J. Phys. Chem. 1995, 99, 15181-15191
15181
Brflnsted Acid Sites and Surface Structure in Zeolites: A High-Resolution 29SiNMR REDOR Study A. L. Blumenfeld, D. Coster, and J. J. Fripiat* Department of Chemistry and Laboratory for Sugace Sciences, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53201 Received: April 6, 1995; In Final Form:August 9, 1995@
In well-crystallized zeolites, the distribution of silicon atoms among Q4(nA1) clusters containing up to n = 4 framework aluminum (FA1) is well documented. Upon replacement of the alkali charge-balancing cations by protons and thermal activation of the original sieve, a solid acid is obtained. There are Lewis sites in nanoparticles of nonframework alumina (NFAl) created by the thermal treatment. The Bransted sites are OH groups bridging framework silicon and aluminum. First, it is demonstrated that it is possible to estimate semiquantitatively heteronuclear second moments M2Si-Hin acid zeolites having chemisorbed ammonia using an adapted 29SiCP REDOR (cross-polarization rotational echo double resonance) technique. Then we show that, on this basis and with comparison with simple models, two kinds of bridging OH groups are distinguishable: (i) one donating its proton to N H 3 , forming NH4+ reorienting isotropically rapidly, and (ii) one operating as a proton donor to a strongly hydrogen-bonded ammonia. The strong Bransted sites of the first kind are mostly located on Q4(nA1) with n = 1 while the hydrogen-bonded species as mostly associated with Q4(nA1) with n > 1. The higher the rate of polarization transfer is, the larger M2Si-H. The largest M2Si-His associated with the sites on which N H 3 is hydrogen bonded. It is suggested that a noticeable fraction of the acidic OHs react with, and therefore are neutralized by, nonframework alumina. Moreover, preliminary information on an exchange process involving NH4+ is presented.
Introduction The acidity of hydrogen-exchanged dealuminated (DH) zeolites, which is at the origin of their use as catalysts, comprises Bransted as well as Lewis acid centers. It is generally accepted that the OH bridging a framework silicon to a framework aluminum (both in 4-fold (IV) coordination) is the Bransted acid center. The simple picture proposed by Uyterhoeven, Christner, and Hall' in 1965 for the generation of the hydrogen by the removal of N H 3 from an ammonium-exchanged Y zeolite is still accepted. Because the nature of the Lewis acidity existing in zeolites is by far not as simply characterized as the Bransted acidity, there has been a tendency to reduce the description of the acidity of zeolites to the sole Bransted centers. There have, however, been numerous studies showing that the catalytic activity in isomerization and in cracking reactions depends on both kinds of acid center^.^-^ Recently, Coster et aLs have tackled the problem of the nature of the Lewis sites in aluminas and zeolites by studying the Al coordination in the surface layer(s) by high-resolution solid state NMR and cross-polarization of 27Al from the protons of chemisorbed NH3. Throughout this paper, the definition of chemisorbed ammonia is the ammonia which remains after outgassing at 110 "C for at least 15 min. This arbitrary definition received some quantitative justification when it was realized that the differential heat of NH3 chemisorption on the Lewis sites on aluminas was equal or highefl than 125 kJ mol-', as was well documented by Dumesic et al.,7 and that the energy corresponding to the irreversibly adsorbed ammonia on Lewis and Bransted sites is larger than 100 kJ mol-'. The results obtained from the N H 3 IH 27AlCP MAS NMR study5showed that two kinds of Lewis sites exist on the surface
-
* To whom correspondence @
should be addressed. Abstract published in Advance ACS Abstracts, September 15, 1995.
0022-365419512099-15 181$09.00/0
of aluminas as well as on nonframework (NFAl) alumina particles in zeolites, namely, a tetrahedral site with an isotropic chemical shift of about 58 ppm and a quadrupolar coupling constant QCC =6 MHz and a 5-fold coordination site with an isotropic shift of about 40 ppm and a slightly smaller QCC. Two distinct vibrational frequencies of chemisorbed CO correspond to these sites, thus, it is possible to determine their concentration.8 Of course, as shown already by Vega and Luzg in N&+-rho zeolite, the cross-polarization of 29Si by the ammonium protons in Si-ON& puts into evidence the silicon bearing the acidic OH. The polarization transfer is only possible in the dehydrated sieves because the translational mobility of W+must be severely restricted. It has been shown by Mestdagh, Stone, and Fripiat'O that, in fully dehydrated N h Y , NI&+ enjoys rapid reorientation but that the surface diffusion coefficient is lower than cm2 s-I at room temperature. The acid proton-hopping frequency'' in Si-OH-Al is less than lo4 s-' below 200 "C, whereas it reaches lo7 s-l at 400 "C. From previous experiments on Lewis acidity one would expect that it is also possible to study the Bronsted sites by cross-polarization of 'H 29Si from chemisorbed ammonia
-
,Si H3N---H---0. 'AI
-
As far as the geometry of the Bransted site is concerned, from a 'H (N&+) 29SiCP MAS NMR it would be possible to assess which among the Si-OH-A1 sites are Bronsted sites. In zeolites with moderate dealumination, there is not only the 29SiQ4(1Al) contribution besides the 29SiQ4(Si-OAl) but also variable amounts of 29Si Q4(nA1) with n > 1. Although the mental representation of Si-OH-A1 is relatively easy, differentiation between the acid sites corresponding to Q4(2Al) or more generally Q4(nA1) is of considerable complexity. A review of surface modifications in various HY zeolites, as revealed by one-pulse 29Sispectra, can be found in Engelhardt 0 1995 American Chemical Society
15182 J. Phys. Chem., Vol. 99, No. 41, 1995 80
F AVSi
Figure 1. Relative signal intensity, Q4(nAl)/C~=,Q4(nA1), for various values of n in dealuminated Y zeolites as a function of the ratio of total framework aluminum to framework Si, FAUSi (from ref 13).
