MOF Linker Extension Strategy for Enhanced Atmospheric Water Harvesting

A linker extension strategy for generating metal–organic frameworks (MOFs) with superior moisture-capturing properties is presented. Applying this design approach involving experiment and computation results in MOF-LA2-1 {[Al(OH)(PZVDC)], where PZVDC2– is (E)-5-(2-carboxylatovinyl)-1H-pyrazole-3-carboxylate}, which exhibits an approximately 50% water capacity increase compared to the state-of-the-art water-harvesting material MOF-303. The power of this approach is the increase in pore volume while retaining the ability of the MOF to harvest water in arid environments under long-term uptake and release cycling, as well as affording a reduction in regeneration heat and temperature. Density functional theory calculations and Monte Carlo simulations give detailed insight pertaining to framework structure, water interactions within its pores, and the resulting water sorption isotherm.


Analytical Techniques:
The powder X-ray diffraction (PXRD) and water sorption data of MOF-303 presented in this manuscript were extracted from previous publications. 1,2 Liquid-state 1 H and 13 C NMR spectra were acquired on a Bruker NEO-500 (500 MHz). 1  PXRD analysis of MOF-LA2-1 obtained through solvothermal synthesis was conducted on a Bruker D8 Advance X-ray diffractometer equipped with a Cu anode and a Ni filter (CuKa radiation) in Bragg-Brentano geometry. The sample was mounted on a zero-background holder and leveled with a spatula. The PXRD patterns were recorded between 3 and 50° with 2303 steps (~0.02° per step) with an acquisition time of 10 seconds per step, thus resulting in ~6.5 hours analysis time.
The PXRD pattern of MOF-LA2-1 obtained through the green synthesis method was measured by using a Rigaku MiniFlex 6G equipped with a HyPix-400MF Hybrid Pixel Array detector and a normal-focus X-ray tube with a CuKa source. The zero-background holder is made of single crystal Si cut on a 310 axis. The PXRD patterns were recorded between 2 and 50° with 4801 steps (~0.01° per step) with scan speed of 0.5° per minute, thus resulting in ~1.5 hours analysis time per measurement.

S4
Scanning electron microscopy images were obtained on an FEI Quanta 3D FEG scanning electron microscope with an accelerating voltage of 15 kV and a working distance of 10 mm. The samples were dispersed on carbon tape on a stainless-steel holder. Energy-dispersive X-ray spectroscopy (EDS) data were collected using an Oxford X-Max EDS system working at an acceleration voltage of 15 kV.
Optical microscope images were taken by using an HRX-01 digital microscope operated in the transmission mode. The sample was dispersed on a glass slide and then placed on a motorized stage for imaging.
Single-crystal X-ray diffraction (SCXRD) measurements were conducted at the beamline 12. data processing was carried out with the APEX3 software package. 3 The data were integrated by using SAINT 4 and corrected for absorption with SADABS. 5 The structural solutions were determined by using intrinsic phasing (SHELXT) 6 and refined by the principle of least squares (SHELXL). 7 Both solution and refinement, as well as visualization of the electron density and the associated structural model were conducted by using the Olex2 software package. 8 The TGA curves were recorded on a Netzsch Jupiter, STA 449 F5 apparatus. Prior to the measurement, the samples were dried by heating to 150 °C at a rate of 1 °C min -1 . The measurement was then initiated after the temperature in the TGA oven decreased to 40 °C. For the TGA measurement, the temperature was ramped from 40 to 800 °C at a heating rate of 1 °C min -1 .
During the experiment, UHP-grade Ar at a flow rate of 60 mL min -1 was used for the balance purge flow; and UHP-grade Ar (inert conditions) or ultra-zero-grade air (oxidative conditions) at a flow rate of 60 mL min -1 was used for the sample purge flow.
Low-pressure nitrogen sorption measurements were carried out on a Micromeritics ASAP 2420 surface area analyzer. The N2 isotherms were measured using a liquid nitrogen bath (77 K

