1120
R. A. GARDNER
tion should show changes in the isotherm. However the isotherm for the N-Hz system is reproducible over successive measurements. PAULH. LEWIS(Texaco, Inc.).-Why does the susceptibility curve increase with the addition of hydrogen and increasingly so a t lower temperatures? n70uld this be due to sintering? R. J. LEAK.-The increase is produced by altering the spin-orbit coupling of the electrons with the lattice which is responsible for the magnetic anisotropy. The activation energy barrier b e b e e n aligned and non-aligned orientation apparently is lowered so as to make it easier for the domains to be oriented with the magnetic field. The effect becomes more pronounced a t lower temperatures because there is less thermal energy available to cause transition over the barrier. A. C. ZETTLEMOYER (Lehigh University).-When Si is oxidized, 0 - ions form on the ?;io. Do these influence the magnetization measurements? R. J. L E A K - T ~0~- ions that are proposed by Chessick, Yu and Zettlemoyer in the latter stages of oxidation are not in contact m-ith t h r metal and would be unlikely to have any effect through an insulating barrier of X O . J. 4.RASHKIN (E. I. du Tont de Xemours 8: Co.).-Can you say anything about bond localization in the adsorptioli of oxygen and hydrogen? You say that the hydrogen atom donates an electron to form a covalent bond with the nickel. Can you reconcile this m ith work done by Handler and others mho have shown that adsorbed hydrogen atoms
Vol. (34
increase the p-type surface conductivity of the sputtered germanium surface, indicating that hydrogen is acting elcctronegatively?
R. J. LEAK.-FrOm the n-ork presented here 1 cannot. However, from high field-low temperature saturation measurements it has been concluded that the hydrogen-nickel bond is localized (R. E. Dietz and P. W. delwood, J . -4ppl. Phys., 30, 1015 (1959)). The formation of a covalent bond by the hydrogen and nickel each donating an electron tells nothing a priori about the polarization of the bond. Other workers have shown that hydrogen forms the negative end of the polarized nickel hydrogen covalent bond (J. J. Broeder, L. L. van Reijen, W. M.H. Sachtler and G. C. A. Schuit, Z Elektrochem., 60, 838 (1956)). PAULH. LEwIs.-The authors comment that the magnetization-volume isotherm versus 02 adsorbed (Fig. 1) is a straight line. They deduce from this that the nickel is undergoing the same process a t the start of the admission of oxygen as a t the end. From the fact that the 02/H? adsorption ratio is 3 . 3 they conclude that the process is bulk oxidation. Is this a valid conclusion? Why wouldn’t the formation of a S i + + - O - bond a t the surface have the same effect as one in the bulk? R. J. LEAK.-sOthing definite can be stated from Fig. 1 about the bonding of oxygen to nickel except that it is the same throughout the isotherm. Proof of a differencz between chemisorption of oxygen and oxidation a t -78 is given in Figs. 4 and 3 with confirmatory evidence quoted a t -130, -183and -196’.
A PREDICTION OF HETEROGENEOUS CATALYTIC REA4CTIONS BY R. A. GARDFEH‘ Contribution from the Chemical and Physical Research Laboratory, T h e Standard Oil Company (Ohio),(’leveland, Ohio Received Match 7, 1960
A “kinetic isotope effect” was predicted for the catalytic oxidation of hydrogen-deuterium mixtures. Kith copper oxide as the catalyst HzO would be formed a t a faster rate than DzO while with nickel oxide as a catalyst the t x o reactions were predicted to occur a t the same rate. This paper presents the methods used to arrive at these predictions and the results of the experimental investigation of these reactions. Reaction mechanisms are proposed and additional predictions arc made which will be investigated experimentally.