and Michel's reference book.I2 At least on Y zeolites it is generally observed that dealumination affects first Q4(nAl) with n > 1. Figure 1 shows the evolution of some Q4(nAl) contributions with respect to the NMR N S i ratio (in our vocabulary, the FAllSi ratio) in 10 dealuminated HY (DHY) obtained by S i a ~ ~ t aby r ' ~either the classical (w)4si& or S i c 4 techniques. As the ratio FAllSi decreases, the sum of the relative intensities of the Q4(nAl) contribution with n > 1 decreases almost linearly, while the relative intensity of Si1Al increases slightly at the beginning of the dealumination and then levels out at about 47% for S Z A l between 3 and 6.7. These observations are in good agreement with those reported in ref 12, p 234. Of course, at higher degrees of dealumination Si-1Al decreases at the same pace as FAllSi. Another question concerns the role played by the aluminum in the Q4(nAl),n > 1, linkages in the formation of NFAl,which contains the Lewis sites. Connected with the preferential removal of (Si-nAl) and the formation of NFAl is the question of the possible condensation reaction between a Si-OH-A1 linkage and an hydroxyl group of NFA1. We know5 that NFAl contains aluminum in 6-fold (VI), 5-fold (V), and 4-fold (IV) coordination. An hydroxyl attzched to NFAlvl is basic, and it could condense with acidic OH. This reaction has been evidenced in DHY by Macedo et al. l 4 The chemical shift of Si-O-Al$ resulting from that condensation should not be very different from that of Si-OH-AI:. It is not obvious which, if any, of the hydroxylated Si-A1 linkages reacts preferentially in that way. Thus, the question of the Bransted acidity is not as simple as one might think, except in a case like HZSM-5 with a high SUA1 ratio and no or few NFAl, such as that studied by Gorte et ~ 1 . (Si/FAl '~ = 40), where it is probable that the one-to-one correspondence between the Bransted site and FAl holds. In other zeolites, we feel that there is a real necessity to study the Si-OH-A1 linkages by the technique of 'H (NI&+)-29Si cross-polarization. Indeed, the attempt to increase our knowledge of the Brclnsted acidity goes far beyond the aim of our earlier study of the nature of the Lewis acidity, because as far as the Bransted acidity is concerned, the Si coordination does not contain the information. Instead, the discrimination could be based on the 'H (NH4)-29Si and/or 'H (NH3)-29Si distances. Pfeifer and his have recognized the signature of acidic bridging OH groups in two 'H lines at 3.9 and 5 ppm (from TMS), assigned by them in HY, to OH in large cavities and hexagonal prisms, respectively. From the analysis of the 'H sideband pattern, Hunger et a1.I8 have recently obtained 27Al-'H distances (which should be at the most 0.01 nm larger than the 29Si-H distances) in LTA (A), FAU (X and Y), MOR, and MFI (ZSM-5) zeolites with different H contents but little amount of NFAl. These distances cover a range between 0.234
Blumenfeld et al. and 0.252 nm and are correlated with the size of the oxygen rings which contain the Al substitutions. The geometry of the Si-OH-A1 sites has been studied by neutron diffraction in HY obtained after calcination of N&Y under vacuum at 300 OC.I9 Ab initio calculations have also been performed.20 Reasonable agreement has been obtained with the 'H-27Al distances indicated above, suggesting Si-H distances on the order of 0.24 nm. The expected distance between the proton and the next nearest Si neighbor should be between 4 and 5 A. Thus, the efficiency of the magnetization transfer from the proton to that silicon should be low. In other words, in a CP dynamics study carried out by observing the variation of the intensity of the Q4(OAl) signal with respect to the time of contact between the proton and silicon "reservoirs", this resonance line should be weak with respect to that corresponding to Q4(lAl). In fact, as shown in Figure 5 , ref 18, there is little difference between the intensity of these lines in CP and one-pulse spectra. The SiQ4(1Al) line is slightly stronger than SiQ4(OAl) in calcined HY and in H-mordenite. The formation of terminal Si-OH, Q3(Si-OH), in the mesopores formed by dealumination or the formation of OH group nests as analyzed in the pioneer work by Engelhardt et d2'may be partially responsible for favoring polarization transfer to the next, nearest, or more remote Si neighbor. Because of the ambiguous information obtained from the CP dynamics on the nature of the environment of the acidic OH bridge, another technique was necessary to achieve the goal of this contribution. Recently, we have advocated22 the use of cross-polarization rotational echo double-resonance (CP REDOR) to obtain information on the 'H-27Al connectivity on the surface of an alumina containing AIIV, AIV,and Al"', such as the nonframework alumina species in zeolites. The advantage of using this technique for understanding the nature of the Bransted sites is the possibility of calculating the heteronuclear second moment M z ~ I - ~In. spite of the fact that M2SI-Hincludes a distribution of distances, some kind of average distance can be estimated, or the second moment itself can be compared with that of some model structures. It will be shown that for model solids, containing a larger number of Si-H pairs, M z ~ ~ - ~ calculated from the structural data coincides with that obtained from 29SiCP MAS REDOR. As far as we are aware, there has been only one SEDOR study of acid zeolites. Kenaston et used it for determining the 27Al-'H distance in HZSM-5 (Si/ Al = 20); that is a solid for which the pair approximation is valid. Their result was in good agreement with the distances mentioned earlier (0.243 nm). For our goal REDOR is more advantageous than SEDOR because it offers the possibility of obtaining M2SI-Hfor silicon with different aluminum environments and, of course, to suppress the proton-proton interaction. In addition, the CP REDOR allows one to work on samples contained in a sealed vial after pretreatment, which is absolutely necessary for this kind of work.I5 Three different dealuminated and acid zeolites will be thoroughly studied here. They belong to the three main families, namely, faujasite, mordenite, and ZSM-5. The first two were used for the earlier studies of the Lewis a ~ i d i t y .In ~ addition, we have tried to explain the features of the I5N Nh4R spectra of the I5NH3 used as probe molecule and proton reservoir for cross-polarization. It will be shown also that the simultaneous use of the CP dynamics and of the 29Si and I5N REDOR experiment in acid zeolites having chemisorbed NH3 reveals new facts about Bransted acidity and the Bransted-Lewis sites interaction. The principles of the rotational echo double-resonance (REDOR) experiment will be explained after the Experimental
J. Phys. Chem., Vol. 99,No.41, 1995 15183
Bronsted Acid Sites and Surface Structure in Zeolites
TABLE 1: Sample Characteristics (All Data in lozo atomdg)
0 1
31 ~
sample USY VG2 HZSM-5
FAI"
NFAIb
(NH3)id
11.1 8.4 5.1
17.5 7.6 1.1
(5) (4.5) (4)
Framework aluminum. Nonframework AI computed from the ratios (SUA1) obtained from 29SiNMR and from chemical analyses. Approximate amount of irreversibly adsorbed NH3.
