not unimportant. It has been pointed out (R. Bader, Ciba-Geigy) that, although only -3 % of all industrial heterogeneous catalysts is used in the synthesis of fine chemicals, the processes involved yield -20 % of

the profits of all industrial heterogeneous catalysis applications. Indeed, catalysis plays an (increasingly) important role in the synthesis of many fine chemicals. Supported noble metals are being used successfully in hydrogenation processes (although selective conversions

(11)

Table 1. Heterogeneous catalysis in bulk chemistry vs. fine chemistry feature bulk chemistry fine chemistry types of products

industry focussed on empirical approach towards

development of catalyst and process

performance products (e.g.gasoline) and/or simple molecular structure (e.g. methanol) pure chemicals with complex molecular structure (regio/stereosel. often required) types of processes types of plants number of processes research budget per process research of chemical continuous largely gasph usually high temperature/p dedicated limited large catalyst and ase ressure process largely batch liquid phase usually mild temperature/pressure multi purpose (partly) high 1imited process (chemistry) combined catalyst usually separated, cat manufacturer and fine chem. ind, resp. of polyfunctional compounds still present problems). Reusable acid catalysts (e.g. ion exchange resins, zeolites) are increasingly applied. The relatively new fields of homogeneous hydrogenation catalysis and enzyme catalysis seem to gain importance rapidly, especially in the field of enantioselective reactions.

Compared to these, oxidation catalysis is developing much less spectacularly. Indeed, in fine chemistry many oxidations are still only feasible non-catalytically, using stoichiometric reagents (esp. chromium and manganese compounds). These non-catalytic reactions often present both economic and environmental problems. Thus there is clearly a need

(12)

for selective oxidation catalysts that would enable a growing use of clean oxidants such as oxygen (air) and hydrogen peroxyde.

In hydrogenation reactions, much attention is presently given to homogeneous catalysts. These often enable enantioselective reactions, and have therefore a distinct advantage over heterogeneous catalysts. In

3

the catalytic oxidation of hydroxyl groups however, sp carbons are 2

converted into non-chiral sp carbons. Here, except in the case of oxidative resolution of racemic mixtures, enantioselectivity is of no importance. So, in this field (easily recoverable) heterogeneous catalyst are more attractive than homogeneous systems.

Whilst in homogeneous catalysis developments are being made on the basis of chemical insight, the present status of supported noble metal catalysis has been reached largely by many decades of trial-and-error research. In the case of hydrogenation, this approach has proved fairly successful, but in the case of oxidation (or: oxidative dehydrogenation) results have been rather poor. This difference is plausibly explained by two factors. The first is that, in the case of oxidative dehydrogenation, two processes must take place simultaneously on the metal surface: substrate dehydrogenation, and hydrogen oxidation. The second factor is the complete thermodynamic instability of the oxidation system: the organic molecules tend to yield carbon dioxide as only stabile organic product, and the catalytic zerovalent metal surface is readily converted into the corresponding oxide.

So, to enable the use of noble metal oxidation catalysts, a thorough understanding of both the inorganic (catalyst) and the organic (reaction) aspects of the system seems necessary. Selective and stabile catalytic processes will in many cases only be feasible when 'designer catalysts', tailor-made for the process in question, are developed through an integrated approach of catalyst engineering.

However, in fine chemistry catalysis, the processes are carried out on a relatively small scale. This inhibits the fine chemical industry from spending the necessary effort on heterogeneous catalysis research. Therefore, in fine chemistry catalysis the preparation of heterogeneous catalysts and the development of catalytic processes are most often strictly separated. By contrast to the situation in bulk chemistry, the fine chemistry catalyst manufacturer produces catalysts for largely unknown applications, whereas the fine chemical industry regards

(13)

supported noble metal catalysts as black boxes. This situation will clearly not lead to the advanced catalysis necessary to solve the problems outlined above.

Universities can play an important role in overcoming these problems by educating chemists with a knowledge of both catalysis and fine chemistry, and further developing catalysis from an art into a science.

The subject of this thesis is noble metal catalyzed oxidation of polyols into carboxylates. To carry out most of the work a model substrate had to be selected. Glucose 1-phosphate (G1P) was used throughout this study. The reasons were that (i) G1P was of interest as a sugar derived intermediate in the synthesis of glucuronic acid and (possibly) vitamin C, and (ii) the G1P oxidation suffered from reactivity and selectivity problems as also encountered with many other substrates. In those cases where the actual structure of the substrate was of importance other starting materials were included in the studies. Upon initiating the research, current state-of-the-art oxidation procedures were found to present several problems:

- commercial platinum catalysts showed poor stability; better catalysts had to be developed,

- literature procedures for the preparation of platinum on activated carbon catalysts were mostly of the cookbook type and yielded unpredictable results,

- the selectivity of the oxidation process was poor,

- the choice of the metal was based on empirical knowledge rather than fundamental insight,

- convenient analytical procedures for the analysis of the reaction mixtures were non-existent.

The present work was aimed at overcoming these difficulties. The character of the work is fundamental. Rather than giving an empirically optimized process for one specific substrate, emphasis was given on studying the basic principles of catalyst preparation, activity, stability, and selectivity. Figure 1 presents the different aspects of the oxidation process, and the chapters in which they are treated.

(14)

CATALYST PREPARATION CATALYTIC PROCESS CARRIER STRUCTURE (CH. 2) IMPREGNATION (CH. 3) REDUCTION (CH. 4) (gasphase)

/

-c^v

(bulk)

o

BOUNDARY EFFECTS (CH. 7) HPLC ANALYSIS (CH. 10) PORE DIFFUSION (CH.6) OXYGEN POISONING (CH. 5) METAL SELECTION (CH. 9) SELECTIVITY (CH. 8)

(15)

2 Catalyst preparation I: carrier

INTRODUCTION

In contrast to most inorganic carriers, activated carbon shows an excellent stability in virtually all liquid media. It also allows a relatively easy recovery of the active phase: Carbon is therefore the carrier of choice in most heterogeneously catalyzed liquid phase processes in fine chemistry. Activated carbon supported catalysts may also be applied to gas phase reactions.

Platinum-on-carbon is used as catalyst for hydrogenation and -to a lesser extent- for oxidation reactions. A large number of patents and papers deal with the preparation of platinized carbon. However, most of this work is empirical in character, and little has been reported about the actual 'chemistry' of the catalyst preparation process.

Such knowledge would greatly facilitate the design and preparation of Pt/C catalysts with desired properties. Therefore, we have investigated the catalyst preparation process in detail, including: (i) the chemical modification and characterisation of activated carbon, ( i i ) . the impregnation of these carriers with platinum compounds, and (iii) the reduction of these impregnates to metallic Pt/C catalysts. In this paper, we will focus on the chemical modification and characterisation of the carrier carbons.

(16)

In the literature (1), several studies have appeared concerning the effect of the carbon texture on the properties of Pt/C catalysts. The pore structure controls the diffusion of reactants and products within the catalyst particle. Thus it affects the selectivity (2), the activity (3), and the stability (4) of the catalyst. The metal particles are located in the (meso (4)) pores of the carrier, and thereby limited in size (4,5). It has been reported that, apart from this crystalline phase, also a (near) atomic form may be present (6), located probably in the micro pores.

