J. Phys. Chem. 1994, 98, 8549-8554
8549
Lewis Acid Sites and Surface Aluminum in Aluminas and Mordenites: An Infrared Study of CO Chemisorption V. Gruver and J. J. Fripiat' Department of Chemistry and Laboratory for Surface Studies, University of Wisconsin-Milwaukee, P.O. Box 413, Milwaukee, Wisconsin 53201 Received: March 16, 1994; In Final Form: June 9, 1994"
-
The absolute number of Lewis sites has been semiquantitatively estimated by measuring the integrated absorbance of CO adsorbed a t 100 K on aluminas and dealuminated mordenites. This has been made possible by using the known amount of Brransted sites in zeolites and the linear regression between the specific absorbance and the frequency. This linear regression as well as the linear regression between the CO adsorption energy and the frequency has been justified on the basis of the perturbation of an anharmonic oscillator by a n electric field. From the absolute number of Lewis sites it is concluded that thenonframeworkaluminum debris in dealuminated mordenite are forming nanosized particles in the zeolite pores. Two strong Lewis sites are observed on the surface of these particles as well as on the surface of aluminas. A possible assignment is suggested.
Introduction Infrared spectroscopy has a long and outstanding record as a useful technique for probing surface sites either from the spectrum of adsorbed species or from a surface IR-active group. A classroom illustration of the potential of'the I R technique is the detection of acid sites on acid catalysts by observing the spectrum of pyridine and the spectrum of the surface OH groups. In the 1950s already CO was used as a molecular probe for studying the surface of dispersed metal, as pioneered by Eischensl and reviewed by Hollis and Pritchard2 in 1980. Little and Amberg3 in 1962, Seanor and Amberg4 in 1965, and Angel and Schaffers in 1966 extended the use of CO to the study of insulators. The last authors found that the frequency of CO adsorbed in the near-faujasite zeolites increased linearly with the field strength produced in the zeolitic pores by divalent (exchangeable) cations. This important finding was not only confirmed by Egerton and Stone: but it was shown that the heat of adsorption is also a linear function of thevibrational frequency. A linear relationship was also found between the heat of adsorption and frequency of CO adsorbed on oxides by Paukshtis and Yurchenko7 and by Soltanov, Paukshtis, and Yurchenko.s The specific absorbance was also a linear function of the frequency. Of course, such a relationship is of practical interest, since it would permit the measurement of the number of adsorption sites identified by a specific frequency domain. Thus, CO may be regarded as a very interesting infrared molecular probe. Its qualities stem from its soft basic character (small dipole moment: 0.1 D and relatively large polarizability). In gaseous CO the fundamental frequency is at 2143 cm-1, and under a pressure of 100 atm the specific absorbance is 5.4 cm/ pmol.4 On metal or dispersed metal, the frequency is significantly smaller than in oxide and zeolites where it goes from about 2090 to about 2250 cm-I. Generally, the vibration band is complex because it contains contributions of CO adsorbed on sites with different acid or basic strengths, and the band has to be deconvoluted to obtain a quantitative picture of the distribution of sites. Ballinger and Yates9 focused the attention on the Lewis sites created by dehydroxylation of aluminas, and they observed two kinds of Lewis sites, the first appearing upon mild dehydroxylation and the second after dehydroxylating above 800 K. On the first kind of site the CO stretching frequency is near 2195 cm-I while on the second kind it is near 2213 cm-l. This contribution will be devoted to the study of aluminas and dealuminated H-mordenite. Our specific goal will be to 0
Abstract published in Aduance ACS Abstracts, July 15, 1994.
