Platinum Metals Rev., 1966, 10, (4), 122
Evaluation of Supported Catalysts for Liquid Phase Reactions
Effect of Experimental Variables on Rates of Reaction
Sufficient information is now available to lead the intending user of a supported noble metal catalyst to that combination of metal and support most likely to catalyse the reaction he desires to perform. However, each combination can be prepared by a number of different techniques, and the resulting catalysts need to be evaluated since their performance in the desired reaction will depend on a number of factors that are still only poorly understood. The purpose of this article is to describe some of the precautions which need to be observed in evaluating catalysts for liquid-phase reactions.
As a result of some six decades of experimental work on metal-catalysed liquid-phase reactions, there is now a vast body of information available to guide the intending user of supported noble metal catalysts to that combination of metal and support most likely to perform the reaction he is interested in (1). The recent publication of two books (2, 3) devoted solely to this field demonstrates the extent of current interest in the technique. Unfortunately, however, even if the literature or experience shows the most favourable metal-support combination there are still a number of methods by which this catalyst can be prepared, a number of grades of the support may be available, and the concentration of the metal on the support may be varied within wide limits. The state of our knowledge is not generally such as to enable us to predict in advance the effect of these variables, and so the intending user is necessarily faced with the task of comparing at least a small number of catalysts in order to find that most suited to his particular purpose. In this process of evaluation there are pitfalls for the unwary, some of which are not widely appreciated. In this article we illustrate the problems by means of our own experience with the liquid-phase hydrogenation of nitrobenzene.
Hydrogenation of Nitrobenzene
The hydrogenation of a nitro group attached to an aromatic ring is one of the most easily catalysed liquid-phase reactions. Nitrobenzene for example is readily hydrogenated to aniline and water at room temperature and atmospheric pressure, when methanol is used as the solvent and supported palladium or platinum as the catalyst (1). Many of the peculiarities of this reaction are shared by other catalysed liquid-phase reactions and our results may therefore be used in a general way to show the effect of variables on rates of reaction.
There are three phases present in a liquid-phase hydrogenation: (i) hydrogen in the gas phase, (ii) the reactant and solvent in the liquid phase, and (iii) the catalyst in the solid phase. It is well established that the overall reaction sequence may be broken down into the following steps:
Diffusion of the reactants to the catalytic surface
Their adsorption, reaction and desorption
Diffusion of products away from the surface.
We have been able to establish in another system that the rate of diffusion of the liquid-phase reactant to the surface and of the liquid-phase products away from the surface exceeds the rate at which hydrogen reaches the surface (4). This is presumably because the hydrogen needs to dissolve in the liquid phase and its solubility in liquids is generally low. Thus, if any diffusion limitation of the reaction occurs, it will be the hydrogen that is responsible. If we are to make full use of the potentialities of the catalyst, that is, if we are to ensure that either adsorption, reaction or desorption is rate-limiting, then we have to ensure that hydrogen is supplied to the surface faster than it is consumed. Now the rate of solution of hydrogen will be proportional to the surface area of liquid exposed to the gaseous hydrogen. We can therefore overcome hydrogen diffusion limitation by vigorously agitating the system to provide a greater area of gas-liquid contact.
The degree of agitation required to produce a diffusion-free system will of course depend on the catalytic rate, which in a given system is proportional to the activity of the catalyst and the weight of catalyst taken. The faster the catalytic rate, the more vigorous the agitation required to overcome diffusion limitation, and vice versa; this is shown schematically in Fig. 2. This figure also makes it clear that the relative activities of different weights of a catalyst, or equally of the same weights of different catalysts, will not be correctly given unless all the rates are diffusion-free.
Ideally therefore catalysts should be evaluated and used under conditions such that only the catalytic reaction is rate-limiting. The two criteria for this are:
The rate should be proportional to the catalyst weight, and
The rate should be independent of the degree of agitation.
It is impossible to over-emphasise the importance of ensuring such conditions before other variables are examined: if this is not done, one runs the strong risk of studying only their effect on the rate of solution of hydrogen.
There are several ways of achieving diffusion-free conditions on a small scale in the laboratory. The most convenient method is to shake the solution of reactant with the catalyst in suspension vigorously in a small flask in the presence of hydrogen. In our experience, diffusion-free rates of up to 400 ml.min.−1 can readily be attained if 20 ml of the reactant solution contained in an 80 ml round-bottomed flask is vigorously shaken in a Griffin or Baird and Tatlock laboratory shaker. Vibration frequencies of about 1200 min.−1 are easily achieved and are adequate to ensure diffusion-free conditions for most reactions. Accurate control of the frequency is unnecessary if it is well in excess of the critical minimum value. Agitation by shaking is of course impossible with large reactors and stirring or some other procedure has to be used instead.
It is important to distinguish between diffusion limitation due to hydrogen deficiency (which can be overcome by agitation) and that due to the slow diffusion of dissolved molecules within the pore structure of porous catalyst particles (which is not affected by agitation). This latter effect is a largely unknown quantity in liquid-phase reactions, although its importance is often suspected, and is usually regarded as one of the effects which determines the activity of the catalyst.
In establishing the optimum conditions for performing a catalysed liquid-phase reaction, the most important variables to consider are catalyst weight, temperature, hydrogen pressure, nature of the solvent and reactant concentration. Increasing any of the first three of these variables is expected to increase the rate of the desired reaction (although there may be some surprises), but each procedure is in some measure expensive and may also lead to decreased selectivity or increased yields of by-products. It is therefore by choosing the best solvent and reactant concentration that improved performance may most readily be attained.
