J . Phys. Chem. 1989, 93, 2590-2595
2590
spectra of adsorbed CO may be assessed. Magic-angle spinning yields precise isotropic shifts and thus could detect subtle changes in the properties of the adsorption site caused by, for example, promoters or coadsorbates. However, magic-angle spinning spectra of adsorbed CO should not be used to exclude the existence of some species. As noted above, the bridge-bonded species do not appear and could be overlooked if the lack of agreement between the center of mass of the broad-line and high-resolution spectra was not noted. The I3C N M R spectra of CO on Rh/silica illustrate another caveat; similar isotropic shifts of the dicarbonyl and linearly bonded CO caused two species to appear as one. To separate these components, it was necessary to decompose the spectra by using sideband intensities predicted from fits to the broad-line spectra. Also, it is tenuous to extract relative site populations from the magic-angle spinning spectra, as seen by comparing the broad-line and high-resolution results in Table I. Measuring both high-resolution and broad-line spectra aids the assignment of component peaks. For example, CO on Ru has at least three isotropic peaks in the range 200-175 ppm, which this study identifies as 199, 195, and 180 ppm. The peak at 199 ppm can be correlated to the Lorentzian peak in the broad-line spectrum by its atypical sharpness (cf. Figure 7) and its lack of sidebands. The other two peaks have sidebands whose intensities and range correspond to linearly bonded CO species. We note that these assignments are inconsistent with an earlier study of CO on Ru/Y ~ e o l i t e which ,~ proposed the following peaks and assignments: dicarbonyl (168 ppm), linearly bonded CO (1 80 ppm), and bridge bonded (203 ppm). On the basis of the results obtained here, we propose alternate assignments: multicarbonyl (203 ppm), linearly bonded (180 ppm), and linearly bonded (168 ppm). The bridge-bonded species is predicted to lie upfield of 203 ppm and we believe was not resolved, for reasons discussed in preceding sections.
V. Summary High-resolution and broad-line I3C N M R spectra of CO on
silica-supported Ru and Rh corroborate previous assignments and provide additional information on the adsorbed states. The high-resolution spectra obtained by magic-angle spinning reveal isotropic shifts consistent with the broad-line components proposed previously to be linearly bonded CO and multicarbonyls formed on isolated metal atoms. In addition, experimental evidence suggests two types of linearly bonded CO on Ru/silica. The sharpness of the linear peaks suggests that some Ru and Rh aggregates have negligible magnetic susceptibility, consistent with raftlike structures. Quantitative comparison of the high-resolution and broad-line spectra of adsorbed CO reveals possible pitfalls of relying on magic-angle spinning spectra alone. First, bridge-bonded CO, whose presence is indicated by the quantitatively correct broad-line spectra, is not sufficiently narrowed to be observed in magic-angle spinning spectra. If the I3CNMR spectrum of the bridge-bonded species is broadened by chemical inhomogeneity, then improved resolution (for example, to resolve 2-fold and 3-fold bridged CO) lies in the realm of sample preparation and not N M R methodology. Second, species whose line shapes are distinctly different in the broad-line spectra, such as the Lorentzian of the Rh dicarbonyl and the powder pattern of the linearly bonded CO, may have the same isotropic shift and be unresolved in magic-angle spinning spectra. While in principle they may be separated by analysis of sideband intensities, this is not practical to do without constraints provided by other techniques. Finally, the relative site populations in the high-resolution spectra are inconsistent with the independently calibrated broad-line spectra. Avoiding these and other traps by using the strengths of several techniquesresolution of species and accurate determination of isotropic shifts from magic-angle spinning, and anisotropies and quantitative measurement of all species from wide-line spectroscopy-to compensate for the weaknesses of the other gives more reliable results. Registry No. CO, 630-08-0; Rh, 7440-16-6; Ru, 7440-18-8.