Studies with graphitized carbon blacks (7) showed that an increase in so called 'surface heterogeneity', caused by burn-off in air, reduced the rate of (high temperature) sintering and (thus) enhanced the dispersion of the active phase. The oxygen structures at the surface, introduced during the burn-off procedure, are thought to retard the platinum migration. This 'surface heterogeneity' is also believed to stabilize CoMoS HDS catalysts (8). In these studies however,

o

temperatures upto 500 C were employed during the catalyst preparation. Much of the carbon surface 'chemistry' is lost then. Inorganic impurities, present in activated carbons that have not been washed properly, have been reported to impede the (high temperature) sintering of platinum crystallites (9).

In figure 1, the structure of the carbons used in our studies is represented. Pure graphite consists of hexagonal carbon basal planes which are stacked parallel, mostly in an ABAB sequence (10). Activated carbons (11) contain crystallographically disordered graphite-like crystallites. These are agglomerated to carbon powder particles, which may be converted into extruded forms. In the extrudates, the carbonized binder renders the product a satisfactory mechanical strength.

Dependent on the pretreatment, the edges of the carbon layers may be occupied by different oxygen containing surface groups (often referred to as 'oxygen complexes'). The quantitative characterisation of these groups (12) is not a simple task. In early surface chemistry studies, the carbons have been treated as -rather complex- organic molecules, containing well defined moyeties, such as carboxyl, lactone, hydroxyl, quinone, etc. Many attempts have been made to quantify these groups by means of classical organic reactions. However, this approach has failed to provide a generally applicable self-consistent method of surface

(17)

o ^

o

O I

1 I 4 mm

ACTIVATED CARBON EXTRUDATES

I I 4 nm

DISORDERED GRAPHITE CRYSTALLITES

Figure 1. Structure of activated carbons used.

characterisation. This may be illustrated (13) by the results of active hydrogen determinations in Graphon, a graphitized carbon black: different methods yielded results which vary over an order of magnitude.

More recently, it has been pointed out (14) that the surface groups in carbons cannot be considered as isolated groups, since they are part of large conjugated systems. As a result, the reactivities of these groups are strongly altered. For example, it has been shown that anhydride groups which are part of large conjugated systems are much more stabile towards alkaline hydrolysis than their small-molecular size analogues (15).

20 Llm | CARBON POWDER WITH I CARBONIZED BINDER |

_ _ _ I

•IHOH C-OOH

VyVo

YWO

OH ^ " H y O

X

c

.o

*0ÓH 1 nm

(18)

Another interesting feature of carbonaceous materials is the posssibility of tautomeric 'interconversion of functional groups. Keto enol tautomerism is pertinent (14), effectively resulting in the interconversion of carbonyl and hydroxyl groups. Similar tautomerisms can be envisaged for other surface groups. Thus, using chemical characterisation methods, the number of any specific type of surface group is always suspected to be influenced by the analytical tool.

Spectroscopie methods (IR, XPS, NMR) are more promising. Infrared spectroscopy has often been applied in the analysis of carbons. Some attention has been given to problems in the interpretation of the spectra, which are again caused by conjugation effects (16). X-ray photoelectron spectroscopy has been applied to carbon fibers and electrodes (17,18). In ideal conditions, a quantitative analysis of surface groups can be obtained (carbon atoms having 0 to 4 bonds to oxygen have been differentiated and determined separately) (18). Nuclear magnetic resonance spectroscopy seems less promising but should not be neglected.

A practical approach towards the surface chemistry is formed by acid and base titrations. Although interpretation in terms of specific types of surface groups seems not possible, the obtained data can be directly related to the carbon surface charge. Electrophoresis has also been used to determine the zero point of charge of carbons. However, this method has recently been disqualified as a means of internal surface chemistry characterisation (19). Electrophoresis yields results which are specific for the chemical structure on the outside of a carbon particle. This 'outside structure' is often different from the surface chemistry within the pore system.

The exact electronic nature of the carbon layer structure may be of great importance. Activated carbons possess semi-conductor properties

(20). The band gap energy is approximately 0.1-0.05 eV for normal activated carbons, whereas graphite is a semi-metal. Conjugation with surface groups depends on the energies involved and thus on the extension of the carbon layer structures. Therefore, generalisations of results are extremely perilous, especially when different types of material are involved (e.g. activated carbons and graphetized carbon blacks).

(19)

In the present work Norit ROX 0.8 was used as starting material. This commercial extruded activated carbon (figure 1) was choosen for its purity, mechanical strength, and corresponding ease of handling. The carbon was modified by different nitric acid oxidations, air oxidation, thermal decarboxylation, or hydrogen reduction. The resulting carriers were submitted to several characterisation techniques, i.e. nitrogen porosimetry, thermal decomposition-GC (CO formation), TGA, acid/base titrations, XPS, IR, and NMR.

MATERIALS AND METHODS

Starting material: Norit ROX 0.8 activated carbon (Norit NV,

Amersfoort, The Netherlands). Peat based, gas activated extrudates (d 0.8 mm, 1 3-5 mm). Macropores 0.2-5 nm (from mercury porosimetry). Total

pore volume 1.0 ml/g. Inorganic impurities before washing procedure (see below): Fe 0.02%, Ca 0.02%, total ash content 3% (mostly Si02).

chemical modification methods

Nitric acid oxidations Carriers N5, N10, N20, and N30 were prepared

by oxidation of ROX 0.8 with different amounts of nitric acid. ROX 0.8 (50 g) and water (230.8, 211.5, 173.1, and 134.6 g resp.) were stirred and heated to 100 C. Nitric acid (65% by weight) (19.2, 38.5, 76.9, and 115.4 g resp.) was added in 30 min, and the reaction was continued for 60 min (100-103 C, stirring). The mixtures were cooled, poured in water, and washed (see below). Small amounts of fines were removed by flotation.

Air oxidation Carrier A was prepared from ROX 0.8 by air oxidation

o

for 24 h at 350 C in a rotating kiln (courtesy Norit NV).

Decarboxylation Carrier D was prepared from ROX 0.8 by heating for

o

1 h at 320 C in flowing helium using a vertical tube oven.

Hydrogen treatment Carrier H was prepared similarly using hydrogen

0 0 gas (3 h, 320 C) followed by passifation in helium (1 h, 320 C ) .

Washing procedure Prior to further use all carbons (50 g) were

slowly percolated with: (i) water (5 1 ) ; (ii) sodium hydroxyde solution (0.5 M, 5 1, final eluate colorless); (iii) water (5 1 ) ; (iv) hydrochloric acid (0.5 M, 5 1, final eluate salt free); and (v) water (10 1, final pH 6 ) , and dried overnight at ambient conditions.

(20)

Characterisation methods

Nitrogen porosimetrv Mesopore volume distributions and BET surface

areas were determined by nitrogen adsorption at -196 C using a Carlo Erba Sorptomatic 1800 apparatus.

Thermal decomposition-GC Dried samples (300-400 mg) were heated

o

(10 C/min) in helium (40 ml/min). The carbon oxides released by the o

sample were quantitatively analysed at 2 min (20 C) intervals by gas chromatography (courtesy Norit NV).

TGA Thermogravimetric analyses were performed on a DuPont TGA

apparatus using the same temperature program (courtesy Norit NV).

Titration curves Samples of the material (200 mg) were mixed with a

potassium chloride solution (5 ml 0.1 M ) . Concentrated potassium hydroxide or hydrochloric acid solutions were added. After equilibration (5 days) the pH values of the solutions were determined. Corrections were made for the capacity of the solution at extreme pH values.

XPS X-ray photoelectron spectroscopy was performed on a Leybold

Heraeus XPS/AES apparatus using MgK radiation and Shirley background correction (courtesy DSM Central Laboratory).

