Site‐Specific Iron Substitution in STA‐28, a Large Pore Aluminophosphate Zeotype Prepared by Using 1,10‐Phenanthrolines as Framework‐Bound Templates

Abstract An AlPO4 zeotype has been prepared using the aromatic diamine 1,10‐phenanthroline and some of its methylated analogues as templates. In each case the two template N atoms bind to a specific framework Al site to expand its coordination to the unusual octahedral AlO4N2 environment. Furthermore, using this framework‐bound template, Fe atoms can be included selectively at this site in the framework by direct synthesis, as confirmed by annular dark field scanning transmission electron microscopy and Rietveld refinement. Calcination removes the organic molecules to give large pore framework solids, with BET surface areas up to 540 m2 g−1 and two perpendicular sets of channels that intersect to give pore space connected by 12‐ring openings along all crystallographic directions.

X-ray powder patterns of as-prepared materials in the 2θ range 5−50° (step size 0.01°, timestep 160 s, 0.04 rad Soller, 45 kV, 35 mA) were recorded on a PANalytical Empyrean automated diffractometer equipped with a X'Celerator detector (Bragg−Brentano geometry, Cu Kα1 X-radiation, λ = 1.54056 Å, via a primary monochromator). For structural analysis and refinement of dehydrated calcined materials, samples were loaded in 0.7 mm diameter quartz glass capillaries and dehydrated at 170 °C under a vacuum of 10 −5 mbar for 10 h before being sealed. X-ray powder diffractograms were then collected in Debye-Scherrer mode over a 2θ range 5-70° (step size 0.1°, time step 80 s, 40 kV, 35 mA) on a Stoe STAD I/P diffractometer with a primary monochromator and PSD detector using Cu Kα1 X-ray radiation (λ = 1.54056 Å).
Single-crystal X-ray diffraction data for structure solution were recorded at ˗148 ºC using a Rigaku XtaLAB P200 diffractometer and Cu Kα1 (λ = 1.54056 Å) radiation. The data were processed using CrysAlisPro (Rigaku Oxford Diffraction) software [S2] and corrected for Lorentz and polarization effects. The structure was solved using direct methods, with refinement carried out using SHELXL Version 2017/1. [S3] All other calculations were performed using the CrystalStructure crystallographic software. [S4] Solid-state NMR spectra were recorded on a Bruker Avance III spectrometer equipped with a 9.4 T wide-bore superconducting magnet (Larmor frequencies of 400.1, 162.0, 104.3, and 100.6 MHz, respectively for 1 H, 31 P, 27 Al and 13 C). Samples were packed into 4 mm zirconia rotors and rotated at the magic angle at a rate of 10 to 14 kHz. The 13 C NMR spectrum was recorded with cross polarisation from 1 H. A contact pulse (ramped for 1 H) of 1.5 ms was used and high-power (υ1 ≈ 100 kHz) TPPM-15 decoupling of 1 H was applied during acquisition.
Signal averaging was carried out for 4096 transients with a recycle interval of 3 s. 27 Al MAS NMR spectra were recorded with a short pulse (ca. 15-30° inherent flip angle) to provide approximately quantitative spectra. Signal averaging was carried out for 128 transients with recycle intervals of 1 s. The 27 Al multiple-quantum (MQ) MAS spectrum was recorded using an amplitude-modulated z-filtered pulse sequence with signal averaging for 24 transients with a recycle interval of 1 s for each of 180 t1 increments of 35.71 s. The spectrum was sheared and referenced according to the work of Pike et al. [S5] 31 P NMR spectra were recorded with signal averaging for 16 transients with a recycle interval of 60 s (as-prepared material) or 32 transients with a recycle interval of 5 s (calcined, dehydrated material). Chemical shifts are reported in ppm relative to Si(CH3)4, 0.1 M Al(NO3)3, and 85% H3PO4, using L-alanine (δ(CH3) = 20.5 ppm), Al(acac)3 (δiso = 0.0 ppm, CQ = 3.08 MHz) and BPO4 (δ = -29.6 ppm) as secondary solid references.