0022-3654/94/2098-8549$04.50/0
distinguish the different kinds of acid sites and to measure their concentration. A debatable question is about the nature of the information obtained from the IR study of adsorbed CO regarding the strength of a site. It has been mentioned earlier that the vibration frequency is a function of the electric field to which the molecule is exposed. This fact will be supported again in the theoretical section of this paper. However, the notion of the strength of an acid site is different from the notion of the electric field strength. A Lewis acid site is an electron acceptor site. A Br~nstedsite is an OH in which the proton is mobile and which will interfere with an electron donor. As pointed out by Beran,lo the difference in frequency between gas-phase CO (2143 cm-1) and CO+ (2184 cm-I) is only 41 cm-l for full ionization. The charge transfer cannot alone explain the large frequency shift observed upon adsorption. For instance, Beran's theoretical calculation, using the C N D 0 / 2 method, of the interaction of CO at 2.55 A of an exposed A1 leads to a total charge transfer of 0.2. The corresponding CO stretch being at 2225 cm-I, it is clear that the field effect is overpowering. The field is a function of the charge on the atom X (X = A1 in the above example) and of the distance X-C. It is also the force driving the electron density on the molecular probe, here CO, toward theelectron acceptor center. As such, it might be considered as an estimate of the acid strength of the center. Besides the direct examination of the IR spectrum of adsorbed CO, one may investigate the stretching vibration of the residual OHs on the surface after thermal activation. Adsorption of CO shifts the latter toward lower wavenumber frequency. Such a trend is observed, whatever the adsorbate,' on silica1' as well as on zeolites,I2 where changes in IR frequency have been used to locate the acidic OH inside the supercage or inside the lattice. The magnitude of the frequency shift has been related to the proton affinity and, thus, to the strength of Br~lstedacidity. This is discussed thoroughly in ref 7. Indeed, variation in the O H frequency after CO adsorption may provide an alternative way to estimate the acid strength of Brrasted site. Upon C O adsorption the silanol O H stretch frequency decreases by about 90 cm-1 whereas the O H frequency of a Brransted site in H Y shifts toward a lower wavenumber by 250 cm-1, e.g., about 3 times as much.7 Theoretical Aspect
As already emphasized, the first goal of this paper is to describe the technique giving access to the specific absorbance of the CO species yielding a vibrational band at a known frequency. The second goal is to associate an energy of CO chemisorption to this 0 1994 American Chemical Society
Gruver and Fripiat
8550 The Journal of Physical Chemistry, Vol. 98, No. 34, 1994
band. Thus, the sites can be ranked with respect to their strength, the highest energy of chemisorption corresponding to the strongest site, and the number of sites within a domain of strength can be calculated from the integrated absorbanceof thevibrational band at maximum coverage (at the lowest adsorption temperature) and the specific absorbance. As it will be shown in the Experimental Section, in the scientific literature there are quite a number of experimental data showing that the energy of C O chemisorption varies linearly with respect to the frequency of, or more exactly with Au. Au is the difference between the observed frequency of CO adsorbed on a category of sites and the frequency of gaseous CO or of physically adsorbed CO, namely, 2143 cm-1. There is also a number of experimental data showing that the ratio of the specific absorbance of chemisorbed CO to the specific absorbance of gaseous C O is also a linear function of Au. From these data, the empirical relationships needed for the practical goal of this work would suffice, since after the deconvolution of the experimental IR spectra into a certain number of bands (three can be identified) both the number of sites and their strength would be known. However, we have thought that the two empirical relationships defined above should be supported by some simple theoretical development, not only in order to understand their physical meaning but also to estimate their limit of validity. In a linear anharmonic oscillatorl3 the difference in frequency (vo) between the n = 0 and n = 1 levels is close to the mechanical frequency uo by a factor in the order of (huo)/D. D is the dissociation energy of CO, that is, 1074 kJ/mol or 4.26 X 10-9 erg/m~lecule.~~Js A more important frequency shift is obtained when the oscillating dipole is in a uniform electrostatic field. This effect and the associated shift in the intensity is known as the vibrational Stark effect. Recently, BishopI6 has reviewed and developed, on the basis of the perturbation theory, equations in whichquadratic terms in field strength areincluded. Interestingly, the theoretical results were applied to CO. The simple firstorder perturbation theory suggested by Coggeshall17 is a more direct application for the present work. We will recall it briefly. The Schroedinger equation for a diatomic molecule with a Morse potential energy function and submitted to the actions of an electric field E, d2#/dz2
+ ( p / h 2 ) (W - D[ 1 - exp(-az)]'-
p,E,)# = 0
where pp = qz is the dipole moment, q being the charge. p is the reduced charge of the oscillator. The approximate solution proposed by Coggeshall yields for the v = 0 u = 1 transition
-
In the absence of electrical field ( E , = 0) Au would be
In these equations a2 = 2r2pu;/D (3) In using the dimensionless ratio u = qE,/aD, the frequency shift due to the combined action of anharmonicity and electrical field is (4)
Equation 4 is valid for an oscillator with a high dissociation energy D and a relatively large huo and in the temperature range where the fundamental level only is populated. With the reasonable assumption that Y O = YO, u is easily obtained from the following
120
3
c
01
80-
IIcm
Figure 1. Calculated and experimental adsorption energies of CO on
insulators: 0, theoretical values predicted by Bates er a/.;'' 0, experimentalvaluescollectedby Soltanov et a1.* The solid line is obtained from eq 6 and D = 1074 kJ mol-I.
