Local Structure and Protons in Non-Stoichiometric Pseudo-Cubic Pollucite Mineral by Multinuclear NMR

: The pollucite structure is considered as a candidate ceramic crystalline matrix for the ceramic immobilization and long-term storage of 135 Cs and 137 Cs ﬁssion products, and thus, their ﬂuid inclusions. species from the Tanco pegmatite and the Mt. Mica pegmatite, with powder XRD patterns fully consistent with that of a cubic Ia -3 d space-group lattice symmetry, were studied by multinuclear MAS NMR spectroscopy at 9.4 and 17.6 T. The 29 Si SP-MAS spectra indicate a disordered Si/Al distribution incompatible with the random and Loewenstein distributions, because of a clear prevalence of the (3Si,1Al) (2Si,2Al) environments. 29 Si CP-MAS spectra show protons are preferen- tially with having low content of Al in sphere of coordination. Spec- imen Tanco has phyllosilicate impurities to with of 27 Al{ H} dipolar Al be discriminated from distance to protons, compatible isotropic d to molecules. spectra resolved the presence of hydrous O-rich impurities Tanco


Introduction
The mineral species of pollucite, (Cs,Na)[AlSi 2 O 6 ]·nH 2 O, where (Cs + n) = 1, is considered a Cs-rich "zeolite" mineral of the analcime structure-type with cubic Ia-3d symmetry according to the International Mineralogical Association [1]. The understanding of the crystal structure of pollucite is particularly important because of the interest in the field of nuclear technology as a possible matrix from the immobilization and long-term storage of 135 Cs and 137 Cs products of nuclear fission. In this application, it is critical to know the hydrogen speciation in order to understand the structural stability and mobility of the cavity cations, i.e. the Cs atoms. It is quite obvious that an open porous structure with zeolitic water cannot be used in this industry. Thus, the local structure of hydrogen atoms in pollucite minerals must be analyzed in detail.
It has been suggested for pollucite the space group Ia-3d, assuming that pollucite and analcime were analogous structures, despite Cs atoms being located at 16b positions instead of the 24c positions [2], a site where the Na atoms in solid solution have been placed [3]. Using infrared spectroscopy, a different chemical formula for pollucite has been recommended as Cs (1−x) Na x AlSi 2 O 6 ·xH 2 O with x~0.3, where the Na atoms and H 2 O molecules occupy the same site than Cs atoms, but probably displaced from the center of the cavity [4]. A similar model in which the analcime framework NaAlSi 2 O 6 ·(H 2 O) was attributed to pollucite Cs x Na y Al x+y Si 48−x−y O 96 ·(16 − x)H 2 O (in which 2y ≥ 16 − x ≥ y), using specimens with Na 2 O between 1.59 and 4.34 wt.%, is available in literature [5]. In this model, Cs atoms are at 16b sites, the water molecules occupy the large voids of this same set that are not occupied by the Cs, and Na atoms are at 24c sites in the positions between the water molecules. Recent models have considered isotropic as well as anisotropic (birefringent) types of pollucite from the optical properties, the first one with cubic Ia-3d symmetry and the second one with tetragonal I4 1 /acd space group symmetry having two tetrahedral sites for Si and Al atoms, one site for Cs atoms, one site for Na atoms and three sites for oxygen atoms, as well as the water molecules at the cavity sites [6]. An additional model has been derived from a Cs-substituted leucite having three T sites for Si and Al atoms and one cavity site for Cs atoms with a tetrahedral symmetry in the I4 1 /a space group, by using high-resolution neutron powder diffraction [7]. Figure 1a shows a lattice model of pollucite structure with cubic Ia-3d symmetry at room temperature from recent single-crystal X-ray and neutron diffraction [8]. This lattice model consists of framework sites formed by Si and Al atoms in a single tetrahedral T site with a disordered Si/Al distribution. Thus, ten different medium-range atomic organizations can potentially be formed for each Si and Al atoms surrounded by other 0 to 4 Si and Al atoms (Figure 1b). A quantitative estimation of the different structural possibilities, to evaluate the order-disorder degree, cannot be done by reciprocal-space techniques. Figure 1a shows Cs atoms also at a single 16b cavity site, whereas Na atoms are located in a 24c position, as already suggested [5]. The analysis of the difference-Fourier maps of the nuclear density suggests the presence of extra-framework water molecules with oxygen sharing the Cs site, having extremely weak hydrogen bonds on the basis of atomic distances and bond angles [8]. However, Yanase et al. (1997) [9] ignored the structural role of water molecules, as it is known that pollucite does not show any loss of weight below 573 K [10] and dehydration appears at 913 K [11]. The relationship between the dehydration's temperature and water is an important datum, because zeolitic (molecular) water is lost between 100 and 150 • C, whereas protons as an OH group resist in most cases up to temperatures higher than 450 • C. Thus, the structural role of water molecules from reciprocal-space techniques and infrared spectroscopy in pollucite still remains highly enigmatic.
