Effect of ring rotation upon gas adsorption in SIFSIX-3-M (M = Fe, Ni) pillared square grid networks

Dynamic and flexible metal–organic frameworks (MOFs) that respond to external stimuli, such as stress, light, heat, and the presence of guest molecules, hold promise for applications in chemical sensing, drug delivery, gas separations, and catalysis.

Inections, steps, and hysteresis have been reported in gas adsorption isotherms in MOFs as a signature of structural transitions upon gas adsorption. Such behavior has been attributed to gate opening and/or a breathing effect, where the MOF backbone exes. [23][24][25] Nevertheless, there are very few reports that have addressed the mechanism of the dynamics of organic linkers in the porous framework and how this can lead to non-Langmuirian adsorption isotherms. [26][27][28][29][30] The dynamics of the organic linker during adsorption and desorption are inuenced by sorbate-sorbent interactions and will therefore be sorbate dependent. Linker rotation is of relevance because it could potentially enhance the selective adsorption of certain guests; 26,31 even modest swiveling of struts affects the pore size and geometry. Herein, we present experimental observations and computational studies of the inuence of rotating pyrazine rings in the pillared square grid platform, SIFSIX-3-M, of formula [M(pyz) 2 SiF 6 ] (M ¼ Fe or Ni; pyz ¼ pyrazine) during the adsorption of various gases such as Xe, Kr, CO 2 and N 2 . These pillared MOFs have been widely investigated because their hybrid and ultramicroporous nature enable benchmark selectivity towards important industrial gases such as carbon dioxide, xenon and acetylene. [32][33][34][35][36][37] The structural changes in such MOFs during adsorption of different gases remain largely unstudied and are addressed herein.

Results and discussion
The permanent porosity of SIFSIX-3-Fe was conrmed by N 2 adsorption measurements at 77 K that revealed a Brunauer-Emmett-Teller (BET) surface area of 358 m 2 g À1 (Fig. S2 in ESI †). Single component gas adsorption isotherms for Xe and Kr were collected at 298 K from 0-1 atm ( Fig. 2 and S3 in ESI †). Xe uptake of SIFSIX-3-Fe at 1 atm and 298 K was found to be 54.9 cm 3 STP g À1 , whereas Kr uptake is 30.8 cm 3 STP g À1 . The sharp increase in Xe uptake in the low-pressure region reveals a high affinity of SIFSIX-3-Fe for Xe (30 cm 3 STP g À1 at 0.1 bar) compared to other benchmark materials (see Fig. S4-S6 in ESI †).  The isosteric heat of adsorption (Q st ) of Xe for SIFSIX-3-Fe was calculated (Viral equation) using adsorption data at 278, 288 and 298 K. The Q st of Xe in SIFSIX-3-Fe was found to be 27.4 kJ mol À1 at innite dilution (see Fig. S9-S12 in ESI †). To put this into perspective, the Xe uptake and Q st at low pressure in SIFSIX-3-Fe is higher than NiDOBDC (22 kJ mol À1 ) 48 and comparable to the porous organic cage CC3 (ref. 43) (31.3 kJ mol À1 ). However, Q st of Xe is lower than in mmo topology nets (Q st ¼ À37.4 and À30.5 kJ mol À1 for CROFOUR-1-Ni and CROFOUR-2-Ni, respectively, at zero loading). 49 SIFSIX-3-Ni exhibits a BET surface area (368 m 2 g À1 ) and pore size (3.66Å) similar to its Fe analogue. However, the single component adsorption isotherm of Xe in SIFSIX-3-Ni is qualitatively different from that of SIFSIX-3-Fe. At low pressures, the isotherm is convex and then transitions through an inection point to concave, resembling a type V adsorption isotherm with no hysteresis (see Fig. 2b, 3 and S7-S8 in ESI †). Interestingly, the Kr adsorption isotherm shows no such inection point, presumably because of its smaller kinetic diameter (see Fig. 3). Xe and Kr uptakes in SIFSIX-3-Ni of 56.2 and 12.6 cm 3 STP g À1 were measured at 1 atm and 298 K, respectively. In addition, we studied the effect of temperature on the inection point; as shown in Fig. 2b, the inection point in the Xe isotherm in SIFSIX-3-Ni become more pronounced and shis to lower pressures as temperature is decreased. The Q st of Xe in SIFSIX-3-Ni at low coverage was found to be 18.9 kJ mol À1 , lower than in its Fe analog ( Fig. S9-S12 in ESI †). Nevertheless, the Q st increases to 21 kJ mol À1 at moderate loadings, and SIFSIX-3-Ni exhibits nearly equivalent Xe uptake to that of SIFSIX-3-Fe at 1 atm ( Fig. 2 and 3).
