Site-directed cation ordering in chabazite-type AlxGa1−xPO4-34 frameworks revealed by NMR crystallography

We report the first synthesis of the mixed-metal chabazite-type AlxGa1−xPO4-34(mim) solid solution, containing 1-methylimidazolium, mim, as structure directing agent (SDA), from the parent mixed-metal oxide solid solution, γ-(AlxGa1−x)2O3. This hitherto unreported family of materials exhibits complex disorder, arising from the possible distributions of cations over available sites, the orientation of the SDA and the presence of variable amounts of water, which provides a prototype for understanding structural subtleties in nanoporous materials. In the as-made forms of the phosphate frameworks, there are three crystallographically distinct metal sites: two tetrahedral MO4 and one octahedral MO4F2 (M = Al, Ga). A combination of solid-state NMR spectroscopy and periodic DFT calculations reveals that the octahedral site is preferentially occupied by Al and the tetrahedral sites by Ga, leading to a non-random distribution of cations within the framework. Upon calcination to the AlxGa1−xPO4-34 framework, all metal sites are tetrahedral and crystallographically equivalent in the average R3̄ symmetry. The cation distribution was explored by 31P solid-state NMR spectroscopy, and it is shown that the non-random distribution demonstrated to exist in the as-made materials would be expected to give remarkably similar patterns of peak intensities to a random distribution owing to the change in average symmetry in the calcined materials.


S1. Further Details of Solid-State NMR Experiments
Table S1 presents further information on the acquisition parameters for the solid-state nuclear magnetic resonance (NMR) experiments carried out in this work.TPPM-15 decoupling of 1 H (ν1 ≈ 100 kHz) was applied during acquisition.
d.A pulse with a short flip angle (β ≈ 9°) was used to ensure quantitative results.
f. Recorded using an amplitude-modulated z-filtered pulse sequence with 128 t1 increments of 71.43 μs.Spectra were sheared and referenced in the indirect dimension according to the convention in Ref. S2.   g.A pulse with a short flip angle (β ≈ 23°) was used to ensure quantitative results.

S2. Additional Characterisation of γ-(AlxGa1-x)2O3 Precursors
The 27 Al MAS NMR spectra of the γ-(AlxGa1-x)2O3 precursors used in this work are shown in Figure S1a.Spectral integration provides the amounts of Al
Figure S3 shows the two-dimensional 27 Al 3QMAS NMR spectrum of AlPO-34(mim) after shearing, along with the sum projection on δ1 and slices extracted parallel to δ2 for each of the three signals.Figure S5 shows the thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) traces for AlPO-34(mim) and the AlGaPO-34(mim) samples.We previously reported the equivalent data for GaPO-34(mim) S1,S4 and data recorded for this work, shown in Figure S5e is consistent with the earlier work.There is a slight difference in the overall mass loss for the two GaPO-34(mim) samples, which is ascribed to different amounts of absorbed water (0.65 per formula unit in Ref. S1 and 0.15 per formula unit here).It is interesting to note that in GaPO-34(mim), the calcination has previously been shown to occur in two stages, S4 with loss of HF occurring at slightly lower temperature than loss of the 1-methylimidazole.Such a two-stage process could be proposed to occur for the most Ga-rich AlGaPO (see Figure S5d), but it was beyond the scope of the present work to investigate the calcination mechanism for AlGaPOs in further detail.
Table S2 shows the final experimental formulas for the five samples studied, determined using a combination of ICP-MS and TGA data.