Infrared spectra have been obtained of carbon monoxide chemisorbed on copper, nickel and cobalt and several compounds of these metals.2 An infrared technique similar t o that developed by Eischens3 mas used to obtain the spectra. Several infrared absorption bands were observed to result from the adsorption of carbon nioiioxide on a single adsorbent. For example, the chemisorption of carbon monoxide on copper oxide produced infrared absorption bands at 2173, 2127 and 2000 em.-’. The vibration frequencies of t’he neutral carbon monoxide molecule and the carbon monoxide positive ion are reported by Herzberg4 as 2143 and 218-2 cm. -l, respectively. The neutral carbon monoxide molecule has 10 valence electrons. The cnrbon monoxide posit’ivc ion has 9 valence electrons. It mas hypothesized2 therefore that an infrared ahsorption b m d for Chemisorbed carbon monoxide a t 2173 em.-’ corresponds to a carbon monoxide spe-
cies which has a non-integral number of valence electrons between 9 and 10. The experimental data of this study were interpreted Usilig the WolkensteiiP hypothesis of the mechanism of chemisorption. This led to the formulation of a relationship between the vibration frequencies of adsorbed carbon monoxide species and the number of valence electrons associated with these species. From this relationship the vibration frequeiic.:es of the adsorbed carbon monoxide species o i l copper oxide were found to correspond to 9.30, 10.30 and 11.30 valelice electrons for the species produciiig the absorption bands a t 2173, 2127 a i d 2000 em.-’, respectively. The concept of fractioiial numbers of electrons associated with adsorbed species has suggested n inetlic,d for t h e qualitativc prediction of riztes, mechanisms nnc! catalysts for certain heterogcncxms catalytic reactions. The hypothesis which has been used in the prediction of heterogeneous catalytic reactions was proposed by Myers.6 He advaiiced the concept that for certain reactions a roincidcnce between the vibration frequencies of the reactaiits is imessary.
(1) Chelnicnl a n d Physical Research Department, Standard Oil Company (Ohio), Cleveland, Ohio. ( 2 ) R . A. Gardnar nnd R. H. Petrucci, “The Chemisorption of Carbon RIonoxide on Metals” (1939) a thesis, submitted for puhlication. (3) R. P. Eischens, 1%‘. A. Plishiri and S. .4.Francis, J. Chem. PILUS., 22, 1780 (1954). ( 5 ) T h . Wolkenstein, “Advances in Catalysis ” 1’01. I X , Academic Press Inc., New York, N Y.,lc)57, p r 807-817 (4) G. Herzherg, “Spectra of Diatomic Molecules,” D. Van Nos72, 3-11 (1958). trand Co., Princeton, N. J., 1955, pp. 522. (6) R . R. Myers, .Inn S. Y . Acad. ASCI.,
A PREDICTION OF HETEROGENEOUS CATALYTIC Rmcrrom
Sept., 19GO
When the electrons of two potential reactants are in the same or harmonic energy levels, these electrons can be exchanged and combined to form a product. The examples which he presents of the oxidation, hydrogenation, halogenation and alkylation reactions of ethylene strongly support his fundamental hypothesis that a coincidence of vibration between reacting species is a necessary condition for certain reactions. The concept of a "fractional electron population" is also used in the prediction of catalysts and reactions. This refers to the lion-integral or fractional portion of the number of valence electrons associated with ldsorbed species on a particular metal or metal conipound. The fractional electron population is hypothesized to be a property of the metal. Since the carbon monoxide species on copper oxide possessed 9.30, 10.30 aiid 11.30 valeiice electrons, the fractional electron population of copper(I1) is designated as 0.30. Therefore hydrogen species on Cu(I1) will possess 0.30 and 1.30 valence electrons, and oxygen species on Cu(I1) may possess 4.30, 5.30, 6.30 and 7.30 valence electrons. On the basis of the above hypothesis and concepts, the prediction of heterogeneous catalytic systems requirw a knonledge of: (1) the vibration frequencies 31-energy lcvels of the gaseous species which may enter into the reaction; ( 2 ) the fractional electron population of the surface atoms of possible cat:ilysts ; (3) the vibration frequencies or energy levels of the chemisorbed species which may enter into the reaction; (4) a reaction mechanism. The particular reaction to be considered is the oxidation 0 : hydrogen-deuterium mixtures catalyzed by copper oxide and nickel oxide. The fundamental vibration frequencies of the gaseous reactants were olitaiiied from data presented by Hersberg.' The vibration frequencies of these gases are: H2* 4159.2 cm.-l; D,, 2990.2; and 02, 1556.2. The fractional electron populations of the two metals are 0.30 for copper as copper oxide and 0.00, 0.93 and approximately 0.90 for nickel as nickel oxide. These values for nickel were determined in 3, manner similar to that used for copper(I1) from our infrared studies of the vibration frequencies of carbon monQxide chemisorbed on nickel and several nickel compouiick2 The greater number of fractional electron populations for nickel results from the possibility of one and two partially filled 3d orbitals. The proposed energy levels of the Chemisorbed species of oxygen, hydrogen and deuterium can be derived on the basis of dissociative chemisorption of the molecular gases forming atomic species. The eiirrgy levels of the chemisorbed atoms can be determined from the following equations which relate the energy levels to the number of valence electrons associated with the particular adsorbed species. The number of electrons is in turn determined by the fractional elertron population of the particular catalyst. The equations are I,,T
=
VI)
=
1 1j(1 2 c ~ 2900 %eD
(7) G . Herzb,?rg, ref. 4, pp. 632, 533 and 560.