1R Infrared spectra of KBr tablets from finely ground powders were recorded using a Beekman 4210 apparatus.

10-Q. Q 5 -1-5 2 0 log Dp/A 2-5

Figure 2. Pore structures of acivated carbons.

(21)

RESULTS AND DISCUSSION

Mechanical strength The modification methods applied did not result in an excessive weakening of the carbon extrudate particles.

Pore structure Figure 2 provides data on the micro and meso pore structures of some selected carriers (the other carriers gave similar results). Air oxidation caused a small increase in BET surface area and meso pore volume (-20%). The other modification methods had not at all a significant effect on the pore structure.

XPS/NMR X-ray photoelectron spectra of ROX and N30 are presented in

figure 3. Both carriers can be seen to consist of carbon, oxygen, (and hydrogen) only. So, nitric acid oxidation did not result in the introduction of significant amounts of nitrogen. A detailed analysis of the carbon Is signal is hampered by the presence of a surface plasmon loss peak (17), a consequence of the semi-conductor properties of these

13 carbons. A similar problem interfered with attempts to obtain C NMR spectra of carbon N20 (21). In this case, the conductivity caused large shifts and line broadening, resulting in a complete loss of signals.

O 260 520 780 1040 1300 KINETIC E N E R G Y (eV)

(22)

IB Both nitric

-1 acid and air oxidations resulted in increased 1710 and 1230 cm" bands, indicating the formation of C00H groups (22). The 1585 cm band, attributed to C=C linkages (16), was also much enhanced. This is probably due to conjugation with an increased number of C=0 groups (16). 2 0 -1-5 o E c o

10-S o

"O o x

of

C O

5-°

V) ' a 0-4 ~i r -6 8 PH - i — 10 12

Figure 4 . Acid and base adsorption capacities of carbons versus pH.

(23)

Titration curves Figure 4 shows that all carriers have both acidic

and basic properties. The oxidations resulted in an increased surface acidity. High temperature treatment in helium and hydrogen (samples D and H, resp.) gave almost identical surface acidity losses. No steps, associated with discrete pK, values, are observed in the titration

a

curves. Since no step is found at p H ^ it would be possible that there is some overlap between the pK ranges of the acidic and basic groups,

a

leading to z w i t t e r i o n i c structures at pH values around pH.. However, in the IR spectra of the carbons no carboxylate ion band (1400 cm (22)) could be detected.

200 400 600 800 isoth. T C O

Figure 5. Thermolysis-GC analyses of carbons. GN30; A N 2 0 ; Q N 1 0 ; V N5; § A ; «ROX; T D .

(24)

TGA Weight losses as determined thermogravimetrically equalled the

sums of CO and CC^ losses, with the exception of small initial amounts (T < 100°) interpreted as physisorbed water.

Thermal decomposition-GC Figure 5 shows the results of temperature programmed CO formation experiments. Nitric acid oxidation increased the intensity of all peaks; so it seems likely that the concentrations of all types of surface oxygen complexes were increased. Air oxidation (carbon A) caused the most dramatic increases in the high temperature CO and C 02 peaks. Since this oxidation was performed at 350°C, no

significant low temperature (T<350 C) peaks would be expected. Indeed, the 250°C CO- peak of carbon A was not present directly after the oxydation. It was introduced during the washing procedure, appearingly by some hydrolytic conversion of surface groups.

In the COp thermograms, three peaks are observed, with maxima at 250, 430, and 650°C. In the literature (13), similar maxima have been reported (250 and 600°C using Spheron 6, and 400°C using graphite). Only the 400 and 600°C peaks are reported to be acidic towards aqueous sodium hydroxyde (13).

Table 1. Thermolysis-GC and titration data.

CO- CO base capacity oxygen T (°C): 0-360 380-620 640-1000 total total 2-12 p H - 1 2 sample ROX N5 N10 N20 N30 A D 0.20 0.31 0.51 0.64 0.87 0.21 0.10 0.12 0.28 0.45 0.55 0.81 0.26 0.13 (mole/kg)-0.06 0.13 0.19 0.22 0.29 0.34 0.07 0.38 0.72 1.15 1.41 1.97 0.82 0.30 1.29 2.05 2.69 2.98 3.78 3.85 1.07 0.92 1.24 1.52 2.00 2.40 1.62 0.95 0.74 1.03 1.70 1.85 2.30 1.42 0.50 (w%) 3.3 5.6 8.0 9.3 12.4 8.8 2.7

(25)

In table 1 the measured surface acidities of the carbons (figure 4) are compared with the amounts of surface complexes yielding C02 and CO

(figure 5 ) . In our carbons, the surface acidity cannot be solely accounted for by the 400 and 650°C C 02 peaks or even by all C02 yielding

surface complexes. It is to be concluded that the CO yielding complexes also contribute to the surface acidity.

We have attempted to further identify the CO desorption peaks related to the surface acidity. Carbon N30 was treated with LiOH/LiCl (final pH 12) to convert acidic species into lithium salts. The carbon thus obtained, denoted N30-Li, was submitted to pyrolysis-GC analysis; see figure 6. Interestingly, the more pronounced effects were observed in the CO desorption pattern, again indicating that (some of) the CO yielding surface complexes contribute to the surface acidity. Table I also shows that, though the total oxygen content of the carbon is not much affected by the lithium hydroxyde treatment, more oxygen is released as C 02 and less as CO from N30-Li. This must be caused by some

conversion of surface groups during the lithium hydroxyde treatment.

200

T(°C)

Figure 6. Thermolysis-GC analyses of carbon N30 and l i t h i u m hydroxyde treated carbon N30-Li.

(26)

If we assume the CO yielding complexes (partly) to be conjugated C=0 and C-OH groups, the 'extra' acidity is easily explained (even a COOH group can be seen as such a conjugated system). Many low molecular weight systems with acidic phenolic or quinonic groups are known (e.g. 2,5 dihydroxy benzoquinone) (23). Considering the pK values of these

a

compounds (23), it can be seen that a continuous scale of acidities can be obtained from both these conjugated systems and 'normal' COOH groups. So, acidity from CO yielding complexes is indeed to be expected, and the absence of steps in the titration curves is easily explained. The smoothness of the curves will be further enhanced by electrostatic effects, as can be concluded from the work of Van der Plas (24).

The basic surface sites are believed to be pyrone-like (25). The pK value of the pyronium ion equals -1.1. However, the extensive charge delocalisation in activated carbons will stabilize pyronium type structures, leading to a much increased basicity of pyrone groups.

Numerical evaluation of data The obtained data can be used to reach a quantitative picture of the activated carbons. The BET surface areas

2

of the carbons equal - 900 m /g. (The use of the exact value of the BET surface is formally not allowed. However, nitrogen and benzene BET measurements give identical results, and from the adsorption of hydroxy-methylfurfural from aqueous solutions (26) a similar surface area is calculated. A useful approximation is therefore expected). This would

2

correspond to an average surface area of approximately 1.8 A per carbon atom The theoretical (one-side) surface area per carbon atom in a basal

2

plane equals 2.6 A (calculated from the closest C-C distance in graphite (1.42 A) or from the density (2.3 g/ml) and the layer spacing (3.35 A) of graphite). So, the effective mean thickness of the carbon layer stacks equals three atomic layers. These thin stacks are packed together, (partly) separated by (slit shaped) pores, forming the disordered graphite crystallites (figure 1 ) .

