Spectra of the Isolated Hydroxyl Groups on Silica in the surface represented by the observed SiOH (2v) band, those constituting the predominant species represented by the frequency maximum are the primary active centers for water adsorption. Finally, it has been suggested in the literature that water may diffuse into the silica and form hydroxyl groups in the interior.lG The average distance a molecule diffuses in 1000 min was estimated, by extrapolating the hightemperature diffusion coefficients for fused silicas,l6 to be 6 A at room temperature. The presently investigated water-HiSil systems did not show any signs of water diffusion into the silica. The adsorption equilibria were rapidly established and all water could be removed by mild degassing without leaving either residual water or increased amounts of‘ hydroxyls in the specimen. These results indicate that the extrapolation of the high-tempera-
1465
ture diffusion data to room temperature does not lead to realistic estimates of the penetration rates of water through the surface layer and that diffusion and rehydroxylation does not influence the room temperature adsorption phenomena to any significant degree.
Acknowledgments. The authors are indebted to the National Science Foundation for support of this work through Grant No. GA-29526. The Auger analyses were carried out by Dr. G. W. Simmons and the gravimetric adsorption measurements on Na-HiSil (650) by Dr. W. C. Hamilton, which contributions are gratefully acknowledged. (15) R . H. Doremus, J. Phys. Chem., 75,3147 (1971). (16) A. J. Mouison and J. P. Roberts, Trans. Faraday Soc., 57, 1206
(1961).
Infrared Spectra of the Isolated Hydroxyl Groups on Silica 6 . A. Morrow* and I. A. Cody Department of Chemistry, University of Ottawa, Ottawa K I N 6N5, Canada
(Received August 23, 1972)
Infrared spectra in the OH stretching region of the isolated SiOH groups on the surface of silica have been obtained. The silica samples had been preheated to 750” and spectra were recorded with sample temperatures ranging from -145 to 400”. Extreme caution has been taken to ensure that virtually all atmospheric water was removed from the spectrometer and with perfect double-beam compensation, no splitting of the SiOH band was observed, nor were any distinct shoulders apparent. This finding is in contradiction to an earlier similar study and only when the double-beam compensation with respect to atmospheric water is not perfect has it been possible to reproduce spectra showing distinct shoulders. The infrared band profile of the SiOH groups observed in this work was virtually identical with those observed from SiOD groups and band fitting techniques have been used to show that both types of band can be fitted with a two-component curve. A general discussion of the validity of using band-fitting techniques for “resolving” closely overlapping bands in infrared spectra is included.
Van Cauwelaert, Jacobs, and Uytterhoevenl have recently reported that the 3750-cm-l absorption band in the infrared spectrum of silica (due to isolated surface hydroxyl groups) has two distinct shoulders to high and low wave number of the main band. The shoulders were more pronounced at higher sample temperatures and their exact position relative to the main band depended on this temperature. Using a curve analyzer they “decomposed” the band contour into a three-component curve which reflects the positions of these shoulders. In an earlier investigation, Hair and Hert12 produced a spectrum which showed that with a sample temperature of 25” the same band consisted of three clearly resolved components, but this finding was later refuted by Hockey3 who claimed that the “splitting” of this band was an artifact due to improper double-beam compensation in this region of strong infrared absorption by atmospheric water. Hockey’s spectrum showed a relatively sharp band centered near 3750 cm-I which was degraded into a tail on the low
wave number side, and similar spectra at this temperature were reported by Van Cauwelaert, et al.,1 and by Morrow and D e ~ i . ~ Van Cauwelaert, et aL,1 claimed that the appearance of distinct shoulders could not be attributed to improper atmospheric compensation because the shoulders would not shift with changes in temperature. However, this would not necessarily be true if the shoulders were due to inverse absorption in an improperly compensated spectrometer if the peak position of the SiOH band changes with temperature. They state, however, that a perfectly straight baseline was obtained in the 3750-cm-I spectral region when the silica sample was removed from the sample beam of the double-beam spectrometer. None the less, it is not (1) F.