We now further address the question of why the Xe adsorption isotherm in SIFSIX-3-Ni exhibits an inection point and why the isotherm in SIFSIX-3-Fe does not. Structural exibility, 7,50 adsorbate-adsorbate attractions, 51,52 pore lling, 53 capillary condensation, 54,55 and commensurate-to-incommensurate adsorption transitions 56 have been known to induce inection points in adsorption isotherms. The location of the adsorbed Xe within SIFSIX-3-Ni was determined with in situ synchrotron-based PXRD (Fig. S23-S25 †). 57 Xe atoms reside in the center of the 1D channel along the c axis of the crystal lattice (see Fig. 4a). There is a slight expansion of both the a/b and c-axis upon Xe binding. According to the Xe-Xe Lennard-Jones potential 58 and the van der Waals radius of Xe, 59 prohibitively large repulsive forces would prevent two Xe atoms from occupying a single cage of SIFSIX-3-M at the positions observed from in situ XRD in Fig. 4a, precluding strong adsorbate-adsorbate attractions and imposing commensurate adsorption (one Xe atom per cage).
Inection points were observed in the CO 2 adsorption isotherms in [Co(HLdc)]$1.5MeOH$dioxane 30 (hysteresis observed) and MIL-91(Al) 27 (hysteresis not observed). In both of these studies, in situ XRD data indicated that the ligand in the CO 2 -loaded structure rotated from its orientation in the activated structure, and signicant differences in the simulated CO 2 adsorption isotherms in the two rigid hosts support that the inection is a consequence of the twisting of the ligands. A similar ligand swing was observed in ZIF-8. 60 To verify whether ligand rotation could also explain the inection point observed here, we performed in situ PXRD measurement at beamline 17-BM-B at Advanced Photon Source (Argonne National Laboratory) on SIFSIX-3-Ni under three distinct environments: Heloaded, Xe-loaded, and under vacuum. The results showed that the pyz rings in all three cases had similar orientations in bulk, where each ring is rotated at around AE16 degrees about the respective crystal axis (Fig. S16, S23-S25 and Table S1 †). This tilt of the rings about the crystallographic planes observed by our XRD study is consistent with our DFT energy minimized SIFSIX-3-Ni structure (see ESI †) and with theoretical and experimental studies of a cousin, SIFSIX-3-Zn. 61 Note that in situ XRD does not provide conclusive evidence of ring rotation, however.
Here, we propose and provide computational evidence that the inection point in the Xe adsorption isotherm in SIFSIX-3-Ni is due to a very different structural phenomenon: a subtle transition in the rotational orientations of the pyrazine rings. In this transition, the pyrazine rings gradually organize their rotational congurations to better accommodate Xe as guests as Xe loading increases. Note this hypothesis does not conict with the synchrotron based in situ powder XRD observations, which only indicate the ring orientations in bulk: individual rings can still ip between +16 or À16 congurations without altering the bulk structure. This AE16 degree rotation enables CH/F interactions of ca. 3.2-3.3Å in this and related structures but is different to that observed in CO 2 loaded SIFSIX-3-M, where the pyz rings are parallel to the c-axis. 33,37 With respect to a particular cage of interest, we will refer to the rotation in which the plane of the pyrazine ring faces into the cage as the "IN" conguration, and the rotation in which the plane of the pyrazine ring faces out of the cage as the "OUT" conguration.
