Third Report of the Committee on Contact Catalysis (Published in the Journal of Physical Chemistry): Reprint and Circular Series of the National Research Council (1924)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS1 BY HUGH S. TAYLOR The following report aims to present a summary of recent investigations which seem to advance our understanding of the nature of contact catalysis, its mechanism and general technique. It is not an exhaustive summary of the catalytic researches carried out in the preceding year. The grouping of the researches has beer made with regard to their bearing on specific points of interest which it is desired to emphasize, not necessarily on a similarity of reaction. Wall Reactions The scope of the subject under investigation is extending continuously. Considerable attention is now being paid, riot only to catalysts deliberately inserted into a reaction system, but also to the acceleration of reactions by the walls of the containing vessel. The work of Hinshelwood, Hartley, and co- workers, mentioned in the previous report with respect to the decomposition of formic acid on glass surfaces and on silver and platinum has been extended to include different glasses and other metals. The type of reaction has been varied to include other typical decompositions of gases including hydrogen peroxide, chlorine monoxide, sulphuryl chloride, and phosphine. One outstanding conclusion from this world is to emphasise anew that true unimolecular reactions are, as yet, conspicuous by their absence. All of the known cases which have been presupposed unimolecular, have proved to be either wall reactions or reactions occurring on collision. It is therefore evident that we do not yet need, for any known reaction, the concept of radiation to give an explanation of the occurrence of unimolecular reactions. Bimolecular reactions between gases, likewise, are being shown to be wall reactions. The two notable cases investigated recently are the combination of ethylene and bromine and the combination of nitric oxide and oxygen, reactions hitherto generally regarded as gaseous reactions, now shown to be tremendously sensitive to the nature of the vessel in which they are contained. It must also be borne in mind that, even when wall effects are demonstrably small, the gas reaction may still be a catalysed reaction. The effect of water vapour on the combination of hydrogen and oxygen, and of carbon monoxide and oxygen shows that these reactions are not simple collision reactions but that water is a contact catalyst, molecularly dispersed, or, if you will, forms intermediate compounds. The insensitivity of dried gases to reaction seems to demonstrate that it is not the energy of collision which brings about inter ~ Report of the Committee on Contact Catalysis of the Division of Chemistry and Chemical Technology of the National Research Council. Written by Hugh S. Taylor assisted bv the other members of the Committee: Messrs. H. Adkins, W. C. Bray, 0. W. Brown, R. F. Chambers, C. G. Fink, J. C. Frazer, E. E. Reid, and W. D. Bancroft, Chair- man.

898 HE-GH S. TAYLOR action, since small quantities of water vapor cannot alter this magnitude. The water vapor molecules, highly polar, must behave in the same manner as do contact catalysts in activating the several molecular species. This point has been raised in a slightly different form by Hinshelwood. It is also dis- cussed by Norrish in some extracts given below. E;nshelwood and Prichardt have studied the thermal decomposition of hydrogen peroxide, and of sulphuryl chloride in the gaseous state, and the thermal decomposition of chlorine monoxide. Diazoacetic ester was also examined, but found to be unsuitable for quantitative measurements, as tarry deposits were formed. The decomposition of chlorine monoxide proved to be homogeneous, whilst the hydrogen peroxide reaction and the sulphuryl chloride reaction were found to be typical heterogeneous reactions. Concerning such heterogeneous reactions the authors write: "When a molecule is adsorbed by a surface, the forces between it and the molecules constituting the surface modify the internal forces in a way which is at present quite incalculable and must be entirely specific. Generally speaking, it must be expected that the stability would be increased as often as it is decreased. Yet, the accumulation of observations showing that almost any gas reaction takes place more readily on a given surface such as glass than in the homo- gerleous phase, raises the question whether the operation of some general cause is not superimposed on the various specific influences. In the case of combina- tions in which two or more molecules are involved, the encounter of two types is obviously facilitated by the more or less prolonged sojourn of one of them on the surface, but this factor is inoperative in the case of the simple unimolec- ular decompositions. It seems relevant, therefore, to ask whether one univer- sa1 factor nary not be simply the second law of motion. Consider a molecule composed of two parts, A and B. the disruption of which constitutes the de- composition of the molecule. Let B receive an impact from another molecule which imparts to it momentum directed away from A. The small inertia of A, however, enables it to follow B. without the development of much strain between the two. If, however, A were firmly enough held to a surface, its inertia might be so great that the accelerating force, instead of drawing A after B. would cause the disruption of the "bond" between them. The re- luctance of homogeneous gas reactions to proceed might thus be due to the small inertia of the different parts of the molecules rendering disruption by collision very improbable. This is only suggested as one of several possibili- ties. That it is a mechanical picture, whilst we now believe "activation" to consist in the passage of an electron to an orbit of higher quantum number, is not a relevant criticism, since the results of work on the collision of electrons with gas molecules show that a definite correlation exists between quasi- rnechanical and quantum processes. "The thermal decomposition of chlorine monoxide proved to be a homo- geneous reaction uninfluenced by the glass walls of the containing vessel. The velocity of reaction increases as the change proceeds. This is not due to autocatalysis, since oxygen and chlorine have no influence on the rate of ~ J. Chem. Soc., 128, 2725, 27~0 (~923).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 899 decomposition, but is attributed to the occurrence of the change in consecutive stages. The rate of reaction at ~3~.3O is inversely proportional to the initial pressure of the chlorine monoxide, and the influence of pressure appears to operate uniformly throughout the course of the change at this temperature. Hence the decomposition depends on a collision effect and is not a spontaneous unimolecular process. "The influence of temperature on the reaction is such that the time re- quired for the change to proceed from 4o per cent, to 80 per cent. is increased 2.03 times for every ~o° decrease in temperature between ~3~.3O and ~o.7O. From the influence of temperature and from the heat of reaction it is shown that explosion waves should be readily propagated in the gas." Norrishi seeks to formulate the Arrhenius concept of active and passive molecules in terms of catalytic activation by either homogeneous or hetero- geneous catalysts. Among the former, water vapor is important in gas reac- tions. The walls of the containing vessel are included in the latter. "Even the most reactive substances become inert upon complete desicca- tion, and will then regain completely their lost activity by the addition of a trace of some polar substance. In other words, all chemical reactions appear to be catalytic in nature. Except in the case of a few truly thermal decompo- sitions of solids or liquids such as potassium chlorate, silver oxide, and lead acetate, the formation of ozone by the electric discharge, and possibly some unimolecular photochemical decompositions, this loss of reactivity oil desicca- tion would appear to be a general rule of chemical reactivity, and, if accepted as such, it necessitates a revision of our views of activation; the resting form of a molecule must be a far more inert substance then hitherto supposed, and require the association of some polar molecule before activation can take place. When we remember that the main characteristic of a polar molecule is its strong unbalanced field of force, it appears very probable that its function as a catalyst is to weaken, by close association, the intramolecular forces of the resting molecule, and to render it more easily disintegrable. "~i e may thus consider those molecules which have formed a close associa- tion with molecules of the catalyst to be, at any rate, partly activated, inas- much as they alone are capable of any further chemical action. Whether this is the complete stage or only a preliminary stage of activation it is not proposed to consider here, but there would appear to be no difficulty in the explanation of all the phenomena of chemical reactivity by the kinetic theory coupled with this view of activation alone, and without recourse to other hypotheses, as, for example, the "radiation theory." "The catalytic effect of traces of polar substances on gaseous reactions Is only one manifestation of a much more general phenomenon, and it may be said, that whenever any strong, local, disturbing force can be applied to a molecule, so as to distort its stable configuration, that molecule becomes more vulnerable to attack. Thus, the very numerous class of reactions which take place in solution probably owe their existence to the action of the solvent, which exerts a weakening effect on the internal molecular forces of the solute, ~ J. Chem. Soc., 123, 3006 (~923).

900 HUGH S. TAYLOR that may result, in extreme cases, in complete ionization. Again, the phenom- ena of surface catalysis, and surface reactions, are manifestations of the same nature, and owe their existence to the high electrical fields of force which must. exist unbalanced at the surfaces of most solids and liquids, and result in the adsorption and weakening of the structure of the reactant molecules. "We may thus regard molecular activation as occasioned by a definite change of configuration or distortion of the molPcillm.. hrnl~ht n.holit her ~1^C:P. ~· . . ~ ~-^~-EVA.-~ ~ MU 1_~t~U I W1~O Such a change of configuration mint association with some polar catalyst. take place with the absorption of ~n~r~v And there the ~tilrOtar1 ~^l^^lil^ ~, will be in a more highly energised state than the resting molecules. 1 ~ - 7 7 v A ~ ~i,, I& `_, ~ V V ~V MU A ~ 'AL '1 ~ ~ 1 ~ ~ AL 1 ~ ~ . "These views are in harmony with those developed by Lowry in his work on the electronic theory of valency. In a comparative study of the reactions of unsaturated organic compounds, he has drawn the conclusion that sub stances containing the double bond usually react as if one of the bonds were a covalence and the other an electrovalence. On this basis, the formation of ethylene dichloride from ethylene and chlorine involves an unsymmetrical instead of a symmetrical process of activation thus: CH2=CH2 and C1-C1 + _ ~ _ give CH2-CH2 and C1 C1 as an intermediate stage, rather than-CH2-CH2- and 2C1 - . The chlorine is here represented as teeing hrokPn int.n l.wo ions in ~^ ~f ~ ~ ~__1_~_ 1 _ ~ _ , 1 ~ · . ~. ~- v ~v ~L^ ~LO. Amp al two pleural atoms, in Ine c.l~rllntlon which millet nrPr~P~P or are_ pany its attachment to the ethylene. At. ~At 'L ~.41 ~ _ _ ~1 _r ~ 1 1 1 ~ - -- -- 1- -- - -_ . . ___ A__ ~ _~ v if, ~v~ ~^ t_~V111 The unsymmetrical rupture or opening All ~1 ~ u~Uu~e oo~(1 ~ one Anyone gives rise to an analogous process of intramolecular ionization. since the two (?hn.r~P`] n.t.~m~ ~ rid rod+ Arms L'`+ ~ ~ ~__~ V-At_ tJ YV ~ w~Ler_l~j4~ ~ U~111O ~1~ 11~0 11~ BUM bound. The final interaction between the two aCtiVat,P,~ PIP is t.h~n up duced to a mere nPl]t,rn.li~nt,ion of ~nn~c~ii" inns + _ Ag with C1 mainly ire that the ions yield covalent bonds on neutralization in-- stead of undissociated ionic pairs. The analogy between the development of an electrovalence on the one hand and the process of activation on the other is so complete as to suggest that the two phenomena are identical. "The view set forth. above, that molecular activation is a catalytic process of a polar character, is susceptible of direct experimental testing in the case of the gaseous reaction of ethylene and bromine, which has been investigated by Stewart and Edlund.i These two authors have shown that (~) ethylene and bromine at 0°, when dry, do not react together in the gaseous phase, but only on the glass walls of the container, and (2) there is no indication of a pre-- liminary gaseous reaction such as might be expected if a few of the ethylene and bromine molecules were already activated in the gaseous phase. "So far, these experiments are completely in accord with the hypothesis that activation of the ethylene molecule is due to polarization induced ir1 the ethylene molecule by association with some polar catalyst; but they are also capable of being explained on a merely physical basis, for example, by adsorp- tion of the two gases on the surface of the glass, without reference to the chem-- cal character of that surface. _ ~ ] ~,t ~L! ` ~J11FJ . It differs from the union of 1 J. Am. Chem. Soc., 45, IOI4 (I923).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 90I "It is, however, evident that if dry ethylene and bromine could be en- closed in a vessel with completely non-polar walls, it might be possible, if the above hypothesis of polar activation is correct, to retard the reaction very greatly, or even to suppress it altogether, although it is by no means certain that in all cases such a suppression could be looked for. "The results which are contained in the experimental section of the paper must be taken. as strong confirmatory evidence in favor of this hypothesis. It, has, for instance, been found that on enclosing the dry gases by a vessel the interior of which is coated with stearie acid, the reaction proceeds ever more quickly than when the glass walls are pare, whilst, when paraffin wax is substituted for stearic acid, the reaction Practically ceases to take place. Now the work of Hardy,1 Harkins2 and Langmnir3 has led us to regard the former of these two substances as a particularly polar molecule, whilst the paraffins constitute probably the best approach to a completely non-polar substance. Thlls, ir1 spite of their great physical similarity, a stearic acid sur- face brings about the combination of bromine and ethylene, whilst a paraffin wax surface does not, and this difference in their behavior can only be at- tributed to difference of polarity ire the surfaces of the two substances oeca- sioned by the marked chemical differences between their molecules. "The importance of the experimental results recorded in this paper is considered to lie in the fact that they provide evidence of a new character in favour of the theory that molecular activation is not only of a catalytic character, but consists in an induced polarization of the reactant molecules by association with some polar catalyst, either in the gaseous, surface, or liquid phase. They also confirm Lowry's deduction that molecules of un- saturated compounds may exist both in a non-polar "resting form" and in a polar reactive form, and further show that the conversion of the former into the latter may be brought about by a polar catalyst. This phenomenon is probably purely electrical in character, consisting simply in the production of an electrovalence from a covalence by a displacement of one of the electrons constituting the double bond, under the action of the electrical field of the catalyst. Rideal writes that "the rate of the ethylene-bromine reaction can be used to test the polarity of certain varnishes used in the industries and that the results obtained parallel the results of surface tension measurements." It is quite possible that phosphorus bichloride and chlorine would display little tendency to react if a suitable container for the two gases were found. In such a container it would then be possible to determine the density of phos- phorus pentachloride without dissociation occurring. The classical example of a bimolecular gas reaction, the hydrogen-iodine eon~bination studied by Bodenstein, cannot be entirely free from the suspicion of catalytic influences. Calculations which purport to express the velocity with which such a reaction occurs, in terms of collision frequencies and critical energy increments would ~ Fourth Brit. Asso. Rep. on Colloid Chemistry, ~ a:, (~9~); also Proc. Roy Soc. 86A, 6Io (1922). 2 J. Am. Chem. Soc. 39, 354, 54I (I9I7-) 3 Met. Chem. Eng., is, 468 (I9I6); J. Am. Chem. Soc., 39, 1848 (I9I7).