Available NMR studies of pollucite structure have shown that synthetic samples [12] have a broader spectra than that of natural specimens [13], suggesting higher local disorder in the former. A 29 Si NMR spectrum in cesian analcime showed only four resonances at −91, −96, −101 and −106 ppm [14], but the line intensities were not interpreted in detail. Ordered Cs leucite has been prepared by Cs-for K-exchange of a leucite precursor, but a monoclinic P2/a space group has been found [15]. In all of these studies, the role of protons in the crystal structure at the local scale in these tectosilicates has not been studied by NMR. In this work, we address the study of two natural mineral specimens, with contrasting degrees of alteration, to resolve the local structure and role of H atoms of pollucite by means of multinuclear NMR analysis at 9.4 and 17.6 T. Due to the different degree of alteration, it was possible to identify the role of -OH groups and of H 2 O molecules in the specimens, and the real features of the mineral structures. and Na atoms (red), according to [8]. (b) At the right side the ten possible chemical environments for the single T site, including Si atoms (blue) in Q 4 n (4-nSi,nAl) for n = 0, 1, 2, 3 and 4, and Al atoms (red) in q 4 (4-nSi,nAl) for n = 0, 1, 2, 3 and 4.

Specimens
Two specimens of pollucite have been selected for this work. Specimen Tanco is a rock-forming mineral from the Tanco pegmatite in Manitoba, Canada. It is a white and opaque sample that shows turbidity and alteration along fractures when observed by optical microscopy, similar to most other pollucite specimens with some alteration to kaolinite and muscovite low-crystallinity products, as it is well known in literature [16]. In the case of the pollucite from Tanco, the alteration has been described as the product of low-temperature hydrothermal fluids at the deuteric stage, giving rise to the formation of veinlets with kaolinite, montmorillonite-illite mixed-later sequences, lepidolite, smectite minerals, and other anhydrous minerals [17,18]. However, specimen Mt. Mica is a singular sample of pollucite as a pristine sample from a pocket of the Mt. Mica pegmatite in Maine, USA. It is a fully transparent specimen without any transformation to secondary minerals and with a lower Na2O content in comparison with most other natural specimens of pollucite in which crystal-structure determinations have been performed. The two specimens are birefringent when examined in a polariscope and also in thin sections by optical microscopy, i.e., their birefringent character is "anomalous" or unexpected if a cubic symmetry or isotropic crystal structure is a priori accepted.

Chemical Analysis
Spatially resolved quantitative chemical analysis of major elements Si, Al, Na, and Cs were obtained by electron microprobe analysis (EMPA), using point counting technique to obtain information on the chemical homogeneity, with a JEOL-8900M electronmicroprobe from ICTS Microscopía Centro de Asistencia a la Investigación de la Universidad Complutense de Madrid, CNME ELECMI-Centro Nacional de Microscopía Electrónica, (Spain). The spot diameter of the probe was circa 2 μm and the final operating conditions were 15 kV and 25 nA. These experimental conditions were first used for the chemical quantification of minerals standards, including albite NaAlSi3O8 in a specimen and Na atoms (red), according to [8]. (b) At the right side the ten possible chemical environments for the single T site, including Si atoms (blue) in Q 4 n (4-nSi,nAl) for n = 0, 1, 2, 3 and 4, and Al atoms (red) in q 4 (4-nSi,nAl) for n = 0, 1, 2, 3 and 4.

Specimens
Two specimens of pollucite have been selected for this work. Specimen Tanco is a rockforming mineral from the Tanco pegmatite in Manitoba, Canada. It is a white and opaque sample that shows turbidity and alteration along fractures when observed by optical microscopy, similar to most other pollucite specimens with some alteration to kaolinite and muscovite low-crystallinity products, as it is well known in literature [16]. In the case of the pollucite from Tanco, the alteration has been described as the product of low-temperature hydrothermal fluids at the deuteric stage, giving rise to the formation of veinlets with kaolinite, montmorillonite-illite mixed-later sequences, lepidolite, smectite minerals, and other anhydrous minerals [17,18]. However, specimen Mt. Mica is a singular sample of pollucite as a pristine sample from a pocket of the Mt. Mica pegmatite in Maine, USA. It is a fully transparent specimen without any transformation to secondary minerals and with a lower Na 2 O content in comparison with most other natural specimens of pollucite in which crystal-structure determinations have been performed. The two specimens are birefringent when examined in a polariscope and also in thin sections by optical microscopy, i.e., their birefringent character is "anomalous" or unexpected if a cubic symmetry or isotropic crystal structure is a priori accepted.