S6
The initial-guess structures of MOF-LA2-1, representing different linker configurations, were constrained to the experimental cell parameters and constructed from crystallographic data obtained for the water-loaded MOF-LA2-1 (see Section S5 for more details). Only the atomic coordinates of the framework were relaxed during the geometry optimization without relaxing the shape or volume of the unit cell.
Low-energy MOF-LA2-1 structures with different linker configurations were then selected to investigate the most favorable water adsorption sites and compare these adsorption environments to that in MOF-303. The initial guess structures for water molecules adsorbed on different framework sites were generated based on chemical intuition and paralleling the known adsorption sites in analogous MOF-303. Both the atomic coordinates and the cell parameters were relaxed during the geometry optimization of the structure similar to our previous approach to determine the water adsorption sites in MOF-303. 1 The adsorption strength of the different sites was determined based on the average binding energy of the water molecules calculated as where ΔE n is the average binding energy per water molecule at a loading of n water molecules per unit cell, E MOF+water is the combined electronic energy of the MOF and the adsorbed water molecules, and E MOF and E water,gas are the electronic energies of the reference pristine MOF and an isolated water molecule, respectively.
Force-field-based Monte Carlo (MC) simulations in the isobaric-isothermal Gibbs ensemble 12,13 were performed using the Monte Carlo for Complex Chemical Systems -Minnesota (MCCCS-MN) simulation software 14 to investigate water adsorption behavior in selected low-energy MOF-LA2-1 structures at T = 298 K. A simulation setup similar to that used previously for investigating the unary-vapor phase water adsorption in analogous MOFs, MOF-303 and MOF-333 was used. 15 The simulation setup consists of two simulation boxes, one for the MOF phase and one for the water gas-phase reservoir pre-equilibrated to T = 298 K and Ptarget, held in thermodynamic contact. A 3 × 2 × 2 supercell of the MOF structure was used and kept rigid throughout the simulation. Rigidbody translation and rotation moves were performed on randomly selected water molecules to maintain thermal equilibrium. Configurational-bias swap moves allowed for transfer of the water molecules between the two simulation boxes to maintain chemical equilibrium. Volume moves S7 performed on the cubic reservoir box were used to maintain the target pressure. In a typical simulation, 1%, 39%, 30%, and 30% of the total moves were distributed into volume moves, swap moves, translation moves, and rotation moves, respectively.
The MOF-water and water-water interactions were described using non-polarizable force fields.
For the water molecules, the rigid 4-site TIP4P model (saturated vapor pressure, Psat = 4.54 ± 0.12 kPa at 298 K) was used. 16 The non-bonded Lennard-Jones (LJ) interaction parameters and partial charges for MOF-LA2-1 were derived from the force field used for MOF-303. 15 Particularly, the non-bonded parameters for the aluminum oxide rod, and the pyrazole and carboxylate groups of the linker were kept the same as that for MOF-303. The parameters for the -CH=CH-group were taken from the TraPPE-UA force field for butadienes. 17 The Lorentz-Berthelot mixing rules were used to determine the unlike LJ parameters for the MOF-water LJ interactions. A spherical cutoff at 14 Å was used for truncating the pairwise LJ and real-space Coulomb interactions. Analytical tail corrections and the Ewald summation method were employed for the long-range LJ and Coulomb interactions.
A total of N = 2000 water molecules were used for the adsorption simulations conducted at T = 298 K and P/Psat = 0.01-1.0. Each adsorption simulation was started from the empty MOF structure-the MOF structure was optimized via DFT calculations in the presence of water molecules corresponding to a given loading, and such water molecules were deleted before the adsorption simulation-similar to the previous approach to simulate accurate water adsorption isotherms in MOF-303. 15 Most simulations were equilibrated for 50,000 MC cycles (1 MC cycle consisted of N = 2000 Monte Carlo moves), while much longer equilibration periods were used in the vicinity of the sharp step in the isotherm. At least another 50,000 MC cycles were used for the production period. To determine the statistical uncertainties, the production period was divided into 4 equal blocks.

Section S2. Synthetic Procedures
No unexpected or unusually high safety hazards were encountered while performing the synthetic procedures described in the following.

