J . Phys. Chem. 1990, 94, 8821-8831
8821
High-Resolution Solid-state NMR of Exchangeable Cations in the Interlayer Surface of a Swelling Mica: 23Na,"'Cd, and 133Cs Vermiculites V. Laperche,+J. F. Lambert,$ R. Prost,' and J. J. Fripiat* Department of Chemistry and Laboratory for Surface Studies, University of Wisconsin, Milwaukee. Milwaukee, Wisconsin 53201 (Received: March 28, 1990)
The isotropic chemical shift of 23Na,"ICd, or 133Csin cation-exchangedLlano vermiculites is directly related to the hydration states. This swelling mica undergoes phase transitions easily detectable from the X-ray reflections along the C axis, perpendicular to the basal plane, the dml reflections. The reference states are the corresponding dilute salt solutions. With respect to this reference, the chemical shift can be accounted for by a general equation obtained by considering the configuration of the lattice and water oxygen ligands and the electron back-donation from the ligand to the cation. The quadrupole coupling constant (QCC) of 23Naat all hydration states and in the dehydrated state is small (QCC < 1 MHz), whereas in the dehydrated state the "ICs QCC is about 6.7 MHz (vQ = 0.48 MHz). For all three cations the asymmetry parameter (either quadrupolar or chemical shift) is close to one.
Introduction
There have been few studies aiming to probe the surface properties of swelling layered lattice silicates by using the highresolution solid-state NMR of exchangeable cations. On the other hand, this tool has received more attention in the field of zeo1ites.l There are several reasons for this lack of interest for smectites, besides the obvious difference in the industrial applications of both kinds of materials. Most often hydrated smectites have an interstratified structure consisting of a random superimposition of layers in different states of hydration, and in addition, smectites are aggregated into irregular stacks, called tactoids, resulting from delamination. Moreover, natural smectites contain paramagnetic impurities in rather large amounts (a few percent Fe2O3, for instance) in their lattices, and the N M R spectra are blurred by the paramagnetism of the lattice. Hectorite (the trioctahedral magnesium analogue of the dioctahedral aluminic montmorillonite) is quite clean in that respect, and Ij3Cs and Il3Cd MAS N M R studies2s3have been devoted to this material with the main purpose of determining the chemical shifts of exchangeable cation adsorption sites from the N M R features associated with chemical exchange processes. Among bidimensional layered lattice with cation-exchange properties, vermiculite (a trioctahedral magnesium silicate with a large number of AI/Si substitutions in the tetrahedral layer) presents unique features. It does not form interstratified structures upon hydration. Instead, for the Na, Mg, and Ca forms, real phase transitions are observed between the two-layer hydrate, the one-layer hydrate, and the dehydrated phases, as evidenced by sharp, intense, and narrow 001 reflections. The possibility of obtaining monocrystals has attracted the attention of crystallographers, and accurate structural determinations have been performed. They were reviewed recently by de la Calle and S~quet.~ Among vermiculites the material of choice is the Texas Llano vermiculite which has a high charge, the cation-exchange capacity being in thc ordcr of 2.30 mequiv/g, and an iron content that is low enough to prevent the undesired perturbations by paramagnetic impurities. I n the pasts-' the study of 'Hand 2H N M R of adsorbed H20and D 2 0 had brought about a large amount of information concerning their preferential orientation with respect to the lattice crystal axes and their motion. In the work presented here a high-resolution solid-state N M R study of 23Na, IIlCd, and 133Csvermiculites has been carried out with the purpose of correlating the chemical shifts of the resonance line with the hydration state. The existence of different adsorption sites for these cations could shed light on exchange selectivities. 'On leave of absence from CNRA, Station de Science du Sol, Route de St. Cyr, Versailles, 78206, France. Present address: CNRA, Versailles, France. *To whom correspondence should be addressed.