Chemical Analysis
Spatially resolved quantitative chemical analysis of major elements Si, Al, Na, and Cs were obtained by electron microprobe analysis (EMPA), using point counting technique to obtain information on the chemical homogeneity, with a JEOL-8900M electron-microprobe from ICTS Microscopía Centro de Asistencia a la Investigación de la Universidad Complutense de Madrid, CNME ELECMI-Centro Nacional de Microscopía Electrónica, (Spain). The spot diameter of the probe was circa 2 µm and the final operating conditions were 15 kV and 25 nA. These experimental conditions were first used for the chemical quantifi-cation of minerals standards, including albite NaAlSi 3 O 8 in a specimen close to the strict stoichiometry, and no alkali migration nor specimen degradation was detected. The wt.% contents of FeO, MnO, MgO, CaO and TiO 2 were also measured, but in all cases these values are below or close to the real detection limit of this technique.

Powder X-ray Diffraction (XRD)
The two specimens were ground to a particle size of less than 10 microns and were studied by powder XRD patterns using a Bruker D8 Advance instrument from Instituto de Cerámica y Vidrio, (ICV-CSIC, Madrid, Spain) with CuKα 1 radiation at 30 kV and 25 mA, step 0.01 • and step time of 92 s, with 2θ between 5 • and 60 • . The two XRD patterns were compared with the calculated powder XRD after structural determinations from neutron and X-ray diffraction studies with available lattice models by using the cc files of the of the ICSD database (https://icsd.fiz-karlsruhe.de/, accessed on 25 February 2022). The X-ray diffraction (XRD) patterns of the two specimens are fully consistent with a cubic symmetry as in Collection Code CC 84321 [9], in the ICSD (Inorganic Crystal Structure Database) database, as will be studied later in detail.

Nuclear Magnetic Resonance (NMR)
High-resolution 29 Si, 27 Al, 23 23 Na, 27 Al and 29 Si nuclei, respectively. The samples were spun at range 14 kHz at 9.4 T. π/2 (4 µs) and π/8 (2 µs) pulses were used to record Si and Al or Na and Cs signals. A recycle delay of 10 s was used in 27 Al NMR signal and 133 Cs NMR spectra. A recycle delay of 1s was used in 23 Na NMR. An 1800 s recycle delay time was used for 29 Si to avoid saturation effects. The cross-polarization experiments (CP-MAS) were performed by using { 1 H} 27 Al CP-MAS, { 1 H} 29 Si CP-MAS NMR and { 1 H} 23 Na CP-MAS spectra were recorded with a 4 mm MASDVT TRIPLE Resonance HYX MAS probe, at ν r = 14 kHz. Relaxation delay was 5 s for 27 Al CP-MAS, and 3 s for 23 Na CP-MAS and 29 Si CP-MAS. The number of scans was 16k, 4k and 24k, respectively. The contact time for the 29 Si CP-MAS spectra in specimen Tanco was from 8 ms to 2 ms by using 34,816 scans. However, after 73,728 scans at 1 ms in this specimen, the 29 Si spectrum (not shown in figures) was still too noisy, despite the Q 3 and Q 4 signal being resolved. Specimen Mt. Mica was studied in 29 Si CP experiments from 8 ms to 1 ms of contact time, with a reasonable signal-to-noise ratio in all cases. Variable amplitude polarization transfer was achieved with a contact time from 0.5 to 4 ms for 27 Al CP-MAS in specimen Tanco, but only 4 ms for specimen Mt. Mica. A contact time of 8 ms was used for the 23 Na CP-MAS spectra of the two specimens with 10,240 scans, whereas 7 ms were employed for the 133 Cs CP-MAS spectra with 10,240 scans. High-power SPINAL 64 heteronuclear proton decoupling was applied during acquisition. Tetramethylsilane, Al(H 2 O) 6 3+ and 1 M NaCl aqueous solutions were used as external references at room temperature. 1 H MAS NMR spectra were recorded with 64 transients with a recycle interval of 2 s. H 2 O was used as external secondary reference. The estimated uncertainties in the measured chemical shift of the NMR components was approximately 1 ppm for the 29 Si and 27 Al spectra, 3 ppm for the 23 Na spectra, 5 ppm for the 133 Cs spectra, and 0.1 ppm for the 1 H spectra at 9.4 T. These figures are higher for the noisy CP-MAS spectra. The 29 Si signals (I = 1/2) were analyzed by comparison of experimental curves with simulated profiles using Gaussian components for the different chemical environments, with different but similar linewidths as there is not a physical basis for assigning identical linewidths to different chemical environments in the simulation of a single spectrum.
