J. Phys. Chem.

1981, 85, 2543-2546 2543

Determination of I R Extinction Coefficients for Linear- and Bridged-Bonded CO on

Supported Palladium

M. Albert Vannice‘ and S-Y. Wang

Department of Chemlcal Englneerlng, The Pennsylvania State University, University Park, Pennsylvania 16802
(Received: March 3 1, 198 1; In Final Form: June 2, 198 1)

Infrared spectra were obtained of CO adsorbed on Pd dispersed on four supports: 7-A1203,SO2,Si02-A1203,

and Ti02. All catalysts were characterized by both H2 and CO chemisorption which allowed the calculation
of the surface coverages of both linearly adsorbed CO, which gives a band above 2000 m-l,and the bridged-bonded
form, which produces a band below 2000 cm-’. From these values, integrated absorbances, A, and extinction
coefficients, e, have been determined for each adsorbed species on Pd for the first time. For individual catalysts,
absorbances and extinction coefficients were always much higher for the bridged-bonded CO than the linearly
adsorbed co, i.e., average values were Ab = 85 X lo’ cm mol-’, A, = 3.3 X 10’ cm mol-’, €b = 4.1 X lo6 cm2
mol-’, and €1 = 0.36 X lo6 cm2mol-’. Because a specially designed IR reactor cell was utilized, spectra were
also taken under steady-state CO hydrogenation reaction conditions to form CH4. The A1203-, SO2-,and
Si02-A1203-supportedPd catalysts still exhibited strong CO bands; however, Pd/Ti02 wafers exhibiting SMSI
(strongmetal-support interactions)showed no detectable CO bands although they were the most active catalyst.
Using the E values obtained in this study, we estimated the maximum surface coverage on Pd to be less than
1%for the linear species and 0.1% for the bridged species.

Since the first application of IR spectroscopy by Eis- coefficients, associated with band maxima, and integrated
chens and co-workers to study supported catalysts,’ this absorbances, which are related to the area under each
technique has been in the forefront of approaches to study band, could be obtained for each of these two major bands.
adsorbed molecules on metal surfaces. Although numerous Such individual values have been published only recently
papers have dealt with CO adsorption on group VI11 for CO adsorbed on Rh and Pt,‘ls and values for other
metals, few studies have quantitatively related the inten- group VI11 metals have not yet been reported, principally
sities of the observed spectra for adsorbed CO with surface because of the difficulty in determining the representative
Only Pt and Rh have been examined fractions of linearly adsorbed CO and of multiply bonded
until now, and no such studies have been reported for CO, which is most likely to be the bridged-bonded (2-fold
palladium. Undoubtedly this is a consequence of the re- coordination) species. In a recent study, we obtained IR
quirement that accurate CO concentrations on the metal spectra for CO chemisorbed on Pd dispersed on four dif-
must be known at the time the spectrum is recorded, and ferent support materials (7-A1203,Si02, Si02-A1203and
the design of most IR cells has not allowed these uptake TiOJ and each catalyst was carefully characterized by both
results to be obtained. In addition, many CO spectra have CO chemisorption and hydrogen chemisorption measure-
been obtained after evacuation, which can alter surface ments.12J3 This allowed the calculation of fractional
coverage and complicate the determination of CO surface coverages of linear- and bridged-bonded CO and the es-
concentrations. timation of extinction coefficients and integrated absor-
More than one IR band for adsorbed CO frequently bances for each CO band. The influence of Ti02 on CO
occurs on metal surfaces, and typically a high-frequency adsorption on Pd was of special interest because of SMSI
(HF) band exists between 2000 and 2100 cm-l along with behavior which can be induced in this catalyst system.14
a low-frequency (LF) band between 1800 and 2000 cm-’ This paper reports these results.
which is apt to be much broader and may appear to con-
tain several p e a k ~ . ~ pThe
~ J ~HF band is usually associated Experimental Section
with linearly adsorbed (single site) CO whereas multiply The Catalysts were prepared with aqueous impregnation
coordinated CO is assumed to give rise to bands in the LF techniques by using PdClz (Ventron Corp.), and after im-
region.’J’ It would be especially informative if extinction pregnation the samples were dried in air at 393 K for 16
h, bottled, and stored in a desiccator. The support ma-
terials used were Grade 57 SiOzand Grade 979 Si02-A1203
(1)Eischens, R. P.; Francis, S. A.; Pliskin, W. A. J. Phys. Chem. 1956, (Davison Chemical Co.), P-25 TiOz (Degussa Co.), and
60,194. 7-A1203donated by Exxon Research and Engineering Co.
(2) Eischens, R. P.; Pliskin, W. A. Adu. Catal. 1958, 10,1.
(3) Darensbourg, D. J.; Eischens, R. P. “Catalysis”; Hightower, J. W., The H2 (99.999%), CO (99.99%), and He (99.9999%), all
Ed.; North Holland: Amsterdam, 1973; p 371. from the Linde Co., were further purified before use. A
(4) Heyne, H.; Tompkins, F. C. Trans. Faraday SOC.1967,63, 1274. Deoxo unit (Engelhard) and an Oxy-trap (Alltech Assoc.)
(5) Shigeishi, A.; King, D. A. Surf. Sci. 1976, 58, 379.
(6) Seanor, D. A.; Amberg, C. H. J. Chem. Phys. 1965, 42, 2967. were used for hydrogen, an Oxy-trap was employed for the
(7) Vannice, M.A.; Twu, C. C.; Moon, S. H. Submitted for publica- helium, and the CO was passed through a molecular sieve
tion.
(8)Duncan, T. M.; Yates, J. T., Jr.; Vaughn, R. W. J. Chem. Phys.
1980,73,975. (11)Bradshaw, A. M.;Hoffmann, F. M. Surf. Sci. 1978, 72,513.
(9) Little, L. H.“Infrared Spectra of Adsorbed Species”; Academic (12) Wang, S-Y., Ph.D. Thesis, Pennsylvania State University, 1980.
Press: New York, 1966. (13) Vannice, M. A.; Moon, S. H.; Wang, S-Y. J. Catal. In press.
(10)Hair, M. L. “Infrared Spectroscopy in Surface Chemistry”;Marcel (14) Tauster, S. J.; Fung, S. C.; Garten, R. L. J.Am. Chem. SOC.1978,
Dekker: New York, 1967. 100,170.