0022-3654/90/2094-8821$02.50/0
TABLE I: NMR Parameters for the Studied Cations" ionic 7% natural eq, I m2 IO6 s-I T I abundance, 7% radius, A Z 23Na 3 / 2 0.12 11.261 100 0.95 11
"'Cd 1 / 2 '33Cs 7 / 2
'H
1/2
-0.003
9.028 5.610 42.575
12.75 100 99.98
0.97 1.69 0.36
48 55 1
"eq is the quadrupole moment, y is the gyromagnetic ratio, and Z is the atomic number. bCovalent.
The 29SiN M R study of the Llano v e r m i c ~ l i t ehas ~ ~shown ~ that three main kinds of 3Q silicon environments exist, namely, )Q(4Si,OAI), 3Q(3Si,lAI), and 3Q(2Si,2A1), and the ordering of the tetrahedral substitutions has been carefully analyzed by Herrero et a1.I0 On the average, each ditrigonal (pseudohexagonal) cavity in the oxygen 001 plane contains 1.2 AI substituting silicon. The cations in the interlayer are exposed to a relatively narrow distribution of adsorption sites that should correspond to a relatively narrow distribution of chemical shifts. In contrast with smectites, where the negative charges originate mainly from isomorphic substitutions in the octahedral layer, the negative charges in vermiculite are localized in the tetrahedral layer. This is the main reason why, as recalled in the next section, water adsorption isotherms are steplike isotherms, each step being characteristic for one phase. The shielding of the nuclei of the exchangeable cations by their electron shells is affected by the electron transfer with the basal oxygen in the dehydrated phases. In hydrated phases the shielding should be less efficient, since the lattice oxygens play a less active role by contributing to a lower extent to the coordination of the exchangeable cation. In two-layer hydrates where the first coordination shell of the exchangeable cation contains water molecules only, deshielding should be maximum. By use of different cations in the interlayer of the same lattice, it will be shown that the chemical shift is, indeed, probing ( I ) Engelhardt, G.;Michel, D. High-Resolufion Solid-State NMR of Silicates and Zeolifes; Wiley: New York, 1987. (2) Kirkpatrick, R. J. In Reuiews in Mineralogy; Hawthorne, F. C., Ed.; The Mineralogical Society of America: 1988, Vol. 18, Chapter 9. (3) Bank, S.; Bank, J. F.; Ellis, P. D. J . Phys. Chem. 1989. 93, 4847. (4) de la Calle, C.; Suquet, H. In Reuiews in Mineralogy; Bailey, S. W., Ed.; The Mineralogical Society of America: 1988; Vol. 19, Chapter 12. ( 5 ) Woessner, D. E.; Snowden, E. S.; Meyer, G. H. J . Colloid Interface Sci. 1970, 34, 43. ( 6 ) Woessner, D. E. J. Magn. Reson. 1974, 16. 483. (7) Hougardy, J.; Stone, W. E. E.; Fripiat, J. J. J . Chem. fhys. 1976, 64, 3840. (8) Thompson, J. G. Clay Miner. 1984, 19, 229. (9) Herrero, C. P.; Sanz, J.; Serratosa, J. M. J . Phys. C: Solid State Phys. 1985, 18, 13. (IO) Herrero, C. P.; Sanz, J.; Serratosa, J. M. J. Phys. Chem. 1989, 93, 4111.
0 1990 American Chemical Societv
8822
Laperche et al.
The Journal of Physical Chemistry. Vol. 94, No. 25, 1990
TABLE If: Order of Magnitude of the Most Intense 001 Reflections cation no. of layers counts (xlo-,) N a+ 2 20-40 Na+ I -10 Na+ 0 -3 Cd2+ 2 8-12 Cd2+ 1 5-8 Cd2+ 0 51 cs+ 0 90% Lorcntzian character. bRelative contribution to the line of the component with the indicated 6. CMinor X R D line. dVery weak X R D line. ‘Shoulder.