It seems resonable to assume that the carbon basal planes have roughly the same dimensions (fig. 1) as the meso-pores, i.e. -3 nm (figure 2 ) . This corresponds to -300 carbon atoms per basal plane, 60 atoms (20 mole%) being located at the (perimeter) edge. Part of these edge atoms form the linkages to other carbon domains. The remainder is available to accomodate the oxygen atoms also present in the activated carbon carriers (upto 12.4 weight% (table 1 ) , i.e. 10 mole%).

(27)

CONCLUSIONS

The chemical modification methods applied to Norit ROX 0.8 resulted in a set of carbon carriers with virtually identical pore structures. The chemical structures vary considerably: a set of carriers was obtained with widely varying (pH-dependent) surface charge. The oxidized carriers have a moderately to strongly increased surface acidity. The acidity is caused by several types of oxygen functionalities, probably including conjugated hydroxyl groups. For the carriers under investigation, the 'classical' approach towards the surface chemistry failed to yield valid insight.

REFERENCES

1. Ehrburger, P., Adv. Coll. Interf. Sci. 21, 275 (1984). Juentgen, H., Fuel 65, 1436 (1986).

2. Parlitz, B., Schnabel, K.-H., Z. Allg. Anorg. Chem. 465, 299 (1974) 3. Kloubek, J., Medek, J., Carbon 24, 501 (1986).

4. Van Dam, H.E., Duijverman, P., Kieboom, A.P.G., Van Bekkum, H., Appl. Catal. 33, 373 (1987).

5. Plavnik, G.M., Parlitz, B., Dubinin, M.M., Dokl. Acad. Nauk SSSR 206, 399 (1972).

6. Rikhter, K., Levitskii, E.A., Kalomiichuk, V.N., Moroz, E.M., Kinet. Katal. 16, 1578 (1975).

7. Ehrburger, P., Mahajan, O.P., Walker Jr, P.L., J. Catal. 43, 61 (1976). Ehrburger, P., Walker Jr., P.L., J. Catal. 55, 63 (1978). 8. Vissers, J.P.R., Lensing, T.J., De Beer, V.H.J., Prins, R., Appl.

Catal. 30, 21 (1987).

9. Rodriguez-Reinoso, F., Morengo-Castilla, C , Guerrero-Ruiz, A., Rodriguez-Ramoz, I., Lopez-Gonzales, J.D., Appl. Catal. \S_, 293

(1985).

10. Kelly, B.T., "Physics of Graphite", p 34, Applied Science Publishers, London.

11. Mattson, J.S., Mark Jr., H.B., "Activated Carbon", Marcel Dekker, New York (1971).

12. Boehm, H.-P., Knozinger, H., in "Catalysis: Science and Technology" (J.R. Anderson, M. Boudart, Eds.) 4, 39, Springer Verlag, Berlin. 13. Barton, S.S., Harrison, B.H., Carbon 13, 283 (1975).

14. Papirer, E., Guyon, E., Carbon 16, 127 (1978). Papirer, E., Guyon, E., Peral, N., Carbon 16, 133 (1978).

15. MacPherson, A.S., Siudak, R., Weiss, D.E., Willis, D., Aust. J. Chem. 18, 493 (1965).

16. Akhter, M.S., Keifer, J.R., Chughtai, A.R., Smith, D.M., Carbon 23, 589 (1985)

(28)

17. Proctor, A., Sherwood, P.M.A., J. Electron Spectrosc. Relat. Phenom. 27, 39 (1982). Proctor, A., Sherwood, P.M.A., Carbon 21., 53 (1983). 18. Cabaniss, G.E., Diamantis, A.A., Murphy Jr., W.R., Linton, R.W.,

Meyer, T.J., J. Am. Chem. Soc. 107, 1845 (1985). 19. Corapcioglu, M.O., Huang, C.P., Carbon 25, 569 (1987).

20. McMichael, B.D., Kmetko, E.A., Mrozowski, S., J. Opt. Soc. Am. 44, 26 (1954).

21. Alma, N., personal communication.

22. Yoshida, H., Adachi, Y., Kamegawa, K., Tanso H I , 149 (1982), CA 98:116066d.

23. Serjeant, E.P., Dempsey, B., "Ionisation Constants of Organic Acids in Aqueous Solution" (IUPAC Chem. Data Ser. 23) Pergamon Press, Oxford (1979).

24. Van Der Plas, Th., "The Surface Chemistry of Carbon" (Thesis Delft University of Technol.), Bronder, Rotterdam (1968).

25. Boehm, H.-P., Vol 1, M., Carbon 8, 227, 741 (1970), ibid. 9, 473 (1971).

(29)

3 Catalyst preparation II: impregnation

INTRODUCTION

In the preparation of supported metal catalysts, the dispersion and distribution of the active phase are largely determined by the

impregnation step. In the case of platinized carbon, the precursor most often used is hexachloroplatinic acid (HpPtClg). This compound usually yields a better metal dispersion than e.g. platinum ammine complexes (1). The latter precursors result in high dispersions only when the carrier is specially pretreated to increase its ion exchange capacity

(2). From aqueous solutions, hexachloroplatinic acid is adsorbed very strongly onto activated carbons (up to approximately 10 weight percent of platinum, depending on the type of activated carbon). The adsorption is less strong when solutions of the platinic acid in organic solvents are used (3).

The actual 'chemistry' of the impregnation process is still not well understood. Early XPS work (4) indicated that, after impregnation of activated carbon with H,,Pt(IV)Cl, and subsequent drying in air, the platinum was present as Pt(0) and Pt(II). This is surprising since highly dispersed platinum metal is very sensitive to air oxidation (5). The effects of the carrier surface chemistry on the impregnation process are unknown.

(30)

It was decided to reinvestigate the impregnation process in detail, including the effects of some additives and the effects of variations in the surfa chemistry of the carrier. Hexachloroplatinic acid was selected as metal precursor.

Materials A commercial activated carbon (Norit ROX 0.8) was modified

and characterised by several methods, as described elsewhere (6). A summary of relevant data is given in table 1.

HgPtC1c adsorption isotherms Aqueous hexachloroplatinic acid

(Johnson Matthey) solutions (5 ml) were added to samples of the carriers (200 mg). After five days of slow agitation at 20°C, equilibrium was reached. The solutions were decanted and analysed spectrophotometrically

(7) for their platinum content.

Acetone extractions The impregnated carbons (200 mg) were extracted with p.a. acetone (5 ml) at 20°C for five days.

XPS X-ray photo electron spectra were recorded using a Leybold

Heraeus LHS-10 XPS/AES apparatus with a HP 1000 E data system. The instrument was calibrated externally. The Cls signal of the carbon was found at E„=284.3 eV, which coincides with the literature value (8). Sample charging was therefore concluded to be absent.

The impregnated activated carbon extrudates were broken to obtain fresh surfaces, stacked vertically in a sample holder, and dried in vacuo in the XPS apparatus. Both before and during irradiation (MgK ) mass spectra were recorded of the remaining gas molecules in the vacuum system, revealing m/e 35 + 37 (Cl) and 44 (e.g. CO-). Although a slight increase in pressure was observed upon starting the excitation, differences between the Pt4f spectra of samples exposed for very short and long times were insignificant. The Pt signal intensity was about fifteen fold lower than expected on a simple weight basis. A similar effect has been reported for rhodium on carbon, and was attributed to the porous nature of the carrier (9).