Solid-state NMR Study Using Trimethylphosphine as a Probe of Acid Sites in Normal and Dealuminated Zeolite Y Jack H. Lunsford,*,t Pierre N. Tutunjian,*-* Po-jen Chu,+ Eshan B. Yeh,? and David J. Zalewsk? Department of Chemistry, Texas A & M University, College Station, Texas 77843, and Shell Development Company, Houston, Texas 77001 (Received: July 22, 1988)
31PMAS-NMR spectra of trimethylphosphine adsorbed in normal and dealuminated zeolite Y have been used to provide both qualitative and quantitative information on their acidic properties. Interaction of the (CH3)3Pwith Bransted acid sites gives rise to the protonated adduct having chemical shifts in the range -1 to -4 ppm. In a partially oxidized sample the protonated form of trimethylphosphine oxide exhibits a resonance at 64.6 ppm. The integrated spectra of these two protonated species provide a reasonably reliable quantitative determination of the Brernsted site concentration provided the concentration is 233/uc (unit cell). Trimethylphosphine also interacts with Lewis acid sites in steam-dealuminated zeolites which were dehydrated at 400 "C, but the Lewis-bound form is in rapid chemical exchange with the liquidlike (CH3)3P. The resulting resonance is at -62 ppm. When the zeolites were heated to 600 'C, additional resonances were observed in the range -31 to -58 ppm. These are attributed to (CH&P at other Lewis acid sites which develop as the aluminum oxyhydroxide in the large cavities progressively dehydroxylates. In the presence of oxygen the Lewis-bound and liquidlike species are rapidly oxidized to the corresponding oxides, which exhibit resonances at 56.4 and 43.5 ppm, respectively.
Introduction
The protonic forms of zeolites exhibit acidic properties which are largely responsible for the catalytic importance of this class of material. In order to relate acidity to catalytic activity and 'Texas A & M University. 'Shell Development Co.
to structural factors such as the silicon-to-aluminum ratio, it is important to develop methods for determining the acidity of both the Brernsted and Lewis acid sites. Jacobs] has reviewed several of the techniques that have been employed to determine acidity. ( 1 ) Jacobs, P. A. In Characterization of Heterogeneous Catalysu: Dellaney, F., Ed.; Marcel Dekker: New York, 1986; pp 367-404.
0022-3654/89/2093-2590$01.50/00 1989 American Chemical Society
Trimethylphosphine Adsorbed in Zeolite Y Recently, solid-state NMR spectroscopy of basic probe molecules has emerged as a promising approach for evaluating not only the types of acid sites on alumina and aluminosilicates but also the number of acid sites. Nitrogen-15 bases, including ammonia, pyridine, and trimethylamine, have been effectively used; however, the small magnetogyric ratio of I5N results in sensitivity probl e m ~ . ~ -These ~ difficulties have been overcome by adopting )IP-containing probe molecules such as trimethylph~sphine.~-*In addition to being a highly sensitive nucleus, 31Pchemical shifts are relatively large, which enables one to obtain information on the types of acid sites that are present in the catalysts. A variety of phosphorus bases, including trialkylphosphines and phosphine oxides, have now been used to characterize acid sites. Among these, trimethylphosphine has one important advantage in that the spectrum of the protonated adduct exhibits coupling, which enables one to probe Bransted acidity with confidence.5s6 Moreover, there is some indication from earlier studies that the chemical shift differences from -2 to -4 ppm might reflect differences in Bransted acid ~ t r e n g t h .Dehydroxylation ~ of H-Y zeolites at elevated temperatures is known to produce Lewis acid sites, and indeed resonances in the region from -32 to -58 ppm suggest that several types of Lewis acid adducts are f ~ r m e d . ~ In the present study it was of interest to examine dealuminated Y-type zeolites because of their importance in petroleum refining as cracking catalysts. Partial removal of framework aluminum reduces the concentration of acid sites in the zeolite, but measurements of catalytic activity indicate that the acid strength increasesg During the removal process part of the extracted aluminum remains in the cavities as oxides and oxyhydroxides. This extraframework aluminum dehydroxylates when heated, and the resulting Lewis acid sites may result in catalyst deactivation via the formation of nonvolatile carbonaceous residues known as coke.I0 Evaluation of )'P N M R as a quantitative technique for