Figure 4. X R D spectra (Cu Ka) of N a vermiculites showing the phase transitions fromthe two-layer hydrate (2L) to the one-layer hydrate ( I L) and to the dehydrated (D) phase. The numbers refer to the 23Na M A S or static spectra in Figures 2 and 3. 150
100
50
0
-50
-100
ppm
Figure 3. *’Na static spectra 1 bis, 3 bis, 5 bis, and 8 bis correspond to the M A S spectra 1 , 3, 5 bis. and 8 in Figure 2B. Solid line spectrum 9 bis is the static spectrum corresponding to MAS spectrum 9 bis shown as a dashed linc.
state in collccting the XRD spectra was more extensive than during the NMR experiment. Results. Figures 2 and 3 represent typical MAS and static spectra as wcll as somc examples of deconvolution. The first important information is that each phase is characterized by a chemical shift tjCG (CG: center of gravity of the line) and by a typical v ’ / ~ . The quantitative data are gathered in Table 111. Somc typical XRD arc shown in Figure 4. All MAS lines arc typically Lorcntzian. All spectra may be accountcd by one or two lines among three possible contributions: one corresponding to thc two-layer hydrate (fiCG = 4 0.5ppm), one corresponding to thc onc-layer hydrate.(dcG = -7 f 1 ppm), and one being the dehydratcd phasc (6CG = -18 f 1 ppm). Their y 1 I 2 a r e in the approximntc ratio I :4:2. Thc sccond important information is that the spinning sidebands (SSB) obscrvcd for the two-layer hydrate are much outside the
full-width (FW) of the corresponding static spectrum. ui12of this static spectrum for the two-layer hydrate is less than twice the v ’ / measured ~ for the corresponding MAS spectra. On the contrary, the static spectra obtained for the dehydrated phase encompass the SSB pattern of the MAS spectra. Spinning at -6 kHz results in SSB’s at -45 ppm from the main resonance. As far as the one-layer hydrate is concerned, the situation is more complex, since, in most cases, the MAS line contains two contributions, the additional contribution being that of the two-layer hydrate (MAS spectra 4 and 6) or that of the dehydrated phase (MAS spectra 5 bis, 7 , and 8). Interpretation. Because of the large quadrupole moment and low spin number of 23Na (Table I), the second-order quadrupolar shift 6,, cannot be neglected a priori.”,’* The observed does not correspond to the chemical shift hCs
where (17) Gerstein, B. C.; Dybowski, C . R. Transient Techniques in NMR of Solids: Academic Press: New York, 1985. ( 1 8) Lippma. E.: Samoson, A.: Magi J . Am. Cfiem.Sor. 1986. /08,1730.
The Journal of Physical Chemistry, Vol. 94, No. 25, 1990 8825
23Na, "'Cd, and 133CsVermiculites
The Na+ autodiffusion coefficients measured by Calvet et aL2' in homoionic smectites by the radiotracer technique were IO* and cm2/s for the two-layer hydrate and one-layer hydrate, respectively, and these orders of magnitude were also found for vermiculite.22 For an average jump distance of -5 A, that would correspond to correlation times between -lo-" and s, Le., smaller than uL-l. In the dehydrated phase Na+ is buried within the ditrigonal cavity, and its unknown diffusion coefficient must be cm2/s. Thus, the rapid cationic motion in the two-layer hydrate suggests that the line at +4 ppm represents an average chemical shift whereas the line at -1 8 i 1 ppm observed for the dehydrated phase does not result from averaging. Yet, as recalled above, there is an increase by a factor of 2 only in the v ~ of/ the ~ MAS spectra, between the situations where averaging is likely and that where it is unlikely. Another interesting observation results from the comparison of the width of the static and MAS spectra. In the two-layer hydrate, the broadening of the static spectrum is contributed mainly by the 23Na-lH heterodipolar interactions, and the static spectrum is not even twice as large as the MAS spectrum. In contrast, in the dehydrated phase where heterodipolar broadening is negligible, the full width (FW) is about 5-10 times larger than the FW observed for the static spectrum of the two-layer hydrate. Such a large multiplicative