H. Van Cauwelaert,
P. A. Jacobs, and J. B. Uytterhoeven, J
Phys. Chem., 76, 1434 (1972). (2) M. L. Hair and W. Hertl, J. Phys. Chem., 73,2372 (1969). (3) J. A. Hockey, J. Phys. Chem., 74,2570 (1970). (4) 6.A. Morrowand A. Devi, Can. J. Chem., 48,2454 (1970). The Journal of Physical Chemistry, Vol. 77, No. 11, 7973
1466
3144.1
B
I
3760
1
3740
I
3720
I
3700
cm-’ Figure 1. A. Single beam spectrum of residual atmospheric water with complete flushing. B. Double-beam spectrum of uncompensated cell. C. Single-beam spectrum with sample chamber exposed to the atmosphere. T h e transmittance scale applies to spectrum C; the other spectra are linearly displaced for purposes of presentation.
clear whether a compensating cell of equivalent pathlength was placed in the reference beam during their experimental runs, and no quantitative indication was given of the extent to which their spectrometer was purged with dry air. In their band-fitting work, no mention is given of the accuracy of their band-fitting procedure, nor is an overall computed spectrum shown. The validity of assuming a constant shape ratio (0.65) for the Lorentzian-Gaussian band function for all temperatures is questionable, as is the basic assumption that because a single OH band in other oxides gave this ratio, that this ratio should apply to the reputedly more complex overlapping band system in the case of silica. Finally their band parameters (Particularly the halfwidth) of the low- and high-frequency components do not change appreciably with temperature. This is particularly surprising for a temperature change of 400”. In the present paper we have reinvestigated the infrared spectrum of the isolated hydroxyl groups on silica and have found no evidence for the existence of distihct shoulders. Also included is a discussion of the validity of using band-fitting procedures for resolving closely overlapping band systems.
Experimental Section The silica samples (Cab-0-Sil, H5) used in this work were utilized in the form of pressed disks 2.5 cm in diameter containing 5 mg of silica per cm2 and were pressed in a stainless steel die at about 1000 lb/in.2 for a few seconds. The disks were heated for several hours at 750” in a quartz cell and they were then transferred to a variable temperature cell5 where they were evacuated for 1 hr (this procedure is similar to that used by Van Cauwelaert, et ~ 1 . 1 )About . 10 Torr of helium was added to the evacuated cell to ensure good thermal conductivity. The spectrometer used was a modified Perkin-Elmer Model 1 3 6 filter-grating double-beam ratio recording infrared spectrometer, and spectra in the OH stretching region were recorded with a spectral slit width of 2.2 cm-I at a scan speed of 4.9 cm-I min-l, this value for the slit width being chosen because of an optimum signal-to-noise ratio for digitization purposes. Band shapes did not alter The Journal of Physical Chemistry, Voi. 77, No. 1 7 , 1973
B. A.
Morrow and I. A. Cody
with a narrower slit width, and identical although “noisier” spectra were observed using a 1.4-cm-l spectral slit width. The chopper is before the sample compartment in this spectrometer so that problems due to sample emission a t elevated temperatures are eliminated. After dry air flushing, the total single-beam absorption due to residual water is shown in Figure 1A and a perfectly straight base line could be obtained in double-beam operation either with or without the sample cell in place, provided all sample areas were totally flushed with dry air. When we refer below to an “uncompensated cell,” this means spectra recorded with the sample chamber (13-cm path length) totally open to the atmosphere and with the sample cell (path length 7 cm) in the sample beam only. The residual “uncompensated” spectrum is shown in Figure 1B and the total single-beam atmospheric spectrum under those conditions is shown in Figure 1C. For computational purposes, the spectra were digitized at 0.2-cm-1 intervals between 3771.0 and 3684.2 cm-I and band-fitting computations were carried out on an IBM 360165 computer using the programs developed by Pitha and Jones6 for fitting a Lorentzian-Gaussian product or an unrestrained Lorentzian-Gaussian sum function to spectral bands. Plots were recorded on a Milgo plotter. For a discussion the various types of mathematical function used to fit infrared spectra, the reader is referred to several original articles on the subject.? The accuracy of our various computer calculations is expressed in terms of the following parameters: DIS (discrepancy) is the root mean squared of the residual differences between the calculated and experimental spectrum in transmittance; WFM is the wave number of the maximum discrepancy; MAX is the maximum discrepancy at WFM. Other parameters are the half-width, the shape ratio (t,he Lorentzian-Gaussian ratio), and the peak intensity in absorbance. All spectra were calibrated against atmospheric water bands and the experimental data reported in this investigation are believed to be accurate to &0.2 cm-’.