determining acid sites in zeolites also was a goal of this research. In earlier studies Haw et al." adopted titration methods to estimate the numbers of Bransted and Lewis sites on an amorphous silica-alumina. In one study n-butylamine was used to displace I5N-enriched pyridine, a weaker base, from the Bransted and Lewis sites. Estimates of the amount of Bransted acidity were obtained from the equivalents of n-butylamine required to completely displace pyridine from the Brernsted acid sites. On a similar acid catalyst trimethylphosphine, triethylphosphine, and tributylphosphine were added until no further increase in the signal of the protonated adduct was observed.6 For the pyridine and the three trialkylphosphines the end points occurred at 0.17, 0.23, 0.19 and 0.15 mmol of H+ per gram of catalyst. Several reasons were offered for the observed discrepancies, but in the absence of a material of known proton concentration it was not possible to distinguish between the alternatives. Unlike the amorphous aluminosilicates, for which the proton concentration is unknown, the proton concentration in the H-Y zeolite, at least in principle, is known from the framework aluminum content. Thus, one has
(2) Earl, W. L.; Fritz, P. 0.;Gibson, A. A. V.; Lunsford, J. H . J . Phys. Chem. 1987, 91, 2091. ( 3 ) Ripmeester, J. A. J . Am. Chem. SOC.1983, 105, 2925. (4) Maciel, G. E.; Haw, J. F.; Chuang, I.&; Hawkins, B. L.; Early, T. E.; McKay, D. R.; Petrakis, L. J . Am. Chem. SOC.1983, 105, 5529. (5) Lunsford, J. H.; Rothwell, W. P.; Shen, W. J . A m . Chem. SOC.1985, 107, 1540. Rothwell, W. P.; Shen, W.; Lunsford, J . H. J . A m . Chem. SOC. 1984, 106, 2452. (6) Baltusis, L.; Frye, J. S.;Maciel, G. E. J . A m . Chem. SOC.1987, 109, 40. (7) Bein, Th.; Chase, D. B.; Farlee, R. D.; Stucky, G. D. In New Deuelopments in Zeolife Science Technology; Murakami, Y . ;Iijima, A,, Ward, J . W., Eds.; Kodansha Ltd.: Tokyo, 1986; pp 311-318. (8) Baltusis, L.; Frye, J. S.; Maciel, G. E. J . A m . Chem. SOC.1986, 108, 71 19. (9) DeCanio, S. J.; Sohn, J. R.; Fritz, P. 0.;Lunsford, J. H. J . Carol. 1986, 101, 132. Sohn,J . R.; DeCanio, S. J.; Fritz, P. 0.;Lunsford, J . H. J . Phys. Chem. 1986, 90, 4847. (101 Nock. A.: Rudham. R. Zeolites 1987. 7 . 481. ( 1 1 ) Haw,'J. F.; Chuang, I.-S.; Hawkins,'B:L.; Maciel, G. E. J . Am. Chem. SOC.1983, 105, 7206.
The Journal of Physical Chemistry, Vol. 93, No. 6, 1989 2591 a standard material for comparison with N M R results.
Experimental Section Zeolite Samples. All of the zeolite samples used in this study were prepared from Union Carbide's LZ-Y62 (Si/Al = 2 . 5 5 ) or LZ-Y52 (Si/Al = 2.38). The Na+ cations were replaced with NH4+ cations by treatment with a solution of either (NH4)2S04 or NH4N03. A typical ion exchange consisted of stirring 5 g of zeolite with approximately 15 equiv of NH4+ in 500 mL of distilled, deionized water. The solution was maintained at 80 OC for 2 h. After washing the sample, the procedure was repeated an additional two times. The last exchange solution was allowed to equilibrate for 12 h before collecting the zeolite. Samples containing various amounts of Na, designated as NaH-Y, were prepared by a single ion-exchange step. The desired amount of N H 4 N 0 , was added to 2 g of zeolite in 500 mL of water. After heating the solution to 80 OC for 6 h, the zeolite was collected and washed with distilled, deionized water. The final Na concentration was determined by inductively coupled plasma (ICP) analysis. Three methods were used to extract aluminum atoms from the framework of the zeolites, thus altering the silicon-to-aluminum ratio. The dealumination was achieved by steaming, by dehydroxylation, and by reaction of the zeolite with SiC14. The steamed zeolite was prepared by heating an NH4-Y zeolite under flowing N 2 / H 2 0 to 600 OC for either 1 or 3 h. The sample was then cooled under steam and ion exchanged with NH4+ (1 M N H 4 N 0 3 , 60 "C, 12 h) to remove residual Na+ ions. This material, referred to as SDY for steam-dealuminated Y-type zeolite, was crystalline and contained 38 and 28 Al/uc for the samples steamed for 1 and 3 h, respectively. An ICP analysis of the SDY (3 h) zeolite indicated that the material contained a total of 53 Al/uc. Thus, the majority of the A1 atoms removed by the steaming process remained inside the zeolite cavities as extraframework AI. Unless otherwise stated, this and subsequent dealuminated samples were heated under vacuum to 400 O C and maintained at that temperature for 2 h. A method