S5. Additional NMR Spectroscopic Characterisation of AlGaPO-34(mim)
Figure S6a shows the δ1 projections of the Al IV signals from the 3QMAS NMR spectra of the Alcontaining AlGaPO samples.While the signal-to-noise ratio is clearly poorer for the sample containing least Al, spectral integration is still possible, assuming that the lineshapes in the isotropic dimension can be described by Gaussian-Lorentzian lines.Figure S6b shows a plot of the ratio of the integrated intensities of the Al2 and Al3 signals as a function of composition, which suggests that neither site is particularly favoured for Ga substitution.Note that the MQMAS experiment is not quantitative, with excitation efficiency varying with CQ, such that even in the AlPO-34(mim) end member the ratio of signal from Al3/Al2 is not exactly 1.
where 0 ≤ n ≤ 4.This model assumes a random distribution of Al and Ga on each of the four nextnearest neighbour sites, and it can be seen from Figure S8a that there is poor agreement with any of the experimental intensities.This is to be expected, however, since the 27 Al, 71 Ga and 19 F NMR spectra showed that the Al and Ga exhibit a preference for the octahedral and tetrahedral sites, For the distribution of Al on the two octahedral sites: For the distribution of Al on the two tetrahedral sites: Therefore, for P1 overall, which, in the limit of xO = xT, is equivalent to Equation S1.
The values of xO and xT can be determined experimentally from the 19 F NMR spectra (for xO) and a combination of xO and the composition of the material from elemental analysis (for xT).

S6. Additional Discussion of the DFT Calculations
DFT calculations using CASTEP generate the diagonalised absolute magnetic shielding tensor in the principal axes system, σ, from which the isotropic magnetic shielding is given by .

Equation S5
To compare calculated shielding with experimental chemical shifts, a reference value, σref, is used, with .

Equation S6
A plot of calculated σiso against experimental δiso should yield a straight line with gradient of -1 and an intercept of σref. Figure S9 shows such plots for 27 Al and 31 P, using experimental values from calcined AlPO-14, calcined AlPO-34, calcined GaPO-34 and GaPO4 berlinite (noting that the latter two contain P but not Al).Note that, since the gradients of both lines differ significantly from -1, Equation S6 was modified to give , Equation S7where m is the gradient of the line.It is also worth noting that the points for 27 Al in AlPO-34(mim), particularly the octahedral Al1, lie on a line significantly different from the reference set.We interpret this discrepancy as possible evidence of either SDA dynamics in the as-made AlPO-34(mim), or that the sample used in the NMR experiments had absorbed some atmospheric water, which was not present in the crystal structure determination (and, hence, the DFT calculation).
Table S3 shows the values of σref and m used here to calculate the 27 Al and 31 P δiso.
For 19 F, calculated shifts were referenced using the experimental shifts of the two end members (-98.0 and -125.0 ppm for the GaPO-34(mim) and AlPO-34(mim), respectively) and the calculated shifts for the two anhydrous materials with the SDAs in the orientations shown in Figure S10.We note that the value of m obtained using this method deviates significantly from -1, as has previously been observed for 19 F in the literature.The SOD program S6 was used to generate two series of AlGaPO-34(mim) models with the SDA orientation either matching that of AlPO-34(mim) or GaPO-34(mim).S7 In both cases, an anhydrous structure was considered (i.e., the experimental structure for the AlPO and with the water molecule deleted for the GaPO).Figure S10 shows the parent structures for each series.The unit cells each contain six M sites (M = Al or Ga), leading to a total of 36 symmetry-distinct arrangements for substituting up to six Al into the GaPO framework (or vice versa) for each series.The main text describes results for series where Al was substituted into the dehydrated GaPO structure (Figure S10b) and Figures S11 and S12 compare the calculated mixing energies (as in Figure 6 of the main text) and calculated 31 P δiso (Figure 7a of the main text) for the two different series.From Figure S13b it could be concluded that the experimental structure of the GaPO has the SDA in the incorrect orientation, as the reverse N models are lower in energy than those with the SDA in the experimentally determined orientation.However, the experimentally determined structure also contains a partially occupied molecule of water and, when this is occupied, as in Figure S14a, the reverse N models are less stable than those with the experimental orientation of the SDA, as shown in Figure S14b.As shown in Figure 8 of the main text, in the orientation indicated in the experimental structure solution the SDA is able to form an N-H⋯O hydrogen bond with the water, which is not possible in the reverse N structures.Therefore, it can be concluded that the orientation of any given SDA cation is likely to be heavily influenced by the presence of water in the same pore.However, notably, the most favoured metal sites for Al and Ga substitution do not change appreciably when the pore contents or SDA orientation are changed.indicating that when Al is on the octahedral site, the M-F interaction is stronger and slightly more covalent than when Ga is on the octahedral site.
Figure S16 shows the calculated Mulliken charges for framework O atoms in the 36 anhydrous structural models of as-made AlGaPO-34 with the SDA in the orientation of the AlPO end member.
There is clear separation between Al-O-P and Ga-O-P linkages, with Al-O-P oxygens being more negative than Ga-O-P oxygens (as would be expected for the slightly more ionic AlPO4 framework).Additionally, for the Al-O-P linkages there is clear separation between Al IV -O-P and Al VI -O-P linkages, with the latter having a lower charge on the O atoms.Such separation is not observed for Ga IV -O-P vs Ga VI -O-P linkages, presumably as a consequence of the more covalent nature of Ga-O-P linkages in general.Table S4 summarises the numerical values of the Mulliken charges presented graphically in Figure S16.The x axis is quantitatively meaningless but qualitatively corresponds to structures with a higher Al content to the left and a higher Ga content to the right.Bearing in mind that, in the calcined material, there is only one T site in the CHA framework, the observed 31 P NMR spectrum is a superposition of signals from P sites that were formerly P1, P2 and P3 in the as-made material.As such, for the same cation distribution as observed in the as-made material, the 31 P spectral intensities in the calcined material would be given by: Again, in the limit of xO = xT, Equation S11 can be seen to be equivalent to Equation S1. Figure 9b of the main text shows the integrated intensities expected using Equation S1, whereas Figure 9c shows the intensities expected from Equation S11.
Figure S1a.Spectral integration provides the amounts of Al IV and Al VI , and these values are consistent with values reported by Cook et al., S3 who used Al( i PrO)3 as the Al source rather than the Al(acac)3 used here.The values reported by Cook et al. (originally presented in Figure 8a of Ref. S3) are shown in Figure S1b (red and blue lines) with the values obtained in the present work shown as green crosses.