Under the constraint that each pyrazine ring can adopt either an IN or OUT conguration with respect to a cage, each cage can adopt one of 2 8 ¼ 256 possible states (see Fig. 5a; black box denes a cage). We constructed rigid cages of each of these states in silico using unit cell parameters taken from DFT-optimized structures and calculated the ensemble average energy of Xe adsorption in each cage using a classical molecular model (see ESI for details Fig. S13-S17 †). The distribution of Xe adsorption energies in these 256 cage states of SIFSIX-3-Ni in Fig. 5b shows that the orientation of the pyrazine rings has a signicant inuence on the Xe adsorption energy. The cage with the most favorable Xe energy of adsorption (Fig. 5a) exhibits a set of four OUT rings on one half of the cage and four IN rings on the other half; a Xe at its minimum energy position in this type of cage is "hugged" by the four OUT rings with their planes oriented more tangential to the Xe atom. Note the three distinct clusters in the distribution in Fig. 5b. All congurations in the cluster with the most favorable (lowest) Xe energy of adsorption exhibit a set of 4 OUT rings as in the minimum energy conguration (see inset of Fig. 5b); this 4-ring conguration is not seen in the two other clusters with higher energies. All congurations in the intermediate cluster contain a 4-ring conguration with three OUT rings (see inset of Fig. 5b). In contrast, the distribution of Xe adsorption energies in the Fe analogue spans a smaller range of energies and displays only two, less distinct clusters (see Fig. S18 †), showing that the rotational congurations of the rings have a lesser inuence on the host-guest interaction in the Fe analogue because of its larger cage size. The rotational congurations of the rings that yield the most favorable Xe adsorption energy for the Fe analogue are analogous to Fig. 5a.
Next, we investigated the magnitude of the effect that the rotational congurations of the rings can have on the simulated Xe adsorption isotherms. We constructed two rigid-host structures with 16 cages: the rotational conguration of each ring in the rst structure is chosen at random; in the second structure, the rotational congurations are chosen so that each cage looks as in Fig. 5a, the cage that achieves the optimal energy of Xe adsorption. Each cage in the structure can achieve this conguration by aligning its pyrazine rings down the c-axis and forming a chessboard pattern from the view normal to the c-axis (see Fig. S19 †). The four OUT rings on one side of this optimal cage construct an optimal binding site for Xe while the four IN rings on the other side provide the neighboring cage with four OUT rings for an optimal Xe binding site. Fig. 6a shows that the Xe adsorption isotherm in the structure with organized ring congurations saturates at a lower pressure as a consequence of its more favorable guest-host interaction. For SIFSIX-3-Fe, however, the difference between the two rigid host isotherms is less drastic than for the Ni analogue (see Fig. S18 †) (Fig. 6a).
The energies of the SIFSIX-3-Fe and SIFSIX-3-Ni crystal structures as calculated using VASP 62,63 indicate that the minimum energy conguration for a vacant corridor corresponds to the IN-OUT-IN-OUT conguration, shown to be suboptimal for Xe adsorption in the simulations described above. Introduction of one Xe atom per cage into the SIFSIX-3-Fe and SIFSIX-3-Ni systems, however, introduces new energetic effects, which cause the minimum energy conguration to shi to the more favorable OUT-OUT-OUT-OUT conformation next to each adsorbed Xe atom. This is consistent with the addition of Xe gradually reorganizing the rotational congurations of the pyz rings, causing the structure to eventually become a more receptive adsorbent.