that involves dehydroxylation of the zeolite at elevated temperatures was introduced by McDaniel and Maher.', Following the method B, a Na-Y zeolite was ion exchanged twice with NH4+ in a solution of (NH4)2S04. The zeolite was washed free of sulfate ions and heated in air to 815 OC for 3 h. After cooling, the zeolite was again ion exchanged twice with NH4+ at 100 "C for 1 h, washed free of sulfate ions, and then dried. According to powder X-ray diffraction analysis the zeolite was crystalline and contained 33 Al/uc. As with the SDY zeolite, essentially all of the A1 removed from the framework remained in the cavities. This material is referred to as USY for ultrastable Y-type zeolite. The SiC1,-dealuminated zeolite (DY) was prepared by first heating ca. 2 g of NH4-Y zeolite to 375 OC at a rate of 0.9 "C min-I under flowing N z (100 mL min-'). The sample was maintained at 375 "C overnight. Subsequently, the sample was exposed to SiCl, vapor in flowing N, at 50 OC. The sample temperature was raised to 275 OC at a rate of 4 "C min-I, the SiC1, was interrupted, and the sample was maintained under flowing N2 for 4 h at 275 OC. After cooling to 25 OC the zeolite was exchanged three times with NH4+ions. The sample was then deamminated at 400 "C, as described previously. The DY zeolite contained 38 Al/uc. Trimethylphosphine (Strem) was adsorbed from the vapor phase (20 Torr) on the zeolite samples (0.5 g) which were at 25, 50, or 80 "C. For all quantitative experiments the trimethylphosphine was added to the zeolite at 80 "C and the zeolite was cooled to 25 OC in 20 Torr of gas. In some cases excess and weakly adsorbed trimethylphosphine was removed by heating the sample under vacuum. Normally, the zeolite was then transferred to the NMR rotor in a glovebox or dry bag filled with either dry nitrogen or argon; however, several samples were sealed in glass ampules which
~I
(12) McDaniel, C. V.; Maher, P. K. In Proceedings of the Conference on Molecular Seiues, 1967; Society of Chemical Industry: London, 1986; p 186.
2592
The Journal of Physical Chemistry, Vol. 93, No. 6, 1989
I
-1
a
Lunsford et al.
a
40
20
0
-20
-44
-60
P PM ,
100
50
0 PPI4
.
.
.
.
,
.
-sa
.
.
.
I
.
-LOO
Figure 1. Proton-decoupled, 90" pulse ,lP MAS-NMR spectra of (CH,),P in zeolites which were dehydrated at 400 "C: (a) H-Y, (b) DY, (c) SDY. Excess (CH3),P was added to the samples at 80 "C, and spectra were obtained at 25 "C. denotes spinning sidebands.
were inserted into the N M R rotors. N M R Experiments. The N M R results reported here were obtained with a Bruker CXP-200 spectrometer that was equipped with a Chemagnetics probe or with a Bruker MSL-300 spectrometer and probe. Unless otherwise noted, magic angle spinning (MAS) at 3-4 kHz was employed along with either cross-polarization or 90" pulse methods. The spectra were collected with and without proton decoupling. Quantitative spectra were obtained by using a 90" pulse-and-acquire sequence. Standards consisting of various amounts of (NH4),HP04 physically mixed with N H 4 N 0 3were used to determine the absolute number of 31Pspins. By varying the time between successive pulses, it was found that both the standards and the samples had T I less than 5 s. Chemical shifts are reported relative to 85% H3P04 with negative shifts reflecting greater shielding (i.e., higher external field required).
Results and Discussion Lewis Acidity in Zeolites. Proton-decoupled spectra of H-Y, DY, and SDY samples, all of which contained excess trimethylphosphine, are shown in Figure 1. The spectra of the H-Y and DY zeolites are remarkably similar. Notably absent from the spectrum of the DY sample are resonances in the -30 to -60 ppm region from Lewis-bound trimethylphosphine. Although the reaction of the zeolite with SiCI, removes part of the aluminum as volatile AICI,, it is evident from previous studies that part of the aluminum remains within the zeolite cavities9 By contrast, in the SDY zeolite essentially all of the aluminum removed from the framework resides in the cavities. Characterization of this extraframework material by Shannon et aI.l3 has led to the conclusion that it is present as y-A100H (boehmite). The resonance at -62.6 ppm in Figure I C is distinctly at a lower field than the resonance at -67.4 ppm for liquidlike (CH,),P in the zeolite c a ~ i t i e s . ' Also, ~ it is significant that there is no resonance at -67.4 (13) Shannon, R. D.; Gardner, K. H.; Staley, R. H.; Bergeret, G.; Gallezot,
P.;Auroux, A. J . Phys. Chem. 1985, 89, 4778.