902 HUGH S. TAYLOR necessarily need revision in case such catalytic influences were found. That the critical energy increment which ~ reaction requires is determined largely by such catalytic influences is very evident from recent work on the photo- chemical combination of hydrogen and chlorine. Trammi has shown that visible light will not cause the combination of hvdrop;en and chlorine when the gases are thoroughly dried. light, can be combined with the aid of the larger energy quanta available with ultra-violet light; Coehn and Jung2 have shown that thoroughly dried hydro- gen and chlorine will combine when exposed to light whose wave-lengths are less than Auto A. Hinshelwood and Hartley3 have continued their work on formic acid decomposition at glass surfaces. Duroglass gave a much higher percentage of carbon dioxide and hydrogen as compared with carbon monoxide and water, than the earlier glass used. Nevertheless it was shown that the temperature coefficient of the carbon dioxide reaction is again with Duroglass markedly The authors calculate the re ~<~7 But. gaseous mixtures. insensitive to Edible higher than the carbon monoxide reaction. spective "heats of activation"4 as ECo= Door cars., and tco2= ~4500 cars., as compared with 16000 and 28000 calories respectively in the earlier work. Carbon monoxide is shown to have no retarding influence on the progress of the reaction at glass surfaces. .. . . .. hi. . Water vapor apparently accelerates the carbon axe reaction. ~ nets may account in part for an observed increase of carbon dioxide percentage with progress of the reaction, a fact originally due to Ber- thelot. To the reviewer this action of water vapor seems to indicate that the glass surface contains centres of activity which promote either the carbon monoxide reaction or the carbon dioxide reaction, belt; not both. Hinshelwood and Topley5 have extended the measurements of Tingey and Hinshelwood on the temperature coefficient of formic acid decomposition. To glass, platinum and silver as catalysts, rhodium, gold and palladium for the carbon dioxide decomposition and titanium dioxide for the carbon monox- ide decomposition have been added. In this latter case, which yields almost exclusively carbon monoxide, the value of ECo is igloo eats., whereas, with glass, Ec`' is i~ooo-16000 cars. A low value for the energy of activation of formic acid to yield carbon monoxide and water is not an inherent property of the formic acid molecule, but is determined in part by the surface accelerat- ing the change. This is further evidence of the composite nature of the temperature coefficient of heterogeneous reaction velocities as emphasized by Pease. (See later section). For the carbon dioxide reaction the values of ECO2 vary between moo and Door for glass, gold, silver, platinum and rhodium. There is no relation between these values and that of surface activity, which increases in the given series from o.os to soo in the order named, platinum being set equal to loo. ~ Z. physik. Chem., 105, 3;,6 (~923). 2 Ber. 56 696 (~923) 3 J. Them. Soc., 123, ~333 (~923). 4 Calculated from the temperature coefficient of reaction velocity by means of the equation d log k/dt=E/RT0. 5 J. Chem. Soc., 123, bond (~923).

THIRD R1DPORT OF THE COMMITTEE ON CONTACT CATALYSIS 903 Palladium is abnormal sho~-ir~g wide variations in activity and temperature coefficient according as it is free frown or contains occluded hydrogen. The latter decreases markedly the activity. Assuming the Langmuir mono- moleeular layer theory and no differences in 'phase' of the formic acid molecule or of modes of adsorption of the molecule, the authors calculated the fractions of the several surfaces covered at coo° and ~ atmospheres pressure. The fractions vary from lo- for silver and rhodium to 5 X ~o-6 for glass. Evidence will be given later to show that this may represent that portion of the surface which is capable of eatalysing the change under discussion. Catalytic Hydrogenation On the basis of experimental Fork by Cantels, Boswelli has sought to interpret the mechanism of catalytic hydrogenation by nickel, taking aeeount of the role played by oxygen in such hydrogenations as first emphasized by ~Tillstatter. The experimental data led Boswell to the following concept of the mechanism. . "Nickel oxide partially reduced at a low temperature consists of particles of nickel oxide surrounded by n~e'callic nickel carrying positive hydrogens and negative hydroxyls alternately arranged on the surface in several layers; thus, with only one layer of hydrogen and hydroxyls represented iNi~ I 11 | - Ni ~ Jar H+ OH H+ -- , OH ~hen this complex catalyses the union of hydrogen and ethylene four reactions oeeur: A very fast reaction Ni "H+ f 11 ~ . Ni 'OH 0 .! ~ H+ | OH A very slow reaction Nits H+ 11 ~ . Ni OH O ,/x HE ~ ~OH _ 3 A very slow reaction Knin KOHL I 11 ~ . Ni OH O ~ x H+ I H ~Ni ~-:E I 11 1 . Ni ,/Ni~` +, = I 11 . Ni +H' ~COW x OH + C2H6+H20 HE OH H~ +H20 | H Ni ~ H+ 11 | . Ni OH- +H20 YO x-I H+  | OH "Reaetion (I) represents the main reaction which occurs. It expresses the mechanism of hydrogenation by an active nickel eatalyser. 1 Proc. Roy. Soc., Canada, 16, Series III (IBM).

~o4 HUGH S. TAYLOR "Reaction (~) represents the slow removal of negative hydroxyls from the surface of the catalyses and the adsorption of hydrogen constantly taking place. "Reaction (3) represents the slow reaction of this adsorbed hydrogen with the unchanged nickel oxide in the interior of the particles. "A fourth reaction also occurs, involving the addition of positive and negative hydrogens from neutral hv,jro~n mol~(?lll~ the rho r`^rnr~l=- ^- the right hand side or reaction Ail, to form the complex on the left hand side of reaction (3~. This fourth reaction represents the mechanism of hydrogen adsorption. ~· , ~. ~. ~ ~. ~vat I ~1~ JO A ~11 "Equations (~) and (3) also represent the reactions which occur on con- tinued reduction of nickel oxide by hydrogen. This continues until all the nickel oxide in the interior of the particles has been reduced and until finally all the hydroxyls on the surface have been removed and only adsorbed hydro- gen, as positive hydrogens and negative hydrogens, remains. Thus the hydro- gen which is taken up in excess of the equivalent of water formed is held on the surface in two ways: (~) as positive hydrogens and negative hydroxyls, and negative hydroxyls, and (2) as positive hydrogens and negative hydrogens. "Evidently the water represented in these equations is not all evolved for if such were the case the catalyser would soon lose all its oxygen and, as will shortly be pointed out, lose almost entirely its capacity for catalysing hydro- genations. This water is only evolved in the free state in relatively small amount, the chief part remaining on the particles as hydrogens and hydroxyls. This is equivalent to saying that in reaction (~) a negative hydroxyl on the surface of the catalyser has a tendency to unite with a positive hydrogen of a neutral hydrogen molecule, thus loosening the bond between the positive and negative hydrogens of the hydrogen molecule sufficiently to permit the positive and negative hydrogens of the hydrogen molecule to unite with a molecule of ethylene. That is, the hydrogenation is Pictured as trio fit t.h~ Of of the particles by means of oscillating hydrogen atoms which are at one in- stant more closely associated with the hydroxyls and hydrogens on the surface of the particles and at the next instant more closely associated with each other in hydrogen molecules. A small portion of the impacts of positive hydro- gen of gas molecules and negative hydroxyls on the surface result in the perma- nent formation of molecules of water which are evolved as such. ~. = ~_ ~v ~vet ~ ~ ~ ~ "Reaction (2) represents a reaction very slow in comparison with reaction (I) and which is constantly ticking nlnn~ flaring t.h~ h~rr~rmcrmn~+i~n l\Tm~r~t;~r~ ~ = ~ _~ ~ ~ _ ~ ~ v^_~ ^~! ~^ ~ Mu ~ ~ A. _ ~ Waco Ul v ~ 1~ __ _1 1 , 1 ret nyclroxyls on the surface are constantly and very slowly being removor3 and hydrogen being adsorbed. O . w _ "Reaction (3) represents the reaction of this adsorbed hydrogen with unchanged nickel oxide in the interior of the particle. Here also the water represented is not all evolved in the free state but partly goes to reform hydro- gen and hydroxyls on the surface.

THIRD :REPO:RT OF THE COMMITTEE ON CONTACT CATALYSIS 905 "Finally, after long use the oxygen remaining on the catalyser either as negative hydroxyls or unchanged nickel oxide in the interior becomes very small and nothing remains finally but nickel particles with adsorbed hydrogen thus- rH+ Ni HE H+ ~ H Nickel in this condition is a very poor catalyst for hydrogenations. The activity of the catalyst is associated with its oxygen content and its activity can be restored by reoxidation and partial reduction. "It follows from the experimental data that the absorption capacity of nickel for hydrogen depends on the method of preparation. If prepared from oxlcte by recluctlon with nyclrogen at temperatures nelow 275°, it should prob ably require many months to remove completely all the oxygen. And as we have seen, the capacity of a nickel catalyst to hold hydrogen depends largely on its oxygen content. By continuous reduction at 275° for only ten hours, a condition is reached where the water evolved in half an hour is relatively very small. Should this be taken as an indication of the attainment of complete reduction an utterly erroneous result would be obtained for the hydrogen adsorption capacity of nickel, for the catalyst would still contain a large per eentage of oxygen. This probably explains the widely varying statements in the literature regarding the amount of hydrogen which nickel can adsorb, varying from on vols. of hydrogen per volume of nickel to a capacity for hy drogen as great as that possessed by cocoanut charcoal. "No meaning attaches to the measurement of hydrogen adsorption by nickel unless the whole history of the nickel is also described in detail. The term, it seems, should be restricted to the amount of hydrogen taken up by a known weight of nickel spread over a definite surface, the nickel having been prepared by the reduction of nickel oxide by hydrogen at a definite tempera- ture until all the oxygen has been removed. "As nickel oxide has an indefinite composition, being; always a mixture of oxides, the completion of reduction by hydrogen cannot be determined by continuing the reduction until the water equivalent of the oxygen in the oxide has been evolved. There appear to be two ways of determining whether re- duction has been complete or not: (~) to continue the reduction in hydrogen until no water is evolved, even after allowing the nickel to stand in the cold in an atmosphere of hydrogen for several hours and subsequently heating in a current of hydrogen; and (2) completely reduce at 400°C. and then oxidize ~ , ~ with a known volume of oxygen at 4oo° and reduce at the desired temperature until the water equivalent of the oxygen adsorbed has been evolved. "From the standpoint of catalysis of hydrogenation, however, the measure- ment of hydrogen adsorption is, as we have just seen, of little importance, as the normal nickel catalyst is never in the condition of holding; hydrogen alone. "Notwithstanding the relatively large amount of hydrogen adsorbed on a nickel catalyst prepared by partial reduction at ~75°, ethylene alone, in

906 HUGH S. TAN-LOP the absence of free hydrogen, does not react at ~so°C. For hydrogenation free hydrogen must also be present. This is also true for nickel prepared by complete reduction at 4oo°. That is, the hydrogen on the nickel catalyst in either of the two states, (~) positive hydrogens and negative hydroxyls and (~) positive hydrogens and negative hydrogens, do not react with ethylene at ,o° in the absence of free hydrogen. "According to the n~echanism of hydrogenation by nickel just described, most of the conflicting views of investigators are, we believe, explained. The conception of definite hydrides as intermediate products in hydrogenating actions is not valid as the Formal catalyst always contains oxygen and func- tions, as catalyst for hydrogenations, chiefly through the hydroxyl groups on the surface. Even where the catalyst carries only hydrogen this can not be said to exist in the form of definite compounds called hydrides of definite proportion of hydrogen to ~icl~el, but rather as complexes in which nickel carries the hydrogen adsorbed on the surface as positive and negative hydro- gens. "Likewise the oxygen present in the normal catalyst is not there as a definite hydroxide of nickel, but as a complex carrying hydroxyl groups nega- tively charged along with h~Tdrogens positively charged. However, although these combinations are "complexes" rather than compounds yet the hydro- gens and hydroxyls react, it should appear, in stoichiometric proportions. The recent researches of Kelberi must be considered as decisive however, in connexion with the question of the necessity of oxygen in nickel catalysts of high activity. Kelber has prepared nickel catalysts bv reduction of nickel cyanide in hydrogen at various temperatures. catalyst preparation is hereby- avoided. . ~· ~, . . ~ v ., The presence of oxygen in the With such catalysts he has demon strated high catalytic activity even in systems which contain no oxygen of any kind. Thus, the reduction of diphenyl-diacetylene in hexane and of azo- benzene in hexane by hydrogen, in presence of oxygen-free nickel from To avoid all objections, the cyanide, went with extraordinary velocity-. Kelber used hexane instead of water as the containing liquid for the hydrogen. Kelber further shows that nickel so obtained has the same eharacteristies as nickel obtained from oxide, in respect to sensitivity to heat treatment. By reduction at ~so°C. the nickel brought about 60ce hydrogen absorption, in 5 minutes; on reduction at 400°C., So minutes were required for the same gas absorption. Kelber concludes that it is the high temperature which causes a change in surface of the catalyst and that elemerllary nickel can effect the activation of hydrogen. Ills tatter and Seitz suggest that the production of tetrahydro-napthalene or the deea-hydro derivative by hydrogenation of naphthalene in the presence of platinum sponge depends upon the oxygen content of the catalyst.2 They suggest that direct conversion of napthalene to deca- or tetra-hydro deriva tives is possible. With oxygen-rich platinum the tetra-derivative is the pre ferred product. An attempt is made to justify this view from the exhaustive iBer., 57, ~36, ~42 (~924). 2Ber., 56, ~388 (~923). See also, Zelinsks-: Ber., 56, ~723 (~923).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 907 experimental data. The tetra-derivative obviously results from hydrogena- tion of one benzene ring only. Its further hydrogenation only occurs slowly. The deco derivative apparently suffers hydrogenation in both rings simultan- eously when little oxygen is present. Is this a case where oriented adsorption occurs, different in the two cases when the platinum catalyst is rich or poor in oxygen ? An interesting contribution to the problem of the nature of nickel hydro- genation catalysts has been made by Schlenk and Weichselfelder,~ who have succeeded in preparing a hydride of nickel, NiH9. They prepare it by the interaction of anhydrous nickel chloride on an ethereal solution of phenyl magnesium bromide in an atmosphere of hydrogen. Four equivalents of hydrogen are taken up in the preparation of the hydride, which may possibly be accounted for by the following sequence of reactions /Ph 2 Am,-Br+NiCl2 = )/lgBr2+MgCl2+ tNiPh2] (NiPh21 + 2H2 = NiH2+ 2C~H6 The hydride is obtained as a black precipitate which, when the solution is decorated, gives, on washing with ether, a blacl: solid. This solid, on decom- position with. alcohol and so per cent. sulphuric acid, gives a hydrogen e`-olu- tion corresponding to the formula NiH2. NiH2+H2SO4 = ~TiiS04+ 2H9 The amount of hydrogen absorbed in the preparation of the hydride is around Moo volumes per volume of nickel which contrasts strongly with even th maximum values obtained in the adsorption studies at Princeton. Its magni- tude seems to rule out the possibility of its being an adsorption complex. The hydride is stable in ether and unstable in presence of alcohol, which fact the authors link with the known ease of hydrogenation by nickel in alcohol solu- tions and the difficulty obtaining with hydrogenations in ether. The hydride is a good hydro~,enation catalyst, not only as to reactions brought about, but also as to the temperature at which it is reactive; it hydrogenated many unsaturated compounds at room temperature. It is, however, a very sensitive catalyst. Oxygen from the air kills its activity for hydrogenation at room temperatures, an observation which is of weight in view of the claims of Willstatter and Boswell already discussed. This active nickel catalyst, at any rate, does not need oxygen for its reaction efficiency. Definite information as to whether the substance is a hydride or adsorptions complex could be obtained from a preparation of the dry substance and a measurement of its dissociation pressure. This measurement, carried out at two temperatures, would give the thermal data for heat of formation, if a compound. Comparison of these with the known data on heat of adsorption of hydrogen on nickel (vice infra) would then materially add to our informa- tion on the mechanism of hydrogenation. 1Ber., 56, 2230 (I923).