Rotational Echo Double Resonance (REDOR) 27 [20] for details about the 27 Al{ 29 Si} HMQC experiment at 17.6 T. Spectra were fitted by using Bruker WINFIT software and DM2011 [21]. Table 1 shows the chemical composition of pollucite in specimen Tanco and Figure 2 gathers the sites of the chemical analysis as well as the effects of low-temperature hydrothermal alteration with a veinlet filled with hydrated unidentified hydrated aluminosilicate minerals. From these analyses, the chemical composition of this specimen is clearly a nonstoichiometric compound with an average Cs 0.83 Na 0. 20  Pristine areas of pollucite, far from veinlets and far from regions with a high microporosity, due to water-rich fluid circulation, alteration and precipitation of new minerals, have a Cs content of about~34 wt.% Cs 2 O, but areas side by side of these alterations can reach values up to~39 wt.% Cs 2 O. Three types of secondary minerals have been found, one with a Si/Al ratio close to 2, the other two with a Si/Al ratio closer to 3, which could be hydrated minerals as total oxide in recorded analysis is lower than~96 wt.%. Table 2 exhibits the chemical composition of specimen Mt. Mica, which is free of alteration and thus has a more homogeneous chemical composition than the other studied specimen, as Cs 0.94 Na 018 Al 1.23 Si 2.78 O 6 . It shows a higher Cs content with~36.5 wt.% Cs 2 O, but a similar Si/Al ratio with an average value of~2.31, and hence a similar value to that found in specimen Tanco, indicating also a non-stoichiometric compound.   Table 1. Analysis nº 1 to 5 correspond with pollu contrast) and nº 6 to 9 with hydrated unidentified minerals (dark contrast). Note the absence of K in our pollucite specimens at the EMPA sensitivity, su these atoms are not easily bonded in this particular open framework. However, N were also allocated in the cavity sites of the pollucite structure, having a higher co  Table 1. Analysis nº 1 to 5 correspond with pollucite (clear contrast) and nº 6 to 9 with hydrated unidentified minerals (dark contrast). Note the absence of K in our pollucite specimens at the EMPA sensitivity, suggesting these atoms are not easily bonded in this particular open framework. However, Na atoms were also allocated in the cavity sites of the pollucite structure, having a higher content in specimen Tanco than in specimen Mt. Mica. Figure 3 exhibits a comparison between the experimental powder XRD patterns of the two used specimens and some alternative calculated powder XRD patterns from suggested space groups in literature for the pollucite structure. Figure 3a shows the pattern from the cubic Ia-3d space-group lattice model, with one T site (48 g) occupied by both Si and Al atoms, for a specimen with CsAlSi 2 O 6 chemical formula, with most of the (hkl) peaks from file CC 84321 [9]. Figure 3b shows the small change of the powder pattern after the cubic-tetragonal phase transition to I4 1 /acd space group, which is mainly observed in the splitting of the (008) peak into (008) and (080) peaks, as shown by file CC 84322 [9]. In this model, one T site (16f) is suggested for Al atoms and one T site (32g) was found for Si atoms. A different tetragonal lattice model, with I4 1 /a space-group symmetry, has been suggested by using a CsAlSi 2 O 6 composition from a Cs-substituted leucite, where the contrast with the tetragonal symmetry is manifested by optic instability [7], i.e., by optical birefringence absent in an isotropic structure, as file CC 66921 in the ICSD database. Important differences appear in the powder XRD pattern in comparison with the I4 1 /acd model, including the emergence of a (101) peak at 2θ = 9.1 • , other peaks with low intensity as (222) at 2θ = 22.5 • , (303) at 2θ = 27.7 • , and some additional peaks with very low intensity mainly at 2θ between 42 • and 49 • (marked with a green square), as well as some peak broadening related with the phase transition (marked with * in Figure 3c), including a sharp splitting for the (008) diffraction into the (008) and (080) reflections. This particular lattice model involves three T sites (16f) occupied by both Si and Al atoms in a 2/1 ratio, to satisfy the chemical stoichiometry of the mineral formula. The XRD pattern of specimen Tanco is fully compatible with the Ia-3d model, despite a small diffraction at~27.7 • that could be interpreted as diffraction from the (303) in the I4 1 /a model. However, the (101) and (222) signals, of the tetragonal structure, were not recorded in this specimen. No peak splitting was found in the (008) reflection. Some additional peaks due to impurities (red squares in Figure 3) were resolved from mineral phases that were not identified.

