Cu(CH3COO)2·H2O, 1,4-benzenedicarboxylic acid (bdc), 2,5-dibromobenzene-1,4-dicarboxylic acid (Br2-bdc), anhydrous n-bromopropane (1BP) and iso-bromopropane (2BP) were purchased from Sigma-Aldrich. All chemicals were used as received without further purification. QCM Sensors (5MHz) were purchased from openQCM.
Methods:
Synthesis of 4,4'-Azobipyridine: 4,4'-azobipyridine was synthesized following a reported method.56
Synthesis of pillared-layer MOF thin films on QCM sensor: 5 MHz (Au coated) QCM-sensors were dipped in an ethanolic solution (20 mM) of 11-mercapto-1-undecanol (MUD) for 24 hours to obtain –OH functionalized surface. These substrates were then thoroughly washed with absolute ethanol (99.99%), dried and used for thin films synthesis. The MOF thin films were prepared on those functionalized substrate via a well-known layer-by-layer (lbl) liquid-phase epitaxial (LPE) method.57 The method consists of four steps to complete a cycle at 60 ºC as: i) dipped in 1 mM copper acetate ethanol solution for 15 minutes, ii) drained the metal solution and washed with fresh ethanol, iii) dipped in 0.2 mM linker solution (mixture of two linkers) in ethanol for 30 minutes and iv) drained the linker solution and washed with fresh ethanol. MOF thin films with varying doping percentage were prepared by varying the linker solution in different Br2BDC proportions. 4,4'-azobipyridine is the only pillar linker used with either 1,4-Benzene dicarboxylic acid linker or mixture of two dicarboxylic acids (1,4-Benzene dicarboxylic acid and 2,5-dibromobenzene-1,4-dicarboxylic acid) for MOF thin films upto 60 cycles.
Characterizations:
Powder x-ray diffractometer (XRD) patterns of thin films were recorded on a Rigaku XDS 2000 diffractometer using nickel-filtered Cu Kα radiation (λ= 1.5418 Å) ranging from 5 to 20 ° at room temperature (voltage 40 kV, current 200 mA). Out-of plane PXRD was recorded in 2θ/θ (step size 0.01, scan rate 0.2 º/s), in-plane in 2θ/φ geometry with grazing incident angle (ω) at 0.5 º and step size of 0.12 with scan rate 0.1 º/s.
Surface morphology of the MOF thin films were characterized using field emission scanning electron microscopy (FESEM), JEOL JSM-7200F instrument with a cold emission gun operating at 5 kV. Energy-Dispersive X-ray spectroscopy (EDS) elemental analysis and mapping were also done on the FESEM (at 15 kV).
IRRA (Infrared Reflection Absorption) spectrum (4000–600 cm–1) was collected under vacuum using Bruker VERTEX 70v with 2 cm-1 resolution and with 128 scan rate.
X-ray photoelectron spectroscopy (PHI versaProbe III) was performed for the MOF thin films under ultrahigh vacuum (10-9 bar) environment.
The adsorption (mass uptake) profiles were measured using a quartz crystal microbalance (QCM) from open QCM, Italy.
Thickness for all the thin films was calculated using J.A. Wollam ellipsometer (alpha-SE). The data was fitted using a B-Spline model including surface roughness.
Analytical reverse-phase (RP) HPLC was performed on an Agilent HPLC instrument using an Agilent zorbax SB-C3 (5 μm), 4.6×150 mm reverse-phase column at a flow rate of 0.9 mL/min using a linear gradient of solvent B in solvent A at 40 °C (solvent A = 0.1% TFA in H2O; solvent B = 0.08% TFA in acetonitrile). The 214 nm UV absorbance of the column eluent was monitored. During sample preparation, compounds were weighed out in their mentioned ratios and dissolved in 200µL of DMSO. From this stock solution, 5µL was taken and diluted with 60µL DMSO and 60µL of B solvent (B = 0.08% TFA in acetonitrile). Then 0.5 µL injected in HPLC for analysis.
MOF thin films deposited on QCM substrate were dipped in 18 mM aqueous solution of Na2EDTA·2H2O (Na2EDTA = ethylenediaminetetraacetic acid disodium salt) to disintegrate the MOF structure and remove Cu2+ from the solution. After pH adjustment to ~6, the clear aqueous solution (obtained by centrifugation) was taken for the (RP) HPLC analysis.
QCM experiments: MOF thin films were activated at 65 °C at 0.1 mbar. Mass uptake experiments were carried out using a constant flow rate (50 sccm) of dry N2, passing through saturated solvent vapors (1 and 2BP).
Analyses of uptake kinetics: Mass-frequency relationship for the QCM measurements is given by Sauerbrey equation;42
Where n denotes the overtone order (n = 3, 5, and 7) and c is the mass sensitivity constant. For a 5 MHz crystal, c has value of 17.7 ng/cm2.
We examine the data with the assumption of Fickian diffusion, that is, we assume a constant diffusivity, D, is independent of the vapor concentration.
Where Mt is the uptake (g) at time t, Msat is the uptake (g) at infinite time (i.e., at equilibrium), D has units of m2/s, and l is the film thickness. Following the Wöll and coworkers report,43 the above equation can be expressed as following:
D is calculated using the τ value, obtained by fitting equation 3.