(1) (2)
vo = 778.1(8
1121
- eo)
(3)
where v = energy level of the adsorbate in ern.-', e = number of valence electrons associated with the particular adsorbate species. Finally, the low temperature catalytic reaction of Hz or D2with O2 to form H20or D20 suggests two types of reaction mechanisms: (1) the reaction of a gaseous hydrogen molecule with an adsorbed oxygen atom, or ( 2 ) the surface reaction of two adsorbed hydrogen atoms with one adsorbed oxygen atom. The prediction of the catalyst and mechanism of a reaction involves a matching of the fundamental vibration frequencies or energy levels of gaseous and chemisorbed reactants. As has been proposed by Myers6 the first overtone frequencies may also be used. The following equations show the reactants with the closest matching frequencies, the energy levels in cm.-l as subscripts, the numbers of electrons associated with the various species as superscripts, and the percentage difference of the frequencies from their average. Ni( 0 ) 8 . 0 0 + 2Ni(H)o 00 + 0.0 0.0 Si(H20)S.OO--e- S i + (H20)*.00 (4) % diff. 0 Ni(O)B.Oo + 2Ki(D)0.00 --+ 0.0 0.0 S i ( D D ) 8 J O OY0 diff.
CU(O)'.'O 2100 (4200)
+Xi
+
(D20)8.00
(5)
+ H22.00--+ 4159
+
CU(0)4.30 D2'.00 2879 2990
C ~ ( H 2 0 ) ' . ~+ 0 0.48% diff. C~(H20)8.30 +CU ( I I Z O ) ~(.a~) ~
+
+C~(D20)6.~0 + l.88y0diff. C~(D20)8,30+CU
+ (DzO)'.''
(7)
The harmonic frequency is shown in parentheses. The electrons in equations 6 and 7 used in the transi~ . ~ ~Cu tion from C U ( H ~ O ) ' .+ ~ ~ C U ( H ~ O ) and (D2O)6.3O +. C U ( D ~ O ) *are . ~ ~available from the catalyst and fractions of electrons remaining after desorption of the product return to the catalyst. In accordance with the Wolkenstein5 hypothesis of the mechanism of chemisorption, the YTarious adsorbed species exist in equilibrium. Desorption from the surface is hypothesized to occur through the adsorbed species that is closest in electronic configuration to the neutral gaseous molecule. For H 2 0and D20on Cu(II), this is the adsorbed species possessing 8.30 valeiice electrons. The metals in the above equations are in the form of oxides. According to the detailed reaction mechanisms presented in the above equations, hydrogen and deuterium react with oxygen over nickel oxide by the surface combination of hydrogen or deuterium ions (€I+ or D+) with oxide ions ( O F ) . Thew are protons or deuterons having no valenw electrons and oxygcii species having 8 valeiice elect roris. The reactions of hydrogen and deuterium iiith osygeii cntalyzed acrording to the above q u a t ions by cwpper oxide procertl by the attack of gasei)us hydrogen and deuterium molecules onto chemisorbed oxygen species which are in a positive electroiiic configuration. The numbers of valence electrons associated with these oxygen species chemisorbed 011
R. A. GARDNER
1122
copper oxide in equations 6 and 7 are 5.30 and 4.30. The energy levels of the electrons of these species can be calculated from equation 3. The oxygen species on copper oxide which reacts with a hydrogen molecule has 5.30 valence electrons with an energy level of 2100 cm.-l. The chemisorbed oxygen species on copper oxide which reacts with a deuterium molecule is more electron deficient having 4.30 valence electrons and an energy level of 2879 cm.