Powder XRD Patterns
The XRD pattern of specimen Mt. Mica was also fully compatible with the cubic Ia-3d model, as the characteristic diffractions of the two tetragonal models were not recorded. The diffraction peak for the (008) plane is very fine without any indication of peak splitting or broadening. No impurities were detected in this specimen.
From this data no indication of lowering of symmetry from the cubic symmetry is detected by conventional powder XRD methods, and thus only one T site for Si and Al atoms and one cavity M site for Cs atoms were expected in the NMR spectra, as well as some other additional signals from impurities in specimen Tanco. Figure 4 reports the 29 Si spectra at 9.4 T of the two pollucite specimens, using singlepulse (SP-MAS) and cross-polarization (CP-MAS) experiments. The degree of order of pollucite can be evaluated with the 29 Si spectra, from the expected occupancies of Si atoms in a Si-Al distribution considering the number (n) of Al atoms around a Si atom in Q 4 n environments, including (4Si,0Al), (3Si,1Al), (2Si,2Al), (1Si,3Al) and (0Si,4Al). Table 3 shows the line positions and intensities of the simulated 29 Si spectra of the two specimens from Tanco and Mt. Mica, in comparison with the expected intensities for a random distribution (RD), and also for a Loewenstein distribution (LD) for Al-O-Al avoidance when the Loewenstein's rule [22] takes place (by using a Si/Al ratio of~2/3 in Tables 1 and 2), and also data from a Cs-exchanged leucite [15]. The pattern from the I41/acd lattice model after file CC 84322 [9] showing the peak splitting of the (008) cubic structure into (008) + (080) peaks of the tetragonal structure. (c) The pattern from the I41/a lattice model after file CC 66921 [7] to show some additional changes to peak splitting as some peak broadening mainly in peaks marked with an asterism, and also new peaks (green squares) including  -3d model in (a). The patterns of (d,e) have been amplified by x3 to show the peaks with a low intensity. The details of the (008)-(080) peak splittings are for 2θ° values between 53° and 54° for the same width parameters, but the intensity of the (008) peaks in the two specimens have been modified to be comparable with that of the used models. Peaks that have a broadening with the phase transition are marked with *.  [9] with the main (hkl) peaks. (b) The pattern from the I4 1 /acd lattice model after file CC 84322 [9] showing the peak splitting of the (008) cubic structure into (008) + (080) peaks of the tetragonal structure. (c) The pattern from the I4 1 /a lattice model after file CC 66921 [7] to show some additional changes to peak splitting as some peak broadening mainly in peaks marked with an asterism, and also new peaks (green squares) including  The relative areas in the 29 Si SP-MAS spectra of the two mineral specimens deviate from the random distribution and Loewenstein distribution, and also from that of the synthetic Cs-leucite, and they have a prevalence of the Q 4 2 (2Si,2Al) and Q 4 1 (3Si,1Al) chemical environments, the expected ones for a local Si/Al ratio ~2.3 if it exists in a uniform dispersion of charges with partial avoidance or some restrictions for the Q 4 0 (4Si,0Al), Q 4 4 (0Si,4Al) and Q 4 3 (1Si,3Al) environments. It must be related to medium-range order polyatomic organization with local charge compensation of charge by monovalent (1+) cations.    The relative areas in the 29 Si SP-MAS spectra of the two mineral specimens deviate from the random distribution and Loewenstein distribution, and also from that of the synthetic Cs-leucite, and they have a prevalence of the Q 4 2 (2Si,2Al) and Q 4 1 (3Si,1Al) chemical environments, the expected ones for a local Si/Al ratio~2.3 if it exists in a uniform dispersion of charges with partial avoidance or some restrictions for the Q 4 0 (4Si,0Al), Q 4 4 (0Si,4Al) and Q 4 3 (1Si,3Al) environments. It must be related to medium-range order polyatomic organization with local charge compensation of charge by monovalent (1+) cations. Thus, one or two Cs atoms must be locally associated to Q 4 1 (3Si,1Al) and Q 4 2 (2Si,2Al) environments, respectively, in a similar way to how K atoms are locally correlated with Al atoms in the K-feldspars [24,25].  The center of gravity (c.g.) of the 29 Si SP-MAS spectra is~−98 ppm in the two studied specimens (Figure 4a,e, Table 3). However, the c.g. in the 29 Si CP-MAS spectra records a broad and irregular resonance centered at circa −85 ppm from Si atoms in Q 3 sites from phyllosilicate impurities, in addition to the Q 4 pollucite signals. The position of this Q 3 resonance does not change for different values of contact time, but the relative intensity of the Q 3 and Q 4 components changes with this parameter, i.e., Q 3 /Q 4 ratio~0.6 for 8 ms and Q 3 /Q 4 ratio~1.04 for 2 ms (Table 4). Thus, the c.g. of the total spectra is slightly displaced to fewer negative values in specimen Tanco (Figure 4b-d) from −94.4 ppm at 8 ms to −93.7 ppm at 2 ms. This displacement of the c.g. of the total spectra is small in comparison with the change in the Q 3 /Q 4 ratio because the c.g. of the pollucite component is also displaced, but to lower shielding, as the contact time decreases up to a chemical shift −101 ppm for 2 ms (Table 4). This displacement at more negative chemical shifts occurs because of a change in the intensity of the Q 4 n signals, showing an increase of intensity for the (4Si,0Al) and (3Si,1Al) environments, with respect to the (2Si,2Al) environments, as indicated by the Q 4 1 /Q 4 2 ratio (Table 4). However, the resonance forms Si atoms at (0Si,4Al) in this impure specimen that cannot be observed due to an overlap with the signal from the Q 3 impurities, although a similar figure to that of the pristine specimen has been used for these semi-quantitative simulations.