factor must be accounted for by the anisotropy of the chemical shift tensor. The FW of static spectrum 5 bis (which contains contributions from the chemical shifts of Na in the one-layer hydrate and in the dehydrated phases with heterodipolar broadening) is already -3 times larger than that of the two-layer hydrate. Thus, the correlation between the state of hydration and the chemical shift must be taken cautiously, since it covers different physical situations. It is worth recalling here the conclusions of earlier IH N M R studies7 of water in the two-layer and one-layer hydrates carried out on pseudomonocrystals of Na vermiculite with all crystallites having their C crystal axis-perpendicular to the ab plane (turbostratic structure), The C axis could be oriented with respect to the static field Ho, y being the angle between these vectors. In the two-layer hydrate a Pake doublet with a 2.7-G splitting was observed at room temperature for y = 0 and the splitting obeyed the (3 cos2 y - 1) rule. It was concluded that a model where (i) Na+ is 6-fold-coordinated to H 2 0 (Na' being in the middle plane in the interlayer), (ii) the water molecules are reorienting very rapidly around their axis of symmetry directed s), toward the center of the octahedron (correlation time and (iii) protons exchange within or between the hydration shell(s) fitted the experimentalobservations quite well.. The 6-fold coordination of sodium is in agreement with the crystallographic study and the pattern ofthe hko Bragg reflections,4 Such a model would also explain the modest broadening of the 23Na static spectrum with respect to the 23NaMAS spectrum (see Table 111). The 'Hspectra in the one-layer hydrate were found to be more complicated and were never explained satisfactorily.13 Two Pake doublets with splitting 7.1 and 16 G (for y = 0) were observed. - 1) rule quite closely. The crystalBoth obeyed the (3 lographic study of the one-layer hydrate suggests that Na+ is still midplane in the interlayer.4 The hydration shell contains two H 2 0 only out of the -4 H 2 0 present per uc. Thus, there should be two populations of water molecules, The larger splittings observed for the Pake doublets (as compared to the splitting observed for the single Pake doublet in the two-layer hydrate) motions, and in addition, be interpreted by slower N a could average two sites within the two populations of water molecules. In each of them Na+ would be coordinated to two water molecules, but the lattice oxygen environments would be different (for instance, above a triad of lattice oxygens forming the basis of a lattice tetrahedron or six lattice oxygens of the rings). Such a suggestion would explain why the MAS spectrum of 23Na in the one-layer hydrate is broader than those of the two-layer hydrate and of the dehydrated phase. The heterodipolar 1H-23Na
-
and (3)
where uL is the Larmor frequency, QCC is the quadrupole coupling constant, and q is the asymmetry parameter of the electric field gradient (EFG) tensor. Note that we shall use the symbol 7,as well, for the asymmetry parameter of the chemical shift tensor. The text indicates clearly whether we deal with qQ or vC. Our goal is to calculate ?iQS in order to obtain 6cs for each characteristic state of hydration. The task is not easy because the resonance lines (and the SSB) do not show singularities, such as quadrupole splitting, either because QCC is small or because q 1. In such a case, and for Gaussian lines, Freude et al.19920 have shown that, in the absence of distribution of chemical shifts due to chemical heterogeneity, residual heterodipolar interaction (for instance, 23Na-1H), and anisotropy of the magnetic susceptibility, the upper bound value of QCC is related to 6,, and that 6Qs (ppm) is directly proportional to ul12 ppm. Since practically all lines observed here are Lorentzian, we cannot calculate bQs in using Freude's final equation, but still we may apply its starting equation which is, for the center lineI9
-
-6QS(j/2)
= 2(M2MAS)1/2
(4)
