INFRARED
ABI~ORPTION OF co ADSORBED ON Si-SUPPORTED Pt
production of hydrogen quantitatively, as well as the other minor products mentioned earlier. The excited states leading to decomposition were assumed to be triplet biradicals, although other types of excited states were not ruled out by the experimental evidence. The quantum yields for the Hg(6*Pl) decomposition of methylenecyclobutane were approximately a factor of tenfold greater than the Hg((iaP,) decomposition of methylenecyclopropane determined in the present study. This would infer that the excited methylenecyclopropane molecule has a lifetime considerably greater than the corresponding excited methylenecyclobutane, since one would not expect the methylenecyclopropane molecule to be a more efficient deactivator than methyleriecyclobutane. Meth?llenecyclopentane. Lack of detectable photolytic products, except for very small amounts of hydrogen and some unidentified residue products, on irradiation of methylenecyclopentane is surprising in view of the ready decomposition of the smaller homologs. This behavior indicates that the photoexcited state must be extremely stable to decomposition processes and is deactivated with high probability to the ground state.
Infrared Absorption at 21
p
327
The valence tautomer, spirobicyclohexane, would be predicted, on basis of ring strain considerations, to be much more unstable than the substrate and, therefore, an unlikely product. Other isomerization reactions all involve hydrogen atom shifts. Necessity for hydrogen transfer in the isomerization reactions of all three of the methylenecycloalkanes investigated could offer an explanation for the absence of products resulting from such reactions in the photolysis products. I n the case of methylenecyclopentane, a split into two smaller decomposition products requires somewhat less likely reactions than for the two other cycloalkanes studied. Double-ring scission reactions can be formulated forming C2H2, CzH4, and C3H4 as possible fragments. Accompanying these products, however, must be strained ring compounds or ones requiring hydrogen transfer reactions. Acknowledgment. This investigation was supported by a grant-in-aid from the National Science Foundation. Appreciation is also expressed to this same agency for funds furnished for the purchase of the mass spectrom eter used in some of the analytical procedures.
of Carbon Monoxide
Adsorbed on Silica-Supported Platinum by J. K. A. Clarke, G. M. Farren, and H. E. Rubalcaval Department of Chemistry, University College, Dublin, Ireland Accepted and Tranamitted by The Faraday Society (June 18, 1967)
An absorption band of carbon monoxide adsorbed on silica-supported platinum has been observed a t 476 cm-l. This band is assigned to the carbon-platinum stretching vibration of the adsorbed species. The results are reported of a force-constant calculation for a potential-energy function, which includes interaction between the C-0 stretch and the C-Pt stretch. The interaction constant is unusually large. The stretching force constants for the C-0 band and the C-Pt are, respectively, larger and smaller than those previously obtained for related systems by calculations neglecting bond-bond interaction. Details of the technique for improving the transmission of silica-supported metal disks beyond 8 p are discussed.
Introduction In spectroscopic studies of carbon monoxide adsorbed on metals, the region beyond 15 p is of special interest, for it is in this region that absorption mag arise due to excitations of vibrational modes which strongly involve the carbon-metal bonds. Such in-
vestigations should therefore be particularly valuable in yielding information about these bonds and about the adsorption Process itself. We report here Our observation of absorption a t 21 p due to carbon monoxide (1) To whom correspondence regarding this paper should be addressed,
Volume 78, Number 1 January 1068
J. K. A. CLARKE, G. M. FARREN, AND H. E. RUBALCAVA
328
19
20
21
22
23
P.
Figure 1. Absorption due to the carbon-platinum stretching vibration of carbon monoxide adsorbed on silica-supported platinum. The curves show the averaged results of nine separate runs, as described in the text: S is the spectrum, B is the background, and S B is the graphical difference.