4
4
-= zQ4(nA1)/z(n/4)Q4(nA1) FA1
,,=o
(1)
n=O
The nonframework aluminum content (NFAl) is the difference between the total content obtained from chemical analysis and FAl, NFAlISi = AI(CA)/Si - FAl/Si
(2)
under the condition that the dealumination treatment does not extract silicon from the lattice. I5NH3 (98 atom % 15N; Aldrich) was chemisorbed on each zeolite according to the following procedure. The glass cell designed to prepare the catalyst has been described in detail el~ewhere.~ A 200 mg sample of zeolite was dehydrated under vacuum, and the temperature was raised slowly to 400 "C. At that temperature, organic residues were burned with 40 Torr of Experimental Section 0 2 overnight. The combustion products were evacuated, and the temperature was raised to 475 "C and kept constant Reference Materials. Different reference materials were overnight. Prior to I5NH3 adsorption, the sample was cooled used to match the Hartmann-Hahn condition and to check the to 115 "C. About 300 Torr of I5NH3 was contacted with the REDOR pulse sequence. For 29Si NMR, well-crystallized sample for 20 min. The excess was condensed, and the catalyst kaolinite, dickite, and KHSi205 were used whereas I5NH4Clwas was evacuated under vacuum under a residual pressure less than the reference of choice for the IH-l5N experiment. Kaolinite Torr for 10 min at 115 "C. The glass cell was designed (AlzSi205(OH)4) has a &octahedral 1:1 layer structure consisting to allow the transfer under vacuum of the pretreated catalyst of alternating aluminum octahedral and silicon tetrahedral sheets. into a glass vial which could be easily sealed off and used It has a triclinic symmetry and belongs to the CI space group. directly in the MAS rotor. Thus, the catalyst was never in The unit cell parameters are 28 a = 5.1554(1) A, b = 8.9448(2) contact with the atmosphere. An estimate of ammonia chemiA, c = 7.4048(2) A, a = 91.700(2)", p = 104.862(1)", and y sorption among the potential adsorption sites is reported in Table = 89.822(1)". Dickite is a kaolinite polymorph which exhibits 2. significant distortion of the ideal kaolinite layer, including a NMR Technique (One-Pulse Spectra). All 29Siexperiments rotation of the silica tetrahedra and a distortion of the octahedral were performed in a static field of 11.7 T under MAS conditions, sheet. Its space is C,, and the unit cell parameters are the spinning rate being 10 f 0.05 kHz (tr= 100 f 1 ps). For a = 5.178(1) A, b = 8.937(2) A, c = 14.738(5) A, and /3 = the one-pulse spectra, the length of the pulse was 3.5 ps, the 103.82(2)". The coordinates of each atom including the hydrogen are known from neutron and X-ray d i f f r a c t i ~ n . ~ ~ , ~delay ~ between pulses was between 2 and 5 s, and the number of acquisitions was on the order of a few thousand. For the The Si-H heteronuclear second moment (M2Si-H)calculated I5N one-pulse spectra, the pulse width was 3-4 ps while the from the structure is 0.812 kHz2 for both kaolinite and dickite. delay between pulses was 15 s. Some one pulse I5N spectra KHSi2O5 belongs to the space group Pmnb (Z= 4).30,31The were recorded between -100 "C and room temperature. The unit cell parameters are a = 8.15 A, b = 12.54 A, and c = 4.7 conditions for the CP and REDOR experiments will be described A. The adjacent tetrahedral sheets are bonded by strong after the following short theoretical development. symmetrical hydrogen bonds, the 0-0 distance being 2.489 Description of REDOR Experiments. In the early days of A, and by potassium in the interlayer. Some characteristics of the solid-state NMR history all spin interactions except isotropic this puckered layered silicate have been published recently.32 chemical shift and scalar J-coupling were considered as irritating The shortest Si-H distance should be 2.54 A. sources of line broadening and, thus, were eliminated with Activated Catalysts. Three different activated dealuminated increasing technical sophistication. However, it became clear H-zeolites (DH) zeolites were investigated, namely, USY, VG2, very soon that the high quality of the well-resolved liquidlike and Z S M J . The sample preparation included two steps; first spectra obtained under MAS or multiple-pulse irradiation the zeolites were activated in a classical way, and afterward conditions was achieved at the cost of a severe sacrifice in ammonia was adsorbed on the catalysts. The second step was valuable information which was lost along with the correspondan "all under vacuum" procedure where the state of the surface ing interactions. For example, the heteronuclear dipolar interacis perfectly under control. tion (HDI) for an isolated nuclear pair has a simple form of The USY and ZSM-5 were the PQ CBVSOO and CBV3020 materials. The two zeolites were activated at 500 and 600 "C K =-0(3 cos2 0 - 1)ZzSz (3) for 24 h, respectively, in a conventional furnace. The VG2 sample is a modified mordenite. The parent material was exchanged four times by 1 M NH4N03, washed with distilled where the dipolar constant D = y ~ s h / rdepends , ~ ~ directly on water, and dried at 120 "C. The sample was activated in a quartz the internuclear distance r. In the case where HDI is the reactor under a flow of dry helium. Then the temperature was strongest spin interaction for the ensemble of isolated I-S pairs, raised to 250 "C at the rate of 5 " C h i n and stabilized at 250 this distance can be easily obtained from the well-resolved "C for 2 h. The temperature was raised to 500 "C at the same dipolar splitting even in a powder sample. rate and stabilized again for 2 h. The general idea of reintroduction of HDI into high resolution The samples characteristics are reported in Table 1. From solid-state NMR spectra emerged in early 1970s in the pioneerthe deconvolution of the 29Si one-pulse spectrum the ratio Si/ ing works of Waugh et al.33 In these 2D static experiments FA1 was obtained by using the classical (and approximate) the evolution of the spin system consisted of two periods. equation During the first period the system evolved under the influence
Section. In particular, it will be shown that this technique can be applied even though the pair approximation is no longer valid. The main messages of this work are that it is more advantageous to study the 1H-29Si than the 1H-27Alinteraction for characterizing the Br@nstedsites and that REDOR is a more reliable source of information than CP dynamics.
