Pressure-Induced Amorphization of Small Pore Zeolites—the Role of Cation-H2O Topology and Anti-glass Formation

Systematic studies of pressure-induced amorphization of natrolites (PIA) containing monovalent extra-framework cations (EFC) Li+, Na+, K+, Rb+, Cs+ allow us to assess the role of two different EFC-H2O configurations within the pores of a zeolite: one arrangement has H2O molecules (NATI) and the other the EFC (NATII) in closer proximity to the aluminosilicate framework. We show that NATI materials have a lower onset pressure of PIA than the NATII materials containing Rb and Cs as EFC. The onset pressure of amorphization (PA) of NATII materials increases linearly with the size of the EFC, whereas their initial bulk moduli (P1 phase) decrease linearly. Only Cs- and Rb-NAT reveal a phase separation into a dense form (P2 phase) under pressure. High-Angle Annular Dark Field Scanning Transmission Electron Microscopy (HAADF-STEM) imaging shows that after recovery from pressures near 25 and 20 GPa long-range ordered Rb-Rb and Cs-Cs correlations continue to be present over length scales up to 100 nm while short-range ordering of the aluminosilicate framework is significantly reduced—this opens a new way to form anti-glass structures.

Pressure-induced amorphization (PIA), the loss of long-range order under pressure, first discovered in ice 1 , is common among open framework structures such as zeolites and metal-organic frameworks (MOFs) and often attributed to softening of low-energy vibrations of the framework and local distortions [2][3][4] . The potential use of PIA to synthesize 'ideal glasses' has been put forward 1,2 . Assessing if and to what a degree materials truly lack long-range order and are amorphous is dependent on the experimental conditions: in the pressure-induced amorphization of anorthite (CaAl 2 Si 2 O 8 ) X-ray Bragg reflections broaden and loose intensities at pressures lower than observed by Williams and Jeanloz using optical birefringence measurements 5 . We use X-ray powder diffraction and High Angle Annular Dark Field Scanning Transmission Electron Microscopy (HAADF STEM) imaging in real space to characterize the extent of long range order that continues to be present after the samples have been pressurized in the presence of silicone oil, a non-pore penetrating pressure-transmission fluid.
In large pore zeolites such as silicalite it has been shown that the incorporation of certain guest molecules such as Ar and CO 2 delays PIA from pressures near 8 GPa up to 25 GPa and at the same time significantly increases the bulk modulus 6 . In AlPO 4 -54-xH 2 O water coordination to the Al 3+ framework has been suggested to initiate PIA 7 . We present systematic studies of PIA in small pore zeolites with natrolite (NAT) frameworks. Our studies investigate the impact different EFC-H 2 O arrangements have on the onset pressure of PIA.
Na-NAT (Na 16 Al 16 Si 24 O 80 •16H 2 O) is a natural zeolite with small elliptic pores having a conjugate diameter of less than 4 Å. However, under pressure in water-containing fluids pressure-induced hydration occurs due to an expansion and pore-opening in this auxetic material 8 . This behavior was established in Li-, K-, Rb-NAT while Cs-NAT reveals a distinct pressure-induced phase transition without pressure-induced hydration 9 . Computational studies by Kremleva et al. using density functional theory were able to optimize the structures of Li + , Na + , K + , Rb + , Cs + containing natrolites at ambient conditions and point to the importance of the EFC-H 2 O interactions and the strain energy of the aluminosilicate framework under pressure 10 .

Results
Structural changes under pressure. When subjected to pressure in the presence of the nonpore penetrating pressure transmitting fluid silicone oil one observes gradual shifts of all diffraction peaks to higher 2-theta indicating normal compression of the unit cell lengths and volume ( Fig. 1a and Supplementary Table S1). As silicone oil only transmits hydrostatic pressure up to near 5 GPa 11 we only used data up to this pressure to determine bulk moduli. These values indicate that the bulk moduli of the NAT I members Li-NAT and Na-NAT and the NAT II members K-NAT and Rb-NAT (P2) and Cs-NAT(P2) are within error the same whereas the P(1) phases of K-NAT, Rb-NAT, and Cs-NAT are significantly different. (Figures 1b and 3a and Supplementary Table S2).
In Rb-NAT and Cs-NAT, as the frameworks are compressed in silicone oil two coexisting phases with smaller unit cell volumes are observed at 2.8 and 0.2 GPa (Figures 1 and 2), respectively. These phases were modeled to be isostructural to the ambient phases but with a more efficient EFC-H 2 O packing in the channels ( Fig. 2 and Supplementary Table S3). Laboratory XRD data up to 18° 2-theta allowed us to estimate that the smaller volume P2 phases have chain rotation angles about 10° larger than the P1 phases which have approximately 100 Å 3 larger unit cells (see Supplementary Tables S1 and S3). The bulk moduli of the dense phases (P2) with 51(7) GPa and 50(3) GPa for Rb-NAT (P2) and Cs-NAT (P2), respectively, are comparable to those of NAT I phases while the P1 phases with larger unit cell volumes scale differently with the cation radius (see Supplementary Table S2). Further increase in pressure leads to PIA as observed by the broadening and weakening of the diffraction peaks ( Fig. 1a and Supplementary Fig. S1).