(31)

Table 1. Characteristic data on carriers used. carrier ROX N5 NIO N20 N30 A D H modification method -nitric acid oxidation (5-30% acid) air oxidation helium 320°C hydrogen 320°C oxygen content (w%) 3.3 5.6 8.0 9.3 12.4 8.8 2.7 n.d. PHi 5.5 4.7 4.1 3.3 2.8 4.7 8.1 8.4 Pt ads. (W/o)

Figure 1. Hexachloroplatinic acid adsorption onto activated carbons with different surface chemistry, (water, 25 C)

(32)

RESULTS AND DISCUSSION

Types of adsorption Figure 1 shows hexachloroplatinic acid

adsorption isotherms on different carbon carriers. It is obvious that the adsorption is not just a simple physical process. Part of the platinum compound (5-10 w% platinum load, dependent on the carrier) is adsorbed very strongly. Also, a weak 'equilibrium' adsorption may occur, which is virtually independent of the surface chemistry of the carrier. This weak adsorption is likely to be a physical process. Physisorption is expected to occur since hexachloroplatinic acid is soluble in both water and non-polar solvents. It may thus act as a "surfactant", reducing the surface tension between the aqueous solution and the carbon basal plane surface. The strong adsorption is regarded as a chemisorption process, and will now be discussed further.

1172-0 11772 1182 4

K ENERGY eV

Figure 2. XPS spectrum of hydrochloroplatinic acid adsorbed onto carbon ROX.

A: experimental spectrum; after deconvolution: B: EB(Pt 4f 7/2) 72.2 eV; C: Eg 74.2 eV.

(33)

XPS analysis The state of the strongly adsorbed platinum was

investigated by XPS analyses. A typical Pt4f spectrum is presented in figure 2. The spectra were not dependent on the duration of the exposure. So reduction caused by the radiation (10) may be excluded. Experimental bonding energies may be compared with the following literature values (8) (Pt4f 7/2): Pt° 71.0-71.3 eV; K , P tnc i . 72.8-73.4

IV

eV; ICPt Clg 74.1-74.3 eV. Deconvolution of the experimental spectrum (figure 2) resulted in two doublets, with Pt4f 7/2 at 72.2 eV (83%) and 74.2 eV (17%), respectively. The former, larger, doublet is to be attributed to a Pt complex with a slightly smaller bonding energy than K-Pt C1-. So, reduction of hexachloroplatinic acid indeed (4) occurs upon impregnating the activated carbon. It may be noted that poly-acetylene is known to reduce Pt in HpPtClg to Pt(II) (11).

The low intensity doublet with Pt4f 7/2 at 74.2 eV corresponds to a IV

2-Pt complex, probably 2-PtClg . Its intensity is increased by drying of the sample in air at 120°C. The presence of this small amount of platinum in the original oxidation state is believed to be caused by re-oxidation during the sample preparation (artefact), although its true presence cannot be excluded.

To obtain more insight into the mode of adsorption of the platinum complex on the carbon, the effects of carrier modifications and additives were studied.

Effects of carrier surface chemistry Since all carriers have

practically identical BET surface areas and pore volume distributions (table 1) (6), differences in adsorption behaviour on differently pretreated carriers must arise from the surface chemistry of the carbons. Figure 1 shows that the strong adsorption is decreased by oxidative carrier pretreatments. This is easily explained by a decreased reductive capacity of the oxidized carriers.

Effects of additives Figure 3 shows the effects of some ionic

additives on the adsorption process. The presence of hydrochloric acid decreases the strong adsorption, which would, at first sight, point to an electrostatic adsorption mechanism: protonation of the carrier, with

II

2-Pt Cl, counterions. However, whilst potassium chloride shows the same effect as hydrochloric acid, methanesulfonic acid (a strong acid with a very weakly coordinating anion) has no effect at all. So, an electrostatic mechanism can be excluded. Since chloride ions counteract

(34)

Pt ads.

yj i , , , ,

0 1 2 3 4

[Pt] (g/l)

Figure 3. Effects of additives O B hexachloroplatinic acid adsorption onto activated carbon ROX (water, 25 C ) .

additive (0.1 M ) : xnone; V HC1; • KC1; A methanesulfonic acid; O Y b ( N 03)3.

the strong adsorption of the platinum species, this process includes the loss of one or more chloride ligands. Indeed, it has been reported (12) that, upon heating hydrochloroplatinic acid impregnated carbon at 200°C in nitrogen, three HC1 were evolved (at lower temperatures the HC1 evolution was not completed within five hours, indicating a strong adsorption of HC1 on basic sites of the carbon (6)).

Addition of ytterbium nitrate also decreased the amount of strongly bound platinum.

Effects of acetone Figure 4 shows the results of adsorption and

desorption experiments in which acetone/water (90/10) was used as solvent. Acetone itself is strongly adsorbed onto the carrier. It thus competes with the strong adsorption of FLPtCl,, and practically excludes the latters weak adsorption. A carrier impregnated with aqueuos hexachloroplatinic acid can be partly depleted from its platinum load by extraction with acetone.

(35)

P t QdS (w %) 15H 1 0 5 -5 n 1 1 r— 10 15 20 25 total Pt (w°/o)

Figure 4. Hexachloropiatinic acid adsorption onto activated carbon ROX, and extraction with acetone (25 C ) .

adsorption: * from aqueous solution; O f r o m acetone solution. a - d: acetone extractions; Oadsorbed after extraction.

1. H2PtI VCl6 + ^C-H + H20

,2--> P tnC l ^ " + 2 Cl" + 4 H+ + \c-OH

PtCl^ + S > S-PtCl3 + Cl

Scheme I. Reductive adsorption of hydrochloroplatinic acid onto activated carbon.

Model A model rationalizing the chemistry of the impregnation

IV II process is presented in scheme I. The carbon reduces the Pt to Pt , which is then coordinatively bound by the carrier. A reducing site on the carrier surface is denoted as ^C-H and }C-OH is its (acidic) oxidation product. The formation of acidic groups during the oxidation of the carrier by Pt is deduced from titration experiments [13]. S denotes a ligand site for Pt . The nature of these sites, and the complexes derived therefrom, will now be discussed.

(36)

-37-Liqand site S Principally, two types of ligands for Pt1 1 are present

on the carrier: C=C structures in the carbon basal planes, and oxygen containing functional groups on the basal plane edges (6).

fl-complex structures The platinum complexes can be repelled from the carrier by organic solvents (3) like acetone. This would point to adsorption on the carbon basal planes (acetone itself is not a stronq

II

ligand for Pt (14)). Activated carbon can be seen as a very bulky C=C ligand. Complexation with bulky ligands is stereochemically unfavourable. However, platinum complexes can be described in terms of bonding and back-bonding: electron donation from a filled 7r orbital of the ligand to an empty orbital of the metal, and back-donation from the

*

metal to an empty i\ orbital of the ligand (15). Activated carbon has a

high valence band edge (-HOMO) and a low conduction band edge (-LUM0) (16), and therefore a very favourable electronic structure for coordination of Pt .

For two reasons, 7r-complex formation would take place preferentially on the edges of the carbon basal planes. Firstly, these sites are sterically favoured. On the analogy of cis-alkene Pt complexes, the platinum ion may bend away from the 'substituents' of the C=C ligand (15) (this is especially so since the basal planes in activated carbon are thought to be somewhat puckered). Secondly, theoretical calculations (17) indicate that the sites with the smallest electron localisation energy are found at the edges. The armchair edge features pairs of reactive carbon atoms. At the zigzag edge, very reactive sites are alternating with particularly unreactive C-atoms (this picture may, however, be altered by the presence of conjugated oxygen surface groups).

Recently, a similar K-complex formation has been suggested for the adsorption of copper chloride complexes onto activated carbon (18).