-l. Reactions 4, 5 and 6 show a matching of frequencies within less than 0.57,. These reactions are predicted to proceed readily. Reaction 4 however, with a percentage difference of 1.88, is predicted to proceed at a slower rate. Therefore, the reaction of a mixture of hydrogen, deuterium and oxygen containing equal molar amounts of Hz and D2 over copper oxide is predicted t o produce HzO more rapidly than DzO a t conditions where the reactants and chemisorbed species are those noted above. The reaction of hydrogen, deuterium and oxygen over nickel oxide is predicted to produce HnO and DnO a t the same rate. These predicted reactions should predominate at the minimum reaction temperature, since they show the closest matching of frequencies of gaseous and chemisorbed react ants. Experimental Method and Results On the basis of the above considerations, mixtures of hydrogen, deuterium and oxygen were passed over copper oxide and nickel oxide. A large excess of oxygen was used to ensure maintenance of the metal oxides. The products were condensed in traps cooled by Dry Ice-acetone mixtures and analyzed by infrared spectroscopy for OD and OH absorption frequencies. The composition of the entering gaseous mixture was determined by mass spectrometry. The data are presented in Table I in terms of the OD/OH ratio in the product divided by the Dz/H, ratio in the feed (the isotope effect) as a function of the temperature of the reaction. As was predicted, the preferential oxidation of Hz in Hz-Dz mixtures over copper oxide increases as the reaction temperature approaches a minimum. The prediction also stated that HzOand DzO would he formed a t essentially the same rate over nickel oxide at the minimum reaction temperature. This is also substantiated by the data in Table I. The necessity for specifying the minimum reaction temperature arises as a result of the existence of alternative reaction mechanisms which can make increasing contributions to the product formation as the temperature is raised. The following equation illustrates an alternative reaction for D 2 0 formation over copper oxide.
+
~ C U ( D ) ' . ~CU(O)~.~O ~ --+ C11(DzO)7.30+ 3888 2100 3.86y0 diff. (4900) Cu(D2O)8.30--+Cu DzO (8)
+
Discussion During work on this system, several interesting observations have been made. For example, the effluent gas after contact with nickel oxide contained HD as well as H2 and D2. The reaction
Vol. 64 TABLE I
THE OXIDATIONOF HYDROGEX-DEUTERIUM MIXTURES CATALYZED BY COPPER OXIDEAND SICKEL OXIDE CUO
Catalyst
Temp., %/HZ OD/OH Isotope 'C. Feed Product effect
252 266 274 282 310 349
1 . 0 0 0.111 0.111 1.00 ,113 ,113 1.00 ,150 ,150 1.00 ,150 ,150 1.00 ,185 ,185 1.00 .210 ,210
NiO
Temp., Dz/H? OD/OH OC. Feed Product
Isotope effect
174 15.8 15.6 0.986 179 1.19 1.13 ,950 204 1 5 . 8 1 3 . 6 ,861
mechanisms for the formation of H 2 0and DzO over nickel oxide presented in equations 4 and 5 show the combination of adsorbed oxide ions (oxygen species possessing eight electrons) with protons and deuterons (hydrogen species possessing no electrons) to produce HzO and DzO. Since nickel oxide is hypothesized2 to adsorb gases as ions and atoms (or molecules), the existence of H+, H, H-, D f , D, D-, 0+2, O+, 0, 0- and 0-2 on the surface of nickel oxide permits the formation of HD, HzO. DzO and HDO as possible desorbed products. In contrast to nickel oxide, no H D was observed after the gas had contacted copper oxide. As shown in equations 6 and 7 the gases adsorbed on copper(I1) exist in electronic configurations that are intermediate between ions and atoms (or molecules). The following reactions compare the formation of H D over copper oxide and nickel oxide. Cu(H)O.30 1247 Ni(H)OJo 0
+ Cu(D)O.30 --+C U ( H D ) ~--+ .~~ 897
16.3y0 diff. Cu
+ Xi(D)oJO -+ 0
(HI))200 (9)
Xi(HD)OJO --+
Oy0diff.