0022-3654/81/2085-2543$01.25/0 0 1981 American Chemical Society

2544 The Journal of Physical Chemistry, Vol. 85, No. 17, 1981 Vannice and Wang

trap. Weight loadings were determined by using neutron I I I I I I

activation analysis. Details on the materials and prepa-
rative technique are provided elsewhere.13J5
Adsorption measurements were performed in a con-
ventional, mercury-free, glass adsorption system capable
of achieving an ultimate vacuum near 4 X torr (5.3
X Pa). Pressures during adsorption runs were mea-
sured with a Texas Instruments precision pressure gage.
This system has been described previo~s1y.l~The IR
spectra were obtained in a specially designed cell which
could be heated above 700 K to allow in situ pretreatment
(reduction) in Hz and could also be operated as a differ-
ential, plug-flow reactor, if desired, to obtain spectra of CO
adsorbed on a catalyst surface under reaction condi-
tions.l3Js All spectra were recorded on a Perkin-Elmer
580 dual-beam spectrophotometer in which a pure support T

wafer was placed in the reference beam and the supported

catalyst wafer was placed in the sample beam. The re-
corded spectra represent the difference in transmission
between the beams passing through these two wafers, and
this technique very efficiently minimized or eliminated
contributions of gas-phase CO and CO physically adsorbed
on the support surface because identical gas mixtures were
flowed through each cell. As a result, spectra under CO
pressures as high as 500 torr could easily be obtained;
however, all results reported here were at a constant CO
pressure of 190 torr (0.25 atm). The total pressure in the
cell was 1atm (101 kPa) with the balance composed of He.
The wafers normally weighed 50-60 mg and were 0.2 mm
thick after being pressed into the stainless steel cylinder
which comprised the cell. The cell design forced gas flow
through the wafer, and high space velocities with low I I L I I I
22011 21100 1800
pressure drops could be attained. Precise details of the
design of this cell and the IR system have been previously w (an-')