Complexes with ligands containing oxygen Addition of ytterbium nitrate (figure 3) decreases the strong adsorption. Hard ions such as Yb do not form fl-complexes. So, the effect of Yb is more easily explained when the strong adsorption of platinum is considered to take place on weak donor ligands containing oxygen atoms (denoted 0-ligands). Although not all very stabile, several complexes of Pt with O-ligands are known, e.g. acetylacetonates, carboxylates, etc (19). Many such ligand structures are present at the edges of the carbon basal

(37)

planes (6). More will be formed in the oxidation of the carbon during

IV II the H-Pt Clg adsorption process. So, formation of Pt 0-ligand

complexes seems possible.

Location of platinum complexes in the carrier Both types of

coordination (i.e. ff-complexation or voxygen' complexation) would take

place at the edges of the basal plane structures. Although the present data do not enable a choice between the two possible types of complexes, it can be concluded that the Pt ions formed during the impregnation are bound at the basal plane edges, i.e. at the walls of the meso pores.

CONCLUSIONS

Upon impregnation into activated carbon, hexachloroplatinic acid is (partly) converted into a Pt complex. This complex is strongly bound at the walls of the meso pores, either as 7r-complex and/or through ligands containing oxygen. Addition of hydrochloric acid decreases the interaction between the carrier and the noble metal compound, and should be avoided in the catalyst preparation process.

REFERENCES

1. Palmer Jr, M.B., Vannice, M.A., J. Chem. Tech. Biotechnol. 30. 205 (1980).

2. Richard, D., Gallezot, P., in "Preparation of Catalysts" (B. Delmon, P. Grange, P.A. Jacobs, G. Poncelet, Eds.) p 71, Elsevier, Amsterdam (1987).

3. Hanika, J., Machek, V., Nemec, V., Ruzicka, V., Kunz, J., J. Catal. 77, 248 (1982), and references therein.

4. Czaran, E., Finster, J., Schnabel, K.-H., Z. Anorg. Allg. Chem. 443,175 (1978).

5. Stokes, H.T., Makowka, C D . , Wang, P.-K., Rudaz, S.L., Slichter, C.P., J. Mol. Catal. 20, 321 (1983).

6. chapter 2.

7. Ayres, G.H., Meyer Jr, A.S., Anal. Chem. 23, 299 (1951).

8. "Handbook of X-ray Photoelectron Spectroscopy", Perkin Elmer Corp (1979); Briggs, D., Seah, M.P., "Practical Surface Analysis", John Wiley, New York (1983).

9. Brinen, J.S., Schmitt, J.L., J. Catal. 45, 274 (1976).

10. Katrib, A., J. Electron Spectrosc. Relat. Phenom. 18, 275 (1980). 11. Cao, Y., Guo, K., Oian, R., Huaxue Xuebao 43, 425 (1985), CA

(38)

12. Uhlir, M., Hanika, J., Sporka, K., Ruzicka, V., Coll. Czech. Chem. Commun. 42, 2791 (1977).

13. chapter 4.

14. Davis, J.A., Hartley, F.R., Chem. Rev. 81.» 79 (1981). 15. Hartley, F.R., Chem. Rev. 69, 799 (1969); 73, 163 (1973).

16. McMichael, B.D., Kmetko, E.A., Mrozowski, S., J. OPt. Soc. Am. 44, 26 (1954).

17. Stein, S.E., Brown, R.L., Carbon 23, 105 (1985).

18. Hirai, H., Wada, K., Komiyatna, M., Bull. Chem. Soc. Japan 60, 441 (1987).

19. Hartley, F.R., "The Chemistry of Platinum and Palladium", Applied Science Publ., London (1973).

(39)

4 Catalyst preparation III: reduction

INTRODUCTION

In the preparation of supported metal catalysts, the reduction of the impregnated carrier is probably the most delicate step. However, even for a common system like platinum-on-carbon, little is known of the actual chemistry of the reduction process. Pt/C catalysts are thus produced using "cookbook" methods based on empirical experience. Using these procedures the catalysts obtained are often far from optimal. Also, it is not possible to predict the effects of changes in the process variables.

The present study was aimed at (i) gaining insight into the chemistry of the reduction process, and (ii) using this insight to obtain catalysts with different properties. One of these (intrinsic) properties is the carrier surface chemistry, which can exert an important effect on the selectivity of Pt/C oxidation catalysts (1).

A reduction method often used in the preparation of Pt/C catalysts is treatment with hydrogen gas at elevated temperatures (e.g. 400 C) (2). At these temperatures, the surface groups of the carrier are not stable

(3). So, this procedure is not suited when the effects of the carbon surface chemistry on the properties of the catalyst are to be studied.

Alternatively, the reduction can be performed using a hydrogen donor (e.g. formaldehyde) in alkaline aqueous solutions (4). The effects of

(40)

this reduction method on the surface chemistry of the carrier are unknown. However, the reactions can be performed at ambient temperature, at which the carrier surface chemistry is at least thermally stabile. We have therefore selected this liquid phase reduction method for our studies.

The use of hydrogen 'donor molecules in alkaline solutions also provided the possibility to study the relevant reactions. During the hydroxyl ion consuming reduction processes, the pH was kept constant using a pH-stat. Plotting the alkali consumption versus time enabled a detailed study of the stoichiometry and the kinetics of the reduction reactions. The products formed from the hydrogen donor were analysed using 1 3C NMR.

Carbons The modification and characterisation of the carbon carriers

has been treated elsewhere (3). Some characteristic data are summarised in table 1.

Impregnates To the carbon (100 g) and water (1000 g) a calculated amount of concentrated hexachloroplatinic acid (Johnson Matthey) solution was added in 60 equal portions at 0.5 min intervals. Between the additions the mixture was agitated to ensure an equal loading of all carrier particles. The mixture was kept for four days at room temperature in the dark, and the excess solution was removed by filtration. The impregnates thus obtained contained -50% water. The (acidic) mother liquor was titrated with potassium hydroxide; the results are included in figures 1-4.

Hvdrothermal treatment The moist impregnates were kept at 95 C in

closed containers.

Reductions The reductions were performed in a 50 ml screw cap

vessel, equiped with a pH electrode, a reference electrode, a nitrogen gas inlet tube, a gas purge outlet, and a burette tip. The home made reference electrode was of the double junction type, enabling stable measurements in the strongly reducing reaction media. The pH was controlled automatically using a Metrohm 632 pH meter, a 614 impulsomat,

(41)

and a 655 dosimat containing 2.00 M potassium hydroxide. The vessel was agitated continuously.

In the reduction vessel (room temperature), the reductor solution (10.0 ml 0.3-10 M) was added to the wet impregnate (4 g, equivalent to 2.00 g final Pt/C catalyst). The system was flushed with nitrogen gas (1 min), and the reduction was started by raising the pH to the desired value. From then on, the pH was kept constant by the pH control system. After the reduction, the catalyst was thoroughly eluted with water (to pH 7 ) , dried (20°C, 50 mTorr, 24 h ) , and submitted to dispersion measurements.

The alkali consumption vs. time of the separate components (e.g. the carbon) and any combination thereof (as given in figures 1 and 2) was obtained analogeously. All displayed HO" consumptions include the amounts necessary to initiate the desired pH value.

Dispersion measurements were performed using a modified Quantasorb apparatus (Quantachrome Corp.). The samples were reduced in flowing hydrogen gas (20°C, 16 h ) . Then, the Pt metal dispersion was determined through the adsorption of carbon monoxide pulsed into the hydrogen stream. A blank experiment showed the absence of CO adsorption onto the carrier.