equation
u is zero when v = vo since E , = 0, and u is either lower than zero when v > YO or larger than zero when u < UO. Theoretical calculations by Bates and Dwyerls show that when the observed frequency ( u ) of adsorbed CO is smaller than the frequency ( Y O ) of gaseous CO, the oxygen atom is directed toward an electron acceptor center that is either a Lewis or a Br~nstedsite or even a silanol group. This result is in agreement with earlier findings by Kustov et ~ l . 1On~ the opposite, when v > UO, the carbon atom isdirected towardanyofthesesites. BothendsoftheCOmolecule are nucleophilic. It will be suggested later that basic Lewis sites could be identified by CO stretching frequency below 2143 cm-1. The electron transfer from the carbon atom would rationalize the bond contraction (and u > UO). The converse would be true for v < uo according to Bates and Dwyer. Since the sign of u is inverted when Y / Y O becomes larger or smaller than one and since qE,/a = uD, there must be an inversion of the sign of charge associated with the dipole as predicted by the theoretical calculations.~0J9Whatever the sign, the work associated with the displacement of the force qE, over a distance 1/ a must be positive, the corresponding enthalpy being negative. Thus, the cosine of the angle between the field and the dipole q/a must go from zero to x and the energy is
To what extent qE,/a contributes to the CO adsorption energy is not known. Another expectedcontribution is thevander Waals energy which should be in the order of the heat of vaporization, e.g., 6.7 kJ/mol. In order to calculate the energy associated to qE,/a, all we will need is the C O dissociation energy and the frequency shift in order to calculate u. Experimental values for C O adsorption energies have been collected by Soltanov et al.' Theoretical adsorption energies by Bates and DwyerI* are also available. Figure 1 shows that the theoretical values of Bates et al. are in good agreement with the calculations from eq 6 as long as Iv - YO[is smaller than 40 cm-1. The experimental values collected by Soltanov et af. are in average 9 kJ/mol higher than those predicted by eq 6. This difference represents a reasonable van der Waals interaction. Thus, the simplistic theoretical treatment described above accounts remarkably well for the experimental observations and theoretical predictions. Adsorption energies of 21 and of 28 kJ/mol have been reported by Ballinger et aZ.9 and by Zaki et a1.,*0 respectively, for CO on Lewis sites, with stretching frequency between 2190 and 2220 cm-l. Of paramount importance for the Experimental Section is the value of the adsorption coefficient. This is the question that we will address now. In the paper already referred to, C ~ g g e s h a l l ~ ~
Infrared Study of CO Chemisorption
The Journal of Physical Chemistry, Vol. 98, No. 34, 1994 8551
8
1
’
0.05
0
0
0
20
40
60 1Icm
80
100
Figure 2. Linear regression betwee the relative absorbance of adsorbed CO to the absorbance of gaseous CO. The data are those collected for
insulator by Soltanov ef
a1.*
had shown that the ratio of the absorbance of a perturbed (Ap) to that of an unperturbed ( A , ) oscillator is
APIA, = P2(v/vo)(q,/q,)2 (7) the perturbation resulting from the application of the electric field. In (7) p2 is the square of the ratio of the matrix elements J&,z$I* d r of the perturbed and unperturbed oscillator or p2
= [K(v/vo) - 31 (K-212 [K(v/vo)- 212 ( K - 3)
(8)
with K = 4pD/hmo. qp and qu are the charge in the perturbed and unperturbed oscillator, respectively. Numerical calculations show that the product pz(v/vo) in eq 7 is close to unity and that APIAu,for all practical purposes, is proportional to (qp/qu)2. One has to realize that the measurement of the specific absorbance or adsorption coefficient is difficult because CO is always adsorbed by more than one kind of sites. The result is a complex band structure from which the different contributions cannot be obtained without using a deconvolution procedure. Therefore, it would be useful to check the domain of applicability of eq 7 in order to find a practical way to estimate the specific absorbance for any CO-surface adduct vibrating at frequency v. Soltanov et a1.8 have given CO integrated adsorption coefficients (vide infra for definition) in a domain of frequency between 21 58 and 2228 cm-I. We have plotted them us Av and normalized so that the CO gas adsorption coefficient is one. The result is shown in Figure 2 by a linear regression. The slope is 6.765 X 10-2 and R2is 0.82, an acceptable correlation considering theexperimental difficulty explained above. From the linear regression, we proceeded as follows. The charge q, in gaseous CO was obtained by dividing the CO dipole moment, e.g., 0.1 12 D by the CO interatomic length, e.g., 1.1138 A. Thus, the charge (q,) is 0.1 X esu. Then from (7)
qp = q”(
+p)1/2(
-$’2
N
qu( +p)ll2
(7’)
where A p / A , is obtained from the linear regression APIA, = 1 + 6.765 X IAvI. The existence of linear regression of the type
APIA, = 1
+ constant-Ap
is, of course, very important for the interpretation of the experimental results, since it allows calculation of the specific absorbance at any wavenumber if one is known at a specified wavenumber. As shown in Figure 2, Bishop theory (eqs 23 and 24, ref 16) also predicts a linear relationship over a 100-cm-1 frequency range, but the constant is 4.3 X l e 3 instead of the experimental 6.765 X 10-2. The origin of the discrepancy is not known. As shown later, the experimental results in ref 8 are supported by others
I 0.05
0.1 0.15 total charge.(C 8 O),calc.