Figure 2. I'P MAS-NMR spectra of (CH3),P in a SDY zeolite which had been dehydrated at 400 "C: spectra obtained with the sample at (a) 25, (b) -28, and (c) -63 "C. The concentration of (CH,),P in the zeolite was approximately equivalent to that of the framework aluminum atoms. denotes spinning sidebands.
ppm in Figure IC, which is surprising since excess (CH,),P was present in the zeolite. The width of the resonance for the protonated adduct also is considerably greater in the SDY sample than in the DY sample (12 ppm vs 3 ppm). Moreover, as the temperature of the SDY sample was lowered the resonances sharpened up. This behavior in the SDY sample is consistent with a rapid chemical exchange between two phosphine species. When a sample of SDY was degassed at 100 "C following the addition of trimethylphosphine, the only resonance was that of the protonated adduct at -4 ppm. An additional manifestation of the exchange process was observed when (CHJ3P was adsorbed on a SDY zeolite in an amount which was approximately equivalent to the framework aluminum content, Le., approximately equal to 24 (CH,),P molecules/uc. At this loading the large cavities were only partially filled. Oxygen was rigorously excluded from this sample. The resulting roomtemperature spectrum shown in Figure 2a exhibited a single, broad resonance at -19.4 ppm. No spinning sidebands were observed in this spectrum. Upon cooling the sample to -28 "C two resonances at -3.7 and -59.9 ppm were observed (Figure 2b), but both lines wcre rather broad. Further cooling to -63 "C resulted in the spectrum of Figure 2c, which was similar to, but better resolved than, that of Figure IC. The resonance at -3.6 ppm, assigned to [ (CH3)3P-H]+,possesses spinning sidebands, which indicates that the protonated adduct is no longer undergoing a rapid chemical exchange process. The remaining resonance at -60.2 ppm does not have any spinning sidebands associated with it. Moreover, the chemical shift value of -60.2 ppm is between that of Lewisbound (see below) and liquidlike (CH3),P. The effect of temperature was completely reversible. These results suggest that, in a partially loaded sample, chemical exchange is occurring not only between the liquidlike and Lewis-bound trimethylphosphine (14) Previ~usly,~ the resonance at -67 ppm was attributed to physisorbed (CH3),P, but here we make the distinction between liquidlike and several forms of the Lewis-bound base, all of which could be referred to as physisorbed.
The Journal of Physical Chemistry, Vol. 93, No. 6,1989 2593
Trimethylphosphine Adsorbed in Zeolite Y
- 58.6
a
28.7
-
IAJ
-
.54.5
b
h
b
- 61.7
C 1
100
.
.
.
.
I
so
.
.
.
.
I
PPY
.
0
.
.
.
I
.
-so
.
.
.
I
.
-100
Figure 3. Spectra obtained 5 days after those of Figure 1: (a) SDY, (b) DY. denotes spinning sidebands.
but also between the Lewis-bound base and the protonated adduct. The former gives rise to the resonance at -60.2 ppm and the latter to the resonance at -19.4 ppm. Even at -63 OC it was not possible to slow down the exchange between Lewis-bound and liquidlike (CH,),P such that two resonances could be resolved. These exchange processes, however, become slower when the zeolite is fully loaded with (CH,),P. Although the oxidation of trimethylphosphine in zeolites will be reported more fully in a subsequent paper, the spectrum depicted in Figure 3a reveals some of the salient features of the process. After 5 days in the rotor, sufficient oxygen leaked into the SDY sample to form trimethylphosphine oxide (TMPO), which gives rise to the resonances at 43-65 ppm. In agreement with the assignments of Baltusis et a1.6, we attribute the resonances at 43.5, 56.4, and 64.6 ppm to physisorbed, Lewis-bound, and protonated TMPO, respectively. The sharp resonance at 26.9 ppm has not been assigned at this time. It is evident from this spectrum and related studies that the weakly bound (CH3),P is much more easily oxidized than the protonated adduct. The DY zeolite after 5 days in the rotor exhibited the spectrum of Figure 3b. Here the resonance of the liquidlike (CH,),P has decreased relative to that of the protonated adduct, and the principal form of TMPO is physisorbed. Only a small amount of protonated TMPO was detected, and none of the Lewis-bound form was observed. The nature of Lewis acid sites in these dealuminized zeolites has been explored further by dehydroxylating the samples at 600 OC. In the previous study of H-Y zeolite^,^ it was demonstrated that Lewis sites began to form upon dehydroxylation of the zeolite at temperatures greater than 400 OC. But this process is complicated by the fact that dehydroxylation is accompanied by dealumination. In the present study the SDY zeolite was already dealuminated; therefore, the higher temperature (600 "C) mainly resulted in the dehydroxylation of the zeolite as well as the extraframework alumina. Whereas H-Y zeolites begin to dehydroxylate significantly only at temperatures >400 0C,15*16 alumina progressively dehydroxylates as the temperature is increased above 100 OC. As reported by Peril' for y-A1203, 56% of the water (present as hydroxyl groups) is removed at 400 O C and 84% is removed at 600 O C . Peril* has developed a model that shows the evolution of oxide (15) Uytterhoeven, J.; Christner, L. G.; Hall, W. K. J . Phys. Chem. 1965, 69,21 17. (16) Ward, J. W. J. Catal. 1967, 9, 225. (17) Peri, J. B. J . Phys. Chem. 1965, 69,211.
