An Infrared Study of Surface Properties of a ... - ACS Publications

by A. Zecchina, E. Guglielminotti, L. Cerruti, and S. Coluccia. Istituto d i Chimiea Fisiea dell'liniversitd d i Torino, Turin, Italy. Publication cos...
0 downloads 0 Views 782KB Size
INFRARED STUDYOF SURFACE PROPERTIES OF CY-CHROMIA

57 1

An Infrared Study of Surface Properties of a-Chromia. IV. Pyridine and Heavy Water Adsorption on Oxygen-Covered Surface by A. Zecchina, E. Guglielminotti, L. Cerruti, and S. Coluccia Istituto d i Chimiea Fisiea dell’liniversitd d i Torino, T u r i n , Italy

(Received J u l y 26, 1971)

Publication costs assisted by Consiglio Nazionale delle Ricerche ( R o m e )

It is observed that pyridine (py) is physically adsorbed onto a-chromia surface hydroxyls via hydrogen bonds and is strongly chemisorbed onto oxygen-free Cra+surface sites. py and water also chemisorb on an oxygenbands covered surface revealing residual cationic unsaturation. py and water adsorption eliminate C-0 typical of sites which still have one coordinative unsaturation. The Cr=O stretching frequency as well as the b o i d order are lowered by adsorption.

Introduction The nature of active sites on 0c-CrzO3 surface has been investigated in this laboratory by adsorption of HzO, CO, COz, and 02.1-3Infrared study of oxygen adsorption a t room temperature led to the following main conclusions: (1) Oxygen chemisorbs dissociatively onto Cr3+ (cus) ions to form Cr=O surface groups. (2) The chromium-oxygen stretching frequency varies with Cr3+ coordination number (4 or 5) and with the nature of the ligands attached to it. (3) Cr3+ (cus) ions with two vacancies (coordination number 4) chemisorb only one oxygen atom; as a consequence oxygen still leaves cationic vacancies on the surface. I n order to investigate the presence of residual surface cationic unsaturations and so confirm previous conclusions, pyridine (py) and water adsorption on an oxygen-covered surface have been studied. The choice of py as a test molecule for residual cationic unsaturation is due to its great ability to form adducts with metal ions. Water is also useful, because it chemisorbs both dissociatively onto chromium-oxygen ion pairs and in a nondissociative form onto isolated and unsaturated cations.’ In the first paper of this series water chemisorption on an oxygen-free surface was investigated, so that it is possible to compare the two experiments. No infrared data are available for py chemisorption onto oxygen-free cr-Crz03 and this point has also been investigated here. py is normally used to characterize spectroscopically the Lewis-Brginsted acidity of electron-acceptor molecules of the MX, type and of surface acidic sites as ell.^-'^ Lewis acidity and surface hydroxyl acidity are usually checked by the shifts of two nuclear frequencies: the Sa (the notation used by Wilmshurst and Bernstein” and by many others is followed here) a t 1583 cm-I and the 19b a t 1440 cm-I. Brginsted sites, if present, are deduced from a band a t 1540 cm-l that is characteristic of pyridinium ion.6 I n this work special care has been devoted to the