0.2
0.25
Figure 3. Relationshipbetween the charge of the perturbed CO oscillator (qp). obtained from eq 7‘, and the total density of charge (carbon and oxygen) calculated by Beran.IoJ8 with variations in the absorption coefficient A , which may reach
afactorof2. ItistoavoidambiguityonA,thatintheexperimental section we use an internal standard to obtain A,. If the relative APIA, obtained from the regression in Figure 2 were correct, values for qp could be calculated and compared to theoretical charge values. Beranlo and Kustov et ~ 1 . 1 9have calculated the total (C + 0) charge density in applying the CNDO/2 method to CO adsorbed on four model clusters for which the CO vibration frequencies are known. These charges are those shown in Figure 3 (abscissa) while the corresponding values of qp are shown in the ordinate of the same figure. The correlation is better than expected (R2 = 0.95), and it supports simultaneously the use of eq 7 for calculating the charge and the use of the linear regression APIAu=f(Au) for obtaining the desired absorption coefficient. An interesting feature in Figure 3 is the point corresponding to a total charge of 0.036 according to Beran’s calculation. It is obtained for the >0--0C interaction with a vibration frequency at2125 cm-1, thatis,inthedomainwherethesignofAvisinverted. This would suggest the absorption coefficients below and above vo = 2143 cm-1 have symmetrical values. This “symmetrical” behavior is reminiscent of the representation shown in Figure 1 for the adsorption energy. Finally, if the adsorption energy varies as suggested in Figure 1, an interesting consequence is predictable. If after lowtemperature (1 50 K) adsorption the temperature rises slowly, the contributions of the CO vibrating at large IAvI (positive or negative) must be favored similarly with respect to the bands observed nearer 2143 cm-1. As reminded in the Introduction, the Brernsted acidity cannot be estimated from the frequency shift of CO adsorbed on an hydroxyl group. Whatever the acid strength of OH, the CO vibrational frequency is always 10-30 cm-I higher than 2143 cm-1. So, in order to obtain information on the strength of Brmsted acid sites, we have to consider the shift of the O H stretch provoked by CO adsorption. An empirical relationship has been established by Paukshtis and Yurchenko2’ between the proton affinity (PA) and the shift of the observed OH (AVOH) with respect to the shift observed for isolated silanol group having absorbed CO. The empirical relationship reads P A (kJ/mol) = 1390 - 442.5 log[AvOH/Av(SiOH)] (9) The lower the PA, the higher is the acidity of the OH group. Equation 9 will be mainly applied to zeolites, the OH on alumina being quasi-neutral. Note that in eq 9 1390 kJ/mol represents the PA of the Si-OH group and that Av(Si0H) = 100 cm-1. Thus, if AVOH on alumina is in the order of 100 cm-1, the corresponding O H group acidity will be in the order of that of the silanol group, that is, a very weak acidity, indeed. Experimental Section Materials. Two aluminas were studied in this work, namely, a commercial y-alumina (surface area 200 m2/g) and an alumina
8552 The Journal of Physical Chemistry, Vol. 98, No. 34, 1994 TABLE 1: Some Phvsical Characteristics of the Zeolites. A (Si/ (m2/g) A l ) c ~ (Si/AlF) VG2 VG3 VG5 VG2600
464 500 452 470
5.2 5.2 6.4 5.2
9.8 11 28 10
Alt A1Y (X1OZ1/g) (X1OZ1/g) AINF
1.6 1.6 1.4 1.6
0.93 0.84 0.35 0.91
0.67 0.76 1.05 0.69
a (Si/Al,)cA and (Si/AIY) are the ratios obtained from chemical analysis and Z9SiMAS NMR, respectively. Alt and AlY are the absolute AI total contents and framework A1 contents per gram of dehydrated zeolite (300 "C),respectively. A is the surface area obtained from the Langmuir isotherm of N2 physical adsorption.
Gruver and Fripiat
I I
1 2100
2150
I
2200
.
,
.
2100
llan
obtained from the limited hydrolysis of tri-sec-butoxide as described by Coster et a1.22(surface area 280 m2/g). The latter alumina contains, besides 4- and 6-fold coordinated aluminum atoms ( A P and AP', respectively) as usual in transition aluminas, a fairly large amount of pentacoordinated aluminum (AlV).The surface composition will be discussed later. Four dealuminated H-mordenites were prepared from the parent Na-mordenite (LZ-MS, Union Carbide). The Namordenite was exchanged four times by 1 M NH4N03 solution, washed and dried a t 293 K, heated to 523 K a t 5 K/min, and kept for 2 h at 523 K for removing hydration water. With the same heating rate the temperature was increased to 773 K, and the material was kept for 2 h at that temperature. The sample prepared in this way is called VG2. VG3 and VG4 were obtained by steaming VG2 for 1 h a t 773 and 873 K, respectively. VG4 was treated by an 1 M HCI solution for 1 h at room temperature, leading to the VG5 sample. VG2 600 was obtained by calcining VG2 at 873 K for 1 h. From the unit cell volume obtained from X-ray diffraction patterns, the crystallinity was 91%, 90%, 8 1%, and 86% in VG2, VG3, VG5, and VG2 600, respectively. The sum of the meso- and micropore volume remains below 9% of the total pore volume. Some physical characteristics of the zeolites are shown in Table 1. For the IR study the powders were pressed into disks weighing 16 mg/cm2 fitting the IR cell sample holder. Before carrying on the CO adsorption the disk was heated in the cell filled with air at 723 K and then in vacuum for 2 h between 748 (zeolites) and 873 K, (aluminas).