published. l6 Figure 1. Transmittance spectra of CO adsorbed on Pd at 300 K, P,

The Pd/A1203,Pd/Si02, and Pd/SiOZ-Al2O3catalysts = 190 torr, P, = 570 torr: (A) 1.98% Pd/A1901, (B)1.86% PdITIO.
were reduced 1 h at 673 K in flowing H2in the adsorption (448 K), (C) 1.53% Pd/SI02; (0) 2.12% Pd/&Ai20~, (E) i.86./0'
system following a specified pr0~edure.l~ Two different Pd/TI02 (SMSI).
Pd/Ti02 catalysts were prepared by using the two re- TABLE I: CO Bands in He at 300 K
duction procedures of Tauster et al.;14 i.e., reduction at 448
K provided a Pd/Ti02 catalyst with normal adsorption HF LF
behavior while reduction at 773 K produced an SMSI state (A) 1.98% Pd/s-A1,0, 2085 1970 (s), 1930
which markedly suppressed Hz and CO ads0rpti0n.l~ (B) 1.86% Pd/TiO, 2096 1926,1890 (s)
Hydrogen and CO uptakes were measured by using the (C)1.93% Pd/SiO, 2085 1985,1938
dual isotherm technique of Benson, Hwang, and Boudart17 ( D ) 2.12% Pd/SiO,-Al,O, 2098 1965
(E) 1.86% Pd/TiO, (SMSI) 2058 1925
to determine chemisorbed hydrogen at 300 K, and the
method of Yates and Sinfelt'* was employed to determine
irreversibly adsorbed CO at 300 K. With the exception Results and Discussion
of Pd/Ti02, these pretreated catalyst powders were used, The spectra obtained at 300 K (and higher tempera-
after passivation in air, to press the wafers for the IR study, tures), along with baseline behavior in the absence of CO,
and the wafers, including the reference wafer, were then are reported elsewhere,13but the transmittance spectra are
reduced in situ at 548 K. After this reduction step, ad- shown again in Figure 1 for comparison purposes. The
sorbed hydrogen was removed by flowing He through the narrower HF band and the broader, more dominant LF
cell a t 548 K. The absence of H2 in the gas phase was band typically found for supported Pd catalysts are clearly
determined by using a gas chromatograph to analyze the evident. The baseline intensity, To,for each support was
exit stream from the IR cell. The wafers were then cooled superimposed on the CO spectra in the region between
under flowing He to 300 K and a baseline was obtained 2100 and 2600 cm-', and relative intensities, TIT,, were
in pure He before CO was introduced into the gas stream. calculated on this basis. The CO peak positions in Figure
The Pd/Ti02 wafer was pressed from fresh, unreduced 1 are listed in Table I.
catalyst powder and IR spectra were obtained after an in Small shifts (5-10% T )were sometimes observed in the
situ reduction at 448 K. The wafer was then further re- baseline obtained in pure He and that measured during
duced at 773 K to induce SMSI behavior and cooled in He the spectrum for adsorbed CO, which introduced some
to 300 K, and the spectra were obtained. uncertainty in the absorbance values. The integrated form
of the Beer-Lambert law is a(v) = (Cl)-' In T o / T ,where
(16)Palmer, M. B., Jr.; Vannice, M. A. J. Chem. Tech. Biotechnol, a(v)is the absorption coefficient, which can be dependent
1980, 30, 205. on the radiation frequency, v, C is the molar concentration
(16)Vannice, M. A.; Moon, S. H.; Twu, C. C.; Wang, S-Y. J. Phys. E. of CO in the sample, I is the distance through the sample
1979, 12, 849.
(17) Benson,J. E.; Hwang, H. S.;Boudart, M. J.Catal. 1973,30,146. through which the IR beam penetrates, and T and Toare
(18) Yates, D. J. C.; Sinfelt, J. H. J. Catal. 1967, 8, 348. the transmitted radiation intensities in the presence and
Linear- and Bridged-Bonded GO on Supported Palladium The Journal of Physical Chemistty, Vol. 85, No. 17, 198 1 2545

TABLE 11: Extinction Coefficientsa and Integrated Absorbancesb for CO Adsorbed on Pd at 300 K
chemisorption on
used sample,
pmol g-’ 10-7x 10-7x io-’X 10-6 x x
A, Al, Ab, fbC EbrC
catalyst H* CO col/cob cm/mol cm/mol cm/mol cm2/mol cm*/mol
1.98% Pd/q-Al,O, 19 23 0.53 25 6.7 35 0.66 2.1
1.93% Pd/SiO, 6 11 10 19 1.4 190 0.16 9.3
2.12% Pd/SiO,-Al,O, 16 27 4.4 6.9 1.9 29 0.26 0.87
1.86% Pd/TiO, (448 K) 16d 41d 6.8
7.6
1.86% Pd/TiO, (SMSI) 2.0 7.0 9.2
1.86% Pd/TiO, (SMSI) 1.5e 5.0e 14
average 14 3.3 85 0.36 4.1
gas-phase CO (ref 21) 0.042 0.00018
E = log (T,/Tmax)/(CZ). A =J u dv/(CZ). At band maxima in Figure 1. Based on fresh, reduced catalyst. e Up-
takes at 548 K.