choice of “analytical” bands of adsorbed py. Due to the wide spectral range in which a-Cr2O3is transparent (4000-800 cm-l) we had several py modes from which to choose the one most sensitive to donor-acceptor bonds and hydrogen bonds. According to Filimonov and Bystrov’s12 viewpoint, modes of A1 symmetry species should be expected to be more affected by the N-14 bond formation. Shifts of skeletal modes are also produced when py N-oxide is formed.13 The breathing mode 1 (992 cm-l) is very sensitive to coordination bond formation; it is shifted upward to 1030 cm-I in strong donor-acceptor complexes like ~y-+BC13.~Other skeletal modes of species AI might be used as well, the most sensitive being: mode 12, which is a t 1033 cm-l in pure py and shifts to 1044 cm-’ in py N-oxide and to 1050 cm-I in py complexes; mode Sa, which goes from 1588 cm-l to 1602 in py Noxide and 1630 cm-I in complexes of the py+BX3 type. The use of the breathing mode as the test-band, however, is preferable to mode 8a, in that the latter could give rise to a Fermi resonance effect with a com(1) A . Zecchina, 8. Coluccia, E. Guglielminotti, and G. Ghiotti, J . P h y s . Chem., 75,2774 (1971). (2) A . Zecchina, S. Coluccia, L. Cerruti, and E. Borello, ibid., 75, 2783 (1971). (3) A. Zecchina, S. Coluccia, E. Guglielminotti, and G. Ghiotti, ibid., 75,2790 (1971). (4) A. Terenin, W. Filimonov, and D. Bystrow, 2. Elektrochem., 68, 180 (1958). (5) N. N. Grenwood and K. Wade, J . Chem. Soc., 1130 (1960). (6) N. S. Gill, R. H. Nuttall, D. E. Scaife, and D. W. A. Sharp, J. Inorg. Nucl. Chem., l 8 , 7 9 (1961). (7) N. S. Gill and R. 5 . Nyholm, ibid., 18,88 (1961). (8) G. S. Rao, Z . Anorg. Allgem. Chem., 304,176 (1960). (9) J. Yarwood, Trans. Faraday Soc., 65,934 (1969). (10) E. P. Parry, J . Catal., 2, 371 (1963). (11) J. K . Wilmshurst and H. J. Bernstein, Can. J . Chem., 35, 1183 (1957). (12) V. N . Filimonov and D. S. Bystrov, Opt. Spectrosk., 12, 31 (1962). (13) P. Mirone, A t t i Accad. N a z . Lincei, Cl. Sci. Fis. M a t . N a t . Rend., 35,530 (1963). T h e Journal of Physical Chemistry, Vol, 76, No. 49 1973

A. ZECCHINA, E. GUGLIELMINOTTI, L. CERRUTI,AND S. COLUCCIA

572 100

100 80

60

eo

1% 40

60

t% so0

1400

1200

XKH] cm-1

800

Figure 1. Infrared spectra of py adsorbed onto a-chromia oxygen-free surface (% transmission us. wavelength in cm-1): curve 1, background; curve 2, after exposure to 15 Torr of py a t room temperature (R.T.); curve 3, after 1hr. evacuation a t R.T.

40

20

noo Figure 3. Infrared spectra of py adsorbed into oxygen covered a-chromia (% transmission us. wavelength in cm-1) : curve 1, background of oxygen-free a-chromia contacted with 40 Torr of oxygen a t R.T. and then evacuated a t R.T. to a final pressure of lo-* Torr; curves 2-4, increasing py coverages; curve 5, py saturation.

1800

1400

1200

loo0

8oocm-l

Figure 2. Infrared spectra of py adsorbed into hydrated a-chromia (% transmission us. wavelength in cm-1): curve 1, background of oxygen-free a-chromia contacted with water vapor ( p = 8 Torr) 2 hr a t 400' and then evacuated for 1 hr at R.T.; curve 2, after exposure to 15 Torr of py a t R.T.; curve 3, after 1 hr evacuation a t R.T.

+

bination band a t 1597 cm-1 (1 6a), which is of species AI and is a t 1602 cm--' in the liquid py spectrum. Since the combination band also can be shifted upward due to the increased frequency of both 1 and 6a modes upon complex formation, we conclude that the behavior of mode 1 is by far the most representative of the formation of donor-acceptor bonds. Dissociative adsorption of water or heavy water (used instead of water because the latter would give an O H bending mode a t 1100-800 cm-l where the Cr=O stretching mode is also present) should give rise to bands of surface OH and OD groups, whose stretching frequencies are expected to be sensitive to the oxidation status of The Journal of Physical Chemistry, Vol. 76, N o . 4, 1978

the adsorbing chromium ions. I n fact OH or OD groups attached to Cr3+ions should be basic or amphoteric while OH or OD groups attached to higher valence ions (Cr5+ or Cra+) should be acidic in nature. On the other hand, H20 or DzO adsorption might influence the stretching frequencies of Cr=O surface groups which are present on an oxidized surface.