Figure S1 .
Figure S1.(a) 27 Al (9.4 T, 14 kHz MAS) NMR spectra of γ-(AlxGa1-x)2O3 prepared using Al(acac)3 as the Al source.(b) Plot of the proportion of Al IV and Al VI in the mixed-metal oxides (green crosses) compared with values reported by Cook et al.S3 for the analogous oxides prepared with Al( i PrO)3 as the Al source (shown in red for Al IV and blue for Al VI ).

Figure S7 shows
FigureS7shows expansions of the aromatic region of the 13 C CP MAS NMR spectra for the samples studied.While small changes are observed in peak positions, as shown in FigureS7b, these are not particularly systematic with composition and likely reflect a combination of SDA orientation effects and the varying degrees of hydration of the samples.

Figure
Figure S8a shows the integrated intensities of the three 31 P signals assigned to P1 in the AlxGa1-xPO4-34(mim) samples, as a function of x.Using the binomial theorem, the relative intensities of the signals for a P with n P-O-Al linkages, p(nAl) can be expressed as: model, assuming a random distribution of Al and Ga in the octahedral sites, with the composition of this site given as xO, and a random distribution of Al and Ga in the tetrahedral sites with composition xT, and bearing in mind that P1 has two P-O-M IV and two P-O-M VI (M = Al or Ga) linkages, leads to the following expressions.

FigureFigure S8 .
FigureS8bshows that Equation S4 yields predicted values for P1(3Al) and P1(4Al) in good agreement with the observed integrated intensities of the two signals at lower shift (compare the green and blue points and lines in FigureS8b).This observation suggests the assignment of the three observed signals as P1(0-2Al), P1(3Al) and P1(4Al), as discussed in the main text.FiguresS8c and S8dshow predicted integrated intensities for P1(0-2Al), P1(3Al) and P1(4Al) using Equation S1 and Equation S4, respectively.