Scanning electron microscopy was performed on a JEOL JSM-5600 SEM fitted with an Oxford INCA Energy 200 EDX analyser. Thermogravimetric analysis of as-prepared samples was carried out on a Stanton Redcroft STA-780 in a dry air flow with a heating ramp of 5 ºC min -1 up to 700 ºC. Elemental compositions were determined on a PANalytical Axios WDXRF (wavelength-dispersive X-ray fluorescence) spectrometer with a 4 kW Rh tube.
Elemental analysis was carried out by Elemental Analysis Service, London Metropolitan University, United Kingdom. Solid state UV-Vis absorption spectra were recorded on a JACSO V-650 UV-visible spectrophotometer with a photomultiplier tube detector. All spectra were obtained over the wavelength range 200-900 nm with a bandwidth of 5.0 nm and rate of 200 nm min -1 . Transmission 57 Fe Mössbauer spectra were collected at 300 and 4.2 K with a sinusoidal velocity spectrometer using a 57 Co(Rh) source. Velocity calibration was carried out using an α-Fe foil. The source and the absorbing samples were kept at the same temperature during the measurements. The Mössbauer spectra were fitted using the Mosswinn 4.0 program. [S6] The broad spectral contributions were fitted using a Blume-Tjon magnetic relaxation model. [S7] The EPR spectra were obtained using a Bruker EMX 10/12 spectrometer operating at ~9 GHz with 100 kHz modulation frequency. Selected samples were contained in 4 mm OD quartz tubes (Wilmad lab-glass). Measurements were performed in an ELEXSYS Super High Sensitivity Probehead (Bruker ER4122SHQE). The EPR spectra were recorded at 20 °C using 1 mW microwave power, a 6000 G field sweep centred at 3500 G with 1024 points resolution, a time constant and conversion time of 40.96 ms each, a modulation amplitude of 1 G and a microwave frequency of 9.839 GHz.
Calcination was performed in a tube furnace in flowing air at 600 °C (ramp 5 °C min -1 ) for 10 h. In an optimised process the air was replaced with flowing N2 during cooling and when the sample reached 50 °C hexane vapour was added to the N2 flow and allowed to adsorb for 20 mins. This stabilised the calcined samples to subsequent exposure to moist lab air. To establish their porosity, samples were heated under vacuum and adsorption isotherms for N2 at −196.15 °C were collected using a Micromeritics Tristar II 3020.
Scanning transmission electron microscopy (STEM) was performed in a XFEG FEI Titan transmission electron microscope operated at 300 kV. The column is fitted with a CEOS corrector, aligned prior to investigations using a gold standard assuring a potential spatial resolution of 0.8 Å. The electron dose was controlled via a monochromator, which allowed the displacement of the crossover position within the monochromator and therefore enabled increase or decrease of the number of electrons in continuous mode in STEM configuration.
For the current experiments, the beam current was usually ≈ 3.5 pA or lower, resulting in an electron dose of around 2000 e -/Å 2 . Prior to observation, samples were crushed using a mortar and pestle and suspended in ethanol. A few drops of the suspension were put on coated holey carbon microgrids and allowed to dry. The inner collection angle of the detector for imaging was 50 mrad.