15184 J. Phys. Chem., Vol. 99, No. 41, 1995
Blumenfeld et al.
TABLE 2: Estimate Distribution of Chemisorbed Ammonia on the Studied Zeolites (All Values in 1Vomoleculedg; B = Bronsted and L = Lewis Acid Centers) NH3(B)
NH3(L)
NH4'
NH3(HB)"
not change the modulation effect of the pulse train (K=ZJ,), but it retums the S magnetization to its initial state at the end of the evolution period. Thus, the ratio (SI - SO)& represents the degree of dipolar dephasing. The plot of this ratio us the evolution time (or the number of rotor cycles during I-spin irradiation) is the so-called REDOR evolution curve. For the isolated I-S pair with a dipolar constant D,being the single fitting parameter, this curve can be calculated as
1
"IT
0"
(-Xn)(Z)
Q
nn nnni (2)
( x ) (Z) (Z)
Figure 2. Cross-polarization (CP) rotational echo double resonance (REDOR) pulse sequence. Pulse and receiver phases $1, $2, and $3 are cycled in order to cancel the breakthrough signal during the accumulation. See ref 22 for additional technical information. of HDI (I3C-'H), while during the second period this interaction was averaged out by the proton decoupling. The conventional 2D data processing resulted in the well-resolved I3C-lH dipolar splittings along one of the axes for each line in the spectrum. However, this static technique works well only for simple and well-crystallized organic solids. When fast magic-angle spinning became a standard tool, the introduction of a controlled dipolar dephasing into the evolution of a spin system had to be done in a different way. The time dependence of dipolar frequencies caused by fast sample spinning had to be accounted for. The nature of this time dependence originates from the fact that fast rotation occurs not only in the Cartesian space but also in the spin space as well, since the spin magnetization is involved into the sample rotation. This results in a time dependence of dipolar frequencies for all nuclear pairs in such a way that their average values over each rotor cycle are zero and no dipolar dephasing occurs. The basic idea of a rotational echo double-resonance (REDOR) experiment is to modulate the time dependence of the dipolar evolution so that its time average over the rotor cycle is no longer zero. This goal is achieved by using the well-known property of strong 180" radio-frequency pulses to change I, (or S), into -Iz (or -SJ. The bilinearity of the dipolar spin Hamiltonian relative to Z,Sz term provides a simple way to do that. As usual, in this kind of experiment, the time evolution of a spin system after strong excitation (either a 90" pulse or CP) proceeds in two steps. During the first one the I-spins (not observed) are irradiated with a train of 180" pulses. To get a predictable modulation effect, these pulses should be synchronized with the period of sample rotation and be positioned at certain moments during each rotor cycles. In this version of the REDOR sequence34we used equally spaced pulses with two of them during each rotor period (see Figure 2). The second step is just a detection of a signal under conventional MAS conditions without any perturbation of S-spins. The extent of the degree of the dipolar dephasing accumulated during the first period is the ratio of the amplitude of the signal at the end of the first period (SI)to that obtained without I-spin irradiation (SO). Since there are other sources of spin dephasing (for example, isotropic chemical shifts or field inhomogeneity), the spin-echo method of signal detection is used. For this purpose one 180" pulse at the middle of the evolution period is transferred from the I-channel to the S-channel (observation). Such transfer does
~ 2 n ~ ' 2 c o s [ 4 J ' Z n Dcos t , a sin /3 cos /3] sin /3 d/3 d a (4) where a and ,8 are azimuthal and polar angles of the intemuclear vector in the MAS reference frame, tris the rotor period, and n is the number of rotor cycles. Some examples of this dependence for isolated 29Si-'H pairs are given in Figure 3. This approach has been extensively used, mostly to determine the I3C-l5N distances in biomolecules and
biopolymer^.^^ Difficulties immediately emerge when the assumption of isolated nuclear pairs is no longer valid. The complexity of the dipolar spectrum for an SIn spin system grows dramatically when n, the number of I nuclei to be considered, increases as shown in the calculation of the REDOR and SEDOR dipolar frequencies for SZ2 36 and S14.37 It is evident that the exact calculations of the dipolar spectrum either increase the number of fitting parameters or require some structural information. Moreover, the number of dipolar frequencies rapidly increases as 2", making the heteronuclear dipolar spectrum untractable. Here a new semiempirical approach to the problem of calculating the REDOR response in the extreme case of SI,,, with large n is proposed. The corresponding heteronuclear dipolar line shape is completely unresolved, but as shown long ago,38 if a spin system is governed by only I& interactions and if the space distribution of nuclei is homogeneous, the corresponding line shape is Gaussian. In that case, the REDOR response can be fitted with a simple parameter only, namely, the Gaussian line width, or the square root of the heteronuclear second moment (M2s1). In the present study the I spins are always protons and S spins are either 29Sior 15N. Let us assume now that the heteronuclear dipolar spectrum