,
,
,
I
2200
2150 1Ian
Figure 4. (A) Vibrational spectra observed for CO adsorbed by VG3 dealuminated mordenite. From top to bottom: 153,193, and 253 K. The deconvolution of the lower spectrum is displayed. (B) Same legend as in (A) but the sample is the superfive alumina. Band Assignment and Full Width at Half-Height
TABLE 2
(fwhh).
samde
assignment
frea (cm-l)
fwhh (cm-l)
alumina a 1um in a H mordenite H mordenite H mordenite alumina all samples all samples all samples all samples
Lewis site Lz Lewis site L1 Lewis site L2 Lewis site L1 Br~rnsted weak Lewis site OH physically ads CO basic site BI basic site Bz
2207 3 2187 f 3 2218 f 2 2193 f 3 2170f 1 2166f6 2154 f 3 2140 f 1 2124 f 4 2106f6
32.5 k 1 22.5 f 1 32.5 f 1 22.5 f 1 21k 1 21 f 1 14 f 0.5 14.5 f 0.5 15.5 f 0.5 33f 1
The assignments are discussed in the text.
At lower frequency, one and sometimes two lines, called BI and Bz,respectively in Table 2, were observable. It will be suggested (vide infra) that they might correspond to basic sites. Their integrated absorbances were not negligible, but since the specific absorbances were not known in this region of the spectra no attempt was made to obtain the corresponding number of sites. For the "high"-frequency region the conversion was carried on as follows. Let A be the integrated absorbance of a band observed between VI and v2 cm-1,
Experimental Procedure
Twenty Torr of CO was introduced in the cell, and the sample was cooled a t about 100 f 5 K. Keeping the pressure constant, the temperature was increased in about 10 steps up to room temperature. The thermocouple barely touching the disk, the temperature measurements were not accurate. At each step, the temperature was kept constant within a few degrees and 400 signals were accumulated. The intensity of the band corresponding to physically adsorbed CO which appears between 2138 and 2145 cm-1 decreased steeply as the temperature was increased. When it was strongly attenuated, the spectra were deconvoluted into 5 or 6 lines using the Peakfit program. Figure 4 shows spectra in the domain between 2050 and 2300 cm-l where the CO vibrations are observed at different temperatures. The bands' assignments and the full width at half-heights (fwhh) are shown in Table 2. There were two bands attributable to Lewis sites, L2 and L1, which were observed in aluminas at slightly lower frequencies than 'in dealuminated zeolites. The band at 2170 cm-l was assigned to the Brernsted sites. A band is always observed at 2166 f 3 cm-1 on aluminas. It has been observed by Zecchina et al.23on a-Alz03and assigned by them to CO adsorbed on weak Lewis sites, eventually associated with AlV1. In fact, between the band assigned to physically adsorbed CO and that assigned t o strong Brernsted sites in the dealuminated H zeolites, we observed, in agreement with 0thers,~39J9a band at 2154 cm-l.
A=c l n ( To/T ) dv
(10)
where Toand T were the transmissions without and with the cell, respectively. A, is the specific absorbance (unit cm/mol) and A , = A/Cr where C is the number of moles absorbed per gram of catalyst (the correction for the gaseous, not adsorbed CO being negligible) and where r is the weight of the wafer per cm2 (typically 16 X 10-3 g/cm2). In the spectral region where the linearity A, us Av has been checked (Av up to 90 cm-I). A, = A,[1
+ (6.765 X 10-2)Av (cm-I)]
(11)
Then, C is A/rA,. Of course, A increases as the temperature decreases. Since the surface becomes more and more loaded with physisorbed CO as temperature decreases, the intensity measurement of the weaker bands becomes increasingly inaccurate. What is needed is the absolute value of the number of C O corresponding to the monolayer loading of a specified adsorption site. The problem was solved in the following manner. The integrated intensities A were plotted against the temperature as shown in Figure 5, and by iteration thedata pointswerefitted to a Langmuir
The Journal of Physical Chemistry, Vol. 98, No. 34, 1994 8553
Infrared Study of CO Chemisorption
Results and Discussion
B C L1 ('4)
-
L2 ('3) 0
-3
5
4
6
7
a
l O 0 O f F (K)
Figure 5. CO thermal desorption from the Brernsted and Lewis (L1) and (L2) acid sites in VG2 600. 5000 I
I
Lattice AI g.atom/g
Figure6. Linear regression (with intercept zero) between the normalized integrated absorbance of CO on the Brernsted site at full loading toward the AI:' content obtained from 29SiNMR in dealuminated mordenites, VG2, VG3, VG5, and VG2 600 (Table 1).