Experimental Section All details of the oxygen-free catalyst preparation and characterization have been described elsewhere. Oxygen-covered catalysts are obtained from oxygen free ones by room temperature reaction with oxygen ( p = 40 Torr). All of our spectra were run on a Beckman ir 12 spectrophotometer. The spectral conditions were those described in the previous papers of this series. Reagent grade pyridine from Carlo ErbaMilano was used after a simple vacuum distillation. Heavy water was ICN reagent grade with 99.75% deuterium. Curve 2 of Figure 1 shows the ir spectrum of 15 Torr of py in equilibrium at room temperature with an oxygen-free catalyst. A 1 hr evacuation at room tem-

573

INFRARED STUDY OF SURFACE PROPERTIES OF Q-CHROMIA 100

f%

100

80

80

60

60 T% 40

40

20 20

1100

m Figure 4. Infrared spectra of Os adsorbed onto py covered or-chromia (% transmission us. wavelength in cm-I): ---, background of or-chromia oxygen-free surface; --, after exposure t o 15 Torr of py and evacuation for 1 hr a t R.T.; ---.-,after exposure to 40 Torr of oxygen; .-e-, after 1 hr evacuation at R.T.

perature to a final pressure of to 10-5 Torr leads to the situation illustrated by curve 3 of Figure 1. Figure 2 refers to a hydrated a-chromia sample. I n particular, curve 1 is the background spectrum of the oxygen-free catalyst exposed for 2 hr at 400" to water vapor and then pumped down a t room temperature for 1 hr. Curves 3 of Figure 1and 3 of Figure 2 are the spectra of irreversibly adsorbed py that cannot be evacuated a t temperatures as high as 400". A particulary strong chemisorption must be involved in order to account for the high poisoning activity of py on several oxides. Figure 3 shows the adsorption of py onto a sample which had been allowed to preadsorb oxygen at room temperature. The full-line curve is the spectrum of 0 2 chemisorbed onto a-Cr203,while other curves refer to increasing py coverages. It is observed that an oxygen-covered catalyst is still capable of adsorbing py and that the spectrum of the chemisorbed oxygen is fairly sensitive to py coverage. Figure 4 refers to an experiment opposite to the previous: py is allowed to adsorb onto an oxygen-free sample at roam temperature and then pumped down to lob4 to low6Torr (line spectrum). Oxygen (40

lo00

900 cm-9

000

Figure 5. Infrared spectra of DzO adsorbed onto oxygen covered a-chromia (yotransmission us. wavelength in cm-*): curve 1, background of oxygen-free or-chromia contacted with 40 Torr of oxygen a t 1t.T. and then evacuated at ILT. to a final pressure of 10-4 Torr; curves 2-4, increasing D20 coverages; curve 5, heavy water saturation.

Torr) is then admitted to the cell and the spectrum recorded. Finally, the gaseous phase is evacuated at room temperature to a final pressure of low5Torr. I n Figure 5 the effect of heavy water adsorption on the spectrum of oxygen adsorbed a t room temperature is reported. The various curves refer to increasing heavy water coverages. The final spectrum was obtained a t an equilibrium pressure of 15 Torr. Figure 6 shows the spectral modifications in the OD stretching region due to exposure of an oxygen-covered sample to heavy water. Only high coverages are considered, for a t low coverages no spectral changes are produced and the intensities are very weak. Equilibrium D20 pressures were 0.15 and 15 Torr, respectively (curves 1 and 2). I n the same figure, spectra of DzO adsorbed on an oxygen-free surface a t the same equilibrium pressures are also reported (curves 1' and 2', respectively). After D20 saturation, the oxygen-covered sample was pumped down at various temperatures increasing from 2 5 O , and the spectra are show in Figure 7. Oxygen bands are easily removed, and at 400" they are completely absent. This is quite surprising, for on a well dehydrated sample (see part 112) Cr=O groups are only slowly removed upon evacuation and are still present a t 400". The Journal of Physical Chemistry, Vol. 76, N o . 4, 1078