rgo8 HUGH S. TAYLOR There is no evidence in the hydrogenation studies of Pease that oxygen is in any way necessary or beneficial in the reaction of hydrogen and ethylene at copper surfaces. '1 · ~ 1 ~ Pease~s preparations are active some I50° lower than Nero recorcrect copper catalysts. lIis work is of importance in that it associates, for the first time, kinetic measurements with adsorption measure- ments on one and the same catalyst. "Because of the known variability among samples of catalytically active material both as regards catalytic activity arid adsorptive capacity, it was onsidered vital to obtain measurements of the two properties on the same sample of material. This has accordingly been done. Measurements of reaction velocity have been alternated with determinations of adsorption isotherms on the same sample of catalyst in such a way that sets of measure- ments of each kind have been "bracketed" by measurements of the other. This has been done in order to take account of any change in activity. "Edith respect to the velocity measurements at or, the velocity (slope of curve) is greatest with a mixture of 2He:~C~H~ and least with a mixture of THE ::C2H4. With a 50 per cent. mixture the velocity is intermediate between the other two. If the reaction were bimolecular, as the chemical equation suggests it might be, the velocity should be the same for the mixtures of THE: ~C2H4 and ~H2:~C2H4, and with a 50 per cont. mixture the maximum velocity should be attained. The observed order of the curves suggests rather that the reaction is more nearly unimolecular with respect to hydrogen and independent of the ethylene concentration. An excess of ethylene actually inhibits the reaction rather than causes an increase in velocity. "heathen the amount of ethylene is constant, increasing the hydrogen eon- eentration 3.9 times causes the velocity to increase 3.o times; and when the amount of hydrogen is kept constant, increasing the ethylene concentration 4 times causes the velocity to decrease to o.6 of its original value; that is, with the same concentration of hydrogen, the reaction velocity increase ~.7 times when the ethylene concentration is decreased to ~/4 of its original value. It seems to Pease that "a reasonable explanation of these observations earl be made in terms of the adsorption theory of catalysis, with the aid of the results of the adsorption measurements. The adsorption of pure ethylene is markedly greater than that of pure hydrogen, being ~.45 cc. at lo mm. pressure against o.3s cc. for hydrogen, and 6.80 cc. at 760 mm. against No ee. for hydrogen. Since, therefore, the adsorption of ethylene at lo mm. pressure is greater than that of hydrogen even at 760 mm. it is undoubtedly true that from almost any mixture of the two, considerably more ethylene than hydrogen will be adsorbed. Further, if we suppose that those active centers on the catalyst surface which are capable of holding hydrogen are among those which can hold ethylene, it follows that when there is a mixture of the two in conta et with the surface they will be competing for these centers and, since the ethylene is the more strongly adsorbed, the hydrogen will occupy relatively few of such spaces. We shall, therefore, be dealing in most eases with a surface largely covered with ethylene, with hydrogen molecules scat 1 J. Am. Chem. Soc., 45, I 196, 2235 (I923).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 90) tered over it here and there. Let us suppose that both ethylene and hydrogen must be adsorbed before reaction can occur. Note have concluded that ethylene will usually be present in large excess on the surface, so that its surface eon- eentration will be of secondary importance, so far as it enters directly into the velocity expression. The velocity should, therefore, depend mainly OF the amount of hydrogen adsorbed. Other things being equal, the latter will increase with the partial pressure of the gas. It also seems reasonable to be- lieve that as the partial pressure of ethylene, and therefore its adsorption, decreases, the amount of hydrogen adsorbed at a given partial pressure will increase. Since, therefore, we have assumed that the velocity depends upon the amount of hydrogen adsorbed, we may expect it to increase with increasing hydrogen concentration and decreasing ethylene concentration, within limits. These are the relationships found by experiment. "The average value of the velocity constant is o.so at o° and ~.32 at Oh. The velocity has therefore increased ~.64 times for a ~o° rise in temperature. This is equivalent to an average temperature coefficient of ~.62 per ~o° rise between o° and 20°. Such a great temperature coefficient effectually disposes of the possibility of diffusion playing a dominant part in the process. The increase in reaction velocity to be expected from diffusion alone would be about ~ per cent. per ~o° instead of the 6~ per cent. found. Moreover, if diffusion were a controlling factor, the velocity should depend upon the con- centration of that reactant which would diffuse most slowly, namely ethylene, whereas actually it depends upon the concentration of the more rapidly diffus- ing hydrogen. "Further information regarding the dependence of the reaction velocity- upon the hydrogen adsorption was obtained in some experiments during which the catalyst was poisoned with mercury. No determinations of reaction velocity were made before poisoning but the magnitudes of the adsorptions indicate that the catalytic activity was somewhat greater than that of the catalyst already described. Several adsorption experiments were made and then a little mercury was run up into the stopcock of the manometer and blown into the evacuated catalyst bulb. The quantity of mercury was esti- mated from the bore of the stopcock to be o.o~s ec. or zoo ma. This would be equivalent to about 20 CC. of vapor at o° and 760 mm. The bulb was then heated to 200° for i/ hour and evacuated. After cooling, the mercury had disappeared and the catalyst was unchanged in appearance. The adsorptions at 380 mm. of hydrogen and ethylene, respectively, were found to be 3.25 cc. and 8.55 cc. before poisoning and ohs cc. and 6.70 cc. after poisoning. The value of dP for a 50 per cent. mixture after poisoning was o.7 mm. and was estimated to be coo mm. before poisoning. "It is evident that the mercury has reduced the adsorption of hydrogen to less than To of its former value but has reduced the reaction velocity to. about I/200 of its former value. The ethylene adsorption has been only moderately diminished. Here again it is evident that the catalyst must be able to adsorb hydrogen as well as ethylene before it can bring about reactions

9IO HUGH S. TAYLOR "In the course of the experiments on the catalytic combination of ethylene and hydrogen, the effect on both catalytic activity and adsorptive capacity of partially de-aetivatin~, a copper catalyst by heating it to 450° in ~ vacuum was determined.) As this gave results which differ somewhat from those obtained by de-activation with mercury, they are also included. The effect of de- aetivating this sample of copper by heating was in a general way similar to the effect of de-activating the other sample by poisoning it with mercury. The curves have been moved over toward the pressure axis to nearly parallel positions, at the higher pressures at least. The heating has, however, decreased the hydrogen adsorption relatively less than the poisoning and the ethylene adsorption relatively more. Thus, at one atmosphere the decrease ire hydro- gen adsorption amounts to 70 per cent. while the decrease in ethylene adsorp- tion amounts to ~: per cent. These are to be compared with decreases of 9~ per cent. for hydrogen and ~4 per cent. for ethylene caused by mercury poison- ing. The absolute decreases at one atmosphere are c.60 for hydrogen and ~.95 for ethylene. It will be seen that these figures are mueb more nearly of the same order than in the case of copper poisoned with mercury. "The decrease in catalytic activity in the ethylene-hydrogen eombi~ation accompanying these decreases in adsorption amounted to 85 per cent. Just as in the case of the poisoning by mercury, one must go to very low pressures to find a corresponding decrease in adsorption, indicating that it is the strong (low-pressure) adsorption which is mainly responsible for catalytic activity." This last observation seems especially important to the reviewer. "It is clear frozen the relative adsorptions of the different gases by active copper that we may at once conclude that ordinary condensation in capillaries is not a sufficient explanation of the results, although it may account for the adsorption of ethane and partially for that of ethylene. The action seems rather to be a specific one between the copper surface and the particular gas. It seems probable, however, that any copper surface will not do, but that the surface must be in ~ special condition. Frown the evidence here presented, taken in conjunction with previous experience in the Princeton Laboratories, it would seem that an active copper surface is one which has scattered over it regions containing atoms whose fields are highly unsaturated. This follows from the feet that heating Betide copper to temperatures as low as 450° caused appreciable sintering besides decreasing the surface activity. Sintering at so low temperatures points to the pre-existenee or the surface of atoms of ~ Interesting results on the effects of heating active copper to successively higher temperatures have been obtained in the course of this investigation. In the present in- stance, the catalyst had been prepared at 200°, and heated to 300° after reduction. It had not thereafter been taken above cool. After the experiments on the active material so obtained had been carried out, the catalyst was heated first to 350° for an hour and then to 4oo° for ~2 hour without a marked change in activity resulting. It was then heated to 450° for one hour after which it was found to have decreased in activity as will be shown. Further heating at 450° for ~ hour was without noticeable effect, however. Similar re- sults were obtained With another catalyst which was eventually heated to 550° to produce a very inactive material. For each rise in temperature ~ noticeable decrease in activity occurred but further heating at the same temperature was without marked effect. There seems, therefore, to be a stable condition of the surface corresponding to the highest tem- perature to which it has been heated. All the healings described above were carried out in a vacuum.

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 9I I high mobility arid therefore in a state of unsaturation. The process of sinter- ing is the process of saturation of these atoms, and since the agency which causes the sintering. also decreases the surface activity, it is reasonable to look upon these unsaturated atoms as the cause of this activity. One would look for atoms of this character in surfaces of high degree of curvature in "peaks", that is to say, on the copper surface rather than in the "valleys" or capillaries. "It seems probable that each of these "peaks" can attach more than one molecule of adsorbed gas. Otherwise it is difficult to see how combination of ethylene and hydrogen, for example, can take place as a result of adsorption. As already pointed out, since each hydrogen molecule that is adsorbed appar- ently displaces an ethylene mole eule, the same point on the copper surface cannot hold ~ molecule of both. The two must, however, be in close juxta- position if combination is to occur. This can be true only if a given peal: possesses more than one possible point of attachment. The aeti~rity is, there- fore, not due to isolated active atoms scattered over the surface but to groups of these atoms." At the higher temperatures with less active copper catalysts conditions were somewhat different. "Measurements of the velocity of combination of hydrogen and ethylene in the presence of copper at Sod, coos and 250° have shown that in this tem- perature region the reaction is more nearly bimolecular, ire contrast to the combination at or, at which temperature the reaction is approximately uni- moleeular with respect to hydrogen and inhibited partially by excess of ethylene. The more nearly normal character of the reaction at the higher temperatures is believed to be due to the fact that under these eireumstanees the reacting gases are not measurably adsorbed) by the catalyst. The tem- perature coefficient is much smaller at the higher temperature and is decreasing. By taking into account the decrease of adsorption with rise in temperature as well as the normal increase in velocity of the surface reaction, these facts have been aeeounted for qualitatively. In connection with the above views of Pease it is interesting to record that Wright and Smith2 and Smiths have studied the sintering of metals. Smith concludes that:- (~) Sinterirlg may take place in crystalline and amorphous substances. (~) The sintering of a crystalline substance is due to a change in the size of the crystals or to the formation of an allotrope. (3) The sintering of an amorphous material is due to the formation arid growth of crystals. The following; sistering temperatures are given: Ppid Pt black, soo°; Pd-blaek, 600°; Pptd Ag' ~80°; Pptd Au. 250°; Pptd Co, 200°; Reduced Cu. soo°C; Pptd Cu. 250°; Ppid Fe, 750°; Ppid Ni, 700°. 1 It is probably more correct to assume that the adsorption is small and approximate- ly proportional to the partial pressures of each gas. H. S. T. 2 J. Chem. Soc., 119, 1683 (I92I). 3 J. Chem. Soc., 123, ~o88 (I923).

9 I 2 HUGH S. TAYLOR Judged by loss of adsorptive capacity of the reduced metal sintering may take place at much lower temperatures than those recorded above. What this means is that loss of adsorptive power is much the most sensitive index that we have at the present time as to change of surface upon heating. An attempt has been made by Dougherty and Taylor1 to gain some in.- sight, by kinetic measurements, into the mechanism of the catalytic reduction of benzene to hexabydrobenzene. The results indicate that the reaction does not occur at all according to the stoichiometric equation, as calculated from gas concentrations, but at rates governed by the distribution of the reacting materials between the catalyst and the gas phase. The trend of the reaction with change of temperature has been studied, and equilibrium values at the higher temperatures have been calculated. The results on the latter show that apparent equilibria in the gas phase, as measured in this way, do not neeessar- ily coincide with those which would be expected on the basis of the ordinary equation representing the reaction. DeLydrogenation becomes marked even in presence of hydrogen above 200°C. Water vapor in small amounts, up to ~ per cent. of the hydrogen volume used in the reaction mixture, had only a slight depressing effect on the reaction velocity. Carbon monoxide in small amounts, about ~ per cent. of the hydrogen volume, had a very marked poison- ing effect, particularly at low temperatures of ~oo° or under. As the reaction temperature was raised the poisoning was less noticeable. In large quantities, however, around 50 per cent. of carbon monoxide, the reaction was completely stopped at ~80°. Hexabydrobenzene, at low temperature, ~oo° or less, had a depressing effect on the reaction velocity. This effect disappeared at higher temperatures, in the neighbrohood of ~ 80°. The observations show that it is necessary to use great care in making comparative measurements on account of the variability of the nickel catalyst. It was found that different catalysts, although prepared exactly in the same manner, might have quite different activities, and that the activity of a given catalyst changed markedly with time and use. The observations also show that quantitative measurements on a reaction of this kind are difficult due to the fact that the actual reactant eoneentrations, on which the velocity of the reaction depends, are those on the catalyst surface; and these concentrations may be independent of, or bear a varying relation to, the reactant concentra- tions in the gas phase. From experiments at 80° and go°C it is shown that the temperature co- efficient of the reaction measured is approximately 3.~:~.9 or ~.65 per lo degree rise. This is evidently the temperature coefficient of a chemical reae- tion as opposed to that of a diffusion process. The experience gained with. this kinetic investigation demonstrated the need for both adsorption and kinetic studies on one and the same catalyst. Continuing his earlier studies2 in which he showed that the catalytic activity of moistened platinum and palladium in the catalysis of hydrogen- oxygen mixtures is determined as regards velocity by reaction and by the ~ J. Phys. Chem., 27, 533, (~923). 2 2nd Report p. 8~4; Ber., 49, 2369 (~9~6); 53, 298 (~920); 55, 273 (~922).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 9I3 pre-treatment of the catalyst whether by hydrogen or oxygen, Hoffman, has shown that with iridium, no variation in rate is induced by prior treat- ment with either gas. The iridium does not seem to adsorb either gas selee- tively, as demonstrated by measurements of its electrode potential. It is equally efficient in acid or alkalir~.e solution. The catalytic behavior of these three metals for this reaction is therefore definitely associated with the ad- sorptive capacity of these metals for the gas mixtures. lIoffman.n's method of attack is suggestive as a method of study of the still-debated question as to whether oxygen is necessary and indispensable in hydrogenation processes. Mitchell and Marshall2 have reinvestigated the work of Anderson3 on the activation of hydrogen by platinum as revealed by the temperature of reduc- tion of copper oxide. They show that with pure hydrogen no such activation occurs and that Anderson's results are to be attributed to the presence of small amounts of oxygen. The nature of the active hydrogen produced under such eireumstanees, i.e., with oxygen present, is still uncertain. The authors lean to the eonelusion that it is triatomie hydrogen.4 Tin has been shown, by Brown and Henke5, to be an excellent catalyst for the reduction of nitrobenzer~e to aniline. It is superior to copper at all rates of gas passage but the lowest tried. It is superior to nickel at all but the highest rates. The catalyst is best prepared from the hydroxide by preeipita- tion with sodium carbonate from a sta~.D.ous chloride solution. Oxidation of the hydroxide prior to reduction increased the efficiency of the result;in.g eata- lyst, the lower the temperature of oxidation the better the resulting catalyst. The lower the temperature of reduction of the oxide the better was the result- ing catalyst. A catalyst in the form of coarse lumps is better than in the powdered form. Tin is a new-comer in the ranks of catalysts for reduction. The mechanism of its action is worthy of study. Is it a hydrogenating catalyst or is its action dependent on alternate oxidation and reduction? Both possibilities have their own interest. Preferential Hydrogenation Rideal has studied the rate of hydrogenation of einnamie and phenyl propiolie acids in presence of colloidal palladium. With the metal sol present in large quantities, solutions of the sodium salts of the two acids are hydro- genated at equal speeds, the rate being governed by the rate of hydrogen supply, and proportional to the square of the shaking speed. The reaction velocity is of zero order. For small quantities of sol the velocity is propor- tional to the concentration of palladium and the phenyl propiolate is hydro- genated at approximately twice the rate of the einnamate. Above certain critical limits the rate is independent of the shaking speed. The reaction velocity is within wide limits independent of the salt eoneentration. ~ Per., 56, ~ ~65 (~923). 2 J. Chem. Soc., 123, 2448 (~923). 3 Ibid., 121, ~ ~53 (~922). Cf. Venkataramaiah: J. Am. Chem. Soc., 42, 930 (~923). 5 J. Phys. Chem., 27, 739 (~9~3). 6 Trans. Faraday Soc., 19, 90 (~923).