desorption isobare, namely a exp(bT')
A = Amax
1
+ a exp(bT')
for each kind of site. a is a constant and b is a temperature coefficient. A,,, was obtained from the best fitting curve parameters. In aluminas the curves of the type represented by eq 12 level at a higher temperature. The average of the data points collected between 140 and 150 K is within 10%of A,,,. For practical purposes it can be used in place of A,,,, considering this margin of error as reasonable in the complicated procedure leading from the spectrum to the integrated absorbances at full loading. Indeed, it is evident that (i) the deconvolution procedure, which includes a base line correction, (ii) the procedure used to reach A , (eq 12), and finally (iii) the approximation to convert A , in the number of chemisorbed CO on a specified type of site cannot yield very accurate results. The margin of relative uncertainty is in the range of &lo%,but the error on the absolute number is unknown. It remains to convert A,,, in a number of sites using eq 1 1 . The main problem here concerns the value of Ao. If we use that suggested by Paukshtis (0.18 cm/pmol) the number of Brcansted sites calculated for the dealuminated mordenite shown in Table 1 is almost exactly double the number of A l r , . Since it has been shown24~2~ that the number of Brcansted sitesisdirectly proprotional the number of Air, we recalculated A0 through the regression shown in Figure 6 using eq 1 1 with Av = 27 cm-I. We obtain, indeed, A0 = 0.385 cm/pmol. This value of A0 will be used all through this work, buttheoriginof thediscrepancywithPaukshtis' value is unknown. Morterra et a1.26 have reported absorption coefficients for CO on Ti02 which fit well in eq 1 1 but using A0 = 0.69 cm/pmol.
-
The results obtained following the experimental procedure described above are condensed in Table 3. If the strength of Lewis sites Lz and L I is assumed to be represented by the C O adsorption energy shown in Figure 1, the L2 sites are stronger than the L I sites. No conclusion about the acid strength of an OH group can be reached from the IR study of C O adsorption. One interesting piece of information contained in Table 1 is the similarity between the numbers of Lewis sites in the nonframework alumina debris in zeolites and those on the surface of aluminas. The two same bands, with somewhat lower frequencies in aluminas, and similar numbers of sites are observed (except for VG5) in both kinds of solids. The first conclusion is that the debris are highly dispersed. Indeed, i f ~ e a c c e pthat t ~ ~thedensityin Alatomsin thesubsurface layer of a spinel-like alumina is 8/nm2, then on a solid with a surface area of 300 m2there are 2.4 X loz1atoms, and out of this number according to Table 3, about 1020are detected as Lewis sites (L1 + L2). This represents about 4% of the total surface aluminum content. From Table 1 we see that in the zeolite we have about 102l NFAl in the alumina debris. Thus, we should detect 4 X l O I 9 (L2 + L1) sites on these debris. In fact, we detect about 4 times as much. In fraction of exposed A1 atoms it is about 1 . Hence, the debris are in the form of very tiny particles with a high degree of dispersion. They do not form chunks of alumina sintered outside the zeolite particles, as could have been expected. The second important piece of information in Table 2 is on the structural similarities between the superfive and NFAl debris on the surface. It had been shown earlierz8vz9that the N M R lines characteristic of the superfive alumina and of the NFAl in dealuminated zeolites have aluminum distributed among three kinds of coordination shells, namely, tetrahedral, pentahedral, and octahedral. For the nonframework species, they are represented by the following symbols: AirF, AIEF, and AlE' respectively. The Brcansted sites are associated with AIF whereas the Lewis sites are associated with coordinately unsaturated sites (CUS) in the nonframework aluminum (NFAl). The distribution of these sites is linked to the nature of NFAl. It is interesting to point out that the ratio L I / L ~is almost constant in the dealuminated zeolites (except VG5, again) and in the superfive alumina. In a paper recently accepted30 it was shown that in aluminas there are two kinds of Lewis sites chemisorbing ammonia, namely, a tetrahedral site with an isotropic shift 58 ppm and quadrupolar coupling constant (QCC) -6 MHz and a pentahedral site with an isotropic shift about 40 ppm and a slightly smaller QCC. Very similar sites exist in zeolites