STEM Modelling
Simulation of (Annular Dark-Field) ADF-STEM images was performed using QSTEM, a computer program written by Christoph Koch. [S8] It is freely available from his website at Humboldt University, Berlin. Parameters matching those for the XFEG FEI Titan were used.
A convergence angle of 17.5 mrad and collection angle 50-200 mrad were used for the Fecontaining sample. The pixel step size chosen was 0.36 Å and the slice thickness 1.36 Å. A source size of 1.7 Å gave a similar atom image size to that in the experiment. The images were oversampled 5-fold and Poisson noise added corresponding to an electron dose 4 times the experimental one.
All experimental images have been ABSF (Average Background Subtraction Filter) filtered [S9] to reduce the noise in these low dose images. Low dose is required to avoid beam damage. These filtered images contain information local to the unit cell. To get an averaged image a mask of small discs surrounding each diffraction spot in the FFT of the ABSF image has been applied and the inverse FFT calculated. Generally the processing is applied to large images (up to 4k by 4k) and small regions extracted for presentation.  (Table S2). Lorentz and polarization effects. The structure was solved by direct methods [S10] and expanded using Fourier techniques. The non-hydrogen atoms were refined anisotropically.

Single-crystal X-ray Diffraction
Hydrogen atoms were refined using the riding model. The final cycle of full-matrix leastsquares refinement [S3] on F [S11] was based on 5184 observed reflections and 397 variable parameters and converged (largest parameter shift was 0.00 times its esd) with unweighted and weighted agreement factors of 0.0770 and 0.2339. The goodness of fit was 1.04 and unit weights were used. The maximum and minimum peaks on the final difference Fourier map corresponded to 0.83 and -0.62 e -/Å 3 , respectively. Neutral atom scattering factors were taken from International Tables for Crystallography (IT), Vol. C, Table 6.1.1.4. S11 Anomalous dispersion effects were included in Fcalc [S12] ; the values for f' and f" were those of Creagh and McAuley. [S13] . The values for the mass attenuation coefficients are those of Creagh and Hubbell. [S14] Powder X-ray Diffraction Rietveld refinement of the structures of as-prepared STA-28 materials was carried out using the GSAS suite of programs and the EXPGUI graphical interface. [S15] The crystal structure of STA-28, synthesised using 1,10-phenanthroline and solved by SCXRD was used as the starting model for the refinements. The instrumental background was fitted automatically by using a Chebyschev function. The peak profiles were modelled using a Pseudo-Voigt function (type 2). [S15,S16] The An overall isotropic atomic displacement parameter was set to 0.002 Å 2 for the framework atoms and 0.001 Å 2 for all C and N atoms. The fractional occupancies were set to 1.00 for the phenanthroline SDA atoms. During the cycles of refinement the x, y, z coordinates of the SDAs atoms were allowed to refine together. Convergence was achieved by refining simultaneously all profile parameters, scale factor, lattice constants, 2θ zero-point, and atomic positional and atomic displacement parameters for the framework atoms and the occupancies of the SDAs atoms. For Rietveld refinement of PXRD data from as-made STA-28 using 4methyl and 5-methyl-1,10-phenanthrolines, the same starting modelling obtained from SCXRD structure solution was used. Methyl groups were added to the phenanthroline molecules following the method depicted in Figure S1. Refinements were carried out following the procedure described above. For 4,7-dimethyl and 5,6-dimethyl 1,10-phenanthrolines, a similar method was used except the background was fitted using a Cosine-Fourier function and the peak profile modelled using a Pseudo-Voigt function (type 3). [S15] In the case of FeAlPO STA-28, the refinement process is the same as described earlier, with the added step of determining the Fe occupancy and location within the framework. The fractional occupancies of the Al atoms were refined one by one. The occupancies of the four tetrahedral Al sites stayed the same while the value for the octahedral Al site increased. This indicated that the Fe in the framework is located at the octahedral site.
Thus, Fe was added into GSAS with the same fractional coordinates as the octahedral Al atom. The fractional occupancies of both this Al and the Fe atom were set to sum to one and subsequently refined. This gave the percentage of Fe present in the octahedral site within the STA-28 framework.
Rietveld refinement of dehydrated calcined AlPO4 and FeAlPO STA-28 was performed in a similar way, with similar restraints on framework bond distances to begin with, which were gradually removed, and using as a starting model the energy-minimised fully tetrahedrallyconnected structures in space group Fddd obtained using the program GULP [S17] (see below).