-
adsorbed on silica-supported platinum ; we assign this band to the carbon-platinum stretching vibration. Figure 1 is a smoothed plot of the nine separate spectra which were obtained as discussed below. This information and that obtained from the 5-p spectra of Cl20l6 and C13016 adsorbed on silica-supported plath u m z T 4 yield values of the stretching and of the stretch-stretch interaction force constants for the adsorbed species. Previous investigators6 of related systems did not include a stretch-stretch interaction force constant in the functions used to describe the potential energy of the adsorbed species. I n fact, this constant is not negligible, hence earlier conclusions regarding the bonding in the adsorbed species must be reconsidered in view of the present results. Experimental Section The adsorbent used was in the form of a compressed disk, containing 10mg/cm2of silica with lo%, by weight, platinum dispersed in ten impregnations according to the methods described earlier.6 Ammonium chloroplatinate, prepared from spectrographically standardized metal as supplied by Johnson, Matthey, and Co., Ltd., was used as the source of metal; the silica was Cabosil (300-350 m2/g), and the carbon monoxide, supplied by the British Oxygen Co., was 99.95% pure. Since the 21-p band was very weak, it was recorded using a high scale expansion, and the resulting signalto-noise ratio was poor. The results reported here are based on three separate experimental sequences. Each involved reducing the adsorbent, recording three background spectra, adsorbing the carbon monoxide a t 1 torr, and recording three spectra with the adsorbed species present. The disk was thoroughly cleaned between sequences. At room temperature, the clean disks transmitted about 5% of the incident 21-11 radiation, but in the presence of carbon monoxide, or of other gases, a t pressures greater than 0.01 torr, they became virtually opaque. We have discovered that the transmission of these disks increases with temperature: The Journal of Phuaical Chemistry
at 250°, they transmit about 40% of the incident 21-p radiation. When the carbon monoxide was admitted into the cell, the disk was cooled and its transmittance decreased, but when the temperature was restored to 250°, the transmission was sufficient to obtain spectra. Concurrent examination of the 5-p region showed a normal high-coverage (about 90%) absorption due to the adsorbed carbon monoxide. The in situ cell is, essentially, of a conventional designI4 and the methods for cleaning the adsorbent are also c~nventional.~The spectra were obtained with a Grubb-Parsons Spectromaster using the slit program Program 10. An evacuated cell was used in the reference beam to compensate for absorption by atmospheric water, and the monochromator compartment was purged with dry air. Discussion We assume that the structure of the adsorbed species is that of a linear XYZ molecule with the carbon atom bonded to the metal. Contributions to the potential energy by displacements parallel to the CO axis are given, neglecting anharmonicity, by eq 1. 2V = FcoSco2 -k FCMSCM~ -I- 2fScoSc~
(1)
SCOand SCM are the CO and the C-Pt stretching displacements, FCO and FCM,and f are the corresponding stretching and stretch-stretch interaction force constants. This system has two axial normal modes of vibration: the high frequency mode, va, is largely a C-0 stretching deformation, but it contains a small out-of-phase contribution from the C-M stretching and absorbs near 5 p;2-3 the low frequency mode, VI, is mostly a C-Pt stretching vibration with an in-phase contribution from the CO stretch. Given values of va for Cl20and for ClSO,and of v1 for Cl20,the three force constants can be calculated by standard methods.? Unfortunately, the value of VI to be used is not immediately obvious. To eliminate the effects of intermolecular interactions between adsorbed molecules, it is best to use the frequencies for the zero-coverage limit; these are known for v3, but the corresponding value for v1 is not known because of the experimental limitations already described. Unpublished work in this laboratorya suggests that the surface-coverage dependence of the frequency of v3 is due to coupling between the adsorbed molecules by interactions which depend on the (2) R. P. Eischens, 8.A. Francis, and W. A. Pliskin, J. Phya. Chem., 60, 194 (1956).
(3) J. K.A. Clarke, G. M. Farren, and H. E. Rubalcava, unpublished data. (4) L. H. Little, “Infrared Spectra of Adsorbed Species,’’ Academic Press, New York, N. Y., 1966. (5) C. W. Garland, R. C. Lord, and P. F. Troiano, J. Phya. Chem., 69, 1188 (1965). (6) J. K.A. Clarke, G. M. Farren, and H. E. Rubalcava, J. Phys. Chem., 71, 2376 (1967). (7) E. B. Wilson, J. C. Decius, and P. C . Cross, “Molecular Vibrations,” McGraw-Hill Book Co., Inc., New York, N. Y., 1955.