of 29Si (or 15N) nuclei is unresolved and Gaussian. This is a critical assumption, especially in zeolites where the closest proton-containing species to 29Sinuclei are NJ&+. The powder-averaged dipolar spectrum of this SZ4 system is not a pure Gaussian. It might, however, be close to it because of the additional broadening caused by the interaction with a proton "background", namely, with protons within the second or near coordination spheres. The need to consider this "background" arises from the unexpected high efficiency of CP excitation of 29Si nuclei without chemisorbed ammonia close by, as discussed later. Moreover, molecular dynamic processes of N H 3 or NJ&+ species proceeding fast at the dipolar time scale also result in the coalescence of residual splittings. To be able to use the standard calculation s ~ h e m e ?we~ ~ ~ ~ have to be convinced that homonuclear proton-proton dipolar coupling does not influence the spin system evolution at the REDOR time scale. As far as we know, several groups have reported well-resolved 'H MAS NMR spectra of surface OH groups in various zeolites, even at rather moderate spin In particular, Fenzke et aL40 were able to determine the contribution of heteronuclear 1H-27Alcoupling to the proton spinning sideband pattem in H-Y and SAPO-5 zeolites, and Haw et aL4' performed double-resonance27Al 'H experiments with HZMS-5 zeolite. In our case, we are dealing with carefully
-
Bransted Acid Sites and Surface Structure in Zeolites
J. Phys. Chem., Vol. 99, No. 41, 1995 15185
- 29Si-1H CP-REDOR*...*...-
1
._.._.__..... -
._...I)
. 0
KHSi20S Kaolinite
........... “0
“0
20
40
60
20
80
60 # of Rotor Cycles
40
100
80
# of Rotor Cycles
Figure 3. Calculated REDOR fraction us the number of rotor cycles ~ 1 kHz or a large for an isolated pair with D = Y H Y S , f i l ~ ; , - = collection of pairs and a Gaussian dipolar broadening A = dM2s1-H.
dehydroxylated samples, with NH3/NH4 groups rotating fast at the NMR time scale. The calculating procedure that we have used here goes as follows. First, we calculate the REDOR evolution curve for some value of 29Si -‘H dipolar constant. We have done it by numerical integration of eq 1 with D value corresponding to the intemuclear distance of 2 A and the rotor period of 100 ps. This curve plotted us the number of rotor periods should be considered as a “universal” curve for 10 kHz MAS. There is no need to calculate the REDOR response for a pair with a different value of dipolar constant; it is immediately obtained by the scaling of the “universal” curve along the x axes. Since the heteronuclear dipolar line shape is inherently inhomogeneous, it can be considered as a superposition of a large number of doublets with splitting 20, weighted by a Gaussian distribution:
w(DJ =
1
0.2
t 0
Q
1
2
3
S
4
I 6
Time (msec)
1 1
0.2 0
exp(-D;/2A2)
Dickite Zr=O.Zmsec * Zr=O.lmsec
0
CP-REDOR
15N-’H
/ 20
. NH4CI 60
40
80
I
100
Y o f Rotor Cycka
Figure 4. REDOR fraction measurements obtained for KHSi205 (W) and kaolinite (0)us the number of rotor cycles. The solid and dashed lines are calculated from the experimental values of M Z ~ I - The ~.
where
For each Di doublet a REDOR curve is calculated, corrected for the Gaussian weighting factor, and all the contributions are added up. Thus, the resulting REDOR curve may be considered as the response of a spins system with a given heteronuclear second moment, M2s1, Some examples of these calculations are given in Figure 3. To test this procedure, we performed ‘H 29Si and ‘H I5N REDOR experiments with some model compounds which have been described in the Experimental Section. It is seen in Figure 4 that in the worst case of N h C l more than 60% of the experimental points are fitted quite satisfactorily by the calculated curve. The only parameter in our calculations was a heteronuclear second moment. The obtained values are in good agreement with calculated ones. These results provide a reasonable justification for the procedure to be applied here. They cover a set of samples with different structures and different dynamical behavior. For example, ammonium ions in m C 1 are in fast isotropic reorientations at room temperature; the intruion second moment is, thus, averaged out, and only interion contribution remains. The 29Si-’H second moments were calculated in the “rigid” lattice approximation. Adjustment Procedure. The following parameters were adjusted separately just before each experiment: (i) the power in the observation channel in order to match the HartmannHahn condition, (ii) the proton and 29Si (or 15N) 180O pulses, and (iii) the synchronization of the 180° pulses in the proton channel with the rotor period. All the adjustments were carried
-
-
example for dickite, which has the same M2Si-Has kaolinite, shows the good agreement between measurements of the evolution of the REDOR fraction at two spinning frequencies (rr-I) of 5 and 10 kHz. The intermolecular contribution to the second moment obtained from the distances (0.387 nm) between I5N in I5N& reorientating isotropically in cubic N&CI was used to predict (solid line) the evolution of the I5N-H REDOR fraction.
out with the model compounds, kaolinite and KHSi2O5 (29Si) or I5NH4Cl (I5N). The power level in the proton channel was kept constant in the range between 34 and 40 kHz. The modified Hartmann-Hahn condition was matched for either n = +1 (29Si)or n = -1 (I5N) spinning sidebands.