A. ZECCHINA,E. GUGLIELMINOTTI, L. CERRUTI,AND S. COLUCCIA 100 7%

80

60

40

20 2800

Figure 6. Infrared spectra of OD stretching range of oxygenfree and covered a-chromia surface (optical density us. wavelength in cm-l): curve 1, after exposure to 0.15 Torr of DaO a t R.T. on an oxygen covered surface; curve 2, after exposure to 15 Torr of DaO at R.T. on an oxygen covered surface; curve l', after exposure to 0.15 Torr of DzO a t R.T. on an oxygen-free surface; curve 2', after exposure to 15 Torr of DtO a t R.T. on an oxygen-free surface.

tloO

1000

900 un-'

800

Figure 7. Infrared spectra of oxygen covered and hydrated a-chromia (% transmission us. wavelength in cm-1): curve 1, background of oxygen-free surface; curve 6, oxygen covered surface contacted with DeO (15 Torr) a t R.T. and then evacuated for, 2 hr a t R.T.; curves 5-2, after outgassing for 2 hr a t 100, 200, 300,and 400'.

Discussion ( 1 ) Pyridine Adsorption onto Oxygen-Free a-Cr203. By comparison of curves 2 and 3 of Figures 1 and 2, the spectrum of physically adsorbed py (Le., removable upon room temperature evacuation) on poorly and highly hydrated samples can be derived. It can be observed that: (a) the bands due to physically adsorbed py are at the same frequencies in both cases, (b) such frequencies do not appreciably differ from those of liquid py (see Table I), (c) equilibrium pressures and other experimental conditions being the same, physically adsorbed py bands are much stronger on the highly hydrated sample. It is therefore concluded that the physical adsorption occurs mainly on surface hydroxyls, which would be engaged in strong hydrogen bonding with the nitrogen. Spectral modification in the OH-stretching and -bending regions (3700-3000 and 1100-800 cm-l, respectively) seems to cordirm the above conclusions. Physical adsorption strongly lowers the OH-stretching frequency (the spectrum is not reported for the sake of brevity) while the OH-bending mode at 800-950 cm-l is weakened in intensity and shifted upward to 1000-1100 cm-l (see Figure 2) as expected on formation of strong hydrogen bonds. The surface hydroxyl-py interaction does not lead to the formation of observable amounts of pyridinium ions, for no absorption is produced a t apT h e Journal of Physical Chemistry, Vol. 76, N o . 4, 1978

Table I : Spectrum of Pyridine in the Phase, cm-l Band no.

Liquid

11 12 18a 15

992 1033 1071 1147

9a, 3 19b 19a 8b

1219 1440 1485 1573 sh

Physisorbed

922 1033 1071 1147 ll22l 11235-1240 1442 1486 1577 1594

Chemisorbed

1010-1020 1046 1073 1149

{ ::%I240 1450 1488 1577 1610

proximately 1540 cm-l, where the characteristic N-H+ bending mode is commonly observed.6 It is therefore concluded that no appreciable amounts of Br@nsted sites are available on a-CrzOa. Curves 3 of Figure 1 and 2 are for irreversibly chemisorbed py and indicate that the frequencies are the same in the two cases but quite different from those of the liquid or physically adsorbed species (see Table I). We think that the observed spectrum could be explained in terms of the formation of a strong bond of the donor-acceptor