NMR Spectroscopy
As specimen Mt. Mica does not have such Q 3 impurities, the 29 Si CP-MAS spectra show more clearly these changes of pollucite with the contact time ( Table 5). The c.g. moved from −99.6 at 8 ms to −101.1 at 1 ms at the same time the Q 4 1 /Q 4 2 ratio increases from 0.76 at 8 ms to 0.91 at 1 ms, values that are far from experimental error after the estimated uncertainties. These changes suggest that protons are preferentially bonded to Si atoms that are surrounded by a few or no Al atoms, giving rise possibly to Si-OH bonds as a Q 3 local framework depolymerization and explaining the easy transformation of pollucite to Q 3 phyllosilicate minerals, such as muscovite, smectite and kaolinite, frequently observed in nature [16][17][18].  27 Al CP-MAS of the two specimens at 9.4 T, with signals from Al atoms in tetrahedral and octahedral coordination in these specimens ( Table 6). The Q 4 tetrahedral Al site of pollucite in the two specimens gives a very similar asymmetric signal in all of the NMR spectra, with a center of gravity circa 58.5 ppm, implying a single site distribution for Al atoms at these experimental conditions. However, the 27 Al CP-MAS spectra in specimen Tanco show Al atoms from octahedral coordination at circa 0 ppm and from tetrahedral coordination in a Q 3 environment at~71 ppm, which can be associated with the phyllosilicate impurities. The intensity of the Al atoms in a Q 4 environment of pollucite in the 27 Al CP spectrum is very low in comparison to the same signal from impurities. In specimen Tanco, the 27 Al CP-MAS spectra were performed for contact times between 0.5 ms and 8 ms (shown in Figure 5b). With the decrease of the contact time, the relative intensity of the resonance at 0 ppm from octahedral aluminum increases, and the resonances of tetrahedral aluminum decreases. Thus, all of these data indicate that Al-OH bonds are much less developed than Si-OH bonds in pollucite. However, no signals from impurities are recorded in the pollucite from specimen Mt. Mica with a CP spectrum identical to the SP spectrum, as expected. The 27 Al{ 29 Si} HMQC spectrum at 17.6 T (Figure 5e), mediated by the scalar J-coupling between Al and Si nuclei, shows the direct connectivity between the Al atoms and the Si atoms with different chemical environments. However, the signal at −107.5 ppm is absent, confirming the attribution of the~−107.5 ppm resonance to Si atoms in a (4Si,0Al) environment, a result that was confirmed by additional (not shown) 27  When the magnetic interaction between 27 Al and 1 H nuclei is analyzed by means of the dipolar interaction, it is found that two different Al signals with similar intensities are recorded at 58.8 ppm and 60.2 ppm, which are needed to account for the full REDOR evolution displayed in Figure 6. The full simulation of the 2D spectra shows that the Q 4 n (4-nSi,nAl) distribution of the silicon atoms connected to the Al site at 58.8 ppm is 9/70/21, whereas it is 32/45/23 for the one extracted at 60.2 ppm for n = 3/2/1 environments, respectively. This confirms the presence of two different spectroscopically distinct structural T sites for Al atoms, as this doublet is very difficult to be explained from a random or a Loewenstein distribution of Si and Al atoms.