Figure S9 .
Figure S9.Plots of calculated σiso against experimental δiso used to reference calculated δiso for (a) 27 Al and (b) 31 P. Black points are for the reference data and red for AlPO-34(mim).Black dashed lines indicate the line of best fit for the reference data and in (a) the line of best fit for the AlPO-34(mim) data is shown in red.

Figure S10 .
Figure S10.Parent framework structures for the series of models for AlGaPO-34(mim) with the SDA in the orientation of (a) the AlPO-34(mim) and (b) the GaPO-34(mim) end members and the (c) AlPO-34(mim) and (d) GaPO-34(mim) "reverse N" models (corresponding to a pseudo-C2 rotation of the SDA about the H3C-N bond).Structures are viewed down the crystallographic c axis, with C = black, N = blue, F = green, M = purple framework, P = grey framework, O and H atoms are hidden.Note the purple framework cation sites may be occupied by either Al or Ga, depending on the exact structural model considered.

Figure S11 .
Figure S11.Plots of Emix against x for the 36 distinct arrangements of Al and Ga in the AlGaPO-34(mim) structural models with the SDA in the orientation matching (a) the AlPO end member and (b) the anhydrous GaPO end member.The dashed grey line indicates Emix = 0 and the solid grey line is the convex hull.Note that unlike Figure 6 of the main text, in this figure Emix = 0 has been calculated separately for each series.

Figure S12 :Figure S13 .
Figure S12: Calculated 31 P δiso for P sites with 0-4 Al NNN in the 36 distinct structural models for as-made AlGaPO-34 with the SDA in the orientation matching (a) the AlPO end member and (b) the anhydrous GaPO end member.Points corresponding to structures on or near the convex hulls (see Figure S11) are shown in colour, whereas the much less energetically favourable structures are shown in grey.

Figure S14 .
Figure S14.(a) Parent structure for the series of models for AlGaPO-34(mim) based on the GaPO-34(mim) structure with water present.The structure is viewed down the crystallographic c axis, with C = black, N = blue, F = green, M = purple wireframe, P = grey wireframe, O and H atoms are hidden, apart from the O atom of water, which is shown in red.(b) Plots of Emix against x for 14 structures based on the experimental hydrated GaPO-34(mim) structure (closed circles) and the corresponding 14 "reverse N" models.The dashed grey line indicates Emix = 0 for the crystallographically determined C and N positions.

Figure
FigureS15ashows the 19 F MAS NMR spectra of the AlGaPOs (reproduced from Figure3aof the main text), overlaid with the calculated 19 F δiso for the 72 anhydrous structural models of the asmade mixed-metal materials generated by SOD as discussed above.The assignment of the signals as Al-F-Al, Al-F-Ga and Ga-F-Ga is unambiguous.FiguresS15b and S15cshow plots of the calculated δiso against the mean F-M bond length (M = Al, Ga) and M-F-M bond angle, respectively.It can be seen that, as proposed in the main text, the Al-F bond lengths are shorter and the Al-F-Al bond angles are slightly larger than the Ga-F and Ga-F-Ga bond lengths and angles,

Figure S15 .Figure S16 .
Figure S15.(a) 19 F (14.1 T, 25-40 kHz MAS) NMR spectra of the AlGaPO-34(mim) series, from Figure 3 of the main text, with the calculated ranges of 19 F δiso (from the two series of SODgenerated structures) indicated.(b and c) Plots of calculated 19 F δiso against (b) mean M-F bond length and (c) the M-F-M bond angle for the 72 optimised structural models of anhydrous AlGaPO-34(mim) generated by SOD.
1 H-13 C cross-polarisation (CP) MAS experiments used a spin lock pulse (ramped for 1 H) of 1 ms.

Table S3 .
Values of reference shielding and gradient for 19 F, 27 Al and 31 P used in this work.

Table S4 .
Calculated Mulliken charges for O atoms in Al-O-P and Ga-O-P linkages.