Results Experimental spectra of the SiOH band of silica at -145, 25, 200, and 400” are shown in Figures 2A and 3A. The total band area and the wave number of the maximum absorption in each case is given in Table I. These spectra are very different from those shown by Van Cauwelaert, et aL,1 a t the same temperatures (except for the spectrum at -145“ which is unique to this work) in that the distinct shoulders on either side of the main absorption peak are absent. Since the spectra observed by Van Cauwelaert, et al., had shoulders to high and low wave number of the main absorption band, a band analyzer was used to decompose this band into three components. They used a Voight type Lorentzian-Gaussian product function assuming a 65% Lorentzian shape ratio in this fitting, and although they do not say what their criterion of a good “fit” was, we will not dispute that a reasonable fit can be achieved. However, there is no a priori reason to assume that the same shape ratio should apply to all bands, nor is one justified in assuming that the same shape ratio should apply over a (5) 6.A. Morrow, J. Sci. instrum., 43,487 (1966). (6) J. Pitha and R. N. Jones, NRC Bulletin No. 12 and 13, National Research Council of Canada, Ottawa, 1968. (7) (a) R. N. Jones, Appl. Opt., 8, 597 (1969); (b) K. S. Seshadri and R. N. Jones, Spectrochim. Acta, 19, 1013 (1963); (c) J. Pitha and R. N. Jones, Can. J. Chem., 45,2347 (1967).
1467
Spectra of the Isolated Hydroxyl Groups on Silica
400°C
200°C
A 05-
C 05
3760
3720
3760
3720
cm-'
cm-'
Figure 2. A. Experimental spectra of SiOH band at -145 and 25". B. Computed spectra using a two-component sum function. C. Computed spectra using a three-component product function.
400". B.
TABLE I : Experimental Data Figures 2A and 3A
TABLE II: Parametersfor
for the Spectra shown in
Tepp,
c
Wave number, c m - l (at max intensitv)
Peak intensity in absorbance
Band area
-145 25 200 400
3751.4 3748.2 3744.7 3740.6
0.97 0.76 0.57 0.41
6.56 6.1 1 6.05 5.63
400" temperature range.? The parameters derived from a
mathematical analysis of closely overlapping band systems depend very much on the type of function used and on the number of components assumed. Therefore, it is doubtful that their calculated band parameters are unique and certainly without stating some measure of the error of their fit, they are a t best a rough guide. As an illustration of the above point, we have carried out band-fitting analysis of our own spectra, using a Lorentzian-Gaussian sum function and a Lorentzian-Gaussian product function.6 Since there is a tail to low wave number of the main band, we tried to fit the spectra to a two-component curve (one strong band a t the peak position and one weak band in the region of the tail) and to a three-component curve (one strong band with two weak bands in the tail). Figures 2B and 3B show the plotted results for the two band sum function and Figures 2C and 3C show the results for a three-band product function. The data obtained from the use of these functions are collected in Table 11. In all cases we consider that a good fit was obtained when the discrepancies were 0.01 or less, this particular value being satisfactory for a very intense narrow band where errors are likely to be large where the band rises most steeply (hence DIS is greater with the low-temperature spectra). The values of the parameters by which we judged the accuracy of a particular fit for the functions used in Figures 2 and 3, and for two other trial functions, are shown in Table 111. DIS is smaller for the three-component curves and we could have improved our
Figure 3. A. Experimental spectra of SiOH band at 200 and Computed spectra using a two-component sum function. C. Computed spectra using a three-component product function.
Function and ( N ) U
Teomp, C
the Computed Spectra in Figures 2 and 3
Wave number, cm-',
intensity absorbance
Halfwidth, cm-'
3750.8 3746.7 3751.2 3748.5 3740.8 3748.0 3742.7 3748.2 3745.4 3735.1 3744.5 3738.5 3744.7 3741 .O 3730.8 3740.7 3733.6 3741.2 3738.1 3730.7
0.84 0.38 0.80 0.33
4.1 8.8 3.2 6.1 17.7 4.7 11.4 4.1 8.7 23.9 5.9 13.8
0.05 0.6% 0.14 0.58 0.22 0.03 0.51 0.13 0.45 0.15 0.02 0.36 0.10 0.30 0.1 1 0.06
5.5 12.1 33.9 7.8 15.7 7.2 9.3 22.6
Shape ratiob
0.0 1 .o 0.54 0.70 0.99 0.1 1 0.88
0.55 0.70 0.99 0.29 1 .o 0.64 0.69 0.99 0.38 0.52 0.79 0.41 0.99
S denotes the Lorentzian-Gaussian sum function and P denotes the L-G product function. The number in parentheses is the number of components which were fitted. b The shape ratio is the fraction of Lorentzian character in a given band.