Molecular dynamics simulation:
At first, the corresponding superstructure is generated from the unit cell of the Cu(bdc)(azbpy) MOF spanning along 3*3*6 dimensions. We have considered two situations with the superstructure framework extended as 3*3*6 (in x, y, z directions respectively) considering : I) high concentration gradient, i.e MOF thin film pores are almost empty (containing only one molecule of each of the alkanes, 1BP and 2BP in the framework), II) low concentration gradient, i.e. ~20% (of saturation amount) filled pores (including multiple molecules of each type of alkane at a time, i.e. 11 molecules of 1BP and 9 molecules of 2BP corresponding to the respective number of molecular uptakes at ~20% loading). Initially, in both the cases molecular dynamics simulation is performed by freezing the MOF (utilizing ‘freeze group’ utility installed in gromacs) as we anticipated only adsorbent-adsorbate interaction driven diffusivity trends. The partial charges over the atoms of MOF are obtained from the quantum calculations (see below) and the alkane molecules are modeled using charmm36 force field parameters.58-60 Each of the simulation both in high and low concentration gradient for each of the haloalkane molecules are performed in gas phase. The entire system was packed in a rectangular box of dimension 3.26 × 3.26 × 9.54 nm3 as per the resultant dimension of the superstructure.
Further, the similar set of simulations was performed in flexible framework. The respective bond, angle and dihedral parameters of the MOF were obtained using obgmx tool and the partial charges were kept unaltered as the set of the previous simulations performed excluding framework dynamics were excluded.
To realize the effect of bromine substitution in MOF on the haloalkane isomer selectivity, Cu(Br2-bdc)(azbpy) has been simulated under the condition of high concentration gradient, i.e MOF thin film pores are almost empty (containing only one molecule of each of the alkanes, 1BP and 2BP in the framework). The simulations in brominated MOF for each of the haloalkane were carried out in flexible framework arrangement where the partial charges of MOF were obtained from the similar quantum calculation (see below) and the bond, angle and dihedrals parameters from obgmx. The alkanes were modeled with charmm36 parameters as utilized before. The dimension of the simulation box was kept fixed corresponding to the similar dimension of the MOF superstructure build along the direction of 3*3*6 (corresponding to x, y and z directions respectively).
Simulation method: Each of the simulation is performed using periodic boundary conditions (PBC) set in all three dimensions. Long-range electrostatic interactions were maintained employing the particle mesh Ewald (PME) method61 with cubic interpolation. For the electrostatic interactions at short-range, the cut-off of 1.2 nm was employed. During freezed framework simulation of the Cu(bdc)(azbpy), the MOF dynamics was excluded using “ freeze group” utility of gromacs. Otherwise for the alkane molecules in the freezed system and the MOF along with alkanes in the flexible framework system, for constraining each the bonds involving hydrogen atoms, the LINCS algorithm62 was applied. At first, the system was energy minimized with the steepest-decent algorithm followed by stepwise equilibration in seven successive steps with gradual increase of temperature (in an interval of 50 K starting from 50 K upto 300 K) of 100 ps each (with time step of 0.0005 ps). During equilibration, the average temperature was kept fixed at the corresponding temperatures by using V-rescale thermostat via coupling the MOF and the alkane molecules separately. Finally, the equilibrated system was subjected to NVT production run for 10 nanosecond. During production simulation also, the average temperature of 300 K was maintained with the help of same V-rescale thermostat. To execute all simulations GROMACS software of version 20x were utilized.
The diffusion phenomenon of the haloalkane molecules were inspected by calculating mean square displacement (msd) utilizing the tool of “gmx msd” and the corresponding diffusion coefficient (D) were approximated. To understand the mode of chemical interactions of the haloalkane molecules with the different parts of the MOF pair correlation function of the alkane moieties (whole molecule, bromine atom and the Cα carbon of the alkane molecule linked to the bromine group) were measured with respect to the specific parts of the MOF (i.e. –N=N-, pyridyl and bdc).
Ab initio Molecular Dynamic Simulation
All AIMD simulations were run using CP2K software63 (version 9.1) with PBE functional64 and each MD time step of 0.5 fs at 300 K. The dimensions of the cell with periodic boundary conditions were 10.886 Å, 10.886 Å, and 15.9041 Å. The valence electrons of H, C, N, and O atoms were modelled with DZVP-GTH basis sets whereas DZVP-MOLOPT-SR-GTH basis set was employed for Br and Cu centers. The core electrons of all the atoms were modelled using GTH-PBE pseudopotentials. All NVT simulations employed Nose-Hoover chain thermostatting.65-67 For the self-diffusion studies, NVT simulations were run for 50 ps. The analysis of the probability distribution of Br (2BP and 1BP) - C (bdc linker) distance involved data collected at an interval of 5 fs during the crossing of the molecule across the pore (1BP: 12.5 ps – 17.5 ps; 2BP: 25 ps – 29 ps) surrounded by the bdc molecules. The MSDs were computed with freud code.68 The calculation of the diffusion constant was done by linear fitting of the MSD data (between 12.5-17.5ps for 1BP and 22-28ps for 2BP).