200
100
0 PPM
-lw
Figure 4. 3'P MAS-NMR spectra of (CH3)3Pin zeolites which were dehydroxylated 10 h at 600 OC: (a) H-Y, (b) SDY after several minutes in the probe, (c) SDY after 2 h in the probe. denotes spinning sidebands.
ions and oxide vacancies on the surface. Oxide vacancies and, thus, exposed aluminum ions begin to occur at temperatures as low as 250 "C, but two or more adjacent oxide vacancies occur only at temperatures above 450 "C. From this model it is evident that the types of Lewis acid sites change with the extent of surface dehydroxylation. Although there are undoubtedly differences between an extended alumina surface and the small particles in the zeolite, the qualitative effects are believed to be similar. In a previous study it was demonstrated that (CH,),P adsorbed on y-A1203which had been degassed at 500 "C gave rise to a single resonance at -56.3 p ~ m . ~ The spectra of (CH3),P in H-Y and SDY zeolites, after the samples had been heated in vacuo to 600 OC for ca. 10 h, are shown in Figure 4. The spectrum of Figure 4a obtained with the H-Y zeolite is similar to that which was reported previously for a H-Y zeolite that had been dehydroxylated 2 h at 700 O C 5 The most prominent feature is the resonance at -58.5 ppm. The proton coupled spectrum (not shown) was almost identical, except that the resonance at -3.7 ppm was significantly broadened. Even in the static mode the resonance at -58.5 ppm was relatively narrow, which suggests that the coordinated (CH,),P is mobile on the N M R time scale. As observed with the SDY sample (Figure IC), the spectrum exhibited no resonance at -67 ppm even though the zeolite contained excess trimethylphosphine. The 31Pspectrum of the analogous SDY sample, dehydroxylated at 600 OC, is shown in Figure 4b. In many respects the spectrum resembles that of an H-Y zeolite which had been dehydroxylated for 2 h at 600 O C ; only the [(CH,),P-H]+ resonance is considerably weaker in the SDY sample. The resonances from -3 1.1 to -54.5 ppm, attributed to Lewis-bound trimethylphosphine, are clearly evident. This rotor, however, was not sealed well, and it is evident that air slowly leaked into the sample. As indicated by the spectrum of Figure 4c, after 2 h in the probe oxidation products were formed, and the intensity of both the protonated TMPO and trimethylphosphine resonances increased at the ex(18) Peri, J. B. J . Phys. Chem. 1965, 69,220.
2594
The Journal of Physical Chemistry, Vol. 93, No. 6, 1989
pense of the Lewis-bound trimethylphosphine. The reversible conversion of Lewis acid to Brernsted acid sites upon addition water has been noted previously with ze01ites.l~ In addition, a new resonance becomes apparent at -61.7 ppm, which is near the position of the high field resonance of Figure IC. Upon longer exposure to air this peak also disappeared. The latter result provides a clue to the origin of the resonance at -61 to -63 ppm. As water is adsorbed by the highly hydroscopic sample, it is evident that the zeolite framework is partially rehydroxylated, and one would expect that the alumina in the cavities also would be rehydroxylated. We tentatively assign the resonance at -60 to -63 ppm to (CH3),P which is weakly adsorbed on partially hydroxylated alumina in the zeolite cavities. This form of trimethylphosphine is undergoing rapid exchange with liquidlike molecules that also must be present in the sample. Here it should be noted that (CH3),P in a Na-Y zeolite exhibits a resonance at -59.5 ppm, which demonstrates that noncoordinating ions also promote a 31Pchemical shift. The extraframework aluminum in the DY material perhaps is in the small cages and thus is inaccessible to the probe molecule. Trimethylphosphine is a soft base and therefore does not interact strongly with hard acids such as tricoordinate aluminum. Even when Lewis acids are present in the zeolite, the coordinated (CH3),P is often characterized by relatively small chemical shifts and rapid chemical exchange. Thus, the demarcation in the NMR spectrum between Lewis-bound and other weakly bound forms of trimethylphosphine is not always obvious. The (CH3),P and (CH,),PO spectra complement one another in that (CH3),P is more sensitive with respect to Brolnsted acidity; whereas (CHJ3P0 provides a more definitive test of Lewis acidity. For example, the resonance at -62.6 ppm in Figure I C does not provide conclusive evidence for Lewis acidity, but the resonance at 56.4 ppm in Figure 3a, obtained after partial oxidation of the (CHJ3P, confirms the presence of Lewis acidity in this sample. Consistent with this interpretation, almost no resonance was observed at -62.6 ppm in the spectrum of (CH3),P on DY (Figure lb) and the spectrum for Lewis-bound TMPO is not present in Figure 3b. These results, together with the spectrum of (CH,),P on H-Y (Figure la), confirm that for samples which have been carefully degassed to 400 "C Lewis acidity is present neither in a H-Y zeolite nor in a DY zeolite, but it abounds in a SDY zeolite. For samples dehydroxylated at T L 600 "C Lewis acidity is present in all protonated Y zeolites, and the extent of dehydroxylation alters the nature of the Lewis acid site. Quantitative Determination of Brolnsted Acidity. For the pure H-Y zeolite each framework aluminum atom is charge-balanced by a proton; thus, from the number of A1 atoms per unit cell one can determine the number of H+ ions per unit cell. A calculation of the number of protons in a dealuminated zeolite is not nearly so straightforward as some of the extraframework aluminum may be present in the cationic form which cannot be removed by ion-exchange techniques. For example, the cationic aluminum may be located in the small cavities in the same manner that lanthanide ions are present in a Y-type zeolite after dehydration at elevated temperature^.'