9I4 HUGH S. TAYLOR There appears to be an aging; effect with the sots. All sols commence with a velocity curve of zero order and terminate in one of the first order. For active sols the portion not of the zero order is very small whilst for aged sols the portion of the first order is relatively large. For inactive sols the curve is of the first order throughout. Furthermore, with aging, the rate of hydro- genation is diminished considerably. Rideal attempted to establish the hypothesis that the salt was adsorbed by the palladium sol. He showed that the sol protected by o.: per cent. gum arable undergoes aggregation when treated with the sodium salts. Ten milli ~. . . . .. . v ~ He showed that the sol protected by o.: grams of a sol aggregated In this manner were filtered through a small filter and washed into a small tube eonr~eeted to a lo ee. hydrogen burette. The aggregated sol and filter paper absorbed 4.35 ees. of hydrogen at 2s°C. A duplicate filter paper through which IO CCS sodium phenyl propiolate had been Altered required a further ~ ce. of hydrogen. Ten ma;. of the sol untreated with salt absorbed ~.53 ce. Hen ee, Rideal eoneludes, the sol had adsorbed salt equivalent to 4.3s-~+~.s3~=~.82ee. of hydrogen. This corresponds to ene molecule of salt to approximately two atoms of palladium, which may or may not be significant. The aging of the sol is attributed by Rideal to reduced adsorptive capacity for the unsaturated salt. The feet that art low sol concentrations the phenyl propiolate is hydro- genated twice as fast as the einnamate, suggests to Rideal that the salt is not desorbed from the sol surface until completely saturated and that the phenyl propiolate takes up two hydrogen molecules from the palladium in the same time as the einnamate takes up one. A number of investigations indicate, however, that this is not necessarily true for all eases of preferential hydrogenation. Most of the work on hydro- genation of oils involves the possibility of preferential hydrogenation and eer- tain of the researches on the subject indicate its existence. Moore, Richter and van Arsdalei indicated that the more unsaturated glycerides were hydro- genated preferentially to the glycerides containing only one double bond. Quite recently, Richardson, Knuth and Milligan2 have confirmed this eon elusion showing that the preferential nature of the process is even more pronounced than had been previously believed. A newer method of analysis of the hydro- genated product revealed, in a typical ease the following percentage of satur- ated, oleie and linolie acid glycerides in the oil before and after hydrogenation. Cotton Seed Oil Saturated Oleie Linolie Acids Before Hydrogenation 22.7 27-5 49.8 After Hydrogenation 24.0 67. ~ 8.9 It is evident that in this experiment the hydrogenation was practically exclusively hydrogenation of linolic acid glycerides and negligible hydrogena- tion of oleie acid compounds. This would indicate almost exclusive adsorp- tion of the more highly unsaturated glycerides at the nickel surface. The authors found that the selectivity of the tydrog;enation appears to be more marked with increasing amounts of catalyst and with increasing; temperatures ~ J. Ind. Eng. Chem., 9, s4r (~9~7). 2 Am. Chem. Soc., September Meeting r923, Milwaukee, Vlis.

THIRD REPORT OF THE COMMITTEE ON CONTACT C.\T\Ll-SIS 9I5 up to an optimum in the neighborhood of ~oo°C. Quantitative measurements on preferential adsorption should prove very interesting, in this ease. As Banerofti has already pointed out, there are almost no quantitative data on selective adsorption in liquid systems. An intensive study of the field will be fruitful alike to colloid chemistry and contact catalysis. Dehydrogenation Dougherty, in the worl: previously cited, attempted to measure the posi- tion of equilibrium in the reaction C6H~+3FI~ = C6~ His results were not very conclusive although they did establish the rever- sibility of the process. He used nickel as a catalyst. Therein, apparently, lay some of his trouble in views- of the following conclusions of Zelinsky. Some- one ought to repeat Dougherty's world using; platinum or palladium instead of nickel. Zelinsky and Pablo ire studies of the deLydrogenation of eyelohexane with platinum, palladium and nickel show that the efficiency of these eata- lysts for a given velocity of vapor passage decreases in the order given; deby- drogenation starts as low as ~so°C. and is complete around 3oo-3so°C. With platinum and palladium, even as high as 400°C., little or no earbonisatior~ occurs. With nickel, deeo~nposition occurs at a much lower temperature; hence, presumably the very inferior behavior of nickel in comparison with the other two. Ze]insky3 show sthat o- dimethyl eyelohexar~e undergoes deLydrogenatior but that I, ~ dimethy1 eyelohexane does not undergo deLydrogena.tion under such conditions. Zelinsky thinks that ~,~ dimethyl eyelohexane is therefore to be regarded as different from the hexahydro aromatic compounds. De- hydrogenation in Zelinsky's view has therefore a selective eharaeter. It is distinctly probable that the free-energy factors concerned will reveal reasons for the absence of dehydrogenation, as they undoubtedly do for the absence of dihydro and tetraLydro derivatives in the hydrogenation of benzene. E. C. Kendall showed some >-ears ago that ditydro-benzene when led through a heated glass tube at 200°C. was completely decomposed to benzene and hexaLydrobenzene. Promoter Action Pease and Taylor's bibliography of the literature on promoter aetion4 showed definitely that little or nothing was known as to the mechanism of promoter action. Beginnings of an attack on this problem are row apparent and elucidation may be expected to follow. It seems essential to take single eases and study them thoroughly and not to generalise too soon. One ease in which the mechanism of promoter action seems definitely to have been obtained is available in the older literature. i "Applied Colioid Chemistry," p. 73 (~92~). 2 Ber., 56, i249 (~923) 3 Ber., 56, 787, ~7~6 (~923); J. Phys. Chem., 24, ~4~ (Woo).

9I6 HUGH S. TAYLOR Elissafoff studied) the action of glass wool and heavy metal salts on the velocity of decomposition of hydrogen peroxide both singly and in conjunction with each other. Elissafoff showed that, together, the glass wool and heavy metal salt effected a much more rapid decomposition of the peroxide than would be anticipated on the basis of additive effects. This case is certainly therefore a reaction velocity at the surface and not a diffusion velocity. Were it possible to make adsorption measurements, other modes of decomposition of hydrogen peroxide would possibly be found to be of the same type. Thus with a decomposition velocity of 0.86 in presence of o.s g. glass wool in so ccs. Of peroxide, and one of ~.63 in presence of ~ ~.54 millimolar solution of copper sulphate without glass wool, a solution with the same copper sulphate con- eentration plus o.s g. glass wool in 20 CCS. gave a decomposition velocity of ~o.8; all the velocity units are expressed in arbitrary units. The concentra- tion of hydrogen peroxide was I 2 millimolar. In this ease, at least the mechan- ism is apparent. It is known that the decomposition of peroxide takes place at the glass surfaces. It was probable that the copper salt was concentrated at the glass surface and so exercised greater effect. That this was so the follow- ir~g observations indicate clearly. The unimolecular constants for two copper ion concentrations of ~ and lo millimols per litre were o.oo~3 arid o.oo~3 re- spectively, in the ratio, therefore of o:~.77. The amounts of copper salt adsorbed from these solutions by Jena glass powder of the same glass were in the ratio of ~ :~.73. It is apparent that the decomposition velocities are pro- portional to the amounts of adsorbed copper salt. An informing contribution to the theory of promoter action has been made by Medsforth2 who has studied the effect of promoters added to a nickel catalyst in the hydrogenation of carbon monoxide and carbon dioxide to yield methane. Reasoning from the simultaneous production of water in the reaction, the addition of eatalytie dehydrating agents to the nickel catalyst was made with material increase in the attainable reaction Velocity for a given conversion of the reaeta~ts. Ceria, thoria, glueina, chromium oxide, alumina, and silica gave a, from Refold to ~-fold, increase in velocity over that obtainable with the straight nickel catalyst. Zireonia, molybdenum oxide and vanadium oxide were somewhat less efficient, though still good, pro- moters. Tin and magnesium oxides, copper and silver metals produced no acceleration over the straight nickel. With the carbon dioxide reaction the increases in velocity effected were somewhat less than those recorded for the monoxide reaction above. The order of effieieDey was exactly the same. The order of e~eieney is roughly that of oxide catalysts recorded by Sabatier in reference to strict dehydration processes. Ir1 explanation of the activity of the promoters, Medsforth assumes the function of the nickel to be to assist the union of the gases to form a 'complex' or intermediate compound of the methyl alcohol type, probably via formal- dehyde. The promoter then functions as a catalytic dehydrating agent on the Z. Elektrochem., 21, 352 (I9I5). - Chem. Soc., 123, 1452 (I923).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 9I7 intermediate compound giving water and a methylene radical whence, imme- diately, methane results. The promoter assists the known dehydrating action of the nickel. This can be schematized thus CO+2H2 ~ H3C.OH The increased activity of the promoted catalyst is therefore ascribed to in- erease in (~) and suppression of Alp. The combined dehydrogenation and dehydration effected by Ipatiew~ with a nickel -alumina eat alyst whereby the conversion of camphor to isocamphene is effected at ~oo°C. in one step is cited as supporting evidence , CH2 , / CH2 C8H14 1 ~C8H WACO \ CHOH ~C8H14 ~1 ~ / CH2 C8H14 ~ ACHE The several steps of this process can be conducted singly with the single catalysts though less efficiently. LIedsforth calls attention to an important feature of promoter action which he has noted, that of selective promotion. It would appear that when two reactions, both capable of being accelerated, take place at the same time in the presence of the same catalyst and the sane promoter, that which is normally slower is accelerated to a greater eo~nparative degree than that which is normally the faster. Support for this statement was obtained in the obser- vation that when carbon monoxide and steam react in the presence of nickel and nickel promoters, whilst carbon dioxide and hydrogen are the main prod- ucts, methane is also formed, the quantity of which is greater when, for ex- ample, alumina is present than when nickel alone is used. Similarly, in the production of methane from carbon monoxide and hydrogen, more carbon dioxide is formed as a by-product due to the simultaneously occurring water gas reaction, when promoters are added to the nickel catalyst, than if this latter is used alone. In discussing the applications of his dehydration hypothesis )ledsforth reviews several cases of promoter action. For catalysis of the water gas reae- tion with iron oxide as catalyst it is significant that the promoters among the most effective are hydrating agents arid oxygen carriers. The action of ceria- thoria in the ineandeseent mantle may also be in part due to combined oxygen earrier-dehydration effectiveness. As a temporary classification of promoters for purposes of discussion Meds- forth gives the following: (~) The promoter decomposes intermediate compounds formed by the catalyst. (2) The promoter causes the reacting substances to combine, the resulting intermediate compound being decomposed by the catalyst. (3) The promoter adsorbs or combines with one of the reacting substances producing a greater concentration of the latter at the catalyst surface. Further contributions to the problem are promised. ~ J. Russ. Phys. Chem. Soc., 44, ~695 (~9~2).

9I8 HUGH S. TAYLOR The origin of the carbon dioxide in the methanatior~ process has been elucidated by Armstrong and Hilditehi who have shown that when purified water gas is passed over nickel at ~oo-3oo°C. the predominating reaction is eCO+~H: = COACH. The reaction is regarded as the sum of two reactions ., CO+H20 = CO~+H2 CO2+ 2H2+ 2H2 = CH4+2H20. the former of which is regarded as oeeurrix~g in the same manner as the reae- tion in presence of copper previously studied by theme, namely via formic acid CO+H20 RECOOK ~E2+CO~. With cobalt, the reaction eom- menees at a lower temperature, ~80°C., but the above reaction is subsidiary to the main methanation process CO+3He=CH~+H~O. Silver is inert, iron almost so, platinum and palladium of minor activity. Mixed catalysts were less efficient than the single catalyst. With nickel at increasing pressures up to 6 atmospheres the minimum temperature of interaction rises. The re- aetion yielding carbon dioxide and methane produces more methane from water gas than any of the other reactions. It may therefore have value as a means of increasing methane content or lowering carbon monoxide content of town's gas. The authors state that CO~+He goes directly to methane and gives no carbon monoxide, so that partial reduction is apparently not taking place. This is at variance with some observations made in the laboratories of the Munitions Inventions Dept., in England, during the war, where methane prepared from carbon dioxide n.nd hvAr~cr~n of~ntn.inPr] ~ small mar_ eentage of carbon monoxide. In contrast to the eonelusions of NIedsforth cited above, Armstrong and EIilditeh3 conclude, with regard to the 'promotion' of a straight hydrogenation process, the simple addition of hydrogen at an unsaturated linkage, in presence of nickel, alumina, silica oxides of iron and magnesium being employed as promoters, that the stimulation observed can be satisfactorily explained on the basis of increased available catalytic surface of the nickel. There is some evidence of the removal or adsorption of eat alyst poisons (sulphates in the precipitated oxides, or traces of impurities in the oil hydrogenated); but these appear as minor influecees compared to the effect on the extent of sur- faee of nickel produced. They have been able to snake an appreciably less amount of reduced nickel effect the sable amount of action whatever the ex- tent of the catalyst in alumina or other 'promoting' oxide. ~_ J ~ A ~ =~ ~ ~ ~ ~TV ^ 1~. ~ ~_~ ~ ~ AL [~ \_ 1 Armstrong and Hilditeh4 showed that the presence of sodium earbon.ate~ effectively promotes the hydrogenation of phenol at nickel surfaces, About 25 per cent. by weight of nickel appears to give the maximum effect. In the presence of carbonate the reaction rate is more nearly linear than in the ab- senee of the carbonate. This factor RIl~.Ct.R that t,pP filno~imn of t.hP mr^~mr ~--~>D ~ v v -, ~ ~ ~ ·, V ~ ~ ~ ~^ V ~ J.w All ~l~l\J U41 IS a protective one to the catalyst, keeping it free of inhibiting impurities. 1 Proc. Roy. Soc., 103A, 25 (I923). 2 Proc. Roy. Soc., 97A, 265 (I920). 3 Proc. Roy. Soc., 103A, 586 (I923). 4 Proc. Roy. Soc., 102A, 21 (I922).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 9I9 The promotion of nickel by the addil;ion of copper, Armstrong and Hil- diteh ascribe to the influecee of the copper on surface area. Catalytic Oxidation and Promoter Action The enhanced activity of manganese dioxide-copper oxide (Hopealite) mixtures as catalysts for carbon monoxide oxidation has received farther attention this year, and progress towards the solution of the mechanism in this case has been made. Bray and Almquist2 state that their results ir~dicate that the mixture, 60 per cent. Mn02, 40 per cent. (DUO is but slightly more active than other mixtures, and that basic copper carbonate as a eor~ponent has little if any advantage over copper hydroxide. The following theory of the mixture effect in this ease proved useful throughout their experimental work. "By the action of carbon monoxide and oxygen a protective film is formed on the catalyst which interferes with further action, unless it can be rapidly desorbed as carbon dioxide. The film builds up to a lesser extent for mixtures than for the one-component catalysts at the same temperature. The slow limiting reaction may be the rate of Resorption of carbon dioxide or the rate of a transformation within the film. We could stop with the statement that this is a question of the structure of the catalyst, but it seemed worth while to seek an interpretation in terms of valence theory. "The porous granules are believed to consist of a network of chains of atoms held together by valence forces.3 The forces that come into play at or in the film are also valence forces. ~ hen the catalyst contains the two oxides they will tend to neutralize each other's valence [orees, since they diver in basieity or polarity, and the strength of the valence forces at the film may be expected to be less than for a one-eomponert catalyst. In other words, an increase in the rate of Resorption, or increase in the rate of a reaction within the film, is attributed to what may be considered a partial chemical reaction between the two oxides. Whitsell and Frazer4 conclude that "manganese dioxide in lIopcalite mixtures is the initial cause of the oxidation at low te~nperat;ures. The active preparations are able to oxidize carbons monoxide extremely rapidly either catalytically or at the expense of their own oxygen. The analytical data show that these active samples have a very low potash content, less than o.s to no ~ Cf. Dewar and LieLmann; U.S.P. :,268,692; 1,~7 ,4c5. 2 J. Am. Chem. Soc., 45, 2305 (~923J. 3 Compare Langmnir: J. Am. Chem. Soc., 38, ~265 (~9~6,. The formation of the firm porous granule from the hydrated oxide or oxides may be thought of as follows. Each hydrated oxide is precipitated in the form of minute particles the size of which is determined by the method of precipitation, but is greater than molecular dimensions. The filter calve, before it is dried consists of particles surrounded by film of water, which enable the relative positions of the particles to be easily changed for example in a kneading process. In the preliminary drying as the water is slowly expelled, the par- ticles are gradually drawn together, and in many places contact is sufficiently close to allow valence forces to come into play between the molecules of different particles. The plastic material has now been transformed into a solid. Finally, as the water of hydration is grad- ually expelled, the body becomes porous. ~ J. Am. Chem. Soc., 45, 2848 (~923).