Computational Modelling
Energy minimisation of AlPO4 and FeAl4P5O20 STA-28 was performed using the program GULP. S17 The simulation was carried out at constant pressure using the potentials derived by Sanders et al. [S18] Starting from the experimentally derived single crystal structure, the 1,10phen OSDA was removed and energy minimisation was performed with no symmetry constraints applied. The best space group for the optimised structure was determined using the 'Find Symmetry' tool within the program Materials Studio [S19] (and found to be Fddd).
The energy of AlPO4 STA-28 was compared with those of other AlPO4 frameworks.
The structure of FeAlPO STA-28 was also optimised via Density Functional Theory using the program CASTEP [S20] with the Generalized Gradient Approximation (GGA) and the Perdew-Burke-Emzerhof (PBE) exchange-correlation function. The structure obtained from Rietveld refinement for the calcined dehydrated FeAlPO STA-28 structure was used as the starting geometry. No symmetry constraints were applied during the calculation. The nearest space group for the optimised structure was Fddd.

S2. Synthesis of AlPO STA-28
STA-28 (AlPO) 1,10-Phenanthroline In an optimised preparation, the STA-28 gel is prepared by mixing 1.16 g (11.86 mmol) of H3PO4 (85%, Alfa Aesar), 0.038 g (0.62 mmol) of fumed silica (0.007 μm powder, Aldrich) and 0.97 g (13 mmol) of Al(OH)3 (Alfa Aesar) in 9 mL (500 mmol) of deionised water. 0.23 g (1.25 mmol) of 1,10-phenanthroline (Aldrich) was then added. The addition of 1.98 g (7.63 mmol) of TBAOH aqueous solution (55 wt% in H2O, Sachem) is used to adjust the gel pH to 7. The gel is stirred for two hours at room temperature and loaded into a Teflon-lined 30 mL stainless-steel autoclave. Crystallisation is carried out at 180 °C for 4 -7 days and the resultant products are suspended in water and sonicated to force separation of crystalline from amorphous solid, with the amorphous material removed by decanting. Then, the crystalline materials are dried under 80 °C for 12 hours. This gives approximately 1 g of an off-white solid. Only traces of Si (< 0.3 wt%) were found in the product. STA-28 was prepared with a range of different 1,10-phenanthrolines, as described above.

S5. AlPO-STA-28(1,10-phen): Chemical and Thermal Analysis; Solid state NMR
Chemical Analysis XRF gave Al/P = 1.03, with negligible Si content. Deviation from 1.00 assumed to be due to the presence of minor mounts of amorphous impurities (seen for example in the NMR). Calculation of the water content from the TGA weight loss ( Figure   Solid State NMR Figure S8 shows the solid-state 13 C, 27 Al and 31 P NMR spectra of STA-28(1,10-phen). The 13 C NMR spectrum ( Figure S8(a)) is consistent with 1,10-phen bound to Al and is assigned as shown. The 27 Al MAS NMR spectrum ( Figure S8(b)) shows a complex signal for tetrahedral Al, a sharp signal for octahedral Al, and a broad signal corresponding to pentacoordinate Al in an amorphous impurity. The 27 Al MQMAS spectrum shown in Figure   S9 allows resolution of four distinct tetrahedral Al sites with isotropic chemical shifts of 47, 45, 41 and 40 ppm, and quadrupolar products (PQ = CQ (1 + hQ 2 /3) 1/2 ) of 1.9, 1.4, 1.6 and 2.3 MHz, respectively. The octahedral site has diso = -4.9 ppm and PQ = 2.0 MHz.
The 31 P MAS NMR spectrum of as-prepared STA-28(1,10-phen), shown in Figure S8(c) contains two signals from the AlPO framework at -26.6 and -30.9 ppm in an approximately 2 : 3 integrated intensity ratio, consistent with the presence of five P sites. While assignment using DFT calculations was not possible (as the 20 H2O per unit cell were not located in the crystal structure), the recently developed DISCO program [S21] can be used to provide an approximate prediction of 31 P iso, based on the mean P-O-Al bond angles and P-O bond lengths. The predicted chemical shifts for STA-28 are shown as red points in Figure S8(c), and it can be seen that three P signals are predicted to be at higher shift with two at lower shift, which is the opposite of the experimental result. Therefore, assignment using DISCO alone is ambiguous in this case.    Figure S13. The refined positions of the methylated phenanthroline molecules within the 12R channels of STA-28, viewed along z.