INFRARED STUDYOF SURFACE PROPERTIES OF a-CHROMIA type between py molecules and surface Cr3+ions which are coordinatively unsaturated. Further observations support our assignment: (a) High resistance to desorption suggests that the py surface interaction must be a strong one in agreement with the “hard” character of both py and Cra+ (cus) ions.’* (b) No carbon monoxide is chemisorbed onto an cr-Cr20asample which preadsorbed py: active sites must therefore be the same in the two cases. (c) The ir spectrum of chemisorbed py is very similar to that of complexes involving 5 dative bond of the py+M type. In particular, the presence of a shoulder a t 1240 cm-‘, next to the 1220 cm-’ band, has been observed by Gill, et u Z . , ~ in 24 complexes of the general formula M(py)zXz, and it was justified as due to the formation of the py+M bond. Also the shifts of a few nuclear bands are typical of the formation of a medium-strong py+Cr bond. I n detail, mode 1 shifts to -1015 cm-’, mode 12 to 1046 cm-l, mode 19b to 1449 cm-l, and mode 8a to 1610 cm-‘. According to Filimonov’s hypothesis12 a shift of mode 1 from 992 to 1010-1020 cm-’ clearly indicates the formation of a mediumstrong dative bond of the type formed with SnCla and AlCla. (d) The amount of cheniisorbed py is higher on less hydrated samples and this is consistent with the previously proposed theory14 that it is the dehydration process which creates Cr3+ (cus) ions on the surface. It has previously been that on the (001) face, five types of Cr3+ ions are present which differ in the coordination number (4 or 5) and in the nature of the ligands. They are five different adsorbing sites with different Lewis acid character and thus capable of chemisorbing py. On the other hand, no different bands are observed in the ir spectrum of the coordinate py: this might be ascribed either to slight differences in acidity among the various sites or to a nonselectivity of the py itself. A similar conclusion was drawn by Gill, et uZ.,‘j on studying py complexes with metals of coordination number 4 and 6. I n fact they remarked that “there are few systematic changes in the positions of the bands with changes in mass, electronegativity or valency of the central atom or with changes in the other ligands.” As a consequence we conclude that py does not distinguish among different unsaturated surface cations. Although the above observations quite clearly indicate that a tr-bond is formad between py molecules and Cra+ (cus) ions, there are a few points in the experiment of Figure 2 that seem to deserve further discussion. I n fact it is quite surprising that a thoroughly hydrated sample (Le., heated with HzO a t 400” and then evacuated at R.T.) still has an appreciable adsorption capacity towards py while no capacity whatever is exhibited towards C 0 . l Spectrum 1 of Figure 2 shows a weak and broad band a t -1590 cm-l that has been assigned to water molecules coordinated to Cr3+ (cus) ions.’ Such a band is no longer detect-

575

able in spectrum 3 suggesting that coordinated water has been removed, possibly through the exchange mechanism

CrJ+

CrJ+

The above mechanism is consistent with the “harder” nature of py with respect to water; such an exchange process could never occur with CO, which is a “very soft” base. We already observed in a previous paper’ the presence of coordinated water on a chromia surface that had been rehydrated a t 400”, but we could not make any quantitative consideration owing to the weakness of the 1590-cm-’ mode. The removal of such coordinated water through the above mechanism, causing the characteristic absorption of coordinated py to appear in the spectrum, indicates that the amounts involved are not negligible. (9) Pyridine Adsorption onto Oxygen Covered Surfaces. Room temperature adsorption of 0 2 onto a-Cr203 has been discussed previously.2 Figure 3, full line curve, shows the jr spectrum due to chemisorbed oxygen. Two bands at 1024 and 1016 cm-‘ have been assigned to the stretching of Cr=O groups in which the metal has coordination number five and thus still has one coordinative vacancy. Three bands a t 995-980 cm-‘ have been assigned to Cr=O gtoups in which the metal ion has the full coordination number. The results reported in Figure 3 fit the above assignment. The adsorption of py gradually eliminates the two bands a t 1024-1016 cm-’ and produces two new absorptions a t lower wave number (-980, -940 cm-l). The following mechanism might account for the observed modifications

Coordinatively insaturated Cr=O groups would be transformed into saturated ones upon py uptake, thus explaining the disappearance of the two bands a t 10241016 cm-’. The complex absorption produced a t lower wave number (980-940 cm-l) could be assigned to the new structures (1 and 2) produced by mechanism (2).

1

2

The shift to lower frequencies may be easily explained by taking into account the sensitivity of the chromiumoxygen stretching frequency both to the metal coordi(14) R. L.Burwell, Jr., G. L. Haller, K. C. Taylor, and J. F. Read, Advan. Catal. Relat. Subj., 19,62 (1969).