920 HUGH S. TAYLOR per cent of K;O, while the partially active sample initially tried contained very likely 3 per cent. or more, as has been shown by other investigators (English). These samples were otherwise quite alike as to physical structure. The commercial sample after partial reduction was able to take up enough oxygen at an elevated temperature to restore its activity for a time. ~ hen it was more completely freed from potash it was able to take up oxygen fast enough to become completely catalytic at lower temperatures. This points to a mechanism of alternate reduction and oxidation of the catalyst. The efficiency is quite evidently dependent on the nature of the active surface as well as on its extent. A sample of preparation No. 4, for example, which had been ignited too strongly, on evaporation became quite dense, resembling the natural product and was entirely inactive, although alkali-free. This meant a packing and possibly a total change of structure of the material, and a reduction and probably alternation of the nature of the surface but without destruction of the porous structure. The "promoted", Hopcalite, sample is active, although it still may contain ~.74 per cent. of K20. To state with certainty the effect of the cupric oxide on the mixture it would be necessary to have data on a sample of manganese dioxide containing this amount of alkali. If the alkali is all associated with the manganese dioxide, as.the Hopcalite is 60 per cent. of manganese dioxide, the latter actually contains Ho per cent. Of K20. As a sample of manganese dioxide containing this much impurity would hardly be completely catalytic alone it seems that the cupric oxide does show promoter action. It is still possible that it cuts down the adsorbed alkali or affects the way it is held so that its poisonous effect is annulled. On the other hand, both cupric and man.ganous ions are catalysts in other oxida- tion processes. These adsorbed ions may, therefore, act as oxygen carriers to the carbon monoxide. The activity seems to be intimately connected with the ability and rapidity with which the substances can take up oxygen, which may be caused by the rapid shifting of electrons in manganese atoms, so the poison or promoter may affect the stray field or the atomic or molecular con- figuration of the catalyst itself. "Attention is next called to the ~anganese-oxygen ratios it) the samples. Here as in the previous investigations the loss of oxygen by manganese dioxide is noticed even at room temperatures and in a wet sample, indicating a disso- ciation pressure of oxygen in the pure manganese dioxide greater than the partial pressure of the oxygen in the atmosphere. English finds that these oxides behave as solid solutions, the oxygen pressure varying with the com- position of the mixture; this is similar to the conclusions of Sosman and Hostetteri in the case of the oxides of iron. The action of promoters and poisons may be due to their presence as constituents of such solutions. The fact that the mixtures lose oxygen at room temperature shows that they have a dissociation pressure greater than the oxygen in the air, and the activity be- cause of this is greatly increased by the fineness of division of the particles. ~ J. Am. Chem. Soc., 3S, So7 (~9~6).

THIRD REPORT OF THE. COMMITTEE ON CONTACT CATALYSIS 92 I The molecules are at a point where electron changes occur with great rapidity, and oxygen evaporates and condenses as readily as molecules do in the case of a liquid at its boiling point. The rapidity of oxidation of carbon monoxide (the time of contact is of the order of o.o~ second, comparable to that in the oxidation of ammonia) shows that the monoxide is Hot held very tightly as such. If it were, it would be its own poison. Failure to effect Resorption of carbon monoxide as such from manganese dioxide points to a rapid re-arrangement and reaction. That carbon monoxide may be adsorbed is Dowry by the experiments with the less active cupric oxide. The course of the reaction would be, then, adsorption and simultaneous oxidation or its adsorption by the catalyst; Resorption of the carbon dioxide or its adsorption by capillary condensation in case the catalyst were not already saturated; and finally reoxidation of the catalyst. The carbon dioxide is inert and chemically inactive and therefore does not poison the catalyst except hymechanically covering the surface and preventing contact of the reactant with the catalyst surface. Bentoni attacked the problem by studying adsorptions of carbon monoxide (and hydrogen.) by various oxide catalysts and mixed oxides. "It will be observed that the order of adsorption of the different gases is the same on each of the oxides. Carbon dioxide is most extensively adsorbed, carbon monoxide is next, followed by nitrogen, then oxygen, while hydrogen is least adsorbed. The order of decreasing boiling points is carbon dixoide, oxygen, carbon monoxide, nitrogen, hydrogen. The corresponding order for melting points is carbon dioxide, carbon monoxide, nitrogen, oxygen, hydro- gen; in other words, the same as the order of adsorption. This relation holds because the adsorptions are largely of the secondary valence type. A glance at the tables will show that for active oxides, the adsorption of carbon monox- ide is abnormally large, and this abnormality increases at higher temperatures. Thus at o° and above, carbon monoxide is held on active oxides mainly by primary adsorption, while at low temperatures the adsorption is largely secondary. "Although the effective surface areas of these oxides are unknown, so that it is not possible to compare them in terms of adsorptions per unit area, yet this difficulty may be overcome to some extent by using ratios of the volumes of different gases adsorbed by each oxide. "The ratio of carbon monoxide adsorbed at-79° to carbon dioxide ad- sorbed at o° is nearly the same for each adsorbent This suggests that both of these cases involve mainly secondary adsorption. For acidic oxides either the adsorption of oxygen is abnormally great, or that of carbon dioxide is abnormally small. Ordinary chemical considerations suggest that the latter alternative is the correct one. Obviously, however, the adsorptions in these cases are principally of the secondary valence type. The deviations from complete uniformity could perhaps be attributed merely to quantitative, ~ J. Am. Chem. Soc., 45, 887, 900 (~923). .:

922 HUGH S. TAYLOR rather than qualitative differences in the forces involved, yet there is no reason why certain of these oxides should not adsorb carbon dioxide or oxygen to some extent by primary valence forces. The large differences in the ratios of carbon monoxide at o° to carbon dioxide at o°, or of the monoxide at o° to oxygen at o°, show the specific nature of earshot monoxide adsorption at this temperature. If the assumption be made that, with adsorption by silica, primary valence forces do not enter the process appreciably, these ratios furnish a means of distinguishing quanti- tatively between the primary arid secondary adsorptions. On this basis the seeo~dary adsorptions of carbon monoxide at o° should in all eases be ~.77 times as great as the oxygen adsorptions, or o.o8 times as great as the carbon dioxide adsorptions. In Table I are given the secondary carbon monoxide adsorptions at o°, calculated in this way, together with the observed total adsorption. The last two rows contain the primary adsorptions, obtained by subtracting the secondary adsorptions from the total. TABLE I Primary and Secondary Adsorption of Carbon Monoxide Co203 Hopca- CuOIII MnO2 Fend ~0 lite Total CO at o ? 4.42 I .66 I . 90 I .62 o. os3 2.662 o.8s 0.0ss (2 . 662) I .42 o. o30 (2.662j 0.77 o. Go (o. o) 0.20 0.023 (o. a) Secondary cale. 02 at o° from CO2 at o° Primary, O2 at o° from CO2 at o° SiO2 0.47 o 44 0.82 o.4g ? 3 98 ? 3 93 0.23 0.60 0.36 o.6g I .43 I . 30 ~ .30 I. 2~ The two methods of calculation do not give identical results because, as already mentioned, the adsorptions of carbon dioxide and of oxygen by certain of these oxides earshot be regarded as purely secondary. The two methods do, however, place the oxides ire the same order with respect to the primary adsorption of carbon monoxide. Similar results are obtained for hydrogen, but since the adsorptions of this gas are very small, the relative precision of the measurements is much less than with carbon monoxide. It should be noted that these calculations are quite independent of any assllmp- tions with regard to the relative effective surface areas. It has, however, beer tacitly assumed that a large primary adsorption has no effect on the secondary capacity. If, as seems likely, this is not strictly true, all the second- ary adsorptions in Table I should be diminished, and the primary adsorptions therefore ir~ereased, by a certain small fraction of the calculated primary ad- sorptions. Obviously this correction could not alter the order of the oxides with respect to primary adsorption. The order of chemical reactivity of these oxides toward hydrogen and carbon monoxides is the same as the order ire which they are listed in Table I arid, therefore, the same as that of the primary adsorption, with the exception of manganese dioxide and cupric oxide, which are reversed. This parallelism ~ As determined in these experiments from the slope of the volume-time curves previously described. Cf. Wright and Luff: J. Chem. Soc., 33, I, so4 (~878).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 923 suggests that primary adsorption is an intermediate stage in the reduction of these oxides, at least al; comparatively low temperatures. In other words' carbon monoxide, on coming in contact with a readily reducible oxide, is almost instantly adsorbed by primary valence forces, forming a surface com- plex which can decompose either into the original substances or into the redue- tion products, depending or the conditions. At higher temperatures the rate of decomposition of the surface complex into the reaction products is extreme- ly rapid but at eomparat~vely low temperatures it becomes so slow that at arty instant a large fraction of the surface is covered with this adsorbed layer of carbon monoxide molecules. At still lower temperatures carbon monoxide is adsorbed less and less by primary vaTenee, arid more and more by secondary. The latter, however, is not a preliminary stage ire the reduction, except in so far as a primary valence union results from the secondary type by a shift of electrons. Relation between Extent of Adsorption and Catalytic Activity The catalytic behavior of these oxides in the combination of carbon ~onox- ide and oxygen has been investigated) by the Chemical Warfare Service, and also to some extent by the Munitions Inventions Department in England The order of catalytic activity was found to be FIopcalite, cobalt sesqui-oxide, cupric oxide, manganese dioxide, ferric oxide. Vanadium pentoxide was not investigated and therefore cannot be placed with certainty, but it is known that silica corpses at the end of the list. All the oxides whose adsorptive eapaci~ ties were measured were prepared and dried by the same processes as those used for the samples whose catalytic activity had beer, determined, except in the ease of Hopcalite. Hopealite similar to that used for the adsorption experiments was founds to be less than loo per cent efficient at temperatures below 4oO, so that this mixture must be put in second place, after cobalt sesquioxide, in the activity series. For eon~renienee of comparison, these facts are collected in Table II, together with the results of the adsorption measure- ment~s. ID the table the properties in question decrease from left to right. TABLE TI Comparison of Catalytic Activity and Adsorption Catalytic activity Co203, Hopcalite, CuO, MnO2, Fe203, V205~?), SiO~ Secondary adsorption sick, Fe203, Mom, Co203, Hopcalite, CuO, V2Q5 Primary adsorption of CO C'o203, Hopcalite CuO, MOON, Fe,O~, V0OD, SiO~. The most obvious conclusion to be drawn from Table II is that no connec- tion whatever exists between the extent of secondary adsorption and catalytic activity for carbon monoxide oxidation. ~ the primary adsorption of carbon monoxide, however, is in exactly the same order as the catalytic activity. ~ Rideal and Taylor: Analyst, 44, 89 (~9~9), Rideal, J. Chem. Soc., 115, 99 ~ (~9~9) Lamb, Bray and Frazer: J. Ind. Eng. Chem., 12, 2~3 (~920); Merrill and Scalione: J. Am. Chem. Soc., 43, ~98~ (~9~). ~Bodenstein and Ohlmer: Z. physik. Chem., 53, ~66 (loos). This statement is; supported by new experiments with precipitated silica. 3 In an experiment by H. S. Taylor.

924 HUGH S. TAYLOR This means that if the total adsorptions as measured are compared with the catalytic activity, no relation will appear, because the adsorption consists in general of two different phenomena, only one of which has a bearing on th activity. The powerful force fields at the surface of silica, indicated by its high meltir~g point, produce a comparatively high adsorption of all gases, but it is a secondary valence adsornt,ion and non~oll~nt.l~r 1~ Blur t.^ ~xrmol: i; any, catalytic effects. ~. . . 1 ~ ~ ~ J ~ "It van ', w~' 11 Charcoal, probably the best adsorbent known, cata- ~yzes few reasons, because the adsorptions in question are largely secondary. Charcoal does catalyze the chlorination of natural gas as well as a number of oxidation reactions, but the adsorption of oxygen certainly is of the primary vale rce type, as is probably also that of chlorine. Secondary adsorption ap- pears to produce, at the most, only comparatively slight activation of the adsorbed molecules. Concerning the actual chemical con position of oxide oxidation catalysts Weiss, Downs and Burnsi make an interesting contribution. They show that in presence of benzene-air mixtures of definite concentration at a given tem- perature the catalyst is, in reality, a definite ratio of two oxides, ~205 and Vim. At 400°C. with ~4 parts by weight of air to ~ part by weight of benzene the catalyst after use was 94.3 per cent V205 and 5.7 per cent ~204 while before use it was a mixture of 60 per cent V20~ and 4o per cent. V,O`. Then the benzene concentration was increased the percentage of ~200 in the used catalyst fell. With 2.2 parts of air to one of benzene the percentage had fallen to 9.~ per Gent. STROP. This adjustment of the oxide ratio to gas concentrations suggests strongly that the mechanism of the catalysis involves an oscillation between ~205 and V204. At temperatures above 400°C. with any given gas concentration: the proportion of ~ToO5 will progressively decrease. The authors have found that complete combustion also increases at the expel se of the partial oxidation product. The opposite is true of the lower temperature range. Hence, the authors conclude that the proportion of complete combus- tion is not dependent on the ratio of ~205 to V204 but upon some other factor, such as the activation of the reacting substances. Dunn and Rideal2 studying the oxidation of nickel sulphide by gaseous oxygen in aqueous solutions show that the process is a heterogeneous sur- face reaction occurring in stages with the intermediate production of basic salts. The oxidation is markedly decelerated by soluble vanadium compounds. The catalytic effect is ascribed to colloidal V(OH)3 and is greatest in weakly acid solutions. Adsorption and Catalysis The general conclusions of the work at Princeton have been summarized in a Gommunicatior~3 to the Colloid Symposium from which the following extracts are quoted. Adsorption is a condition precedent to catalytic change. - v ~ The data ob- tained by Taylor and Burns on hydrogenation catalysts showed marked adsorption of gases which take parl in hydrogenation processes. Low ad Ind. Eng. Chem., 15, 965 (~923) . ~ J. Chem. Soc., 123, ~42 (~92~). 3 Colloid Symposium Monograph, p. for et seq., Madison (~9~3).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 92 ~ sorptive capacities were found with relatively inert catalysts. Pease studied this relationship in detail with ethylene and hydrogen on copper showing that high catalytic activity was paralleled by high adsorptive eapaeity for both gases. Pease further showed that by suppressing the adsorption of hydrogen by partially poisoning the copper catalyst with mercury the cata- lytie activity was likewise suppressed. Adsorption of both reactants is th~re- fore a eon dition precedent to efficient catalysis in this ease. Benton showed marked adsorption of carbon monoxide and, to a lesser degree, oxygen by oxide catalysts capable of effecting the combination of these gases. Dough- erty and Taylor demonstrated the adsorption of benzene vapors by nicl~el. Taylor, Benton and Dewt have measured ammonia adsorption on a variety of metals which eatalyse the decomposition of ammonia. Taylor and Beebe2 have showr~ that hydrogen chloride is adsorbed by the copper chloride catalyst of the Deacon chlorine process. The Form of the Catalyst and Adsorption: The extent of adsorption per unit weight of catalyst is determined by the method of preparation, distribu- tion or inert supports or by subsequent treatment of the surface by eataly-st poisons or by heat treatment. Variation in adsorptive capacity with variation in the methods of prepara- tion, may be illustrated from the work on copper, on nickel and on an oxide such as cupric oxide. These results are strikingly displayed in the following tables. Temperature of Reduction of CuO. ~ ature of CuO Adsorptions on Copper Time re- quired for ~ eduction Few hours 2so°C. Ignited nitrate 200°C. Kahlbaum's granules 3o-4o furs. iso°C. Kahlbaum's 4 days ~5 5 Adsorption per loo g. Cu at C)°C and 760mm H2 C'H4 0.2 2.85 3.o 8.o OT'se~7ers T and Bu p T alla D Adsorptions on Nickel Temperature Nature Time required Adsorption H: per Observers of Reduction of NiO for Reduction No g. at 25O C. Of NiO. and 760 mm. 300°C F,x nitrate ~ hours 47 ecs. T and Bu 300°C Ex nitrate ? to ecs. G and T 300°C Ex nitrate 2 days To ees. T and Be This catalyst probably more finely divided than the first two. Adsorptions on CuO Nature of CuO Adsorption per ~ Go g. CuO at ~ 5°, 7 60 rum. CO2 0 ~CO Observer Strong ignition of Cu o.ols o.oos o.o~: Benton  Calcination of Nitrate a. ~32 o.oo o. ~80 Benton Pptn of hydroxide 36.2 to C) Go (o°C) 13.3 (o°C) Benton ~ Unpublished work. 2 J. ~m. Chem. Soc., 46, 45 (~924).