^ In addition, it is possible that extraframework aluminum, present as oxybydroxides, may physically block certain acid sites by partially filling the large cavities. The value of trimethylphosphine as a quantitative probe of Brernsted acidity in zeolites was evaluated by absorbing the base at 80 "C, cooling the sample to 25 "C, and determining the number of [(CH,),P-H]+ species from the resonance at -1 to -4 ppm. In those cases where partial oxidation of the phosphine occurred, the integral of the resonance at 64.6 ppm, which resulted from protonated TMPO, was added to the integral of the resonance for [(CH,),P-H]+. It should be noted that the oxidation of the physisorbed (CH,),P was much more rapid than that of the protonated adduct. Thus, the potential problem introduced by chemical exchange in the SDY sample could be avoided by determining the Brernsted acid concentration from the spectrum of Figure 2a, rather than Figure IC. (19) Marynen. P.; Maes, A.: Cremers, A. Zeolires 1984, 4, 287.
Lunsford et al. TABLE I: Quantitative Evaluation of Bransted Acid Sites in Zeolite Y
[(CH,),P-H]+/ Al/uc" H-Y SDY(1 h) SDY(3 h ) USY DY
54 38 28 33 38
gX
(2.6)b (4.0) (6.9)
17
8.6c
(5.8)
9.5' 13.5'
(4.0)
9.9
[(CH3)3- %[(CH3)P-HI+/ P-HI+/ total uc A1 P/uc 24.6 15.4 20.6 28.6 20.5
46 41 74 87 54
37 38 33 45 33
a Framework AI per unit cell (uc). *%/AI ratio given in parentheses. 'Obtained from [(CH3),P-H]+ plus [(CH3)3PO-H]+.
TABLE 11: Effect of Sodium Ion and Proton Concentrations on the Forms of Trimethylphosphine in Zeolite Yo [(CH3!,- [(CH,),7% H'
P-H1 NaSl-Y Na49H8-Y Na31H20-Y NaZ2H3,-Y Na, ,H,,-Y Na6H4*-Y
0 5 16 24
21 25
P-Nal' 24 22 16
(CHd,P TMPO detected 62 1 13
2
13 12
80 69 63 52
"All concentrations expressed as molecules per unit cell The results obtained for five samples are summarized in Table I. It is evident from these data that the concentration of the protonated adduct is less than that of the framework aluminum. When trimethylphosphine was adsorbed on the H-Y zeolite dehydrated at 400 "C, the concentration of [(CH,),P-H]+ was determined to be 24.6 Al/uc. This value represents only 46% of that expected from the aluminum content. This material contained 6.3 Na+/uc; therefore there were actually 48 H+/uc. Hence, 52% of the protons were detected by NMR. With respect to reproducibility it is interesting to note that a second sample, run on a different instrument by a different operator, gave a value of 23.5 [( C H ~ ) ~ P - H ] + / U C . The inability of the probe molecule to sample all of the protons in this zeolite appears to be the result of filling and steric factors. The filling of (CH3),P in the large cavities was determined by several methods, and it is evident that the maximum loading is about 40 molecules/uc or 5 molecules per large cavity. Bein et al. arrived at an identical number.' This value is consistent with a measured liquid density of 0.70 g/mL for (CH3)3P. A H-Y sample was weighed before and after exposure to (CH3),P, and the amount adsorbed was determined to be 38 molecules/uc, which is in good agreement with the value of 37 molecules/uc determined from the N M R spectrum (protonated adduct plus liquidlike material). As indicated in Table I, all of the zeolites had a total loading of 33-38 molecules/uc with the exception of the USY sample, which contained somewhat more phosphine. This relatively constant loading, even in the dealuminated zeolites, suggests that partial blocking of the cavities of extraframework aluminum is not a serious problem. If one considers a model of faujasite containing 54 Al/uc, it becomes evident that a substantial number of aluminum ions are present as next-nearest TO., tetrahedra, along with their charge-balancing protons.20 There is insufficient volume for two trimethylphosphine molecules to exist adjacent to these proximal acid sites. Thus, even though there is adequate volume for ca. five molecules per large cavity, only three of these are sufficiently near to the acid sites for protonation in the normal H-Y zeolite. In order to further explore this limitation on the use of (CH3),P to evaluate quantitatively the number of protons in a zeolite, a series of NaH-Y zeolites were investigated, and the results are given in Table I1 and Figure 5. As the H+ concentration increased, the loading level of (CH,),P generally increased. At a level of 20 protons/uc (Na3,Hzo-Y) 80% of the protons were (20) Beagley, B.; Dwyer, J.; Fitch, F. R.; Mann, R.; Waiters, J. J . Phys. Chem. 1984, 88. 1744.