. Initial Prediction of the Pore Volume and Water Adsorption Properties
We first constructed a hypothetical MOF, MOF-LA2-1-FO (FO, full optimization), from the parent MOF-303 wherein the PZDC 2− (1H-pyrazole-3,5-dicarboxylate) linkers of MOF-303 were replaced with PZVDC 2linkers containing an extension by a vinyl group. Without any a priori knowledge of the crystal structure of this MOF, we constructed a DFT-optimized structure of this MOF, where the contact angle between the aluminum oxide rods and linkers was similar to that in MOF-303 with the pyrazole moieties forming an alternating pattern of hydrophilic-hydrophobic pockets. In this arrangement, the vinyl group extension allowed for a more than 30% increase in pore volume compared to the parent MOF (0.598 cm 3 g -1 versus 0.452 cm 3 g -1 ). Force-field-based Monte Carlo simulations in the isobaric-isothermal Gibbs ensemble (see Section S1, Computational Methods for more details) were used to predict the water adsorption isotherm of MOF-LA2-1-FO at 298 K ( Figure S3). The simulated adsorption isotherm for MOF-LA2-1-FO showed a steep step at a relative humidity of ~18% and an overall water uptake of 0.6 g g -1 -a 1.5-fold increase compared to the uptake of MOF-303-FO that was generated by using the same procedure. It should be noted that the position of the step and the saturation loading for      Figure S6f). We note that in this linker configuration, the N sites of S18 the linkers can adsorb subsequent water molecules, thereby leading to a higher number of favorable framework sites for H2O adsorption compared to MOF-303.  Section S3. 4

. Simulation of Water Adsorption Isotherms
We next probed the dependence of the water adsorption behavior on the different linker configurations of MOF-LA2-1. Force-field-based Monte Carlo simulations in the isobaricisothermal Gibbs ensemble (GEMC) were used to compute the water adsorption isotherms at 298 K. Considering the similarity of the primary adsorption sites in MOF-LA2-1 and MOF-303 (Section S3.3), the simulation setup was chosen to be similar to our previous study focusing on the prediction of water adsorption isotherms of MOF-303 (see Section S1, Computational Methods for more details). 15 Rigid framework structures of MOF-LA2-1, optimized in the presence of 1 H2O molecule per asymmetric unit that were deleted prior to the GEMC simulations, were used for these calculations. This arrangement led to an expanded hydrophilic cavity, thus accounting for the structural flexibility of the MOF, which was previously shown to be important for obtaining an accurate initial water uptake in MOF-303. 15 Using the above-described procedure, the water adsorption isotherms of MOF-LA2-1 in the ZUS(w)-trans,trans; ZUS(n)-cis,trans; ZUS(w)-trans,cis; and ENT(w)-trans,cis linker configurations were simulated ( Figure S7). Noteworthy, the ZUS and ENT linker configurations exhibit significantly different water adsorption behavior. In agreement with the measured adsorption isotherm, both the ZUS(w)-trans,trans and ZUS(w)-trans,cis configurations, in which the pyrazole rings are present on the wider side of the hydrophilic cavity, show an initial water uptake of ~0.6-1 H2O molecule per asymmetric unit at already 5% RH and a sharp step in the isotherm at 28-30% RH, slightly shifted compared to the experimental isotherm. We note that these two linker configurations differ only in the orientation of the vinyl groups, and the similar adsorption behavior of these two linker configurations suggests that the orientation of the vinyl groups (cis or trans) does not significantly influence the overall adsorption isotherm. On the other hand, the ZUS(n)-cis,trans linker configuration, in which the pyrazole rings are present on the narrowed side of the hydrophilic cavity, does not exhibit the initial water uptake at < 10% RH observed in the experimental isotherm, even though the framework structure used for this linker configuration was optimized in the presence of 1 H2O molecule per asymmetric unit. This is consistent with the observation that the water molecules did not adsorb at the strong adsorption sites during the DFT optimization, as calculated for the other ZUS linker configurations. Instead, the adsorbed water molecules move out of the plane of the two pyrazole linkers into the MOF pore, thereby not expanding the cavity significantly upon water adsorption. This linker configuration S21 displayed a steep step in the isotherm at ~22% RH, thus exhibiting a larger deviation from the experimental isotherm than the ZUS(w) configurations.
In contrast to the steep step observed in the adsorption isotherms for the three investigated ZUS linker configurations, the ENT(w)-trans,cis linker configuration exhibited a more gradual increase in its water uptake. The pyrazole functionalities are more distributed across the hydrophilic cavity,

Section S3.5. Simulation of Water Desorption Isobars
We also probed the dependence of the water desorption behavior on the different linker    Figure S9. Scanning electron microscopy coupled with energy-dispersive X-ray spectroscopy capturing a representative fraction of the bulk material of MOF-LA2-1.