S8. Synthesis and analysis of FeAlPO STA-28
In attempts to replace Al in the STA-28 AlPO4 structure, Al (  XRF analysis on the sample prepared with (Fe/P)gel = 0.2 gives an empirical framework formula of Fe0.78Al4.22P5O20. FeAlPO STA-28 was also obtained if ferrous Fe(OAc)2 was replaced with FeCl3 in the gel ( Figure S15).

S9. Spectroscopy of FeAlPO STA-28 (Fe/P 0.20 in gel, 0.155 in product)
UV-visible Figure S16. Solid-state UV-vis absorption spectra for AlPO (left) and FeAlPO (right) STA-28. Figure S17. EPR spectrum for FeAlPO STA-28 Figure S18. Mössbauer spectra (black lines) obtained at 300 and 4.2 K for FeAlPO STA-28. Fitted lines are given in colour. Table S12. Mössbauer parameters of FeAlPO STA-28 derived from spectral fitting in Figure 23. Solid State NMR Figure S19a shows the 27 Al NMR spectra of FeAPO STA-28 prepared with Fe/P gel ratios of 0.05, 0.10, 0.16 and 0.20 and Figure S24b shows a plot of the ratio of tetrahedral/octahedral Al (Al [IV] /Al [VI] ) against Fe/P. As the Fe content of the material increases, the 27 Al resonances broaden owing to both increased disorder and paramagnetic-induced nuclear relaxation enhancement. Adding more Fe appears to reduce the amount of tetrahedral Al preferentially.

Mössbauer
Given the results of the other characterisation techniques presented in the main text, which show that Fe substitutes almost exclusively on the octahedral Al5 site, this appears counterintuitive. A likely explanation is that the paramagnetic Fe 3+ is generates a "sphere of invisibility", whereby paramagnetic relaxation effects result in signals from 27 Al near to the Fe being broadened beyond detection with the current NMR experiments (any through-bond hyperfine effects resulting in paramagnetic shifts are expected to be small as the shortest bonded pathway from Al to Fe is Al-O-P-O-Fe). The relaxation effect would be more pronounced on the Al [IV] sites, since there are many more of these in close proximity to any given octahedral site -there are 12 Al [IV] sites within 5.2 -6.2 Å of a Fe [VI] while the nearest Al [VI] would be further than 9 Å away. Substituting any one octahedral site with Fe completely removes the signal from one Al [VI] but would also significantly reduce (or remove) the signals from many nearby Al [IV] sites. This suggestion is supported by the fact that the resonance lines can be seen to broaden significantly as the Fe content increases, and also the total signal intensity per transient drops significantly with increasing Fe content (for the 0.05 Fe/P sample, the spectrum shown was recorded with 128 transients and moderate receiver gain, whereas for the 0.2 Fe/P sample 4096 transients were required with maximum receiver gain). The observation of some octahedral Al even with Fe/P = 0.2 is in agreement with the Rietveldrefined Fe occupancy of 0.73.   Figure S21. Restrained Rietveld refinement of as-made, hydrated FeAlPO STA-28 (Fe/P=0.155) in space group I 2/a using the AlPO structure determined by single crystal as a starting model.   Table S16.