INFRARED
ABSORPTION OF
co ADSORBED ON Si-SUPPORTED Pt
transition moments associated with v3. If the same is true for VI, then the required frequency at the zero-coverage limit will be only slightly less than the observed value of 476 cm-'. A small difference is reasonable in view of the fact that such shifts decrease with a decrease in transition ]moments, and these vary as the square root of theintegrated intensity.? Since vl ismuch weaker than VI(, for which the zero-coverage frequency lies 35 cm-' below the high-coverage limit,zva 10 cm-1 is used as a conservat,ive estimate of the amount by which the frequency of vII decreases as the surface coverage changes from high to low. Accordingly, we have used the observed values of 2038 and 1991 cm-' for ~ a ( C ' ~ 0and ) v3(C130) and an estimated value of 466 cm-' for vl(C120)in the calculation of the force constants. These are given in Table I. However, the conclusions discussed below are not critically dependent on small changes in the frequencies used in the calculations; the force constants were evaluated, using both high and low coverage values of vg as a function of v1 a t 10 cm-' in-
Table I : Force Constantsa of Carbon Monoxide Adsorbed on Platinum and of Related Metal Carbonyls
a
System
FCO
FCX
f
Ref
CO-Pt(Si02)
17.16
3.51
1.34
CO-Pt(Si02) CO-Pt MO(CO)6 Ni(C0)d
16 15.6 17.0 17.55
4.5 4.1 1.82 2.09
...
This study 2 5 9
...
0.63 0.33
9
Units are mdynes/A.
tervals from 426 to 496 cm-' without substantial change. The effective mass of atom Z was also varied to correspond to one, two, or four platinum atoms, with a resulting variation in the force constants small enough to neglect. The results given are those for the case of four platinum atoms. Table I includes results from two earlier studies of , ~ the 5-p related systems. Eischens, et ~ l . examined spectrum of a mixture of Cl20and C130 adsorbed on silica-supported platinum, and from the results predicted that v1 should occur near 20 p . Subsequently, Eischens and Pliskin* reported absorption a t 21 p by carbon monoxide adsorbed on platinum dispersed on potassium bromide. Garland, et u Z . , ~ observed absorption a t 21 p by carbon monoxide adsorbed on platinum films formed biy evaporating the metal in the presence of the gas. Both groups assigned the 21-p absorption to the carbon-platinum stretching vibration. The difference between our set of force constants and those obtained by the previous workers, even though there is little differencie in the frequencies used, is due to the manner in which the interaction constant f affects the
329
energy difference between the asymmetric and the symmetric vibrational levels. An increase in f shifts v1 upward and vg downward. Thus, if FCO and FCM are evaluated under the assumption that f is zero, the resulting values for these two stretching force constants will be, respectively, smaller and larger than the values which result when f takes its best value, ie., when f is positive. Ideally, our three force constants should be compared with a similar set for molecular platinum carbonyls of known structure, but we have found no published investigations which give such information. The force constant calculations of Jonese for Mo(C0)e and for Ni(C0)4 appear to be the best available for the present purpose; the relevant constants from his calculations are also given in Table I. The value obtained here for Fco is approximately equal to those obtained by Jones, indicating greater similarity between the C-0 bond of the adsorbed species and that of molecular metal carbonyls than had been suggested by earlier results.l~5 The values of FCM and f are substantially higher for the adsorbed carbonyls than for the molecular carbonyls. This larger value of FCM implies, insofar as force constants are measures of bond energies, that the C-Pt bond of these chemisorbed species is stronger than those of the molecular metal carbonyls discussed by Jones. Similarly, the stretchstretch interaction is considerably larger in the adsorbed species. These differences may well be due to the fact that on metal crystals the bonding electrons may, in effect, come from a large number of metal atoms, whereas in the case of the molecular metal carbonyls considered the bonding involves one metal atom. I n considering the validity of the foregoing analysis, a t least two questions arise. The first is whether the assignment of the 21-p band is correct. If VI absorbs a t a much lower frequency, ie., beyond the region where measurements were possible, then the resulting ~ of f would be similar to those disvalues of F C and cussed by Jones. We are inclined to reject this possibility for the following reasons. The spectrum re, ~ a relatively sharp abported by Garland, et ~ l . shows sorption a t 477 cm-', and a broader weaker band a t 570 cm-' which they assigned to the Pt-C-0 bending modes. This assignment is in agreement with recent spectroscopic studies of molecular metal carbonyls."J We saw no absorption near 570 cm-l, although a very weak broad band could have easily been lost due to the poor signal-to-noise level which prevailed. The second question concerns a more fundamental matter: our treatment, and those of Eischens, et U Z . , ~ and of Garland, et U Z . , ~ neglect interactions between (8) R. P. Eischens and W. P. Pliskin, "Advances in Catalysis," Vol. X, Academic Press, New York, N. Y., 1958. (9) L. H.Jones, J. Chem. Phys., 36, 2376 (1962). (10) R. W.Cattrell and R. J. H. Clark, J . Organometal. C h m . , 6 , 167 (1966), and references cited therein.