If the dipolar evolutions 27 Al{ 1 H} of the two pollucite specimens are compared with that of gibbsite Al(OH) 3 , i.e., a natural aluminum hydroxide with well-developed Al-OH bonds, a much slower rising under REDOR evolution appears, related to a weaker Al/H dipolar coupling and thus larger Al-H distances and/or less H surrounding the Al sites. The marked slower increase of the REDOR curve observed in specimen Tanco with respect to specimen Mt Mica follows the same reasoning and shows a lower amount of H and/or longer Al-H distances for the former. It can also be seen that the site with a chemical shift of 58.8 ppm shows a slower REDOR dephasing for both samples and hence is less protonated than the component at 60.2 ppm. Thus, as the two pollucite specimens are birefringent, a possible explanation is related to a local structure with two sites for T cations, and thus a lower average symmetry than cubic symmetry, as suggested by Frank-Kamenetskaya et al. (1995) where two T sites were inferred in a tetragonal structure with a I4 1 /acd space-group symmetry [6], from an incipient displacive transition [7]. Figure 7 exhibits the 133 Cs SP-MAS and CP-MAS spectra, as well as the 23 Na SP-MAS and CP-MAS spectra of the two studied pollucite specimens, indicating that the cavity cations are allocated in broad site distributions. The line shape for 23 Na with its sharp rising left edge and low decaying right wing is characteristic of the so-called "Czjzek" line shape [26], which applies in the case of a fully random distribution of quadrupolar interaction (i.e., electric field gradient at the Na site). As the Si/Al distribution is not random but some charge dispersion exists, perhaps local charge compensation could be at work, giving rise to polyatomic Cs(Na)+Al medium-range order schemes. In addition, as the 133 Cs atoms display a smaller quadrupolar interaction, the line shape of the 133 Cs spectra must be dominated by a distribution of chemical shifts, giving rise to a Gaussian shape. The SP and CP experiments give very similar signals in shape, linewidth and position indicating no particular association with hydrogen, as the line intensity decreases in CP experiments with shorter contact times. The protons (for instance, as water molecules) thus are not located in the immediate neighborhood of the alkali atoms.  When the magnetic interaction between 27 Al and 1 H nuclei is analyzed by m the dipolar interaction, it is found that two different Al signals with similar intens  Figure 8 displays the 1 H MAS NMR spectra of the two pollucite specimens. The 1 H spectrum of the altered specimen from Tanco exhibits two signals at 2.35 and 4.6 ppm, from -OH groups and H 2 O water molecules, respectively. The pristine specimen from Mt. Mica has a 1 H spectrum with a major signal at 2.35 ppm and a minor resonance at 6.52 ppm, derived from -OH groups and probably H 3 O + , respectively. The H 2 O signal from specimen Tanco could be related to the phyllosilicate impurities, mainly in smectites, which are known in the alteration products of pollucite [18]. An alternative explanation could be related to liquid water from fluid inclusions derived from the hydrothermal alterations that are clear in specimen Tanco. However, note that this signal is absent in specimen Mt. Mica where fluid inclusions exist. If this water were from "zeolitic molecules", it should be recorded in the two specimens, but this signal is fully absent in the pristine pollucite specimen. The -OH signals at 2.35 ppm in the two specimens can be related to the reinforcement of the Si atoms in tetrahedral coordination with three oxygen atoms and one -OH group, perhaps as O 3 -Si(3Si,1Al)-OH and O 3 -Si(4Si,0Al)-OH sites, as well as -OH groups from the Q 3 impurities in the one from Tanco. If literature data are considered, these signals should correspond to structurally bound molecular water, i.e., located in the site of Cs atoms. However, this chemical shift cannot be interpreted in this way, as structurally isolated molecular water appears at higher chemical shifts, typically at 4.6 ppm [27]. Hence, no zeolite molecules of water were clearly identified in our specimens of pollucite. This interpretation is compatible with available 1 H NMR spectra in the available literature [28,29]. If the dipolar evolutions 27 Al{ 1 H} of the two pollucite specimens are compared with that of gibbsite Al(OH)3, i.e., a natural aluminum hydroxide with well-developed Al-OH bonds, a much slower rising under REDOR evolution appears, related to a weaker Al/H dipolar coupling and thus larger Al-H distances and/or less H surrounding the Al sites. The marked slower increase of the REDOR curve observed in specimen Tanco with respect to specimen Mt Mica follows the same reasoning and shows a lower amount of H and/or longer Al-H distances for the former. It can also be seen that the site with a chemical shift of 58.8 ppm shows a slower REDOR dephasing for both samples and hence is less protonated than the component at 60.2 ppm. Thus, as the two pollucite specimens are birefrin- 133 Cs atoms display a smaller quadrupolar interaction, the line shape of the 133 Cs spectra must be dominated by a distribution of chemical shifts, giving rise to a Gaussian shape. The SP and CP experiments give very similar signals in shape, linewidth and position indicating no particular association with hydrogen, as the line intensity decreases in CP experiments with shorter contact times. The protons (for instance, as water molecules) thus are not located in the immediate neighborhood of the alkali atoms.  