T-O, C-N/C-C bond distances (Å) and O-T-O/T-O-T angles (°) obtained from
Rietveld refinement against PXRD data for as-made FeAlPO STA-28. In the [100] projection the Fe sites are located 0.8 Å from P sites. These two projected sites are not resolved in the images presented where the achieved information transfer/spatial resolution is 2.0 Å. Hence the additional intensity due to Fe relative to Al sites appears shifted towards the nearby P sites.

T-O, C-N/C-C bond distances (Å) and O-T-O/T-O-T angles
For the FeAlPO sample, Figure S22 shows an example of the raw data in S22(a)   FeAlPO STA-28 is a similar image to that in the main paper, but expanded.

S12. Characterisation of calcined STA-28
The first attempt at calcination was in flowing air in a tube furnace at 600 °C. The sample was exposed to lab air before a PXRD was measured. This gave rise to a broadened pattern (shown in Figure S25, compared with a simulated pattern using an energy-minimised structure). 27 Al MAS NMR (Fig. S26) gave broad resonances that suggested some structural decomposition, and the chemical shift indicated the Al was mainly tetrahedrally-coordinated. Figure S25. Powder X-ray diffraction patterns for simulated (lower) and experimental (upper) calcined AlPO STA-28 (1,10-phen). N2 adsorption indicated microporosity, but the isotherm was not purely Type I ( Figure S27). Figure S27. N2 adsorption isotherms (at ˗195 °C) on calcined STA-28 samples. Triangles, AlPO STA-28 calcined in air and exposed to moist air before activation for adsorption; squares, AlPO calcined in air then stabilised by hexane adsorption on cooling in N2 before being activated, circles, FeAlPO calcined in air at 575 °C and stabilised by hexane adsorption upon cooling in flowing N2 before being activated. Closed symbols, adsorption branch, open symbols, desorption branch. Pore volumes were estimated by uptake at p/po=0.2.

S13. Energy minimised structures of AlPO and FeAlPO
AlPO STA-28 The structure of AlPO STA-28 was energy minimised using the GULP program, [S17] taking the as-prepared I2/a structure as a starting point and removing the 1,10-phenanthroline, which leaves all Al atoms four coordinated. A 2 × 2 × 2 block of the I2/a structure was allowed to relax at constant pressure to an energy minimum position without symmetry (space group P1) and afterwards the 'Find Symmetry' routine used to identify possible symmetry elements from the atomic coordinates. Two unit cells and symmetries were identified, which gave the same energy (to within 6.9 x 10 -9 eV per TO4/2 group).  (3) in the Fddd cell.
The lattice energy of the fully tetrahedral STA-28 was calculated and compared with the lattice energies calculated in the same way for some known AlPO4 polymorphs (GULP, constant pressure [S17] ). The values obtained are consistent with those of Simperler et al. [S24] These values are shown graphically in Figure S28 along with the framework densities of the energy-minimised structures, and in Table S17. Figure 28 (right) plots the lattice energy per TO2 unit, relative to berlinite (the quartz polymorph of AlPO4), in the fashion of Henson and Gale. S25 The energy per TO2 unit was found to be ca. 4 kJ mol -1 less stable than frameworks with similar framework densities such as AST and CHA.

FeAlPO STA-28
The optimised (template-free) structure of FeAl4P5O20 STA-28, in which Fe replaces Al in every Al(5) site of the original I2/a structure was observed to adopt the Fddd space group, with the structural parameters given below.  Figure S29. Rietveld plot of calcined AlPO4 STA-28, after calcination to remove template, stabilisation by addition of hexane vapour when cooled in N2 to 50 °C, followed by removal of hexane under vacuum at 50 °C. Figure S30. Rietveld plot of calcined FeAlPO4 STA-28 (Fe/P = 0.155), after calcination to remove template, stabilisation by addition of hexane vapour when cooled in N2 to 50 °C, followed by removal of hexane under vacuum at 170 °C for 8 h.   Table S21.