Trimethylphosphine Adsorbed in Zeolite Y
The Journal of Physical Chemistry, Vol. 93, No. 6, 1989 2595 the volume of the large cavities. As depicted in Figure 5 , the resonance observed for (CH,),P in Na5,-Y (-59.5 ppm) is much broader than the resonance observed in Na6H48-Y (-67.5 ppm), with the peak width at half-height being 438 and 146 Hz, respectively. This broadening may result from a weak interaction between the base and Na+ ions ( I = 3 / 2 ) . The effect of diluting the protons in the zeolite is further illustrated with the dealuminated samples of Table I: The number of framework A1 atoms having no next-nearest aluminum tetrahedra reaches a maximum at about 33 Al/uc. It is not surprising, therefore, that the SDY(3 h) and USY samples, which had 28 and 33 framework Al/uc, respectively, exhibited the maximum percentage of [(CH,),P-H]+/Al. The value for SDY( 1 h) is anomalously small, and this may reflect the incomplete development of acid sites after short periods of steaming. As noted previously, some of the extraframework aluminum may be present in a cationic form which reduces the number of protons required for charge compensation. The concentrations of protonated adducts measured in the SDY(3 h) and USY samples may indeed reflect the true concentration of protons in these materials.
Conclusions
50
0
P PM
-
50
Figure 5. ,'P MAS-NMR spectra of (CH,),P in NaH-Y zeolites which had been dehydrated at 400 "C. denotes spinning sidebands.
accounted for by the [(CH,),P-H]+ adduct, but at a level of 48 protons/uc (Na6H48-Y) only 52% of the protons were accounted for. The results of Table I1 show that the concentration of the protonated adduct increased to a level of ca. 25 molecules/uc and subsequently remained essentially constant. In the Na-rich zeolites the total amount of (CH3),P was less than in the protonated zeolites, which may reflect the presence of the rather large Na+ ion in the site I1 positions.21 That is, Na+ ions partially occupy (2 I ) Mortier, W. Compilation of Extraframework Sites in Zeolites; Butterworth Scientific Limited: Guildford, 1982; pp 19-31.
Trimethylphosphine is an effective probe molecule for studying both Brernsted and Lewis acid sites in acidic Y-type zeolites. Normal H-Y zeolites and samples prepared by dealumination with SiC14 exhibited almost no Lewis acidity after dehydration and deammination at 400 OC, but a steam-dealuminated zeolite had Lewis acid sites which adsorbed (CH3),P weakly. These Lewis acid sites apparently are associated with the extraframework aluminum which is present in the large cavities as an oxyhydroxide species. There is rapid chemical exchange between the phosphine adsorbed at Lewis acid sites and the liquidlike phase. In these same zeolites chemical exchange also occurs at 25 "C between the Lewis-bound (CH3)3Pand the protonated adduct. The determination of Brmsted acid site concentrations in zeolites using trimethylphosphine is limited by both the capacity of the large cavities and the steric effects of probe molecules at adjacent acid sites. The technique is most accurate when the proton concentration is 133/uc, and even in this case the estimated error is approximately 20%. In general, the method underestimates the proton concentration. Acknowledgment. This research was supported in part with funds provided by the Board of Reagents of Texas A&M University and the U S . Army Research Office. We thank Dr. W. P. Rothwell for his valuable comments and discussions. Registry No. (CH3)3P,594-09-2.