However, before considering these points in greater detail, the effect of increasing temperature needs to be discussed a little further. This effect is often complex, as is demonstrated by the Arrhenius plot (Fig. 3) of our results for nitrobenzene hydrogenation at a palladium/charcoal catalyst. The filled points are the observed rates: a maximum rate is observed at about 45°C, below which temperature the apparent activation energy is 4.0 kcal.mole−1. We interpret these results by saying that as the temperature increases the vapour pressure of the solvent methanol also increases, hence progressively reducing the partial pressure of hydrogen. We may then correct the observed rates on the assumption that the rate should be proportional to the hydrogen pressure: the open points are the corrected rates. The maximum rate is now at about 49°C, and the activation energy is increased to 6.3 kcal.mole−1. The rate at the highest temperature (55°C) is still, however, less than expected. We believe this is because the solubility of hydrogen in the liquid phase will decrease with increasing temperature, and at 55°C the concentration of dissolved hydrogen is insufficient to maintain diffusion-free conditions.
These results emphasise the importance of adequate temperature control when evaluating catalysts, and also the need to examine the effect of temperature before deciding on optimum operating conditions.
The choice of the solvent in which the reaction is to be performed is often important: it can influence not only the rate of the reaction (for reasons which are poorly understood) but also the manner in which other variables affect the rate. For example, the solubility of hydrogen in ethanol is less than that in methanol; consequently under diffusion-limited conditions rates are slower in ethanol and more efficient agitation is required to obtain a diffusion-free rate (see Fig. 4).
Having fixed upon a suitable solvent, the next task is to decide on the best reactant concentration. A number of factors will affect this decision, but we are only concerned with the way in which the rate is affected by changing reactant concentration. It is usually found that the rate decreases with increasing reactant concentration, as is the case with our results on nitrobenzene hydrogenation shown in Fig. 5. Although such a decrease in rate could be caused by toxic impurities in the reactant, such is not the case with our results. The reasons for the effect are likely to be complex.
Most organic compounds contain traces of impurities which may poison supported metal catalysts. The most convenient way of determining whether toxic impurities are present in the reactant or not is to examine the effect of catalyst weight on the rate. If no poisons are present, the points will lie on a straight line through the origin. If a poison is present, the line will intersect the catalyst weight axis at a point corresponding to the weight of catalyst inactivated by the poison present. If it is necessary in larger scale operation to use a reactant which contains some poison, then it is advisable to use a higher than normal catalyst loading.
We have found that there are two impurities present in commercial nitrobenzene: one is dinitrobenzene and the other is an unidentified condensation product which is produced photochemically. Both are, however, easily removed on double vacuum distillation, although the latter is quickly formed again unless the nitrobenzene is dissolved in methanol. Nitrobenzene so purified was used for the experiments described in this article.
Poisoning by products of the reaction leads to a progressive decrease in the rate as the reaction proceeds, but since this may result simply from the decreasing reactant concentration it is necessary to be able to distinguish between the two causes. This is best done by comparing conversion-time plots of two experiments in which different catalyst weights are used. If the ratio initial rate/rate at 25 per cent conversion (or any other suitable greater conversion) is greater in the experiment in which the lesser amount of catalyst was used, then product poisoning is strongly indicated: if the ratio is independent of catalyst weight then product poisoning is probably absent. In extreme cases poisoning by a product may cause the reaction to stop before completion: the hydrogenation of nitrobenzene is, however, only weakly retarded by the water formed.
Supported noble metal catalysts are available with a range of metal concentrations (1). The effect of varying metal concentration on the rate of nitrobenzene hydrogenation over palladium/charcoal and platinum/charcoal catalysts is shown in Fig. 6. At low metal concentrations the rate is proportional to the metal concentration, but above a certain critical value (5 per cent for palladium and 10 per cent for platinum) additional metal is progressively less fully utilised. It is clearly important to work in the region where the metal is being efficiently used, and since these critical concentrations may vary from one reaction to another, the rate dependence on metal concentration needs to be examined for each new reaction studied.
The effect of these experimental variables on the rate of the hydrogenation of nitrobenzene is typical of catalysed reactions carried out in the liquid phase. However, in this reaction only one final product is possible: the problem of catalyst evaluation becomes more severe when, as in the hydrogenation of alkynes or unsaturated aldehydes, more than one product is possible and the catalyst must be chosen for its selectivity as well as its activity. These additional problems will be dealt with in a later article.
We gratefully acknowledge the experiment assistance of other members of the Catalyst Research Group, especially A. J. Bird and W. G. Gunn.
- 1 G. C. Bond and E. J. Sercombe, Platinum Metals Rev., 1965, 9, 74 ; “Platinum Metal Catalysts”, Johnson Matthey publication 7681/2, 1965
- 2 D. V. Sokol’skii, “Hydrogenation in Solutions”, Oldbourne Press, London, 1964
- 3 R. L. Augustine, “Catalytic Hydrogenation”, Edward Arnold Ltd, London, 1965
- 4 G. C. Bond and J. S. Rank, Proc. 3rd Internat. Cong. Catalysis, North Holland Publishing Company, Amsterdam 1965, p. 1225 ; see also F. Nagy,, Á. Pethö and D. Móger, J. Catalysis, 1966, 5, 348