NFAl, in addition to the Brcansted sites. These results were obtained using the polarization transfer from protons of chemisorbed ammonia to surface aluminum, namely, the 27Alcross-polarization technique under high-speed magic angle spinning (CP MAS NMR) and in a high magnetic field (1 1.7 T). The ratio of the tetrahedral to pentahedral sites in NFAl is in the order of 5. Of course, the number of Lewis sites evidenced by a weak electron donor such as CO or a strong base, such as NH3 are not directly comparable, but it seems that the L1 and Lz sites could possible be associated with the tetrahedral and pentahedral sites, respectively. If this assignment is proven correct, then the combination of high-resolution solid-state N M R spectroscopy of surface A1 and of C O adsorption infrared spectroscopy would provide the geometry and the density of the Lewis sites. As outlined earlier, if there is a general agreement for the assignment of the CO vibrational bands to L1 and L2 sites (Table 2), in the low-frequency domain the situation is more confused. The band near 2130 cm-1 could be assigned to CO chemisorbed on an electron d0nor.1~We have observed bands at -2130 and
r;
Gruver and Fripiat
8554 The Journal of Physical Chemistry, Vol. 98, No. 34, 1994
-
TABLE 3 Number of Brornsted Sites (B) and Lewis Acid Sites L1 and Lz per gram of Dehydrated Material (Margin of Uncertainty 10%) sample B Ll L2 total L, L, LiII-2 L I + L2/B
+
-
VG2 1.1E + 21a 9.2E + 19 4.3E + 19 1.24E + 21 1.4E 20 2.1 0.127
VG3 7.4E + 20 1.2E + 20 5.3E + 19 0.9E + 21 1.7E 20 2.3 0.223
+
+
VGS 3.2E 20 4.OE 19 2.6E 19 3.86E 20 6.6E 19 1.5 0.206
+ + + + +
Vg2600 8.7E 20 1.1E 20 4.8E 19 1.03E 21 1.6E 20 2.3 0.184
+ + + + +
Supers( 1)
Y-Al203
6.8E + 19 2.7E 19 0.9E 20 0.95E + 20 2.5
1.OE 20 2.2E 19 1.22E 20 1.22E 20 4.5
+
+
+ + + +
"Read as 1.1 X loz1.
TABLE 4: Main OH Stretching Band (cm-1) Observed in the Dealuminated Mordenites before and after Contact with 20 Torr of CO at -150 K before
after
PA (kJ mol-')
3255 f 30 3556 I3662 3460 f 20
1144f20
dealuminated
3615
mordenites
3665f2
VG2-VG2 600
3728 f 2 3741 f 1
3743 f 3
369 1 3737 3792
3609 3702
1340 1390
3705 3745 3788
3569 3613 3641
1330 1335 1315
y-A1203'
superfive Al2Op a
Calcined at 873 K.
2100 cm-1 in all the samples studied here which might result from the interaction of CO with basic oxygen. Ballinger et d e 9 have drawn the attention to the resultsof an XPS study by Datta,31 who showed that upon dehydroxylating alumina at 500 OC (in the high-vacuum chamber) the binding energy (BE) of Ozp decreases by 0.7 eV, while the Alzp Be increases by 1.6 eV with respect to the BE'S measured for samples heated below 400 O C . The oxygen with an excess positive charge should be basic, and electron transfer could occur between theresonant structure where the oxygen on CO is positively charged. As pointed out very early by P a ~ l i n g ,CO ' ~ has a very large (83 kcal/mol) resonance energy, and the +CO- structure is as probable as the -CO+ structure. Therefore, we would suggest assigning the -2100 and -21 30-cm-I bands to CO chemisorbed on basic sites Bl and B2, B2 being more basic than B1. Observation by Alexandrova et al.32 that fluorination of alumina enhances the band in the region of 2130 cm-1 supports the assignment of bands in this region to basic sites since the fluorine is more negative than oxygen. Finally, it should be mentioned that the OH stretching region before and after CO adsorption has also been examined. Table 4 shows the main bands as well as the proton affinity calculated using eq 9 for the dealuminated H-mordenite and the two aluminas. As dealumination of the zeolite proceeds, the intensity of the 3615- and 3725-cm-I bands decreases while the intensity of the other bands do not change appreciably. In agreement with others (see refs 19, 33, and 34) who studied ZSM-5, we assign the band at 3615 cm-I to the Brornsted sites. The band at 3747 cm-1 does not shift while the 3728-cm-I band disappears upon CO adsorption. After CO adsorption the two main bands are at 3460 and 3255 cm-I. The proton affinity has been calculated assuming that the most intense O H stretch at 3615 cm-1 before CO adsorption is also the strongest after CO adsorption. Concerning the aluminas, we show in Table 4 the vibration frequencies of the O H groups surviving calcination at 600 K. As expected, they all shift upon adsorption of CO and the PA of these O H are about 200 kJ/mol higher than those observed for the strong Brornsted sites in the mordenite (or ZSM-5).