Table 1. Characteristic data on carbon carriers used.

carrier modification carrier negative charge method at pH 10 (mole/kg) R0X N5 N10 N20 N30 A D -nitric acid oxidation (5-30% acid) air oxidation 320°C helium 320°C 0.43 0.62 0.87 1.23 1.60 0.82 0.27

(42)

13

C NMR spectroscopy Nuclear magnetic resonance analysis of liquid samples were performed using a Nicolet NT 200 WB apparatus. The relaxation conditions chosen ensured a (semi) quantitative analysis of all reductor derived compounds.

10-cons. (meq) 8- 64 2 -t (h)

Figure 1. Alkali consumption versus time of basic catalyst preparation processes.

a: ROX (1.90 g ) ; b: ROX + methylene glycol (10 ml 0.1 M ) ; c: H2PtCl6/R0X (0.513 mmole Pt); c': idem, 40 h time offset;

(43)

RESULTS AND DISCUSSION Basic processes

Platinum absent Figures 1 and 2 (curves a and b) show the titration

curves at pH = 11 for the parent carbon carrier (Norit ROX 0.8), and the carbon plus the formaldehyde solution used as reductor. The acidic groups on the carbon surface naturally consume HO". This process is not

very fast, and (at least partly) controlled by the diffusion of hydrated potassium ions to those acidic groups that are located within the carbon micropores. (Hydroxyl ions and protons migrate very fast by the inter-conversion of covalent and hydrogen bonds.) In diluted aqueous media formaldehyde is principally present as methylene glycol (H2C(0H)2) (5),

denoted as MG (1 M solution: -90% MG, ~10%dimer). At pH 11, the reductor (MG pKa ~ 13 (5)) did not consume a significant amount of

hydroxyl ions; also the present conditions do not induce a noticeble homogeneous Cannizzaro reaction. The combination ROX + MG showed a slightly enhanced hydroxyl ion consumption, indicating a slow (cross) Cannizzaro type reaction involving the carbon (either as reactant or as catalyst).

0 30 6 0 9 0 120 t (min)

Figure 2. Alkali consumption versus time of basic catalyst preparation processes.

(44)

HJPtClg/ROX The titration of ROX impregnated with hexachloroplatinic

acid resulted in a HO" consumption which was ultimately 2.5 meq above the ROX titration value (figure lc). We have shown previously (6) that during the impregnation H2Pt(IV)Clg is reduced to a platinum (II)

complex by the carbon. From the present data (scheme I) it is concluded that, during this oxidative impregnation of the carrier, acidic groups are formed. Acidic groups are also formed using other oxidators (3).

H-PtClc/R0X + MG As demonstrated in figure 2 (curve d ) , the actual

catalyst preparation process proceeds autocatalytically. Initially, the curve follows the sum of the H-PtClg/ROX titration and the consumption of hydroxyl ions through the interaction of the reductor and the carrier: [H2PtCl6/R0X + MG] = [H2PtCl6/R0X] + [ROX + MG] - [ROX]. Then,

the actual platinum reduction becomes noticeable and proceeds at an increasing rate. This acceleration may be explained as follows. Initially, very small Pt(0) crystallites will be formed (see below). On these nucleation centers the reductor (MG) will adsorb dissociatively. This results in the formation of chemisorbed hydrogen atoms and (desorbed) formate ions (and possibly some carbonate, see below). The platinum(II) complexes are reversibly adsorbed by the carbon (6) and therefore mobile (through liquid or surface diffusion). When they reach the hydrogen covered nucleation centers, the complexes will be reduced to Pt(0). This leads to an increasing metal surface, and thus to an accelerating Pt(II) reduction rate. So, the platinum crystallites catalyse their own growth. (Similar effects occur in electrodeless plating and in photographic development.)

The reduction of platinum complexes to Pt(0) by hydrogen covered Pt crystallites is a known process (7).

Initial Pt(0) nuclei The formation of initial Pt(0) nuclei is a key

factor in any reduction mechanism leading to platinum crystallites. From the steepness of the fast autocatalytic growth (fig. 2 ) , it follows that (i) fast growth occurs after the crystallites reach a critical size, and, since disperse catalysts can be obtained, (ii) a large number of crystallites reach this critical size at the same time.

■The occurrence of a critical size logically follows from the absence of sub-surface hydrogen adsorption sites in crystallites smaller than six atoms. Assuming that at least two sub-surface sites are necessary to accomodate, in a reactive mode, the two hydrogen atoms supplied by the

(45)

dehydrogenation of the reductor, a critical crystallite size of ten atoms is deduced. Smaller crystallites can adsorb hydrogen only in a form comparable to the hydrogen on edge atoms of larger crystallites (which will result in a different reactivity).

PROCESS: HO CONSUMPTION (H0"/Pt) (meq)a Impregnation: H2Pt(IV)Cl6 + k - H + H20 ---> Pt(II)Cl^ + 2 Cl" + V o " + 5 H+ 5 2.57b Pt(II)Cl~ + S ---> PtCl3S" + Cl" -Reduction: PtCl3S" + H2C(OH)2 + 3 HO" ---> 3 1.54 Pt(0) + 3 Cl" + S + HCOO" + 3 H20 Over-all: H2Pt(IV)Cl6 + V - H + H2C(0H)2 + 8 HO" ---> 8 4.10c Pt(0) + 6 Cl" + ^C-O" + HC00" + 7 H20 Hydrogen transfer:

2 H2C(OH)2 + HO" ---> HCOO" + H-jCOH + 2 HgO

a: based on 0.513 mmol platinum; b: see figure 1; c: see figure 2.

(46)

Many crystallites reach their critical size at approximately the same time. The nucleation of the crystallites is thus also likely to take place at the same time, logically when the reduction proces is started. This implies that the nuclei are either already present in the impregnate, or that a fraction of the platinum(II) is in a reactive form, yielding nuclei as soon as the reduction is started.

The number of nuclei will normally be small compared to the total number of Pt atoms in the system (e.g. when the metal dispersion is 0.4 the crystallites contain approximately 1000 atoms), and well below the detection limits of e.g. XPS. Therefore only indirect data on the number of initial nuclei can be obtained.

HO0 cons. (meq) 2 4 6 8 1 0

-\ l %

\ 150 C N. Éd ^ I 150 I I I b ,0. I i 100 50 ppm b a "100 50 ppm I i 10 15 20 25 t(h) 13

Figure 3. Alkali consumption versus time and C NMR spectra of liquid phase samples: auto-redox reaction of formaldehyde.

H2PtCl6(0.513 mmole Pt)/ROX (1.90 g) + methylene glycol (10 ml 0.1 M ) .

(47)

t(h) H O

cone ( m e q )

7-t(h)

Figure 4. Alkali consumption versus time of catalyst preparation process: different reductors.

H2PtCl6(0.513 mmole Pt)/R0X (1.90 g) + reductor (10 ml 0.1 M ) .

reductor (pH): a: hydrogen (10); b: methylene glycol (10); c: formate (11); d: methanol (10) (upper time scale!).

Pt(0) catalyzed hydrogen transfer Figure 1 shows that after the reduction of the platinum complexes is completed, a consecutive reaction takes place. The experiment was repeated and samples of the liquid phase were analyzed by 1 3C NMR (figure 3 ) . The consecutive reaction is

identified as a (Pt(0) catalyzed) hydrogen transfer from one molecule of (hydrated) formaldehyde to another, yielding formate ions (HO cons.) and methanol. This autoredox process is included in scheme I. It strongly resembles the earlier reported hydrogen transfer from glucose to fructose (8).