Ni(HD)2J0-e- ?;i
+ (HD)2Jo
(10)
As can be seen from the percentage differences of the vibration frequencies of adsorbed H and D on each catalyst, the formation of H D over nickel oxide is greatly favored over the formation of H D over copper oxide. The presence of the above mentioned ions on the surface of nickel oxide also leads to the formation of HDO from Hz, Dz and 0 2 by surface reactions. Since the prediction concerned the oxidation of hydrogen and deuterium and the product analyses were made by the infrared absorbances of the OH and the OD bands, there was no determination of the concentration of HDO. The reaction mechanisms shown in equations 6, 7 and 9 preclude the formation of HDO from Hz,D, and 0 2 over copper oxide a t the minimum reaction temperature. -4consideration of the vibration frequency of the H D molecule indicates that it should he even less reactive to oxidation o\-er copper oxide than Dz. The three reactions with the electronic configurations of the adsorbed sprcies, their energy levels, and the percentage differencc of the matched frecluencies are presented
+
Cu(O')5.30 Hz2.00 +C~(H20)'.30+ 2100 4160 0.4870 diff. (4200) C~(Hz0)*,3" ---+CU
+ H:O
(11)
-1PREDICTION OF HETEROGEKEOUS CATALYTIC REACTIONS
Sept., 1960
2100 (4200)
3628
7.37, diff. C U ( H D O ) ~--+ . ~ ~ CU
+ HDO
1123
Acknowledgment.-The author wishes to acknowledge the advice and assistance of Dr. H . A. Strecker and Nrs. R. Krizan in this study, and to thank the Standard Oil Company (Ohio) for permission to publish this work. (13)
At 660" F. percentage differences of 0.48 and 1.88 correspond I O a 5 to 1 reaction rate difference. Therefore it can be predicted that a mixture of H2, D, and H D containing an equal number of moles of these gases with excess oxygen when contacted with copper oxide a t the minimum reaction temperature will form the oxidized products a t the following relatire rates : 1120 > D,O or HDO. The experimental inr-cstigatioi, of this prediction will be a part of the continuing examination of these concepts aiid hypotheses. Conclusion The foregoing has described a tentative method for the prediction of heterogeneous catalytic reactions. This has been applied to the prediction of a prefereihtl oxidation of Hz in H2-D2 mixtures catalyzed by copper oxide-a "kinetic isotope effect"-and the absence of this effect in the reactions catalyzed by nickel oxide a t the minimum reaction temperatures. On this basis details of the reaction mechanisms have been advanced. The experimental investigation of these predictions has produced data which support these hypotheses. I n addition it has been show1 that predictions can be made concerning the reactivity of H D toward oxidation over copper oxide and nickel oxide. The applicability of these concepts will be probed by future experiments.
DISCUSSION D. J. C. YATES(Colunihia University).-The gas phase spectrum of nickel carbonyl under resolution normally used in this type of work gives a very strong and very sharp symmet,rical band a t about 2060 cm.-'. Small quantities of this gas could have given the band shown on adding CO to evaporakd nickel films. 11. -4. GARDSER.--kttCnlptS were nlade to condense s nickel carbonyl wlim carbon monoxide was ronacuum evaporated nickel film. ?;o indicanickel carbonyl Tvas observed: however, not pursued exhaustively since the 2060 em. -1 band was not^ used to establish the relationship betlvccn the vibration frequency of CO units and the number of valence electrons associated v i t h t h r CO units (Ref. 2 aiid R. A. Gardner and R. H. Petrucci. J . A m . Chenz. SOC.,82, in press (19GO). A . C. ZETTLEMOYER (Lehigh Universit take care of heterogeneity of surface alteration with increasing coverage? R. A. G.mDmR.-.According to the Kolkenshtein hypothesis there are discrcte donor sites and acceptor sites. The ratio of those being occupied depends 011 thc surface coverage. .'1 A. LEX'IS (Tesaro Rcsearrli Centc~r~.-Hoxv do you divorce the frequency shifts observed from iiond length rather than numbers of valence electrons? id length, bond ordrr and strctchdepend on the nnml)cr of (~lectrons lecule. For carbon monoxide thci bond length decreases and the vibration freqwnry incrrases as electrons are removed from tht, molecult..