where M 2 is the second moment of the resonance line. By definition" and for a Lorentzian line shape
For a Lorcntzian line M 2 diverges and the integral must be truncated. If u1I2= 2 / a is not very large, truncation in a domain of frequency in the order of, or larger than, IO yields a finite value of M 2 and numerical simulations show that
(Mzl = 0.987 (s-') V I / ' (s-')
(6)
Then, from eq 4, there follows that 6Qs (ppm) = -( 1 . 9 8 7 / ~ L ) ( ~ ~ ~ ~ ) I ~( 7~)
Since the maximum v 1 I 2 observed for the 23Na resonance lines is
= !dOol P
Equation 22 can be expressed as a function of x = M/doo,and of k / p where p = 0.9 according to the structural composition, and k has to be estimated. For doing this let us consider the particular case of the dehydrated Na sample. In the pseudohexagonal hole, the cation is surrounded by six oxygens per uc, and doolis 9.7 hi. Thus, there are 12 orbitals available per I /2 uc, and for 1 /2 uc/p this number is 12/0.9. Hence, the proportionality coefficient k is 12/9.7 = 1.237, if the basal spacings are expressed in angstroms. Hence. the numerical expression of eq 22 becomes
8 (kHz)
0:
I In
X
1.237 - x
+
where x = M/dool (A) and where M = M , Mo. In order to check the validity of eq 24, a blind program of trials and errors was set up in such a way that starting from the experimental M obtained from the experimental M , and from the calculated M,,, the fitting values of M were obtained. Values of the "best fitting" M were those for which the correlation coefficient of the linear regression between 8,, (kHz) and I In (x/( 1.237 - x) was higher than 0.99. The result is shown in Figure 10. RZ is 0.998, and the numerical cxprcssion of 8 is 8 (kHz) = 3.25I In M / ( N - M )
10
,/
/
8 -
,
/'
/
6 -
-/ d
/'
Mw/ 5uc calc
!
0
1
2
I
3
I*Ln(M/(N-M)) Na
+
Cd
o
Cs
- RsgI
Figure 10. Observed chemical shift 8 = BlsavOvs I In M / ( N - M). The solid line is the linear regression observed from eq 25 and the best fitting M shown in Table V . The insert shows the regression between the best fitting M,,,i, and the observed number of hydration water molecules, Mwobs?
per
'12
uc.
where N = 1.237dw, (hi) and where I is expressed in megajoules (MJ). The best fitting values of M are shown in Table V. From "best fitting" M , M,/IIZ uc can be recalculated and compared to the experimental M,/1/2 uc. The correlation coefficient of the linear regression between both sets of values (shown in the insert in Figure 10) is 0.97, if the experimental M , observed for the two-layer Na hydrate is dropped. The origin of the large discrepancy between M,,,,,, and Mw,obsobtained for this sample might be that the two-layer hydrate for which the measured 6,,, was -4 ppm is stable only in the presence of an excess water. Equation 25, fitted with the theoretical values of Mw,yields standard errors on a,, of 1.3, 1.7, and 2.77 ppm for the Na, Cd, and Cs vermiculites, respectively, while the standard error on the regression between M,,,, and M,,,, is 0.5 water molecule. Thus, the average relative error on the average water content is 14%. Considering the uncertainty on the range of water contents related to specified dool(as shown in Figure I ) , this error may be considered as reasonable. I t may be concluded that eq 25 is well-founded for the vermiculites studied here. It cannot be claimed that it can be generalized to any other phyllosilicates. However, there are two examples which show that, in first approximation, eq 25 correctly predicts the chemical shift. According to eq 25, the isotropic shift of 39K in muscovite (dm, = 10.1 hi) should be about -5 ppm. In this mica the coordination number Mo of K+ with respect to the lattice oxygens is 6, that is, three oxygens from each pseudohexagonal hole between which K+ is squeezed, account taken to its ionic radius (1.33 A). The ionization energy of K is 0.416 MJ. A preliminary MAS N M R study of muscovite, performed so far without the pulse sequence suggested by Kundar et al.35in order
-
(35) Kundar, A . C.; Turner, G.L.; Oldfield, E. J. J . Magn. Reson. 1986,
69. 124.