We believe that our main conclusion is about the high dispersion of the nonframework alumina debris in zeolites. This high dispersion should favor the interaction between the CUS sites that they contain and the Brransted sites located in the framework. Such as interaction should be a t the origin of the synergy between both kinds of sites and, therefore, at the origin of the high acidity of dealuminated mordenites. Acknowledgment. The financial support of DOE Grant DEFG02-90 ERl430 is gratefully acknowledged. The data in Table 1 were mostly obtained by Mr. Yong Hong, and we want to thank him. References and Notes (1) Eischens, R. P.; Francis, S. A.; Pliskin, W. A. J . Phys. Chem. 1956, 60, 194. (2) Hollis, P.; Pritchard, J. In Vibrational Spectroscopies for Adsorbed 137;American Species; Bell,A.T.;Hair, M.T.,Eds.;ACSSymposiumSeries Chemical Society: Washington, DC, 1980 p 51. (3) Little, L. H.; Amberg, C. H. Can. J . Chem. 1962, 40, 1997. (4) Seanor, D. A,; Amberg, C. H. J . Chem. Phys. 1965,42, 2967. (5) Angel, C. L.; Schaffer, P. C. J . Phy. Chem. 1966, 70, 1413. (6) Egerton, T. A,; Stone, F. S. J. Chem. SOC.,Faraday Trans. I , 1973, 69, 22. (7) Paukshtis, E. A.; Yurchenko, E. N. Russ. Chem. Rev. 1983,52,242. (8) Soltanov, R. J.; Paukshtis, E. A,; Yurchenko, E. N. Kinet. Katal. (Engl. transl.) 1980, 23, 164. (9) Ballinger, T. H.; Yates, J. T., Jr. Lungmuir 1991, 7, 3041. (10) Beran, S. J . Phys. Chem. 1983,87, 55. (11) McDonald, R. S. J . Am. Chem. SOC.1957, 79, 850. (12) White, J. L.; Jelli, A. N.; AndrC, J. H.; Fripiat, J. J. Trans. Faraday SOC.1967, 63, 461. (13) Morse, P. M. Phys. Rev. 1929, 34, 57. (14) Pauling, L. Nature ofthe Chemical Bond, 3rd ed.; Cornel1University Press: New York, 1960; p 194. (15) Herzberg, G. Molecular Spectra and Molecular Structure. I . Spectra of Diatomic Molecules; D. Van Nostrand: Princeton, 1979; p 481. (16) Bishop, D. M. J . Chem. Phys. 1993, 98, 3179. (17) Coggeshall, N. J . Chem. Phys. 1950, 18, 978. (18) Bates, S.; Dwyer, J. J . Phys. Chem. 1993, 97, 5897. (19) Kustov, L. M.; Katzansky, V. B.; Beran, S.; Kubelkova, L.; Jitu, P. J . Phys. Chem. 1987, 91, 5247. (20) Zaki, M. J.; Kndzinger, H. Mater. Chem. Phys. 1987, 17, 201. (21) Paukshtis, E. A.; Yurchenko, E. N. React. Kinet. Catal. Lett. (Engl. Transl.) 1981, 16, 131. (22) Coster, D.; Fripiat, J. J. Chem. Mater. 1993, 5, 1204. (23) Zecchina, A.; Platero, E. E.; Arean, C. 0.J . Coral. 1987,107,244. (24) Biaglow, A. I.; Gittleman, C.; Gorte, R. J.; Madon, R. J. J. Catal. 1991, 129, 88. (25) Biaglow, A. I.; Adamo, A. T.; Kokotailo, G. T.; Gorte, R. J. J . Catal. 1991, 131, 252. (26) Morterra, C . ; Garrone, E.; Bolis, V.; Fubini, B. Spectrochim. Acta 1987, 43A, 1577. (27) Knozinger, H.; Ratnasamy, P.Catal. Rev.-Sci. Eng. 1978, 78, 31. (28) Chen. F. R.; Davis, J. G.; Fripiat, J. J. J . Catal. 1992, 133, 262. (29) Chen, F. R.; Fripiat, J. J. J. Phys. Chem. 1992, 96, 819. (30) Coster, D.; Blumenfeld, A. L.; Fripiat, J. J. J . Phys. Chem.,accepted. (31) Datta, A. J . Phys. Chem. 1989, 93, 7053. (32) Alexandrova, N. V.;Bursian, N. R.; Georgievskii, V. Yu.; Gruver, V. Sh.; Zverev, S. M.; Semenskaja, I. V.; Oranskaja, 0. M.; Minachev, Kh. M. In Primenenie tseolirov v katalize; Nauka: Moscow, 1989; pp 62-64. Soltanov, E. A.; Paukshtis, E. N.; Yurchenko, E. A,; Dovdashev, E. N.; Mamedov, S. E.; Gasymov, B. A. Kinet. Katal. (Engl. Transl.) 1984,25,729. (33) Romotowski, T.; Komorec, J.; Paukshtis, E. A.; Yurchenko, E. N. Zeolites 1991, 11, 497. (34) Zecchina, A.; Bordiza, S.; Spoto, G.; Scarano, D. J . Chem. SOC., Faraday Trans. 1992, 88, 2959.