absence of the adsorbate, respectively. From a plot of In

(To/T)vs. v, such as those shown in Figure 2, the total
intensity of a given band was obtained by graphical inte-
gration using a planimeter. For each catalyst, an overall
integrated absorption coefficient, A = ~,,‘a(v)dv, was
obtained based on all adsorbed CO species under both
bands using CO uptakes per gram determined in the ad-
sorption experiments and the diameter of the wafer (di-
ameter = 15.9 mm). The sample thickness, 1, is not needed
because C1= Um/A where U is uptake in mole of CO g-l,
m is the wafer mass, and A is the cross-sectional area of
the wafer.
Integrated absorption coefficients for the HF band, A,,
and the LF band, Ab, and extinction coefficients deter-
mined at the frequencies corresponding to these two band

-
maxima, el and q, where E = 1/C1log (To/T-), cannot be
determined unless the fractions present as linearly ad-
sorbed CO, COI,and bridged-bonded CO, cob, are known.
The assumption that these two forms of adsorbed CO
predominate on Pd is supported by the model of Bradshaw
and Hoffmann, which proposes that CO in threefold or
higher coordination should have frequencies below 1880
cm-l.ll Since intensities are quite low below 1880 cm-l
(Figure I), this assumption seems justified; however, this
is the only assumption which allows an estimation of these <El

four parameters. Because hydrogen adsorption was also 2100 2000 1900 1800
measured on these catalysts and its adsorption stoichiom- u (Cm-1)

etry is established, i.e., H,d/Pd, is near unity,” where Pd,

is a surface Pd atom. This technique provides a value for Figure 2. Absorbance of CO adsorbed on Pd at 300 K, P,, = 190
the total number of surface Pd atoms (Xp,J. It is assumed torr, PHe= 570 torr: (A) 1.98% Pd/Ai,03, (B) 1.86% PdlTIO, (448
K), (C) 1.93% Pd/SiO,, (D) 2.12% Pd/Si02-Ai,03, (E) 1.86% Pd/TlO,
that a surface Pd atom interacts with no more than one (SMSI).
CO molecule, either in the bridged form (COb/Pd, = 1/2)
or in the linear form (COl/Pd, = 1). The assumption that The overall integrated absorbances, A , are in good
only one- and two-site CO adsorption occurs produces two agreement, vary by a factor less than four, and have an
straightforward algebraic equations involving CO and H2 average value of 14. Considering that adsorption on the
uptakes from which surface fractions of linear and used Pd/Ti02 sample reduced at 448 K should be some-
bridge-bonded CO can be determined, i.e. what lower than on the fresh sample, the actual A value
2(Hz uptake) = Xpd, for this catalyst is probably a little higher than that re-
ported. The similarity of these A values indicates that no
dramatic differences seem to exist in the interaction of
adsorbed CO with incident IR radiation in these different
[cob]+ [CO,] = CO uptake Pd catalyst systems. In addition, it is clear that A and E
values for adsorbed CO are orders of magnitude greater
The values of AI, Ab, E,, and tb based on these fractions than those for gas-phase CO, a fact which has been rec-
are listed in Table 11. Because some decrease in the Pd ognized2p9and which facilitates the application of this
surface area occurred during the course of these studies, spectroscopic technique to characterize CO adsorption on
adsorption values on used samples, listed in Table 11, were metal surfaces.
employed in these calculations. The altered state of ad- Integrated absorbances and extinction coefficients could
sorption on Pd/Ti02 samples, especially H2 adsorption on be calculated for both the HF and LF species of CO on
the SMSI sample, prevented calculation of values for in- the three catalysts exhibiting “normal” adsorption be-
dividual bands and only an overall A value could be de- havior. For the HF band, A1 values showed a fourfold
termined for these two catalysts. variation while a larger spread occurred in Ab values. The
2546 The Journal of Physlcal Chemistry, Vol. 85, No. 17, 198 1 Vannice and Wang