Figure 8 displays the 1 H MAS NMR spectra of the two pollucite specimens. The 1 H spectrum of the altered specimen from Tanco exhibits two signals at 2.35 and 4.6 ppm, from -OH groups and H2O water molecules, respectively. The pristine specimen from Mt. Mica has a 1 H spectrum with a major signal at 2.35 ppm and a minor resonance at 6.52 ppm, derived from -OH groups and probably H3O + , respectively. The H2O signal from specimen Tanco could be related to the phyllosilicate impurities, mainly in smectites, which are known in the alteration products of pollucite [18]. An alternative explanation could be related to liquid water from fluid inclusions derived from the hydrothermal alterations that are clear in specimen Tanco. However, note that this signal is absent in specimen Mt. Mica where fluid inclusions exist. If this water were from "zeolitic molecules", it should be recorded in the two specimens, but this signal is fully absent in the pristine pollucite specimen. The -OH signals at 2.35 ppm in the two specimens can be related to the reinforcement of the Si atoms in tetrahedral coordination with three oxygen atoms and one -OH group, perhaps as O3-Si(3Si,1Al)-OH and O3-Si(4Si,0Al)-OH sites, as well as -OH groups from the Q 3 impurities in the one from Tanco. If literature data are considered, these signals should correspond to structurally bound molecular water, i.e., located in the site of Cs atoms. However, this chemical shift cannot be interpreted in this way, as structurally isolated molecular water appears at higher chemical shifts, typically at 4.6 ppm [27]. Hence, no zeolite molecules of water were clearly identified in our specimens of pollucite. This interpretation is compatible with available 1 H NMR spectra in the available literature [28,29]. Mica with one major resonance at 2.35 ppm with 0.9 ppm of linewidth from -OH groups, and a minor signal at 6.52 ppm, perhaps from H3O + .

Conclusions
Two birefringent (i.e., optically non-isotropic), non-stoichiometric (Si/Al ratio ~2.3) specimens of pollucite mineral species from the Tanco pegmatite and the Mt. Mica pegmatite, with powder XRD patterns fully consistent with that of a cubic Ia-3d space-group lattice symmetry, were studied by multinuclear MAS NMR spectroscopy at 9.4 and 17.6 T. The 29 Si SP-MAS spectra indicate a disordered Si/Al distribution incompatible with the random and Loewenstein distributions, because of a clear prevalence of the (3Si,1Al) and (2Si,2Al) environments. The 29 Si CP-MAS spectra show that protons are bonded preferen-

Conclusions
Two birefringent (i.e., optically non-isotropic), non-stoichiometric (Si/Al ratio~2.3) specimens of pollucite mineral species from the Tanco pegmatite and the Mt. Mica pegmatite, with powder XRD patterns fully consistent with that of a cubic Ia-3d space-group lattice symmetry, were studied by multinuclear MAS NMR spectroscopy at 9.4 and 17.6 T. The 29 Si SP-MAS spectra indicate a disordered Si/Al distribution incompatible with the random and Loewenstein distributions, because of a clear prevalence of the (3Si,1Al) and (2Si,2Al) environments. The 29 Si CP-MAS spectra show that protons are bonded preferentially with Si atoms, having low content of Al in the second sphere of coordination. Specimen Tanco has phyllosilicate impurities that were useful to monitor changes in the 29 Si spectra with a decreasing of the contact time, and thus with proton proximity to Si atoms. The 27 Al{ 1 H} dipolar evolutions indicate that two Al site distributions can be discriminated from their distance to protons, which are not compatible with the available models for an average structure with an isotropic cubic Ia-3d symmetry. The 23 Na and 133 Cs spectra indicate site distributions for these atoms, with no local bonding to water molecules. The 1 H spectra resolved the presence of hydrous H 2 O-rich impurities in specimen Tanco, and -OH groups in the pollucite structure of the two specimens, but not zeolitic water. The absence of zeolitic water molecules explains the lower leachability of Cs atoms in the pollucite structure, in comparison with other zeolite-based alternatives, and the high temperature necessary to observe any loss of weight during heating experiments. All of these data suggest that the structure of pollucite from NMR differs from the available lattice models derived from reciprocal-space techniques and from water speciation after infrared spectroscopy. Therefore, the two specimens of pollucite can be described from NMR data as an essentially anhydrous, pseudo-cubic and non-stoichiometric mineral, lacking a strictly lattice periodicity. Pollucite it is just a long-range disordered pseudo-periodic mineral structure, with a strong dispersion of tetrahedral changes for the framework cations and site distributions for the cavity cations.