The Journal of Physical Chemistry, Val. 76, N o , 4$ 197.9

A. ZECCHINA,E. GUGLIELMINOTTI, L. CERRUTI,AND S. COLUCCIA

576

nation number and to the nature of the ligands attached to it. Similar behavior has also been observed in the VO2+ group in complexes where electron-donor groups are present.lb In fact the 31-0 covalent bond has a strong p n + dn back donation from the oxygen to the metal that is sensitive to the presence of electron donors. The formation of py adducts changes the coordination number and increases the electron density of the metal d orbitals and thus decreases the pn d n back donation. The bond index is expected to decrease, as does the M=O stretching frequency. The surface oxocomplexes 1 and 2 differ from the octahedral complexes 3-5 already present on the surface after oxygen adsorption (see part TI of this series) only in the presence of a new ligand.

3

4

(3) Oxygen Adsorption onto a Sample Pretreated with Pyridine. The experiment of Figure 4 is in some ways the opposite of that discussed in section 2. py adsorption onto a bare surface does not entirely hinder the surface activity toward oxygen, meaning that not all of the cationic surface vacancies are saturated. We might suppose that chromium sites with one coordinative vacancy are readily saturated by py and chromium sites with two vacancies adsorb only one py molecule, because a cationic site is quite unlikely to form two strong cis acceptive bonds. On the other hand, only few unstable complexes in which two py molecules are coordinated in a cis configuration to the same metal ion are reported in the 1iterature.lG According to this scheme we might expect py covered surfaces still to have cationic vacancies available for oxygen chemisorption, through the mechanism

5

The stretching frequency of Cr=O groups in structures 1 and 2 might be either higher or lower than in structures 3-5, according to the electron-donor ability of py compared to that of OH- and 02-. It is difficult to decide a prioyi, based only on what is known of similar complexes, because their number is small. I n addition, the properties of the 02-ligand are not well known and, finally, it is not clear whether surface OH- and 0 2 groups are bonded through covalent, ionic, or mixed bonds. The experimental results reported in Figure 3 show that py adsorption produces two bands a t -980 and -940 cm-l, which should correspond to the newly formed structures 1 and 2. Nevertheless from our data a more detailed assignment cannot be deduced and, as a consequence, a comparison between the electrondonor ability of the three surface ligands (py, OH-, and 02-groups) cannot be made. As Figure 3 shows, when py coverages are high the frequencies of all the bands due to adsorbed oxygen are shifted to lower frequencies. This fact is quite normal, as a polar environment commonly lowers the stretching frequency of surface groups. Electronic colligative properties of the solid might be involved as well. I n fact the adsorption of a strong electron-donor group directly affects the electronic properties of the adsorbing site but can also be expected to have a minor effect on the properties of all neighboring sites as well. The Cr=O bond of all the neighboring surface complexes will thus be wealiened by py adsorption. The spectrum of py adsorbed on an oxidized sample does not really differ from that obtained on an oxygenfree sample. All nuclear bands are nearly at the same frequencies in the two cases and are only slightly shifted with different coverages. Appreciable differences are only obse;ved as regards the relative intensities of some peaks. T h e Journal of Physical Chemistry, VoZ. 76, N o . 4, 1978

The overall result of mechanism 3 is the same described by mechanism 2 with production of structures 1 and 2. The presence of two bands at -980 and -940 cm-' demonstrates that the previous hypothesis is correct. Nevertheless the proposed mechanism seems to be an oversimplification. It accounts for the appearance of the two bands a t 980-940 cm-l due to chemisorption of oxygen onto ions that had already adsorbed py and still had a coordinative vacancy. But it does not explain the appearance of a band a t 1033 cm-I upon oxygen chemisorption, which is typical of physically adsorbed py (see Figure 4 and Table I). This fact suggests that other processes might be taking place as well. Among them seems to be likely a ligand displacement reaction of the type

(OH-)OL-02 -(OH-)

I/

-Cr=O

I'

+

py(physisorbed) (4)

I n fact this reaction explains well the production of physically adsorbed py from chemisorbed py. (4) Heavy Water Adsorption onto Oxygen-Covered Surface. The experiment of Figure 5 shows that water is still adsorbed onto a sample which preadsorbed oxygen and that the spectrum of adsorbed oxygen is greatly influenced. As in the case of py adsorption, the bands at 1024 and 1016 cm-l are completely eliminated, thus confirming that these bands involve Cr=O groups in (15) J. Selbin, L. AM, Holmes, Jr., and S. P. MoGlynn, J . Inorg. h'ucl. Chem., 25, 1359 (1963). (16) C. S. Kraihanzel and F. A. Cotton, Ilzorg. Chem., 2,533 (1963).