Volume 79,Number 1
January 1968
330
M. RIGBYAND J. M. PRAUSNITZ
lattice modes of the platinum and molecular modes of the adsorbed carbon monoxide. Grimleyll has treated the general problem of the vibrations of adsorbed molecules by a method which includes the interactions between lattice modes and molecular modes. In applying his treatment to the data obtained by Eischens' group218 he estimated, also neglecting intraniolecular interactions, that v1 was shifted about 30 cm-l upward by platinum lattice modes at 156 cm-l, Le., Grimley's
calculation suggests that the unperturbed value of v1 lies at about 446 cm-l. When our value of v1 is corrected by the same amount, the values 17.01, 3.01, and 1.09 mdynes/A are obtained for FCO, FCM,and f, respectively. These values tend toward those obtained by Joneslgbut they still suggest a strong interaction between the platinum crystallites and the adsorbed carbon monoxide. (11) T. B. Grimley, Proc. Phys. SOC.(London), 79, 1203 (1962).
Solubility of Water in Compressed Nitrogen, Argon, and Methane by M. Rigby and J. M. Prausnitz Department of Chemical Engineering, University of California, Berkeley, California 94720 (Receiwed August 8, 1967)
The vapor-phase solubility of water was measured in compressed nitrogen, argon, and methane at 25, 50, 75, and 100" and at various pressures between 20 and 100 atm. The volumetric properties of the vapor mixtures were described by the virial equation of state, and second virial cross coefficients were obtained from the solubility data. The solubility of a liquid in a gas at low pressures may be calculated from the vapor pressure of the liquid. Raoult's law yields an expression for the mole fraction, yl, of the liquid component in the gaseous phase y1
=
(1
- XZ)PS P
where x2 is the mole fraction of the gaseous component dissolved in the liquid, Pais the vapor pressure of the (pure) liquid, and P is the total pressure. At low pressures, x2 is negligibly small and the solubility is given directly by the ratio of the vapor pressure of the liquid to the total pressure. At high pressures, approaching the critical pressure of the mixture, the nonideality of the liquid phase becomes important in determining the vapor-phase solubility. However, in the intermediate pressure range with which this work is concerned, the critical factor determining the solubility is the nonideality of the vapor phase. The equilibrium of a binary system consisting of a heavy (liquid) component 1 and a light (gaseous) component 2 is governed by the equation ftL
the experimental quantities pressure, P, temperature, T,and the liquid and vapor compositions x and y to give
= ffV
where i = 1,2,f f is the fugacity of component i, and the superscripts L and V refer, respectively, to the liquid and vapor phases. The fugacities may be related to The JoUTnal of Physical Chemistry
where & is the vapor-phase fugacity coefficient, yr(P3 is the liquid-phase activity coefficient, ffocpr) is the reference fugacity of component i at T and at the reference pressure P', and :?z is the liquid partial molar volume. In the temperature range 25-lOO", the solubility in water of argon, nitrogen, and methane is very small, and to a good approximation we may take yl('I) = 1 and OIL = 2rlL (pure). I n the pressure range under consideration here, liquid water is essentially incompressible. The mole fraction of water in the gas may therefore be written
Since x2 is very small compared to unity, the vaporphase solubility is determined primarily by the fugacity coefficient (bl. This may be calculated from the virial equation of state.