926 lIL-GH S. TAYLOR The effect of a catalyst support on the adsorptive capacity per unit weight of catalyst is well illustrated by the work of Gauger and Taylor with nickel from the calcified nitrate and with nickel spread on a diaton~ite bock. H. adsorbed per g. 15-i at Ho mm. arid Cataix-st 8~. so Unsupported Ni o. 69 a. 63 Hi on diatomi'Ge 25O 75° ~ 84O 200° 2 ~ So 250° 0.53 o 84 5 2 5 ~4 /3 The best quantitative data on the effect of poisons on catalyst adsorption obtained in the Princeton world are those obtained by Pease on copper. Ad- sorptions of hydrogen and ethylene on loo g. Cu were made before and after the catalyst was poisoned with mercury, the quantity of poison being estimated at roomy. Adsorption at 0°C., and 380 Ann. He C2H4 Before poisoning3 25 CC. 8 55 ~C. After poisoningo. IS CC. 6.70 cc. The striking disparity in the influence of the poison on the adsorptive capacities of the two gases is worthy of study. The hydrogen adsorption is reduced to less than 5 per cent. of its initial value. The ethylene adsorption, on the other hand, is still approximately 80 per cent. of its ir)itial value. At the present time, we are inclined, taking these data in con junction with others on the effect of heat to be presented below, to attribute this phenomenon to differing capacities of surface atoms to adsorb hydrogen and ethylene. The mercury vapor, on this hypothesis, would be preferentially adsorbed or those portions of the surface which have hydroge~-adsorbing capacity. Heat treatment of an active catalyst preparation is now our standard method of preparing catalysts with controlled adsorptive capacity or catalytic activity. From a variety of experiments, we may choose the following as indicative of the effect produced by heat treatment. Catalyst Heat treatment Adsorption at Observer 0° and 760 mm. H. C,H4 o g. C. Active A. Active Cu. No heat beyond reduction of oxide at 200° C. A. heated to 4so°C. for I . 5 hours Obtained by reduction of oxide at 300°C. C. heated at 400°C. for 4 hours. ~ 6 Beebe 3.70 cc. 8.45 cc. Pease ~ ~5 6.85 Pease 3 5 Beebe The same abnormal depreciation of the hydrogen adsorption on copper is to be noted here as in the poisoning experiments. This evidence we would interpret thus: A smaller fraction of the surface is capable of adsorbing hy- drogen than ethylene. The greater adsorptive force required by surface atoms in order to hold hydrogen is, in our view, to be regarded as possessed by those

. THTRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 9~7 atoms in the surface which have a greater degree of freedom from the normal crystal lattice of the solid catalyst. These atoms haste a lesser fraction of their electron shells surrounded with neigh bouring copper atoms. They therefore possess a greater surface energy. They would also possess a higher vapor pressure. With the moderate heat treatment accorded to the ea~alyst in the above mentioned eases these atoms distil to positions of lesser surface energy more readily than do atoms of less freedom in the solid lattice. It is these atoms of high surface energy which will be most affected by heat treatment; they should be the preferred positions of attachment of catalyst poisons. The Specificity of Ccrtalytic Adsorption:-Freundlich points outi that "since in adsorption by ehareoal, the physical characteristics of the adsorbed gas are of far snore importance than the specific effect between gas and adsorbent, it is not remarkable that also with adsorption by different absorbents the in- fluence of the special properties of the adsorbent is strongly suppressed (stark zuruektritt). It can be said with a certain approximation that oftentimes gases are adsorbed, independently of the nature of the adsorbent, in the order of their compressibilities." This thesis is entirely inapplicable to eatalytie adsorption. The ratio 0~/~'! for the adsorption of two gases by absorbents A, B. C, etc., which on the basis of Freundlieh.'s statement would be approximately constant for each adsorbent, A, B. C, etc., may vary quite widely for catalytic absorbents. The large differences in the ratio of adsorption of carbon monoxide at o°C. to carbon dioxide at o°C. obtained by Benton show the specific nature of carbon monoxide adsorption at this temperature for a variety of oxide cata- lysts. Hopealite CuO MnO~ Fe203 V2Os SiO2 Deco ICON o 72 0.37 0.22 0.09 o. ~4 o.o8 The same ratio at 2s°C. for a few metallic catalysts is obtainable from Burns' measurements Cu. Co. Fe Pd Pt Black taco cocos ~o.o 3.6 ~.8 ~88 ~o.6 It is very evident, since this ratio varies from on to 3oo, that the Freund- lich relation is entirely untenable for such eases as we are dealing with here. It has only a very circumscribed applicability, namely, to chemically inert absorbents and easily liquefiable gases. A most striking ease of the specific behavior of catalytic nickel is to be found in Freundlieh's book (p. 203) in his discussion of some unpublished work by Ziseh on the decomposition of nickel earbonyl at nickel surfaces. As Freundlich points out one might expect, on the basis of the higher critical temperature of nickel earbonyl as compared with carbon monoxide, a much higher adsorption. Actually, carbon monoxide even in minute quantities, exerts a powerful retarding action on the deeompo- sition, indicating marked preferential adsorption. Our present knowledge with respect to the structure of nickel earbonyl and its stable configuration on - ~ "Kapillarchemie." ~nd Edition, p. ~78 (~9~).

928 HUGH S. TAYLOR the basis of the Lewis-Langmuir theory of structure immediately suggests the chemical reasons for this specificity of adsorption; unexplainable on the basis of physical eharaeteristies. Other striking variations in ratio of adsorbed gases are to be found in the records of the Princeton work. Consideration of the preceding section on the inRuenee of catalyst poisons and of heat treatment on adsorptive capacity will show furthermore that the ratio of adsorption of gases by a single catalyst is also variable with variation in the preparation and of treatment of the catalyst. The rule as to nonspeeifieity of adsorber~ts must be discarded when cognisanee is taken of the data on eata- lytie adsorhents. Specificity of Adsorption and Specificity of Catalytic Activity:-The irlflu- enee of specific adsorption in determining specific catalytic activity is best demonstrated by work dealing with the preferential catalytic combustion of carbon monoxide admixed with hydrogen. As is well known, metallic oxides may be used to catalyse the combination of carbon monoxide and oxygen present in equivalent concentrations in a large excess of hydrogen. The mechanism of this preferential oxidation is at once apparent from the `adsorption ratio of the two gases at atmospheric pressure on various oxides at-7 g°C., as determined by Benton . Oxide= Hopcalite MnO2 CuO Co203 Fe203 V20~ SiO2 occO 33 loo 34 ~9 35 ~7 28 of co2 For exac+ con~parisor~ with preferential combustion data adsorptions at low partial pressures of carbon monoxide should be compared with those of hydrogen at approximately atmospheric pressure. The results cited, however, show marked preferential adsorption of carbon monoxide. With metals the preferential nature of the combustion process is less pronounced. With nickel and platinum the hydrogen is freely consumed; with copper a fair preferential eombust~ion may be attained. Note the following data on adsorption ratios of the two gases at various temperatures and atmospheric pressure and contrast them with the oxide data. taco Ni Pt. Black Cu . ~E2 O. 87 (184 ) 3.3 (IOO ) I2 The data cited are also of interest in connection with the problem of specificity of adsorbent discussed in the preceding section. Variation of Adsorption with Pressure and the Heat of Adsorption: As is . well known, the variation of adsorption with pressure on absorbents such as charcoal is approximately given by the Freundlieh equation or=k C I!n where c~=amount adsorbed, k and n are constants the latter being always equal to or greater than unity. The data on the variation of adsorption with gas pressure with metallic catalysts as absorbents are few; some of these, however, show striking eharae- teristies. Gauger and Taylor's data on the adsorption isotherms of hydrogen

THIRD REPORT OF THE COMMITTEE ON CONTACl' CATALYSIS 929 on nickel are the most completely studied thus far. The curves obtained at a variety of temperatures 2s-30~°C., show the characteristic shape of normal adsorption isotherms so far as absence of discontinuities indicative of eom- pound formation are concerned; they show, however, this distinction that at a certain pressure at each temperature, a definite saturation capacity of the surface is apparently reached. This saturation capacity is reached at very low partial pressures, 4o mm. at 2s°C., and approximately 250 mm. at 3os°C. Beyond these pressures, further ir~crease in gas pressure up to atmospheric pressure (i.e. 76~0 = Ig fold increase in pressure at 2s°C.) adds to the amount of gas adsorbed so little as to be within the error of measurement. The same observation' is true in the recent results of Pollardi, employing hydrogen, and, to a less extent, carbon monoxide on platinum. The amount of adsorbed hydrogen in this case does rot sensibly increase beyond a gas pressure of loo mm. Pease's data on the adsorption of hydrogen by copper show a similar if less pro'nouneed attainment of saturation capacity. The adsorption of hydro- gen at 380 mm. pressure was go per cent. of that at atmospheric pressure. Similar behavior with respect to carbon monoxide on copper is shown in some data obtained by Jones and Taylor on the adsorption isotherms of carbon monoxide and carbon dioxide on copper at o°C. and 80°C. Earlier work on absorbents of the charcoal type has not indicated the attainment of saturation capacity of the surface even at pressures well beyond atmos- pheric pressure. A further distinction is also noticeable. Gauger and Taylor's results show that the adsorptive capacity of hydrogen on nickel at satura- tion is, at 3os°C., as much as 60 per cent. of the saturation capacity at 2s°C. Some recent data obtained by I)ew on copper show adsorptions of hydrogen in the ratio of lo to 8.7 at o° end ~o° C. and atmospheric pressure. Con- trast this with the data concerning adsorption on charcoal. The adsorp- tion of carbon monoxide at 4oo mm. and 46°C. is only 8 per cent. off that at-78°C., this temperature interval being about the same as that obtain- ing in Dew's case and less than one-half of that recorded above with nickel and hydrogen. The adsorption of carbon dioxide on charcoal at ~so°C. and atmospheric pressure is less than 7 per cent. of that at-78°C. These strik- ing differences both in the pressures at which saturation is attained arid in the variation of adsorption with temperature are undoubtedly of funda- mental importance in the study of catalytic absorbents. Data on adsorption isotherms may be utilised to evaluate the heat of adsorption of gases on the adsorbent surface. Gauger and Taylor using the minimum pressures at which saturatior~ is reached at the several temperatures and substituting these ire the equation . T1 T: 4 57 T2-T P., log p obtained a value for, X, the heat of adsorption of Moo calories. This cal ~ J. Phys. Chem., 27, 365 (~923).

93o HUGH S. TAYLOR culation is in error since the equation should be applied) to the pressures Pi and P2 at which equal amounts of gas are adsorbed, or in other words, equal fractions of the surface are covered. The data of Gauger and Taylor do not lend themselves readily to such computations if accuracy is desired, as the pressures at which equal fractions of the surface are covered at different tem- peratures are small and consequently most liable to error. From the best available data however, calculated in the correct manner, a value for the isosteric heat of adsorption of ~sooo+3000 calories was obtained. Rideal and Thomas showed that the adsorptive capacity of three different samples of fuller's earth for methylene blue is DO criterion of its capacity to catalyse the decomposition of hydrogen peroxide. The adsorptive powers were ire the ratios of ~.54, 2.~8 and I. The catalytic actions were in the ratios ~.38, o.s8 and ~.5~. The iron content of the three earths is possibly the gov- err~ing factor in the catalysis. Adsorption and the Influence of Support Materials Palladium, spread on active charcoal, with the object of utilising the adsorptive capacity of the support material in addition to the catalytic activ- ity of the metal has been employed by.Foster and Prudes in a study of carbon monoxide decomposition to yield carbon and carbon dioxide at temperatures as low as ~oo°C. Hydrogen had no influence on the change. The reaction went erred with the smallest amounts of water vapor present which points to direct reaction 2CO = CO2+C With silica gel as support material this reaction was accompanied by the reac- tion CO+H20 = CO2+H, since hydrogen was present in the effluent gas. This work is in disagreement with previous claims of Orloff4 who stated that hydrogen and carbon monoxide yield ethylene in presence of nickel-palladium catalysts. Foster and Brude obtained no unsaturated compounds and state that it is safe to assume that ethylene has not as yet been produced by reduction of carbon monoxide. There should be some information forthcoming from American sources on this point. Rosenmund and Langer5 have shown that the nature of the support material is of importance in protecting the catalyst against poisons as well as in influencing the catalytic activity. With palladium catalysts on various supports the influence of arsenious oxide and carbon monoxide as poisons was studied, in the reduction of cinnamic acid. Kieselguhr-palladium catalysts showed the least activity and greatest sensitivitvio poisons. Blood charcoal gave the most active and most resistant preparations. In these two cases activity and resistance run parallel. Barium sulphate supports are more active than pumice; the latter are more active in presence of the poisons. ~ See Freundlich: "Kapillarchemie," p. ~82 (~92~). 2 J. Chem. Soc., 121, 2~9 (~92~). 3 Ber., 56, 2245 (~923) 4 Ber., 42, 893 (~909). 5 Ber., 56, 2262 (~923).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 93I The effect of the supports is evidently a function of the adsorptive capacity of the support for the poison. It acts in these cases as a purification agent in the catalyst system. Adsorption by support materials has proved to be of importance in the measurement of adsorption by contact catalysts spread on supports. Dr. R. A. Beebe has shown that asbestos suitable for use as support for platinum in con- tact mass catalysts adsorbs o.7g, o.~o and o.o4 cc. nitrogen and 53.0, ~.3 and 2.0 CC. sulphur dioxide per gram at o°, ~o° and 2~8°C respectively. Russell, in Princeton, has shown that pumice used as a support for nickel adsorbs ~. ce. nitrogen per gram at ~ ARC. This possibility has always to be looked for in adsorption studies. Heats of Adsorption Benton's paper previously cited indicates the existence of primary and secondary adsorptions the former, only, of which parallels the catalytic activ- ity. The thermal magnitudes accompanying such adsorptions should reveal if arty extensive change in the valence forces has occurred during such ad- sorption. It is for this reason that particular interest attaches to the measure- ments of heat of adsorption which are now being made. Forestii has measured the heat of adsorption of hydrogen on nickel and showed that the heat of adsorption, Qv, at atmospheric pressure was ~500+ too calories. Beebe and Tayrlor2 have shown by direct measurement that the heats of adsorption of hydrogen on active catalysts composed of nickel and of copper are respectively about ~500 and 9600 calories. The great disparity between these values and that of the heat of liquefaction of hydrogen, 450 calories, is the first striking feature of these results. The adsorption is very definitely not a simple condensation process. Taken in conjunction with the variation of adsorption with pressure, as elucidated by the work of Gauger and Taylor and of Pease, the heats of adsorption may be utilised to demon- strate the difficulty of formation of multi-molecula~ films of such gases. From the equation pi ~ T2-T. Loo = P' 4. 57 T2T~ calculatiox1 may be made of the pressure pi at which, at temperature T2, the same weight of gas may be adsorbed as is taken up by the adsorbent at To at a pressure pi, the heat of adsorption being, X. The data of Gauger and Taylor show that at 2s°C and 4o mm. pressure a given sample of nickel adsorbed 8.7 ccs. of hydrogen Utilising the directly observed value for \= ~3soo cars., we may now calculate with the aid of the above equation the pres- sures at which this quantity of gas will be~adsorbed at various higher tempera- tures. ~ Gazz. chim. ital., 53, 487 (~923). 2 J. Am. Chem. Soc., 46, 43 (~924)