Distribution of EFCs in type-I and -II natrolites. Kremleva established that the EFC-H 2 O interac-
tion energies decrease with EFC size and can be grouped into type-I natrolites (NAT I ), Li-and Na-NAT, where the H 2 O molecules interact strongly with the EFC and type-II natrolites (NAT II ), K-, Rb-and Cs-NAT, where EFC-H 2 O interactions are weaker 12 . The structures of these two types have a distinct EFC-H 2 O topology: in NAT I , the H 2 O molecules are in close proximity to the aluminosilicate framework and Li + and Na + are found near the center of the pores, aligned along the c-axis alternately binding water molecules. In NAT II containing K + , Rb + and Cs + , the EFC and H 2 O molecules switch places and the H 2 O molecules are now located close to the center of the pores and the EFC near the aluminosilicate framework (see inset in Fig. 3a and Supplementary Fig. S2). In NAT II systems the H 2 O-EFC chains are disordered and more than one isomer can exist. DFT calculations in the case of Cs-NAT reveal that the energy differences between different isomer structures are about 2 kJ mol −1 pointing to a very flat potential energy surface 11 . The strain energy of the natrolite framework varies significantly as a function of EFC size as the T-O-T (T = Al, Si) angle increases from 130° in Na-NAT to almost linear at 175° in Cs-NAT. The strain energy increases as the volume increases with the size of the EFC. As a consequence of this one needs to first exchange Na + by K + and then insert the larger EFC as the energies required to directly exchange Na + by Rb + and Cs + are too big 13 . Bulk moduli, pressure-induced amorphization, and medium-range order after PIA. In our studies we defined the pressure at which the sum of all Bragg intensities I < 5% I O , I O being the sum of all Bragg intensities at ambient conditions, as the onset pressure of PIA (P A ) ( Supplementary Fig. S3). We observe distinct changes of P A with EFC size for both NAT I and NAT II (Fig. 3b): in the NAT I materials, Li-and Na-NAT, which have stronger EFC-H 2 O interactions and the EFC located near the center of the pores, PIA occurs at 12.7(1) and 15.8(1) GPa, respectively. In earlier work Goryainov 14 used a water free methanol-ethanol pressure transmitting medium and reported PIA in Na-NAT to be completed at 10 GPa, a value almost 6 GPa lower than what we found in silicone oil. This could point to an important influence of the pressure-transmitting medium in PIA as well as the fact that our pressure-conditions above 5 GPa are non-hydrostatic. In the NAT II systems where K + , Rb + and Cs + are located closer to the zeolite framework and the H 2 O molecules near the center of the pores, a different PIA onset pressure dependence with the EFC size is observed. The onset pressures increase from 14.3(1) to 19.6(1) and 26.0(1) GPa, respectively (Fig. 3b). The peaks in the diffraction pattern of Rb-and Cs-NAT after PIA indicate that in these materials more than a short-range order persists, which is in contrast to what is found in Li-, Na-and K-NAT after PIA (Fig. 1a). After pressure release the recovered samples were heated to 250 °C over a period of 5 days and no recrystallization was found to occur ( Supplementary  Fig. S4). PIA in natrolites appears to be irreversible and we do not observe any significant changes in the XRD patterns over time.
HAADF-STEM imaging using an aberration-corrected JEOL 2100F revealed that medium-range order up to 100 nm persists in both Rb-and Cs-NAT after PIA (Fig. 4). The contrast observed in the STEM images scales with approximately Z 2 , Z being the atomic number, and is predominantly due to the Rb-and Cs-cations. The inset in Fig. 4b-2 shows the Fourier transform of the STEM image of Cs-NAT after PIA and clearly indicating the presence of significant order beyond short-range correlations. In both Cs-and Rb-NAT samples both T-T (T = Al, Si) and EFC-EFC correlations are persistent as indicated by the HAADF-STEM real space images and intensities at appropriated-spacing in the XRD    Table S3). The rotation angles of the chain (Ψ) increases from 2.7(1)° in P1 to 13.7(1)° in P2. The Cs-water distributions have not been considered due to the limited 2-theta range.
pattern. Cs-NAT despite having a higher onset pressure for PIA has retained a larger degree of order. Both samples displayed rapid beam damage when we attempted to image at higher resolution. This is not surprising as the presence of water within the zeolite cages has been shown to lead to radiolysis under electron beam irradiation which will be the initial key damage mechanism 15,16 . However, our real-space images clearly show that in Cs-NAT translational symmetry exists over length scales in the 100 nm range.