Our hypothesis is that, under vacuum, the rings are exploring their microstates by ipping between AE16 degrees, adopting approximately random/uncorrelated rotations with respect to one another down the c-axis, while primarily residing in IN-OUT-IN-OUT congurations in the a/b plane. As Xe atoms are introduced into the structure, the rotational congurations of the rings gradually rearrange to the OUT-OUT-OUT-OUT conguration with organization along the c-axis to achieve more favorable guest-host interactions as in Fig. 5a. One can envision that, in the process of the gradual structural transition, the Xe adsorption isotherm transitions from the bottom curve in Fig. 6a to the top curve through an inection point. To test if freely ipping pyrazine rings can induce an inection point in the Xe adsorption isotherm of SIFSIX-3-Ni, we simulated Xe adsorption in the grand canonical (mVT) ensemble while allowing each pyrazine ring to ip between the two rotational congurations with no intra-host energetic penalty (see ESI † for details). The simulated Xe adsorption isotherm in SIFSIX-3-Ni with freely ipping pyrazine rings displays a pronounced inection point (Fig. 6b) as the rings organize (Fig. S16 †) to achieve a more favorable guest-host interaction (Fig. S20 †). The simulated Xe adsorption isotherm in SIFSIX-3-Fe with ring ipping does not exhibit an inection point as a consequence of the smaller effect that the rotational congurations of the rings have on the Xe adsorption energy. In concordance with the experimental results in Fig. 3, the simulated adsorption isotherms for Kr and CH 4 in SIFSIX-3-Ni with ipping rings do not exhibit an inection point (Fig. S21-S22 †), indicating that the presence of an inection point due to ipping rings is sensitive to the match between the adsorbate and cage size.
Notably, a more gradual ring ordering effect similar to the one postulated here is reported for CO 2 adsorption in SIFSIX-1-Cu, [Cu(bpy) 2 SiF 6 ] (bpy ¼ 4,4 0 -bipyridine). 64 It was observed that simulations of CO 2 adsorption in a rigid structure corresponding to the lowest energy ring rotational conformation drastically underestimated the experimental adsorption due to orientational constraints on the adsorbed CO 2 molecules, while the simulated uptakes in various structures with higher energy ring orientations produced results that were in good agreement with experiment. We speculate that, relatedly, pyrazine ring rotation may be involved in a suspected phase change in SIFSIX-3-Zn 32,65 and gate opening in Fe(py) 2 [Pt(CN) 4 ]. 24 According to our simulations, the inection point in the Xe isotherm in SIFSIX-3-Ni broadens and shis to higher pressures with increasing temperature (Fig. S21 †). The broadening is consistent with the notion that the rings have an entropic incentive to dynamically ip to explore their microstates; as Xe adsorbs, the rings organize to achieve a greater host-guest interaction at the expense of entropy. At higher temperatures, entropy begins to dominate the free energy, and the inection broadens. The experimentally measured isotherms at 313 K, 298 K, 288 K, 275 K and 195 K affirm the inection point shiing toward higher pressures and broadening as the temperature increases (see Fig. 2b and S8b †). Remarkably, the inection point occurs at $0.5 Xe adsorbates per cage. We postulate that this is not a coincidence; when 1/2 of the cages are occupied by Xe, the rings cannot adopt orientations independent of one another without interfering with a host-Xe interaction, requiring a long-range organization. This also explains why the inection point shis to higher pressures as the temperature increases.
Interestingly, Kanoo et al. 61 found that SIFSIX-3-Zn adsorbs more carbon dioxide at 298 K than at 195 K. Consistent with our XRD studies and DFT calculations, they found the pyz rings in SIFSIX-3-Zn adopt 17 tilts about their respective crystallographic planes. Furthermore, their Raman spectroscopy studies implied that the structure becomes more symmetric upon the adsorption of CO 2 , likely due to changes in the alignment of the pyz rings. Their spectroscopic data in the SIFSIX-3-Zn analogue is consistent with our hypothesis of a disordered to ordered transition of the pyz rings in SIFSIX-3-Ni as xenon adsorbs.

Conclusions
In summary, we report a new isostructural porous pillared square grid net, SIFSIX-3-Fe, that exhibits high isosteric heat of adsorption of Xe and preferential adsorption of Xe over Kr. We attribute this behavior to the optimally tuned pore size that is commensurate with the size of Xe atom. An inection in the Xe adsorption isotherm in SIFSIX-3-Ni arises, and we attribute this behavior to a disordered to ordered transition of the rotational congurations of the pyrazine rings as opposed to other phenomena such as guest-guest interactions or breathing. In this transition, the rings organize their rotational congurations to achieve a greater guest-host interaction. To our knowledge, such dynamic behavior has not been suggested previously as the origin of an inection in gas adsorption. Our understanding is a step towards the loy goal of engineering MOFs with moving parts to harness these dynamics for applications in gas sensing and separations, drug delivery, and catalysis.