I I I I I I frequency2J1V2lis therefore expected for the multiply

bonded CO.
This picture was supported by desorption experiments
during which the HF band always disappeared first.12
Carbon monoxide is invariably assumed to chemisorb via
an interaction between the carbon atom and the metal
surface,2*22and, as a consequence, a stronger adsorption
bond might result in a larger dipole moment derivative and
produce a larger extinction c o e f f i ~ i e n t . ~Based
~ on this
argument the predicted order of extinction coefficients
would be 6b > el >> which is the trend found in this
study.
The usefulness of these values for A and 6 is demon-
strated by in situ studies in which simultaneous kinetic
data and IR spectra were obtained for Pd methanation
catalysts under well-defined, steady-state reaction con-
ditions after H2was substituted for flowing He at 548 K.13
High band intensities similar to those at 300 K were ob-
served for typical Pd catalysts under reaction conditions
at 548 K, indicating high surface coverages of CO; however,
the Pd/Ti02 (SMSI) catalyst exhibited markedly different
behavior, no IR bands were detectable under reaction
conditions, as shown in Figure 3, despite its high activity.
The results in Table I1 show that A values are very similar
for CO adsorbed at 300 and 548 K and are not strongly
dependent on temperature. Therefore, if the average q and
€b values in Table I1 are also assumed not to be strongly
dependent on temperature or on adsorbed hydrogen,
maximum surface coverages on Pd under reaction condi-
1890 tions can be estimated. The first assumption is supported
I 1 I I 1 I by the spectrum of CO in He obtained 548 K for the
2200 2000 1800 Pd/Ti02 (SMSI) catalyst. The CO uptake was 5.0 pmol
v (cm-1) g-l at 548 K resulting in an A value of 14 X lo-’, as shown
in Table 11, which is very close to values obtained a t 300
Figure 3. IR spectra of CO adsorbed on Pd under reaction conditions; K. The second assumption is supported by the study
, = 190 torr, PH = 570 torr: (A) 1.98% Pd/AI O3 ( N
T = 548 K, P
= 3.3 X s-‘); (6)1.86% $d/TiO, (448K) ( N = 2.0 X lo3 s-‘); Conrad, Ertl, and Latta, which showed that essentially no
(C)1.93% Pd/SiO, ( N N 0.1 X lo-’ s-I), (D) 2.12% Pd/S102-Ai203 interaction occurs between adsorbed CO and hydrogen on
( N = 1.8 X s-’), (E) 1.86% Pd/TiO, (SMSI) ( N = 40 X s-’). Pd single-crystal surfaces.24 With a minimum sensitivity
The number N in parentheses represents the turnover frequency of of 1%transmittance for each peak, the minimum de-
methane formation (molecules (Pd site)-’ s-‘) occurring during the tectable coverage of linearly adsorbed CO is 0.39 pmol g-l
recording of the spectrum. and the minimum coverage of bridged-bonded CO is 0.033
pmol g-l. Assuming the total Pd surface area measured
unusually high value of Ab for Pd/Si02 may well be due after the 448 K reduction is not changed by the 773 K
to the very low surface fraction calculated for the reduction, less than 1% of the Pd surface is covered by
bridged-bonded species. Gas uptakes were low on this the HF CO species and less than 0.1% is covered by the
catalyst and even a small error in the Hz uptake would LF CO species. The evidence that Pd surfaces of very
create a large variation in the calculated amount of C o b active methanation catalysts, such as Pd/Ti02 in the SMSI
However, the major finding of this study is that Ab values state, can have very low CO coverage has led us to conclude
are always much greater than AI values, and 6b is much that only a small fraction of Pd surface atoms constitute
higher than €1 for any particular catalyst. This conclusion “active sites” for methane formation,26and illustrates the
is true even if more than one form of multiply coordinated usefulness of in situ IR characterization of catalysts under
CO comprise the LF band; however, the Ab values calcu- well-defined reaction conditions.
lated here represent an average of all the species producing
the LF band. This is the first quantitative determination Acknowledgment. Equipment used in this study was
of these values for each of the two major CO species on purchased under Contract No. EG 778-02-4463, Depart-
Pd and it confirms the estimate of Palazov, Chang, and ment of Energy, Division of Basic Energy Sciences. Ad-
Kokes that the extinction coefficient of the LF CO species ditional funds were provided by a grant from the Research
on Pd is about an order of magnitude higher than that for Corporation. Support for S-Y. Wang was provided by the
the H F species.lg Doyen and Ertl found that multiply government of the Republic of China in Taiwan.
coordinated CO species on Pd have a stronger metal-
carbon adsorption bond, i.e., a higher heat of adsorption (21) Twu, C. C., Ph.D. Thesis, Pennsylvanic State University, in
progress.
than linearly adsorbed C0,20and a lower C-0 stretching (22) Blyholder, G.; Allen, M. C. J. Am. Chem. SOC.1969, 91, 3168.
(23) Norton, P. R.;Goodale, J. W.; Selkirk, E. B. Surf. Sci. 1979,83,
189.
(19) Palazov, A,; Chang, C. C.; Kokes, R. J. J. Catal. 1975, 36, 338. (24) Conrad, H.; Ertl, G.; Latta, E. E. J. Catal. 1974, 35, 363.
(20) Doyen, G.; Ertl, G. Surf. Sci. 1974, 43, 197. (25) Wang,S-Y.; Moon, S. H.; Vannice, M. A. J . Catal. In press.