Discussion
Our experiments unequivocally show that materials with the same aluminosilicate framework but different H 2 O-EFC topologies have different onset pressures of PIA and linear dependencies on cation size. Furthermore, the P1 phases of the NAT II materials Rb-NAT and Cs-NAT have higher bulk compressibilities, both transform to two denser phases, and subsequently amorphize at higher pressures than NAT I ones. We suggest that the closer proximity of water to the aluminosilicate framework in NAT I materials appears to promote coordination with Al 3+ which has been proposed as one mechanism for PIA in large pore materials 7 . The persistence of long-range order in Cs-and Rb-NAT after pressure release observed in X-ray powder data and HAADF-STEM images suggests an anti-glass structure. This structural model where translational symmetry is preserved over larger length scales while being disrupted at shorter ones was originally forwarded by Troemel for tellurium-oxide-based systems 17 18 . Redfern argued that in metamict zircon (ZrSiO 4 ) such an anti-glass structure is generated as the result of the α -decay of uranium and thorium located on Zr sites, which destroys local short-range order and over time forms areas of highly disordered material that separate ordered regions which remain registered over larger length scales 19 . The structure stabilizing role of EFC cations is well established in zeolite chemistry and the collapse of the framework structure around them due to the changes in T-O-T angles does not have to destroy their long range order. Our in-situ high-pressure powder diffraction experiments and the large strain energy found in DFT calculations of Cs-and Rb-NAT under pressure suggests a local disordering of the framework structure 10 . Subsequently the EFC are trapped and can no longer readily migrate to neighboring pores. This confinement within a disordered microporous aluminosilicate matrix can still 'trap' long-range order and explain the persistence of cation-cation correlations at higher q observed in both Rb-and Cs-NAT at high pressures as well as after pressure release.

Methods
Sample preparation. Ion-exchange procedures and characterizations of Na-NAT, Li-NAT, K-NAT, Rb-NAT, and Cs-NAT are reported in detail in Lee's papers 13,20 . The procedures of ion-exchange was made by 2 steps. First of all, K-natrolite (K-NAT) was prepared using a 4 M KNO 3 (ACS reagent grade from Sigma-Aldrich) solution and a grinded natrolite mineral (ideally Na 16 Al 16 Si 24 O 80 •16H 2 O, San Juan, Argentina) in a 100:1 weight ratio. The mixture was stirred at 80 °C in a closed system for 24 hours. The solid was separated from the solution by washing method using membrane filter (cellulose acetate 0.45 μ m, Advantec). The dried powder was used for the third exchange cycles in the same conditions. The final product was air-dried, and from the elemental analysis (Jarrell-Ash Polyscan 61E Inductively Coupled Plasma), over 99% K-exchange was confirmed. Second, Li-natrolite (Li-NAT), Rb-natrolite (Rb-NAT) and Cs-natrolite (Cs-NAT) were then prepared using K-NAT as a starting material. Total three cycles of exchange were performed as for the K-exchange procedures, and nearly full Rb-and Cs-exchanges were confirmed, quantitatively, by SEM EDS method (the residual sodium and potassium ions were detected by only very small amounts, i.e., < 0.1%). Li-, Rb-, and Cs-NAT were not directly exchanged cations from natrolite mineral (Na-NAT).
X-ray diffraction. X-ray diffraction was performed using MicroMax-007HF (Rigaku Corp.) equipped with rotating anode Mo-Kα (λ = 0.7093 Å) with multilayer optics (VariMax-Mo, Rigaku) and imaging plate (IP) detector (R-axis IV ++ , Rigaku). X-ray generator was operated at 1.2 KW (50 kV, 24 mA), and X-ray beam size was reduced to 100 μ m using pin-hole collimator. Sample to detector distance was about 120 mm, and typical exposure time was about 10 min. IPanalyzer v3.551 and Crystalclear v2.0 software were used to calibrate and process the 2D images. In the case of K-NAT, data were measured using synchrotron radiation at beamline 9A of Pohang light source (PLS). Experimental parameters were 20.07 KeV X-ray (λ = 0.6178 Å), 100 μ m beam size, 15 sec of exposure time, and CCD detector (Rayonix SX165, 80 μ m in pixel size) with S-to-D of 218.577 mm.

HAADF-STEM measurement. High-Angle Annular Dark Field Scanning Transmission Electron
Microscopy (HAADF STEM) was performed using a JEOL 2100F 200 kV FEG-STEM/TEM equipped with a CEOS Cs corrector on the illumination system to image the recovered Rb-and Cs-NAT. The geometrical aberrations were measured and controlled to provide less than a π /4 phase shift of the incoming electron wave over the probe-defining aperture of 17.5 mrad. High angle annular dark-field (HAADF) STEM images were acquired on a Fischione Model 3000 HAADF detector with a camera length such that the inner cut-off angle of the detector was 75 mrad. The scanning acquisition was synchronized to the 60 Hz AC electrical power to minimize 60 Hz noise in the images, and a pixel dwell time of 24 μ s was used.