23Na, ' I ICd, and 'j3Cs Vermiculites to eliminate acoustic ringing, shows a rather narrow peak slightly upfield shifted with respect to a 0.01 M KCI reference. Other preliminary data obtained for a thoroughly dehydrated beidellite indicate that the uncorrected chemical shift of Z3Nais -1 8.5 ppm, e.g., the same shift as that observed in OL N a vermiculite. This last result is particularly interesting, since it shows that the number of AI substitutions in the tetrahedral layer does not play an important role. Indeed, the (Al/Si)IV ratio is 0.12 in beidelite, whereas it is 0.45 in the Llano vermiculite. Thus, the higher negative charge brought by the basal oxygen in vermiculite, as compared to beidellite, does not influence noticeably the 23Na chemical shift. Hence, this observation suggests that electron back-donations from lattice oxygen or from water are hardly distinguishable, and it answers to the inevitable criticism that these contributions are not considered as different in the present development. Still the simplicity of eq 25 is probably apparent. The sign of the shift, with respect to the reference dilute solution, is positive when M / ( N - M ) > 1 and negative when M / ( N - M) < 1. Since there are two full orbitals per ligand, the shift is positive when the numbcr of sites available (or explored) per one cationic charge is smaller than the number of available orbitals, that is, for N C 2 M . In that case, the electron back-donation in the interlayer is more efficient than in the reference solution, where N = 2M. The shift is negative when the number of sites available per cationic charge is larger than the number of ligands, that is, for N > 2 M . Per site randomly explored by one cationic charge, the number of orbitals is too small for efficient electron back-donation. If the interlayer water molecules are preferentially oriented with respect to the cation, the number of full orbitals shared with the cation must be smaller than the overall number of orbitals theoretically available. In this situation the shift should be positive. The two-layer hydrate of Na vermiculite fulfills this criterion. Preferential orientation should also be observed for the two-layer hydrate of Cd vermiculite. The main oxygen-exchangeable cation interaction in the theory presented here is assumed to be electron back-donation. This hypothesis is supported not only by the success of eq 25 in predicting thc isotropic chemical shifts and the number of hydration water molccules but also by the huge positive shift observed for the '"Cd-OH+ cationic species obtained in the NH,-dehydrated
The Journal of Physical Chemistry, Vol. 94, No. 25, 1990 8831
phase of Cd vermiculite. Indeed, in Cd-OH+, the hydroxyl must be covalently bound to Cd2+. In part, the possibility of calculating the size of the assembly ( N ) and, in part, the homogeneity of the interlayer surface, as evidenced by the step isotherms observed for water adsorption,"-I3 render the statistical treatment proposed here easily tractable. It would be most interesting to extend it to the exchangeable cations in zeolites. Tokuhiro et aLB have suggested that the shift observed for Cs close to a six-member ring in zeolite A is more paramagnetic than the shift observed for the same cation close to the eightmember ring by a weaker electron back-donation from the eight-ring than for the six-ring. The task of estimating N a n d M would be more difficult in zeolites for the cations occupy different lattice sites where they are coordinated to variable numbers of hydration water molecules and oxygens. The discussion of the experimental results, so far, has focused on the chemical shift. The other N M R parameters, such as quadrupole coupling constant (QCC) and 7,have been discussed in the interpretation sections, and they do not call for special comments. The large QCC observed for I3'Cs in vermiculite is probably related to the distortion of the coordination shell made from 9 to 12 lattice oxygens. For the three studied cations the asymmetry parameter is, in all cases, close to 1. In particular, no quadrupole splitting has been observed for 23Naand '"Cs, as also seems the case in zeolites. In summary, the sole merit of this work is in showing that, on a homogeneous surface, each hydration stage is related to a defined isotropic chemical shift and that this shift contains information on the configuration of the coordination shell. Understanding the effect of hydration of interlayer cations in a medium where anions are excluded could also profit solution c h e m i s t r ~ . ~ ~ . ~ ' Acknowledgment. An N S F grant (DIR-87198088) and N I H grant (RR04095), which have partly supported the purchase of the GN-500 N M R instrument, are gratefully acknowledged. We thank Dr. D. Tinet, who was in Milwaukee on a NATO fellowship, for interesting discussion. V.L. thanks the C N R A (Versailles, France) for financial support. (36) Bloor, E. G.;Kidd, R. G.Can. J . Chem. 1968, I S , 3425. (37) Popov, A. Pure Appl. Chem. 1979, 5 1 , 101.