932HUGH S. TAYLOR TemperatureCalculated Pressure at which 8.7 cc. IIc are adsorbed 80°C~ . 85 atm. BLOCgsatm. 218°C4~4 atm. 305OC3342 atm. The experimental measurements show, however, that, for example, at ~84OC., as much as 8.o ecs. of gas are already adsorbed at ~50 mm. pressure. It is therefore evident that a further increase in pressure from Comb. to 95 atmospheres only results in the further adsorption of o.7 ecs. This result is in entire agreement with that found experimentally, that the adsorption be- tween ~50 mm. and 760 mm. at BLOC., was, within the error of measurement, constant. The calculated variation should be 7= somewhat less than 95 o. oo8 ccs. Similar considerations hold to a more marked degree at the higher temperatures. At 3os°C and ~ atm. pressure the adsorption is already some 5.5 ees. The adsorption of an additional 3.c ces. would require a gas pressure of 3342 atmospheres. In a similar manner utilising the heat of adsorption of hydrogen on copper' \= 9600 cars., it may he calculated that the quantity adsorbed at o°C and ~ atmosphere would be adsorbed at ~o°C at ~6~ atmospheres. Now, actual test has shown that al ~ ~o°C and ~ aim., pressure the adsorption of hydrogen by an active copper is already 87 per cent. of that at o°C and ~ atmosphere. An increase in hydrogen pressure of ~6~ atmospheres would only result there- fore in an increase of ~3 ~/ Io° = 15 per cent. in the adsorbed gas. We regard the slight variation of adsorption with pressure after the initial strong adsorption at the lower partial pressures in the eases herein studied as the strongest evidence in favor of the Lax~gmnir theory of a unimolecular layer. There is evidently in these eases little or no tendency to build up several layers of adsorbed molecules on such surfaces. There is evident a similar inability to build up several layers of adsorbed gas in the case of carbon monoxide on copper as first observed by Jones and Taylori and recently more thoroughly investigated by Pease.2 The same is apparently true for the cases of hydrogen and carbon monoxide on platinum as the recent studies of Pollard3 show. In all these cases there is evident rapid saturation of the surface at low partial pressures and then subsequent slight increase of adsorption with pressure. The available data on heats of adsorp- tion for these latter eases confirm this view. Mond, Ramsay and Shields4 value for the heat of adsorption of hydrogen is ~3760 cars. Langmuir5 has calculated by an indirect method that the heat of adsorption of carbon monox ~ Colloid Symposium Monograph, p. ~o8 (~923). 9 J. Am. Chem. Soc., 45, 2296 (~923). 3 J. Phys. Chem., 27, 365 (~923). Z. physik Chem., 25, 657 ( ~ 898) . 5 Trans. Faraday Soc., 17, 64~ (~92 ~ ).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 933 ide on platinum is 3 ~,ooo cars. The heats of adsorption of oxygen and chlorine on charcoal are known to be high), and charcoal is an oxidation and halogena- tion catalyst. One other feature of these data on heats of adsorption other than the actual magnitudes involved is of importance. Beebe's measurements of the variation of heat of adsorption with pressure, over the range 0-760 mm., show the magnitude to be constant. This is in sharp contrast to the results pre- viously obtained with absorbents of the charcoal type. In this latter case the heat of adsorption decreases steadily with increasing, pressure, the final values approaching those of the heat of liquefaction of the adsorbed gas. This has been interpreted as showing that the adsorption is really a liquefaction phe- nomenon, the excess heat, over and above that of liquefaction, being accounted for by the high compression supposed to obtain in the first liquid layers of adsorbed gas, such pressure diminishing as more and restore layers form, until, finally, the straight heat of condensation is obtained. The constancy of the values obtained in our studies and their wide divergence from the heat of, liquefaction (igloo and 9600 calories as compared with 4so calories heat of liquefaction) tends to indicate trait in the cases we have studied no multi- molecular layers form. It is interesting; to note that, from Pease's data on ethylene, a gas whose isotherm at o°C. on-copper is much more reminiscent of isotherms on char- coal, the value for the heat of adsorption deduced from the isosteres at o°C., and 20°C. (5.5 ccs. adsorbed at 480 and 760 mm. respectively) may be cal- eulated to be 3750 calories, which is exactly what would be deduced from Trouton's rule. From the isosteres at lower pressures, higher heats of adsorp- tion are calculable. From the isosteres for 3.85 ccs. (200 and 3oo mm. re- speetively) the calculated value is Moo calories. This is in accord with pre- vious data on non-specific or capillary adsorption. Much of our evidence tends to show that ethylene may be adsorbed in capillaries in some of our cop- per samples. With other samples, notably one obtained in Princeton recently where the adsorption of ethylene at one atmosphere pressure was only one- half that of the hydrogen adsorptions, capillary adsorption seems to be less evident. We incline to the belief that the high initial values of heats of ad- sorption should be ascribed to the heat of the adsorption complex, acisorben~- adsorbate, for example, Cu-C,H~; with capillary liquefaction, the heat meas- ured becomes more and more that of the liquefaction, C2H4-C2H4. Under such circumstances the variation of the heat of adsorption with pressure would provide a definite criterion of the formation of ~nulti-molecular ad- sorbed gas films. Interface Phenomena Further evidence of interface phenomenal in chemical reactions is revealed by reaction velocity curves autocatalytic in nature. Sieverts and Theberath4 studied the dissociation of silver permangan ate and obtained such a reaction ~ Unpublished world. M.I.T. 2 See also Pease: loo cit. . 3 Cf. Second Report, J. Phys. Chem., 27, 827 (~923). 4 Z. physik. Chem., 100, 463 (~922).

934 HUGH S. TAYLOR process. Small amounts of impurities inere~sed the velocity of decomposition. Hinshelwoodi thereupon promptly called Sieverts attention to his own work upon both inorganic compounds and explosives such as tetryl in which an auto-acGelerated change oceuls. It is not quite clear whether Hinshelwood accepts the interface theory as accounting for his experimental results. "Now since the rate of reaction in the solid state is only about 50 to loo times less than that in the liquid (in the ease of tet,~-1 at ~zo°C) we must assume that even after the most careful purification the tetryl still contains traces of im- purity which either give rise lo liquid or exert some catalytic effect." Is it not possible that these "traces of impurity" may simply be weak spots in the tetryl crystal lattice, from which reaction starts and spreads outwards. Otto and Fry2 thought that their results showed the decomposition of potassium chlorate to be ~ unimolecular process. Anyone can see that they are in reality a beautiful example of an autoaccelerated process. Iron oxide promotes the decomposition. Recently they have shown that potassium chloride does the same thing.3 The presumptionis the refore stror~g that the process is an interface phenomenon. Neville has added4 to this reaction an interesting case of promoter action. Impure pyrolusite was more effective than pure manganese dioxide. The pyrolusite contained 8 per cent. iron oxide. A mixture of 8 per cent. iron oxide anti 92 per cent. pure manganese dioxide had the same efficiency as the impure pyrolusite. The action of the mixture was more than additive of the effects of the two oxides separately; hence the promoter action, elucidation of the mechanism of which was not achieved. It will probably be quite complex. Jones and Taylor have shown that the low- temperature reduction of copper oxide by carbon monoxide is art interface phenomenon, inhibited by carbon dioxide and by oxygen. The catalysis of the carbon monoxide and oxygen reaction by copper oxide appears to be alternate oxidation and reduc- tion. On copper the process is oxidation of adsorbed carbon ~nonoxide. In this factor it differs markedly from the catalysis of hydrogen and oxygen on copper, the mechanism of which appears to be alternate oxidation and reduc- tion. Gas-Liquid Reaction Velocities and Catalysis Norrish and Ridea16 have recently studied the reaction, H2+S (lid.) HeS. They emphasise that the solubility of hydrogen sulphide in liquid sulphur may have vitiated the earlier work performed by means of a static method, since the gas would be liberated on solidiDeation of the sulphur and would therefore be added to the equilibrium quantity of hydrogen sulphide measured after cooling the reaction bulbs. Furthermore, N-orrish and Rideal point t Phil. Mag.,40,:6g (~920); Proc. Roy. Soc., 99A, 203 (~92~); J. Chem. Soc., 118, 72i (~92~). 2 J. Am. Chem. Soc.,45,r~34 (~923) 3 J. Am. Chem. Soc., 46, 26g (~g24j. 4 J. Am. Chem. Soc. 45,2330 (~923). 5 J. Phys. Chem.,2i, 623 (~923). 6 J. Chem. Soc., 123, 6g6 (~922).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 935 out, it is uncertain whether in the earlier work the walls of the vessel acted catalytically; also, the abnormal temperature coefficient obtained by Boden- stein, ~.34, between 234° and 283OC and ~.77 between 3~o° and 3s6°C does not seem to be in harmony with Bodenstein's conclusion that the reaction is homogeneous and confined to the gas phase. Norrish and Rideal, by employing a dynamic method, have found it possible to show that the combination of hydrogen and sulphur takes place by way of two reaotiolls, a gaseous and a surface reaction, the former being predominant, under the conditions of their experiments at 28s°C., and up- wards, the latter being the more important below this temperature. They showed also that the temperature coefficients of the two separate reactions were constant but widely different in value. These conclusions were reached by an analysis of the reaction velocity measurements.. These revealed that the logarithms of the total reaction velocity plotted against temperature did not yield straight lines. The curves obtained confirrred Boden.stein's result of an increasing; temperature coeffi- cient with increasing temperature. From measurements at different tempera- tures with two different hydrogen pressures, pl. and pi, it was found that a plot of the logarithms of the differences of correspon.dir~g velocities for the two pressures against temperature gave a straight line. This fact led the authors to the conclusion that a surface reaction and a gas reaction were proceeding concurrently and that the former, assumed independent of the gas pressure and therefore constant, disappeared on taking the difference of the correspon.d- ing velocities. In other words, the straight line obtained as stated is in reality the graph of the gas reaction velocity occurring at pressure pup. Assuming; that the gas reaction velocity was proportional to the pressure and the surface reaction independent of the gas pressure, the observed curve for total reaction was resolved into two straight line curves of logarithm of velocity plotted against temperature for the two reactions taken separately. :From the slope of these lines the temperature coefficients were obtained. That for the surface reaction was found to Le ~.48, that for the gaseous reaction was 2.26 which after correcting for the variation of the vapor pressure of sulphur with the temperature reduced to :.~9. Corresponding to these coefficients, by applying the Arrhenius equatiorl, d log V dT = A RT2 where V is the reaction velocity, the values of A, the heats of activation of the gaseous and surface reactions, were found to be respectively 52400 and 26200 calories. Norrish and Rideal call especial attention to the fact that these are in the ratio 2 :~. By varying the size of the reaction vessel they showed also that the surface reaction was directly proportional to the internal surface area of the cresses and independent of the quantity of sulphur in the bulb. The respective reac- tion equations would therefore be: V (gas) = kit X CE2 X CS (surface) = k2 X Surface area,

936 HUGH S. TAYLOR where Cut refers to the hydrogen concentration, Cs that of the ~nonatomi~ sulphur. Norrish and Rideali showed also that oxygen functions catalytically in the union of hydrogen and liquid sulphur. The phenomenon is in reality quite complex. With rise of temperature and increase of oxygen concen- tration beyond lo per cent. at a temperature of ~6s°C., and beyond 7 per cent. at :S:°C., the catalytic action becomes a poisoning action. The observed effects were separated into a strong poisoning effect in the gaseous reaction between hydrogen and sulphur at all temperatures and a ca-talytie effect on the surface reaction which only becomes observable at the lower temperatures (265 and 28s°C.) where the surface reaction is of greater relative importance. This surface catalytic action rises to a maximum with increase of oxygen con- centration in the hydrogen and then falls away again, finally becoming a poisoning action for concentrations of oxygen beyond lo per cent. Simultane- ously, sulphur dioxide is formed at a rate directly proportional to the oxygen concentration. From the known velocities of the several reactions occurring, it was deduced that the effects observed may be quantitatively explained by postu- la~ir)g a gradual preferential adsorption of oxygen by the sulphur surface, all the hydrogen being; displaced when the gaseous concentration of oxygen has exceeded lo per cent, and by ascribing to the oxygen a catalytic activity pro- portional to the number of molecules adsorbed per square centimetre of sur- face. From these assumptions Norrish and Rideal calculate the composition of the adsorbed gas films in equilibrium with a given atmosphere. It is evident from this work end 'that of Pease2, that experimental determinations of such adsorptions would be instructive. In a concluding section of the paper Norrish and Rideal consider the mechanism of troth gaseous and surface reactions. The thermal value found from the temperature coefficient of the gaseous reaction, the "critical incre- ment" of the radiation theory, 52400 cars., is in agreement with Budde's values for the heat of dissociation of S2 molecules into atoms, and thus corre- sponds to the energy required to sever two sulphur bonds. The critical incre- ment of the surface reaction similarly corresponds to the breaking of one sulphur bond and is equal to that required to sublime a molecule of Ss from the surface, which also involves the breaking of one hoed. The surface reac- tion is considered to take place in two stag;es: (~) adsorption of the molecule involving the breaking of one bond and (~) removal of the molecule of hydro- gen sulphide involving breaking of the second bond, the critical increment measured corresponding to the slower of the two processes. The authors also assume that the surface contains mainly So molecules of which a few are opened by the rupture of a linkage and thus polarised. The adsorption of the hydro- gen and the oxygen is assumed to occur at these positions. The catalytic effect of the oxygen is attributed to simultaneous adsorption of oxygen and 1 J. Chem. Soc., 123, I 689 (I923). 2 loc cit. 3 Z. anorg. Chem., 58, I69 (I9I2).