+
wls = w , ~2nm,
(6)
These choices corresponded to a power level in the observation channel of 50 or 24 kHz, respectively. The choice of the sideband was inspired by the probe performance in different frequency regions; in the REDOR and CP experiments we tried to keep the proton power at the highest possible level. The synchronization of the proton pulse train with sample rotation was a critical point for the adjustment. In all our experiments the rotor period was 100 ps while the combined duration of the two 180° pulses during one period was between 22 and 30 ps, far from a d-function approximation. For this reason the following procedure was used. The REDOR sequence with a fixed number of rotor cycles (between 5 and 10) was applied to a standard compound. Then the time intervals between pulses were slightly changed to get the maximum degree of the dipolar dephasing. The adjustment
Blumenfeld et al.
15186 J. Phys. Chem., Vol. 99, No. 41, 1995
USY
ZSM-5
VG-2
I 0.8 0.6 0.4 0.2
0
0
IO
20 30 40 Contact Time (mrec)
50
I
-80
-90 -100 -110 -120 %i chemical Shift (ppm)
Contra Time@“)
Contra Time (mrec)
1
-130 -80
I
-90 -100 -110 -120 29Si Chemical Shin (ppm)
-130 -80
.
,
I
-100
-110
-120 29SiChemicnl Shift (ppm)
-90
-130
Figure 5. Cross-polarization dynamics and corresponding 29SiCP MAS (top) and one-pulse (bottom)NMR spectra. The fitting of the experimental intensities obtained from the deconvolution of the CP spectra (top), as shown for VG2, is obtained using eq 7 and the parameters in Table 3.
turned out to be rather crucial; the deviation of the time interval from the best value by 5% resulted in 10% decrease of the REDOR effect. The check was always made with twice the number of rotor cycles to be sure that inevitable misadjustments do not accumulate during the long pulse sequence. The Processing of the REDOR Curves. For each number of rotor cycles two FID’s were obtained: with and without proton irradiation. After processing with the same scaling factor and integrating the areas within the same limits, the experimental REDOR responses ASIS0 were plotted us the number of rotor cycles, A special software was written that allowed to simulate the REDOR evolution according to the procedure described above. The input consisted of two files (the experimental curve and the “universal” REDOR curve for a given value of rotor period) and the set of adjustable parameters. The latter included the number of different silicon species (three at the maximum), the Si-H second moments, and the weights of each term. No automatic search for the best fit was used; instead, the parameters were varied manually until the best value of the sum of square deviations was achieved. There was no requirement that the sum of the weights should be 100%; however, it was always close.
Results Since the magnetization transfer operates from chemisorbed paramount importance to know its distribution among the various chemisorption sites. A fraction of the ammonia is chemisorbed on Lewis sites or hydrogen bonded to OH, while the other combines with acidic OH yielding The corresponding differential heats6 of chemisorption are I100 & 5 kJ/mol at a degree of coverage 19 5 1, I9 being defined as the ratio of chemisorbed to the total irreversibly adsorbed NH3. The amount of irreversibly adsorbed NH3 was determined for zeolites similar (but not identical) to those studied here. This information is contained in Table 1 together with the most important sample characteristics. As can be seen, the amount of irreversibly adsorbed NH3 is lower than the number of framework aluminum. For HZSM-5 the difference is not large. Thus, only one fraction of the Bronsted sites are strong enough to retain chemisorbed NH3 with an energy larger than ammonia, it is of
m+.
100 kJ mol-’. Equally important would be an estimate of NH3 retained by the Lewis and Bronsted sites. In ref 6 it has been shown that, for the dealuminated H-zeolites used in the present work, the NH3 retained by the Lewis sites that are detectable by the CO infrared technique had a differential heat of adsorption decreasing linearly with In 8. From this observation the ratio of the ammonia retained by the Lewis sites to the total ammonia irreversibly adsorbed can be estimated as well as the NH3 chemisorbed on the Bransted site. It is shown in Table 2. By comparing the estimate of the ammonia adsorbed on Bronsted sites (NH3)B to FA1 content, it is obvious that there is less chemisorbed ammonia than FAl. The highest ratio NHd FA1 is obtained, not ~ n e x p e c t e d l y , ~ for ~ ’ZSM-5. ~ Data by Macedo et a1.I4on NH3 adsorbed by HY have been reanalyzed by assuming that the irreversibly adsorbed NH3 corresponded, as in ref 6, to molecules with differential heats of chemisorption larger than 100 kJ/mol. For these samples again the degrees of neutralization of the Bransted sites by ammonia are in the range which can be calculated from (NH3)B/FAl, that is, about 30%. One-Pulse vs Cross-Polarization. Conventional ‘H 29Si CP MAS NMR was performed on the three zeolite samples. Because of the 29Sibackground of the glass vial which contains the activated sample and chemisorbed ammonia, it is impossible to obtain information from one-pulse experiments. The CP dynamics, that is, the evolution of the signal with respect to the contact time, together with the 29SiCP and one-pulse spectra are shown in Figure 5 . The one-pulse 29SiNMR spectra of the rehydrated samples exhibit partially resolved peaks corresponding from right to left to the Si @(nAl),n = 0, 1,2, contributions. An example of deconvolution is given for the mordenite VG2. If one compares the one-pulse (rehydrated zeolite) spectra to the cross-polarized (from chemisorbed N H 3 ) spectra, three major differences are observed. First, the spectral resolution is lower in the CP spectra because of line broadening. This effect is due to arrays of 29Sichemical shifts for Si sites with an equal number of neighboring Al. It can be explained in terms of a distribution of the T-0-T angles or Si-0 distances. For the hydrated sample the resolution is better because the framework contraints are released by filling
-
J. Phys. Chem., Vol. 99, No. 41, 1995 15187