fit by increasing the number of components, but this is meaningless for curves where no distinct shoulders are apparent. We could also have obtained a satisfactory fit with a weak high wave number band, since any true Lorentzian or Gaussian band can be decomposed into as many components as desired. We consider that the two-band fit in itself is not significant because the tail t o lower wave number is probably just the manifestation of intramolecular perturbations which are bound to be present in a sample of this type The Journai of Physical Chemistry, Voi. 77, No. 1 1 , 1973
1468
B. A. Morrow and I. A. Cody
TABLE Ill: Computed Parameters which Reflect the Accuracy of Some Trial Functions Function and (A')=
Temp, "C
DISb
WFMb
MAXb
-145 -145 -145
0.014 0.014 0.009 0.004 0.01 0 0.01 0
3747.0 3752.2 3745.2 3752.2 3749.2 3746.8 3749.4 3746.8 3749.0 3766.4 3751.6 3749.8 3742.4 3742.6 3765.4 3751.4
0.039
-145 25 25
25 25 200
200 200 200 400 400 400 400 See footnote a, Table I I. of these terms. a
0.008 0.005 0.007 0.006 0.006 0.003
0.005 0.005 0.007
0.003
0.033 0.021
0.013
0.038 0.031 0.021 0.014 0.016 0.014 0.016 0.008 0.016 0.015
0.015 0.008
0.0 W
z 0 a + t J,
a
+-[L .05 I
I
I
See Experimental Section for description
where the minute silica particles are in intimate and random contact with their neighbors. This view has been discussed by Hambleton, et al.,s and has been expressed by Van Cauwelaert, et al.1 Further, no significance is to be attached to the shape ratios listed in Table 11. Our band fitting programs vary the shape ratio so as to obtain a best fit, and one may note that for the product function our shape ratios are near those used by Van Cauwelaert, e t a l . , I whereas for the sum function, the main band was always given a high Gaussian character, and the bands in the tail, a Lorentzian profile. The latter is not unexpected when an intense band is degraded to one side as in our case, since for equal height and area, a Gaussian function has a greater breadth near the peak.7b Nor is a Gaussian profile unexpected for very narrow intense absorption bands because the instrumental distortion factors essentially introduce a Gaussian perturbation on the true band shape.7b Therefore, as the half-band width approaches the spectral slit width (particularly true a t low temperatures) then it is to be expected that the Gaussian contribution to an experimentally obtained profile will increase. Our spectra for deuterated silica in the SiOD stretching region (2760 cm-I) were the same as those reported by Van Cauwelaert, et al.1 These were fitted by the latter to a two-band component curve. We also obtained spectra of mixed isotopic species by partially deuterating the sample and again, an excellent two or three band fit to the experimental data could be achieved.
Discussion As a result of this investigation, the following observations have been made concerning the infrared spectrum of the isolated hydroxyl groups on highly dehydroxylated silica: (1) the spectral band profiles attributable to isolated surface OH or OD groups are similar, (2) no distinct shoulders appear in the SiOH band profile as the temperature is changes, (3) both the SiOH and the SiOD band are slightly asymmetric to the low wave number side of the main band, and (4) both the SiOH and the SiOD spectra can be fitted using band-fitting techniques to a two- or a three-component curve, as can the spectra of the mixed isotopic species. In conclusion, there is no spectroscopic The Journal of Physical Chemistry, Vol. 77. No. 11, 1973
3760
3740
3720
3700
cm-' Figure 4. Experimental spectra of the S O H band with an uncompensated cell: A, 400"; B, 200"; C, 25'; D, -145". The scan speed used was 20 c m - l / m i n and the peak positions are slightly shifted relative to the slower scans used for Figures 2 and 3. The dashed line corresponds to the position of the 3744.5-cm-' band in Figure 1. The transmittance scale applies to spectrum D, the other spectra have been linearly displaced for convenience of presentation.
evidence to suggest that the infrared spectrum of the SiOH groups is in any way different from t,hat of the SiOD groups. Our failure to detect distinct shoulders in our spectra of the SiOH band on silica is in contradiction to the experimental results of Van Cauwelaert, e t al.1 In an attempt to try to duplicate their spectra, we considered the possibility that because gaseous atmospheric water so strongly absorbs in this spectral region, then by intentionally introducing a slight imbalance in the double-beam compensation we might be able to reproduce spectra resembling theirs. The spectra shown in Figure 4 were recorded under conditons in which complete atmospheric compensation was not achieved ,(see Experimental Section). These spectra clearly show shoulders on both sides of t,he main band and it can be seen by comparing the spectra in Figures 4 and 1 that the position of the bands of maximum intensity in the atmospheric water spectrum' (or in the uncompensated double-beam spectrum) coincide with the positions of the,main shoulders in the spectra shown in Figure 4. A dashed line has been drawn in Figure 4 to indicate the position of the strongest band a t 3744.5 cm- I. In the experimental spectra shown in Figures 2A, 3A, and 4, the peak position shifts to higher wave number, the peak intensity increases, and the band width decreases as the temperature is lowered. Since the positions of the atmospheric water bands do not change with temperature, the shoulders in the uncompensated spectra apparently (8) F. H. Hambleton, J. A. Hockey, and J. A. G. laylor, Trans. Faraday Soc., 62, 801 (1966).