INFRARED STUDY OF SURFACE PROPERTIES OF CX-CHROMIA which the chromium ion still has one coordinative vacancy. Two adsorption mechanisms may be proposed

?

The first one is completely similar to mechanism 2 and involves a coordinative chemisorption of water; in the second one a molecule of water is dissociated onto a Cr3+ (cus) 02--(cus) pair. The first resulting structure is analogous to the 2 structure and so we expect it to have a similar spectral behavior. The second structure, involving a CrO group in octahedral coordination with surface OD- ligands, is one of the three octahedral structures formed upon oxygen adsorption on a partially dehydrated surface (see part I12). Therefore we assign the band a t -940 cm-' to the product structure of mechanism 5 and the increased absorption a t -970980 cm-I to the formation of the structure arising from mechanism 6. It is surprising that in the first structure coordinated water has practically the same effect that py has in decreasing the Cr=O stretching frequency, when it is well-known that py is a harder base than water. We think that the large frequency shift induced by water adsorption might be explained by the formation of hydrogen bonds between the OD groups of coordinated water and the Cr=O groups on the same sites. Such a hydrogen bond would contribute to the lowering of the Cr=O stretching frequency and the overall shift will be larger than is expected. The above considerations definitely indicate that the band a t -940 cm-l is due to the product of mechanism 5. Neverthelesa, the reaction is, as usual, an oversimplification in that it completely disregards the fact that any new ligand enters, a t one time, the coordination sphere of more than one unsaturated metal ion. On the other hand, what mainly matters here is that two bands disappear while two more are produced at lower frequencies. As Figure 6 shows, the amount of heavy water chemi-

577

sorbed onto an oxygen-covered (curve 1) surface probably is smaller than that adsorbed on the oxygen-free (curve 1') surface a t the same equilibrium pressures. The spectrum of OD groups in the 2800-2000-cm-1 range widely differs in the two cases. I n the first one the bands a t -2700 cm-', characteristic of free OD groups on reduced chromia, are partially absent. This fact might be due to the different acidity of these newly formed OD groups and to the formation of hydrogen bonds favored by the high oxygen coverages. I n order to confirm the occurrence of mechanism 5 or 6 we studied the desorption of adsorbed water at increasing temperatures. I n fact coordinated water should be desorbed first and, as a consequence, the band at -940 cm-' should disappear and bands a t 1016-1024 cm-' appear again. Figure 7 shows that the band at -940 cm-' is completely eliminated at 200" while a band at 1016-1024 cm-l is produced. Another surprising result is also obtained: all of the bands due to adsorbed oxygen decrease with increasing desorption temperature and a t 400" are completely eliminated. We conclude that oxygen adsorbed on a hydrated sample is less resistent to desorption than oxygen on a well dehydrated one. In paper I' we observed that the degree of dehydration of an a-CrzOa surface is influenced by the presence of adsorbed oxygen and we think that the two phenomena are not independent, although any explanation would first require more experiments and probably different investigational techniques.

Conclusion From the above discussion it follows that: (1) py is physically adsorbed onto a-chromia by means of hydrogen bonds to surface hydroxyls. (2) py is adsorbed onto chromium ions through a medium-strong acceptordonor bond, whose spectral behavior is well characterized. (3) py chemisorbs onto a fully hydrated sample by substitution of coordinated molecular water. (4) py and mater adsorption onto an oxygen-covered surface reveal the presence of a residual surface cationic unsaturation. ( 5 ) py and water adsorption onto an oxygen-covered surface eliminate the Cr=O bands at 1024-1016 cm-' confirming that they involve chromium ions with still one coordinative unsaturation.

Acknowledgment. This research has been supported by the Consiglio Nazionale delle Ricerche.

T h e Journal of Physical Chemistry, Vol. 76, N o . 4, 1973