Bransted Acid Sites and Surface Structure in Zeolites the cages. NH3 chemisorption on the activated sample could modify the T-0-T angle to some extent. The broadening effect is observed for the three samples, but the ZSM-5 is less affected. It is also the most stable lattice with the smallest A1 content. Second, the relative intensity of the different Si resonances is modified in the CP spectra compared to the one-pulse (rehydrated) spectra except for ZSM-5. The CP should favor Si sites with the largest density of rigidly bound protons. A Si Q4(OAl) unit is separated from the closest lattice OH or N b + by at least 0.3 nm, whereas Q4(lAl) and Q4(2Al) have a proton and potentially a NH3 in close vicinity. Consequently, it is expected that the Q4(OAl) intensity will be depressed in CP. In fact, this is observed for USY and VG2 samples; the apparent Si/Al ratio in the CP spectra is 5 and 6, respectively, as compared to the values of 7 and 11 obtained from the onepulse spectra. Yet, the 1P and CP signals are surprisingly similar. Two of the warnings made in the Introduction must be recalled. NH3 adsorbed on the Lewis sites located on the nonframework aluminum (NFA1) could be close enough to lattice Si to be used as an ‘H reservoir for the CP excitation. This could increase the CP signal of Q4(OAl). Part of the Bronsted site could be neutralized by NFAl and would not bear a proton or ammonium anymore, in which case the CP reponse should be damped. Third, the chemical shift for each Si resonance in CP MAS is not the same as that observed in the one-pulse spectrum. For USY, the lines are shifted by -2 ppm in the CP spectra compared to the one-pulse spectrum; the corresponding shift is f 1 . 5 ppm for VG2, and no shift is observed for ZSM-5. This effect might again result from different T-0-T angles and/or Si-0 bond lengths in the rehydrated compared to the activated samples. Among the three sieves studied here, the least stable is USY. Cross-Polarization Dynamics. The evolution of the 29SiCP MAS NMR signal with contact time was studied in the range 2-50 ms. Each spectrum was deconvoluted as the sum of three Gaussian lines corresponding to Q4(2Al),Q4(lAl), and Q4(OAl). For each sample, the width of the three lines was fixed to a reasonable value so that all the CP spectra could be deconvoluted similarly. The relative amplitudes were the only adjustable parameters. The time evolution of the three lines is shown in Figure 5, together with a nonlinear fit. The dynamics of the build up of 29Sisignal intensity was described as
\
llP/
where Zmax is the theoretical maximum intensity. This equation is the simplest solution of the phenomenological differential equations that describe the variation of proton and silicon spin temperatures under spin-locking condition^.^^ It was obtained under the assumption that the contact time dependence of silicon magnetization is governed by only two processes of energy exchange, namely, (i) between silicon and proton spin systems (Tcp process) and (ii) between proton spin system and the lattice (TlpHprocess). The complete solution includes additional dynamic processes that were neglected in eq 7. This is equivalent to the following assumptions: (i) the heat capacity of the proton spin system is much larger than that of the silicons, because cross-polarization has very little effect on the proton spin temperature; (ii) the exchange between the silicon spins and the lattice is slow on the cross-polarization time scale; TlpSi >i TCP;and (iii) the Hartmann-Hahn matching condition is fulfilled. Unfortunately, neither of these assumptions could be checked independently in this work.
$
‘t
0.8
1a
0.6
4
0.4
“0
50
100
150
200
# of Rotor Cycles
Figure 6. Data points represent the experimental increase of the REDOR fraction with the number of rotor cycles. The solid line is the predicted evolution using the parameters in Table 4.
What is more important is that high-speed MAS changes the nature of the cross-polarization process, so that the simple thermodynamic model of the energy exchange is no longer valid. This problem has received much attention during the past several years.40 Therefore, we will discuss the results obtained from the study of CP dynamics only at a qualitative level. The results of the CP fitting procedure are summarized in Table 3. The cross-polarization time Tcp of the Q4(nAl) lines increases with decreasing values of n. As it could have been anticipated, the cross-polarization rate is higher in the spectral region where the number of protons is larger. In an attempt to overcome the difficulties linked to the deconvolution, an altemative data processing has been checked. A three-dimensional file is generated with the first dimension accounting for the chemical shift, the second being the contact time, and the third is the signal intensity. A plane perpendicular to the chemical shift gives the time (intensity versus contact time) evolution for each chemical shift. Each slice is fitted with eq 7, and the values of TtpHand TCPare plotted versus the chemical shift. The two fitting parameters show continuous (and no discrete) variations over the spectral width. This trend will also appear in the REDOR reponse. The most puzzling fact about the CP dynamics comes from the comparison of the TCPvalues. There are large differences between samples compared to relatively small differences within a sample. The TCPvalues for VG2 are 4.0 f 0.1 times shorter than those observed for USY and the USY, TCPvalues are 1.46 f 0.11 times shorter than those observed for Z S M J . From the comparison of Tables 1 and 2, it is clear that the distribution of NH3 among the various Brgnsted sites is specific of the zeolite and that a proton pool not associated with the framework is involved in CP. It could be argued that eq 7 does not describe the CP dynamics correctly and that only “apparent” time constants can be obtained. Discrepancy between direct measurement of Tcp and TlPand figures obtained from the fitting of CP dynamics seems to be the rule instead of the exception, even for a simple system.27 The absolute value of the TCPparameter cannot be related to structural information, but as shown later, REDOR and CP will show a similar trend, from which some information can be gained. REDOR Experiments. The experimental REDOR evolution curves are presented in Figures 6 and 7 together with the best fit simulation curves obtained by the procedure described above. For each zeolite the minimum number of components (two for ZSM-5 and three for VG2 and USY) was taken to obtain a reasonable accuracy. The fitting parameters were the Si-H second moments (or the line widths which are related to the second moments by A = M2’12) of separate components and their weights. The best fit parameters are given in Table 4.It is clearly seen that the REDOR evolution for the ZSM-5 zeolite
15188 J. Phys. Chem., Vol. 99, No. 41, 1995
Blumenfeld et al.