Spectra of the Isolated Hydroxyl Groups on Silica shift with temperature. The spectra in Figure 4 closely resemble those shown by Van Cauwelaert, et aL,l in their Figure 1, except that the “dips” in our spectra due to improper compensation are quite large. Van Cauwelaert, et al., have assured us in a private communication that, as stated in their paper, their own atmospheric compensation was perfect and that the irregular fluctuations to either side of the main band in their spectra are simply due to random electronic noise. Van Cauwelaert, et al.,l also studied the effect on the band shape of adsorbing benzene and triethylamine on silica. With benzene, they reported that the overall band intensity decreased but that the peak position did not change, whereas the peak position changed on adsorbing triethylamine and the relative intensity of the three-components changed. We have also observed these effects, only we found that with an uncompensated spectrometer the relative intensity of the shoulders varied because again, as the peak shifts, the dips due to atmospheric water remained fixed, as they do with temperature changes. With a perfectly compensated spectrometer, no shoulders were observed, although the strong central band decreased in intensity at a faster rate than did the tail to lower wave number. This is not unreasonable since a physically adsorbed species would be expected to interact preferentially with the totally free hydroxyls, as opposed to those which might be perturbed due to intramolecular interactions and hence absorb a t slightly lower wave number. It is difficult to account for the differences in the spectra observed by Van Cauwelaert, et a l . , I and ourselves on chemical grounds. The silica samples used in both investigations were prepared by similar methods and the overall band widths at half-peak height are comparable at each temperature. It is always possible that our samples were quite different owing to different pressures used for preparing disksg (Van Cauwelaert, et al., did not state their conditions). The fact that the relative positions of shoulders observed by ourselves do not exactly coincide with theirs is possibly just a manifestation of the sensitivity of the SiOH peak positon to conditions of preparation.8 Alternatively, this could be due to slight differences in sample temperatures, since the peak position shifts by about 0.20 cm -1/10” change in temperature.
1469
As for the question of the peak shift and changes in band width with temperature, a discussion of this is not warranted here. Such questions are somewhat involved from the theoretical point of viewlo and even for well-ordered systems (gases and crystals) theory and experiment still do not exactly agree.11 However, peak intensification and band narrowing with decreasing temperature is not unexpected.12 We have only just started an investigation of the band shape problem for adsorbed species from a theoretical point of view and we will discuss this topic when the work is complete. Finally, there are two points to be stressed concerning the general use of band-fitting techniques. One is a plea for caution in applying such methods for the “resolution” of overlapping band envelopes. It is most important to indicate one’s criteria for a “satisfactory fit,” since an infinite number of such fits can be obtained if the permissible discrepance (DIS) is large enough. This is particularly importa-nt when the bands are closely overlapping. Secondly, when the spectral slit width approaches the half-band width of any component, the instrumental distortion may determine the nature of the experimentally observed band pr0file.7bJ~>~3 Unless the instrument function is known accurately, such analyses can be subject to error.
Acknowledgment. We are grateful to the National Research Council of Canada and to the Department of National Health and Welfare for financial support for this work. We also wish to thank Dr. R. N. Jones of the National Research Council for many helpful discussions concerning the utility of his band-fitting programs, and Mr. L. W. Thomson for help with the experimental work. F. H. Hambleton, J. A. Hockey, and J. A. G. Taylor, Nature (London), 208, 138 (1965). K. H. lllinger and D. E. Freeman, J. Moi. Spectrosc., 9, 191 (1962); D. P. Chock, J. Jortner. and S. A. Rice, J. Chem. Phys., 49, 610, (1968); J. Horiuti and T. Toya, “Solid State Surface Science,” Vol. 1, Marcel Dekker, New York, N. Y., 1969, p 43. J. C. Breeze, C. C. Ferriso, C. B. Ludwig, and W. Malkmus, J. Chem. Phys., 42, 402 (1965); M. P. Lisitsa and Yu. P. Tsyashchenko, Opt. Spectrosc., 10, 79 (1961). D. A. Dows, “Physics and Chemistry of the Organic Solid State,” Vol. 1, Interscience, New York, N. Y., 1963, pp 675 and 682. R. N. Jones, R. Venkataraghavan. and J. W. Hopkins. Spectrochim. Acta, PartA, 23,925, 941 (1967).
The Journal of Physical Chemistry, Vol. 77, No. 1 7 , 1973