THIRD REPORT OF TTIE COMMITTEE ON CONTACT CATALYSIS 937 hydrogen at the two ends of the ruptured Ss molecule. The strong attractor of the oxygen for the sulphur at one end of the chain will cause a weakening of the force by which the sulphur is held at the other end and thus lower the critical increment of energy necessary to remove the end sulphur atom as hydrogen sulphide. Rideal and Norrishi show the union of liquid sulphur and oxygen to occur at the liquid surface and also on the walls of the containing vessel, and to proceed as well on the latter as on the former, pointing to the existence of a liquid film covering the whole surface of the vessel. The reaction is propor- tional to the oxygen pressure; the temperature coefficient is ~.63 composite of two reactions faith coefficients respectively ~.48 (A) and ~.77 (B). The reaction. A is independent. of pressure beyond o.4~ atm. The reaction 33 is proportional to the oxygen pressure as far as ~ atm. pressure. The critical increment of A is cs7so cars. (cf. preceding); that of B is 37450 cars. Two types of sulphur molecules are presumed to be present in the surface layer giving rise to the A and B reactions. Inhibition Phenomena Sieverts and Lueg2 have investigated the effect of various poisons on the rate of solution of metals in acids. Alkaloids such as nicotine, cocaine, cin- chonine were effective, naphthoquinolines, strychnine, brucine, narcotine and quinine were very effective. The extract consisting of the ether soluble basic constituents of crude anthracene was most effective. For slight amounts of poison, temperature increase reduces the inhibition; for large amounts it is without effect. No obvious connection between inhibition ant! increase in overvoltage could be found. The authors consider that adsorption of the poison on the metal surface accounts for the inhibition which is expressible by an empirical equation K = aCb where To is the uninhibited velocity, Kc that with poison concentration c; a and b are constants. Simultaneous Action of Catalysts and Radiation Rosenmund, Luxat and Tiedemann3 have investigated the influence of ultra-violet light on the reactivity of halogen ring compounds in presence and absence of catalysts. From the preparative standpoint an indication of their results may be gleaned from the results in the following case, the reaction between brom-benzol and sodium ico-amylate dissolved in isoamyl alcohol. 12 hours heating, with copper, without radiation 5 . 2~ Halogen as NaBr 12 hours heating, without copper, with radiation 34. 8C7o ~ 2 hours heating, with copper and radiation 76 . 9~O ,. .. .. ,. .. .. Sodium bromide and the corresponding ether were produced, the temperature being the boiling point of the solution. Comparison measurements of the velocity constants were made: (~) with copper alone; (2) with ultra-violet light alone; and (3) with both agencies, in the case of reaction between p brom-benzene sulphonic acid and potassium 1 J. Chem. Soc., 123, 3202 (Ions). 2 Z. anorg. Chem., 126, I93 (I923). 3 Per., 56, I9jO (I923).

938 HUGH S. TAYLOR hydroxide solution, under the chosen conditions and at as low a temperature as possible. They showed that (~) was extremely slow, (2) was much faster, and (3) was at least twice as great as would be calculated from addition of (~) axld Aid. The explanation of (2), the acceleration by light alone, is to be attributed to the action of the light in activating the halogen compound. For (3) several possible explanations can be given (a) Radiation activates the catalyst (b) The molecule activated by radiation offers more favar- able working conditions to the catalyst. This assumption is favored by Rosenmund and co-workers. They had previously shown a more powerful influence of copper in thermal reaction when the stability of the halogen in the molecule had been reduced somewhat. They investigated a number of compounds in which the stability of the halogen in the compound increased until they found in p chlor-benzoic acid a substance of which the chlorine is held so fast that under the given working conditions, (t= ~o4OC) copper alone was inactive. In this case the velocity when radiated was the same with or without copper. They therefore conclude that the radiation does not influence the copper. Schwarzi thought that Ror~tgen rays activated contact platinum in the contact process, and in the decomposition of hydrogen peroxide. More re- cently, Schwarz and Klingenfuss2 attribute the greater efficiency of the plati- num to a photolysis of the water present giving rise to ~ more active form of oxygen, probably in the form of a peroxide of platinum. The suggestion of Elli~ge~3 that the increased activity was to be attributed to the taking up of electrons by the metal from the radiation, thereby facili- tating oxidation processes, is rejected by Schwarz who points out that this should hold true equally well whether water were present or absent. The presence of water was shown to be necessary. Poisons The elucidation of the mechanism of the action of poisons may result in a further contribution to the problem of mechanism in contact catalysis itself. Some beginnings of great significance are- already evident. Arm- strong and Hilditch pointed out some Rears ados that "the nmollnt of trivia ~ ~ ~ ~ A: 1 ~ ~-- ~1 _ I _ 1 O ~ ~ . . ~ ula~erla1 necessary for total suppression of catalytic activity is far below that required for stoichiometrical combination with even the surface layers of the catalyst. The probability that an 'active catalyst' is merely an average term expressing a surface on which a number of patches of maximum activity occur (the greater part of the surface being of quite a low order of activity) offers a simple explanation of the discrepancy, selective adsorption of the catalyst poison at the relatively few points of maximum activity causing the disap- pearance of practically all catalytic effect." ~ , Ber., 55, Solo (~922). 2 Z. Elektrochem., 29, 470 (~923). 3 Z. physiol. Chem., 123, 257 (~922). 4 Faraday Society Symposium, Discussion 17, 670 (~922).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS 939 This point of view seems to be capable of experimental verification and the results of Pease, already cited floe. en') are the beginnings of quantitative exploration of this idea. Pease showed that an amount of mercury which reduced the adsorptive capacity of hydrogen to ~/zath of its value before poisoning and which reduced the adsorption of ethylene but little, brought about a reduction in velocity to Tooth of its value before poisoning. A little eonsJderation will show that this is iD good agreement with Armstrong and Hilditeh's conclusion. Pease is now extending this quantitative study to the determination of the action of carbon monoxide on the combination of ethylene and hydrogen in presence of copper. His results to date demonstrate that the carbon monoxide mole rules required to suppress the reaction, under the conditions of the experiment, are markedly less than the number of hydrogen molecules capable of being adsorbed by the copper eatalvst. This seems to be quantita- tive evidence that only a fraction even of the surface which is capable of ad- sorbing hydrogen is capable of accelerating the reaction. It will be noted that this idea is in agreement with conclusions from other investigations on the adsorption of catalytic agents, notably those of Benton previously discussed, on primary and secondary adsorptions of carbon monox- ide on oxide catalysts and the parallelism between the catalytic efficiency and primary adsorption data. The conclusion suggests that attention should be concentrated or adsorptions at low partial pressures and on the heats of ad- sorption at these pressures. This complicates the experimental problem involved but it seems to be a necessary eomplieation. Co-actions and the Mechanism of Reduction Prinsi teas eontril~uted a suggestive paper on the meeha~ism of reduction and of oxidation reactions where a third substance is present in addition to the essential reaetant.~. "Oxidation reactions. Lead peroxide and manganese peroxide are insolr~- ble in weak acids, but react easily if a third component is present which can co-operate in the attack on the peroxide, by reacting with the surplus of oxygen." Such a substance may be art aldehvde or any other easily oxidisahle compound. For the same reason an acid which, moreover, can fun etion as an aldehyde, like formic acid, dissolves these peroxides easily: other examples are acids with one or more OH groups, CO groups ete. On the same principle, a weak but easily oxidised acid can supplant a much stronger one if, under the eireumstanees the latter contains active oxy gen. Thus the nitrates, chlorates, etc., of the heavy metals are converted into formates by the action of formic acid: the simplest method of preparing nickel formate consists in adding a saturated solution of nickel nitrate to about 85 per cent formic acid, which is heated on a water bath. A violent reaction takes place with evolution of carbon dioxide and nitric oxide and the resulting nickel formate separates nearly quantitatively. ~ Rec. Trav. chim., 42, 473 (~923).

,940 HUGH S. TAYLOR Reduction reactions. The co-action of three components is of special importance in the reduction of organic substances with a metal and an acid, or with a ferro-, stanno- salt etc., which takes the place of the metal in the reduction. It is a well-known feet that metals which are practically insoluble in a certain acid, become markedly so if this acid eo-aets with an oxidising suL- stance, but in most cases the course of the reaction is uncertain. It; is often supposed, that the oxidising agent first forms an oxide and this reaction is succcecled by a reaction between oxide and acid. But from the fact that compounds like nitrobenzene act just as powerfully without being able to form an oxide, it is certain that the primary formation of an oxide is not a necessary phase in the reaction. Moreover, the action of an oxidising agent is often selective, which can be shown even with test tube experiments: e.g. silverfoil is only slowly soluble in a mixture of an inorganic acid and hydrogen peroxide, much quicker in a mixture of the same acids and potassium bichromate but it dissolves almost immediately in a mixture of potassium permanganate and even a weak organic aeicl, such as acetic acid. This is not caused solely by the instability of the oxidising agent, because neither the unstable hydrogen peroxide, nor the unstable perehromie acid have the same influence; it is, obviously, not caused therefore by oxygen in state naseendi. If this were true, hydrogen peroxide would react most powerfully. In order to study the eo-aetion in eases where the primary formation of a metal oxide was excluded, we chose some years agog, the co-action between a metal, an acid and reducible organic substances such as nitrobenzene and benzaldehyde. Nitrobenzene accelerates in some eases more than a thousandfold. In these eases it was necessary to expose the metal in the other flasks long after that in the nitrobenzene solution kind T(lt~llv v~.ni~h~H in order to. ~+. ~ `~7mirrh_ able loss. TO I _ · l 1 ~ ~ ~J . We % ~^ I_ ~ EVE a' ~-V U <a ~2 In some ~nstanees, benzaldebyde shows the same property, although in a lesser degree, whereas in other eases benzaldehyde causes a decided decrease in the velocity of reaction. This retardation, is probably due to adsorption either of the benzaldeLyde or of its reduction products whereby part of the surface of t.~e math h~om~ inactive. · ~ _ ~ ~ ~ ~ ~ ~ 7 ~ ~ ^~ ~ ~ If ~ ~ ~ 4 ~ ~ ~ ~ ~4 ~ ~ 11 ~11 1~, ~1 ~ ~111~ An acceleration of about the same order of magnitude takes place in a non-ionizing medium like parafin-oil or in one in which the ionization is very small. Consider the reaction between the metal and the acid. It is usually sup- posed that the metal forms positive metal-ions in the liquid, the hydrogen ions taking up the liberated electrons to form a hydrogen atom or mole rule. The difference ir, strength of the acids is usually expressed as the magni- tude of the ionization under comparable eireumstanees, but the cause of this difference must be sought in the chemical character of the anion. It is, there H. J. Prins: Akad. Wetensch. Amsterdam, 23, 9 (~92~).

THIRD REPORT OF THE COMMITTEE ON CONTACT CATALYSIS g4r fore, to be expected that the anion plays an important part in the reaction between an acid and a metal; their difference in behavior cannot be explained' by the concentration of the hydrogen ions, because the latter is, in its turn, a function of the acid radicle, which retains the electron and expels the hydrogen Ion. If we consider the case from the purely chemical viewpoint, we have to account for the different results obtainable with different combinations of metal and acid and for the obvious co-action between the three components. This may be done as shown by Print, by considering the reciprocal changes of the equilibrium between atomic and link-energy caused by the collision of two molecules. The oxygen atom and the metal atom coming within their mutual sphere of action lose potential energy, which is partially taken up by the chain O=C-O-H. causing a disturbance of the equilibrium between atomic and' link energy, which finally causes are increase in the atomic energy of the hy- drogen atom. An analogous activation occurs in the chain of metal atoms on the surface of the metal. In this way an activation is reached by asymmetric complex formation. If the compound is split up by tahing up energy (for example through collision with ~ third molecule), the components leave the compound in a state of increased activity.2 The initial reaction is caused by the affinity between the unsaturated' oxygen atom and the metal and the phenomenor~ will be a purely chemical orate in a non-ionising mediums. If at the same time a reducible compound is present, the reaction proceeds further: the reducible compound takes up the activated hydrogen and a metallic salt is formed. As opposed to this the action of an acid upon a metal in an ionising medium consists in the discharging of a hydrogen ion by the metal, a reaction which is probably hampered by the adsorption of the undissociated molecule upon the metal. The fact that the reduction can take place with the nor-ionized molecules explains the fact that even in 80 per cent acetic acid, the reduction reaction can attain enormous velocities, notwithstanding the small concentration of the hydrogen ions. The reduction reaction is then a col~molecular reaction with at least three components: acid, metal and reducible substance. An acceleration in the hydrogen evolution between for example zinc and 80.3 per cent acetic acid is only caused by phenyl hydroxylamine in the presence of nitrobenzer~e, if the concentration of the latter is small. A pie ce^ of zinc, after etching with dilute hydrochloric acid gives a marked evolution of hydrogen in acetic acid So.3 per cent at 52°. This evolution stops immedi-- ately if so much ~itrGbenzene is added that the solution becomes o.7s molar. A molar solution of phenylLydroxylamine does nol; stop the evolution, but Chem. Weekblad 14, 68 (I9I7) . 2 H. J. Prins: Rec. Trav. chim. 42, 25 (I923). 3 The same activation takes place with other oxygen containing substances e.g. nitrobenzene, ketones, aldehydes and the effect upon the hydrogen evolution depends upon the velocity with which the substance is reduced in connection with the magnitude of the adsorption. In accordance with this view with more negative metals like sodium the activation of the metal is not the last stage of the reactions with active oxygen containing: substances, but proceeds tall a metal atom Is extracted from one surface. . . . . ... , . ~, . 1 r 1 l r_

94~ HUGH S. TAYLOR if so Much nitrobenzene is then added, that -the solution becomes o.7s molar with regard to nitrobenzene, the en olution of hydrogen stops again immediate- ly; if these substances are added in the reverse order the result remains the same. Aniline in molar concentration has r?o appreciable influence Spore the hydrogen evolution.) From this we may conclude: I. that the hydrogen ion if it is adsorbed upon the surface of the zinc, can be supplanted by the nitrobe~zene or its immediate reduction product, if the nitrobenze~e is present in a su~eient amount. 2. that the nitrobenzene molecules or their immediate reduction prc~d~et;s either cover the surface totally or cannot be supplanted by the phenylh;~-droxylamine molecules. 3. aniline is not adsorbed to an appreciable extent. It is therefore probable that nitrobenzene is strongly adsorbed and rapidly reduced, phenylhydroxylamine less strongly adsorbed and aniline practically not at all? the cause of this must be sought ire the diminishing chemical ~CGiV- ity of the eharaeteristie groups NO~, NHOH, NH2 towards zinc, thus exhibit- ing the close relationship between adsorption and chemical reaction The adsorption obviously turns the oxygen atom towards the zinc; with removal of the oxygen atom the adsorption vanishes. ' The action of phenvlhydro~vlamine mav be demonstrated as follows: Etched zinc is heated in a test tube with about 80 per cent. acetic acid to boiling, then the tube is cooled till the hydrogen evolution becomes imperceptible, addition of a small quantity of phenyl hydroxylamine directly causes a marked hydrogen evolution. In an analogous way the action of nitrobenzene may be demonstrated.