TABLE 3: Fitting Parameters (in ms) for the Cross-Polarization Dynamics USY
VG2
Tcp
TI,,"
T~~
T~,H
~ 4 ~ 4 1 ) 9.9 12.9 Q4(lAl) 17.2 Q4(OAl)
52 52 52
2.4 3.4 4.3
276 103
29Sisites
-
USY -
HZSM-5
T~~
3
1.s'u
E
e
T,~H - I
20.0 23.6
ma
I
I
m'
i
- 0.55
ma
%
...*..
means larger than 1000 ms.
..e.,
-100
-90 -.,-,. A=
410
. . e . . ,
.( .._ e.
-120
*9Si Chemical Shift (ppm)
2.13 ~ H Z
,.
9 .
VG-2
1
-'"
_ _ _ _ _ _ _ - - - - -- - - 0
20
40
60
80
"."
# of Rotor Cycles -90
Figure 7. Data points represent the experimental increase of the REDOR fraction with the number of rotor cycles. The solid line is the predicted evolution using the parameters in Table 4. In addition, the evolutions calculated for each of the three contributions with the parameters indicated in Table 4 are shown. The solid line is the sum of them.
-100
.110
-120
29Si Chemiul Shift (ppm)
,
r
0.2
TABLE 4: Heteronuclear Second Momemt (M2Si-H,kHz2) and Contribution to the REDOR Fraction USY ~ ~ s 1 - H
contr, % M~SI-H
contr, % M~SI-H
contr, %
2.13 27 4.54 (1.96)" 20
0.07 (0.1 1)" 39 VG2 0.1 1 41 HZSM-5 0.44 23
0.01 32
-90
-100
-110
-120
29Si Chemiul Shift (ppm)
0.03 (0.01)" 29 0.01 76
The numbers in parentheses represent values which can still fit the REDOR fraction curve, under the condition to modify the weight of the contribution by -10%. They are shown to outline the margin of uncertainty. These symbols are those used in Figure 8.
proceeds on the longer time scale than that for the VG2 and USY. This corresponds to the absence of the short component (with a large second moment) in the deconvolution of the REDOR curve obtained for ZSM-5. The question immediately arises whether the three components found in the overall REDOR evolution correspond in any way to the three lines in 29SiCP MAS spectra of zeolites. This idea seems acceptable since the weights of the components follow the distribution of the N M R lines in such a way that the larger second moment corresponds to Si sites with more A1 neighbors. However, it tumed out,that this simple assignment is not correct. The REDOR evolution of individual lines of 29Si spectra manifested the same multicomponent pattem like we have seen in the case of the overall spectrum. The origin of this behavior is clearly seen from Figure 8 where spectral distribution of silicons with different values of Si-H second moments are presented. The procedure for obtaining these data was the following. For each number of rotor cycles the REDOR spectra were cut into slices with the width of several ppm. For each slice the average REDOR responses were calculated, and the corresponding REDOR evolutions curve was fitted with the numbers of components
Figure 8. Variation of the three (USY and VG2) or of the two (ZSM5) contributions w , (solid lines) to the REDOR fraction evolution in slices cut along the time dimension (or number of rotor cycles) at the chemical shifts indicated by the dots. The corresponding CP spectrum is superimposed, and the weight-averaged M z ~ I = - ~E,wtM21S1-H is represented by the heavy dashed line. The M Z , ~ 'are - ~ those in Table 4.
and the values of second moments that were obtained from the overall evolution curves. The only adjustable parameters remaining in the procedure were the weights of the components. They are plotted in Figure 8 together with the weighted sum of second moments. One can clearly see from these plots that the spectral distributions of silicon atoms with different proton surroundings are much wider than the distributions of silicons with 0, 1, and 2 Al atoms close by. Another important point is the overlapping of the distributions. As a consequence, the average second moment decreases from the left end of the spectrum in a more or less continuous way. Since we are unable to make assignments of the REDOR components in a straightforward way, we have made an effort to relate the obtained values of second moments with structural fragments that are in the vicinity of Si atoms. For this purpose, we invoked the well-known method of moments. In a powder, the so-called "rigid" heteronuclear second moment is
where the sum is extended over all the heteronuclear pairs S-H, r S H is the internuclear distance, and OSHis the angle of the
Brcdnsted Acid Sites and Surface Structure in Zeolites
J. Phys. Chem., Vol. 99, No. 41, 1995 15189 TABLE 5: Computed Second Moment and Line Width A for the Different Models Shown in Figure 9
1
Si
AI
M z ~ - "(kHz*)
A (kHz)
model
10.3 40.3
3.2 6.3 0
L:NHs D NH4+ isotropic
0
A
b H Si
AI
AI
Si
Figure 9. Structural models used to calculate the second moments (eq 9), given in Table 5. A is the basic structure with Si-0 = 0.162 nm, AI-0 = 0.172 nm, 0 - H = 0.105 nm, A i - 0 - A I = 141.3" and LHOSi = LHOAl = 108.5". The parameters are kept in B and D. In C, the angle HOSi is strongly distorted, namely 83.3". The W+cation is reorienting isotropically in B and C. In D, NH3 rotates about C3" parallel to OHN. See text.
S-H vector with an external magnetic field. If reorientation occurs rapidly about the axes making angle @'SH with rSH' vectors, then
(9) For instance, in the case of 15NH3chemisorbed on a Lewis site and reorienting rapidly about its C3v axis 1 - 3 cos2 109")2
M2motion = 29.6 ( kHz2 : , 3
(10)
4rNH
where 29.6 kHz2 is '/3ys2 y~~ fi2 s(s -k I), if rSH is expressed in angstroms. For 29Si***H this constant is 112 kHz2. Thus, the line width A narrowed by the motion in L:NH3 is _
_
_
A (kHz)
0.63 2 1.25 1.94 ~0.6
0.79 1.42 1.1 1 1.4