Figure 3 also shows that in the actual Pt(II) reduction process methylene glycol is converted into formate ions. Only a small fraction of these ions is further oxidized to carbonate.

(48)

Other reductors The catalyst preparation was also performed using

other reducing species, i.e. methanol, potassium formate, and (dissolved) hydrogen. See figure 4. As expected, the autocatalytic effect was observed in all cases, whereas the hydrogen transfer reaction was specific for methylene glycol reductions.

When hydrogen gas was used as reducing species, the reduction rate was probably controlled by the supply of hydrogen. The low solubility of this gas in water is expected to result in a diffusion controlled reaction at the outer part of the carbon extrudate. This would lead to a scale type catalyst with a small metal dispersion (see below).

Carbon surface chemistry An interesting point is the effect of the

catalyst preparation on the carrier surface chemistry. It may be expected that a platinum crystallite is able to modify catalytically the carbon surface groups in its surroundings. The platinum is, after reduction, located (mainly) in the mesopores (9). The walls of these pores partly consist of carbon basal plane edges, i.e. carbon surface groups.

The titration data in figure 4 revealed a small excess hydroxyl ion consumption during the catalyst preparation. So, ionizable acidic groups are formed. Also, during the preparation of (0.5%) platinum on ROX, the oxygen content of the carrier (as determined by thermolysis-GC (3)) was found to decrease from 3.3 to 2.3 % by weight. The number of surface groups is thus decreased.

At the pH values used in the catalyst preparation, most of the acidic groups are ionized. These groups are expected to be stable towards reduction. Carbonyl groups however may be hydrogenated catalytically to hydroxyl groups (which will then (partly) be ionized in the alkaline medium (3)), or completely removed. More decisive conclusions would require an analytical tool which is able to focus on the carbon surface chemistry surrounding the platinum crystallites.

Metallic dispersion

Kev factors The dispersion (D) of Pt/C catalysts prepared by the

liquid phase reduction method will be governed by: (i) the number of crystallites reaching the critical size at the same time, and therefore the number of initial nuclei, (ii) the rate of transportation of Pt

(49)

complexes to the growing crystallites, and (iii) the rate of reduction of the complexes on the metal surface.

Rate limiting step When the Pt(II) reduction on the crystallite

surface is the rate limiting step (i.e. the Pt(II) diffusion is relatively fast), the growth rate of a crystallite will be proportional to its surface area. So, large crystallites will grow faster than small crystallites. This finally results in a small number of very large crystallites, i.e. a low metal dispersion.

When the Pt(II) diffusion to the crystallite surface is rate limiting (i.e. the Pt(II) reduction on the Pt(0) surface is relatively fast), all crystallites will grow at an equal rate. This will result in a high metal dispersion.

The autocatalytic effect (figure 2) occurs in all preparation reactions. In the beginning of the reduction process the actual reduction is rate limiting. Then, after the critical crystallite size is reached, the reaction rate on the growing the Pt(0) surface increases rapidly, and the amount of unreduced Pt(II) complexes decreases. So, at one point the diffusion of Pt(II) complexes becomes the limiting factor, and the over-all rate decreases again. An optimal dispersion will thus be obtained when (i) a high number of initial nuclei is present, (ii) the Pt(II) reduction on the crystallite surface is fast, and (iii) de Pt (11) complex diffusion is slow.

Table 2 shows the effects of variations in the reduction conditions and in the type of carbon carrier on the lag time before the fast growth phase, and on the metallic dispersion. This dispersion is expressed as the molar ratio of carbon monoxide adsorbed (as determined by CO adsorption in hydrogen gas at room temperature) and platinum present. The results can be discussed on the basis of the key factors mentioned above.

Reductor concentration The use of methylene glycol concentrations

higher than 1 M resulted in lower metal dispersions. A concentration of 1 M seems sufficient to ensure an optimal reduction rate, i.e. to keep the crystallite surface covered with hydrogen. Too high a reductor concentration may have two negative effects. The reductor is adsorbed by the carrier, competing with the Pt(II) adsorption (cf. the effect of acetone on the adsorption (6)) The reductor may also compete with the Pt(II) complexes in the adsorption on the crystallites, thus decreasing

(50)

Table 2. Effects of reduction conditions on metal dispersion and lag time before fast growth phase.

variable [reductor] (M) PH carrier reductor value 0.3 1.0 3.0 10.0 10 11 12 D ROX A N5 N10 N20 N30 HC00" H2C(OH)2° H3C0H H2 dispersion (C0/Pt) 0.38 0.40 0.29 0.20 0.40 0.40 0.25 0.40 0.40 0.36 0.27 0.26 0.21 0.10 0.35 0.40 0.13 0.14 lag time (min) 180 90 55 35 90 50 15 60 90 50 95 80 80 70 50 90 250 c

a: 5% Pt load in all cases. Reference system: carbon ROX, 1M methyl ene glycol, pH 10. b: pH 11. c: see text.

the Pt(II) reduction rate (Langmuir-Hinshelwood). So, a higher MG concentration leads to an enhanced Pt(II) mobility and perhaps a decreased Pt(II) reduction rate , both leading to a lower dispersion.

EÜ Increasing the pH leads to a higher carrier surface negative charge (3). The Pt(II) surface complexes are also negatively charged. So, increasing the pH will lead to an increased electrostatic repulsion between the carrier and the Pt(11) species. This results in a higher mobility of the complexes and therefore a lower metal dispersion.

(51)

Carrier surface chemistry The carrier surface negative charge is

also dependent on the carbon pretreatment (table 1) (3). Again the expected decrease in the dispersion with increased surface negative charge is observed. However, the surface chemistry may also affect the number of initial nuclei. The stronger oxidized carriers could produce a smaller number of (zerovalent) platinum nuclei, which would also lead to the observed effect.

Reductor (figure 4) Methylene glycol and formate show comparable

reactivities and yield similar dispersions. Methanol is much more difficult to dehydrogenate and therefore less reactive. Also it is less polar and therefore better adsorbed by the carrier. Methanol as reductor thus yields bad metal dispersions. When using hydrogen, the reduction also proceeded slowly and largely as a zero order process, probably because of the low solubility of hydrogen in the liquid phase. It is likely that only crystallites in the outer shells of the catalyst extrudate particles benefit from the reduction. The resulting dispersion is very low indeed.

Table 3. Effect of hydrothermal treatment of impregnate on dispersion of catalyst. temperature (°C) c 65 95 » i -95 i > duration (h) -16 43 3.5 22 65 164 -20 70 93 reductor HC00" ) Ï > » H2C(0H)2 ) 5 dispersion (CO/Pt) 0.35 0.34 0.43 0.40 0.49 0.54 0.60 0.54 0.40 0.38 0.52 0.58 0.57 a: 5% Pt on carbon ROX, 1 M methylene glycol, pH 10.

Carbon supported noble metal catalysts in the oxidation of glucose 1-phosphate and related alcohols

Carbon supported noble metal catalysts

in the oxidation of glucose 1-phosphate

and related alcohols

(gasphase)

(bulk)

O

TR diss

Carbon supported noble metal catalysts

in the oxidation of glucose 1-phosphate

and related alcohols

PROEFSCHRIFT

Contents

1

Introduction

/

-c^v

o

2

Catalyst preparation I: carrier

o ^

o

O I

_ _ _ I

VyVo

YWO

X

.o

10-S o

of

5-°

3

Catalyst preparation II: impregnation

4

Catalyst preparation III: reduction

-\ l %