Selective sorting of polymers with different terminal groups using metal-organic frameworks

Separation of high-molecular-weight polymers differing just by one monomeric unit remains a challenging task. Here, we describe a protocol using metal-organic frameworks (MOFs) for the efficient separation and purification of mixtures of polymers that differ only by their terminal groups. In this process, polymer chains are inserted by threading one of their extremities through a series of MOF nanowindows. Selected termini can be adjusted by tuning the MOF structure, and the insertion methodology. Accordingly, MOFs with permanently opened pores allow for the complete separation of poly(ethylene glycol) (PEG) based on steric hindrance of the terminal groups. Excellent separation is achieved, even for high molecular weights (20 kDa). Furthermore, the dynamic character of a flexible MOF is used to separate PEG mixtures with very similar terminal moieties, such as OH, OMe, and OEt, as the slight difference of polarity in these groups significantly changes the pore opening kinetics.


Measurements
Powder X-ray diffraction (PXRD) data were collected using a Rigaku SmartLab Diffractometer with Cu K radiation. The particle size distribution was obtained by Horiba Partica LA-950 laser diffraction particle size analyzer. Adsorption isotherms of N2 at 77 K were measured with a Belsorp-mini equipment. Before the adsorption measurements, samples were treated under reduced pressure (10 -2 Pa) at 373 K for 5 h. Differential scanning calorimetry (DSC) was conducted with a Seiko Instruments DSC 6220 under an atmosphere of N2. Scanning electron microscopy (SEM) images were collected using a Hitachi S-3000N SEM system operated at an accelerating voltage of 30 kV. Samples were placed on a conducting carbon tape attached to an SEM grid, and then coated with platinum. Gel permeation chromatograph (GPC) measurements of the PEG was performed in CHCl3 at 40 C on three linear-type polystyrene gel columns (Shodex K-805L) that were connected to a Jasco PU-980 precision pump, a Jasco RI-930 refractive index detector, and a Jasco UV-970 UV-vis detector set at 256 nm. 1 H NMR spectra were obtained using a JEOL A-500 spectrometer operating at 500 MHz. Solid-state NMR measurement was performed on a 9.4 T Bruker solid-state NMR instrument with an Advance III 400 MHz spectrometer and a double resonance 4 mm magic angle spinning probe. 1 H-13 C heteronuclear correlation (HETCOR) with frequency-switched Lee-Goldburg (FSLG) homonuclear decoupling was conducted. The HETCOR spectrum was obtained using a recycle delay of 2 s with a spinning rate of 10 kHz. FSLG contact time of 1 H-13 C crosspolarization was 2 ms. A single crystal of 2 containing H−PEG−H (0.6 kDa) was mounted using MiTeGen's MicroMount TM . Intensity data were collected at 103 K in flowing low temperature nitrogen gas on a Rigaku XtaLAB P200 with VariMax Mo Optic with MoKα radiation (λ = 0.71075 Å) and a confocal monochromator. The structure was solved by direct methods and refined by fullmatrix least-squares cycles in SHELX2018/1. 1,2 All non-hydrogen atoms were refined with anisotropic thermal parameters. Hydrogen atoms attached to C were located at geometrically calculated positions and refined with isotropic thermal parameters. The guest PEG molecules included in the channels showed severe disorder and the SQUEEZE command in PLATON 3,4 was used in the structure refinement.

MD simulation
All the molecular dynamics (MD) simulations were executed using the pmemd module of the Amber12 package, which was modified to handle covalent bonds across the periodic boundary. All the density functional theory (DFT) calculations were executed at the level of M06-2X/6-31+G(d) using the Gaussian09 package. The guest PEG molecules are modeled as X-(O-CH2-CH2)14-O-X (terminal groups X are H atom, Tr group, or Me group) by general Amber force field (GAFF) and their atomic charges are assigned by the Merz-Kollman scheme. We use previously reported force field parameters for the MOF framework model of 1b. 5 For the 1a model, we use the same parameters for the Zn dimers and triethylenediamine in the model of 1b and newly prepare those for the 1,4naphthalenedicarboxylate. By using DFT calculation results of Li + capped 1,4naphthalenedicarboxylate neutral model, the atomic charges are assigned by the Merz-Kollman scheme and intramolecular parameters (bond, angle, and dihedral terms) are determined to reproduce the rotational barrier of the naphthalene group rotation.
The MD simulation models are composed of the host MOF framework (1a or 1b) and guest PEG molecules with different terminal groups (H-PEG-H, Tr-PEG-Tr, or Me-PEG-Tr) under periodic boundary conditions. We executed the simulation of four systems, with the combinations that follow: (1) 1a with H-PEG-H, (2) 1a with Tr-PEG-Tr, (3) 1a with Me-PEG-Tr, and (4) 1b with Tr-PEG-Tr. The MOF framework models are built by aligning the unit cell 4  4  10 with the unit cell size 10.921  10.921  9.611 Å 3 (1a) or 10.948  10.948  9.804 Å 3 (1b) along the a-, b-, and c-axis direction. The MOF framework models are periodically connected along the aand b-axes and each of the terminal Zn atom on the [001] surface are capped by a CH3 group.
We executed MD simulations with the following simulation parameters. The integration time step is 1 ps and the SHAKE algorithm is applied to keep the distances of bonds involving hydrogen atoms. The temperature is controlled to be at 373 K by the weak-coupling algorithm under constant temperature and volume ensemble.
The initial structures in the MD simulation systems were created by placing 24 PEG chains on one side of the [001] surface of the MOF framework with the periodic boundary box size 43.684  43.684  200.0 Å 3 (1a) or 43.792  43.792  200.0 Å 3 (1b). Then, the PEG molecules on the MOF framework surface were equilibrated by a 300 ps MD simulation with the constraint of the atomic positions of the MOF framework by belly algorithm. Finally, production MD simulations were executed for 500 ns.

Synthesis of MOFs
Synthesis of 1a: 1a was prepared according to literature. 6 3.36 mmol of Zn(NO3)2•6H2O, 3.36 mmol of 1,4-naphthalenedicarboxylic acid, 1,67 mmol of triethylenediamine were dissolved in 20 ml of DMF, then heated at 120 °C for 48 h in a steel autoclave. After the reaction, the resulting white powder was collected by centrifugation and washed several times with dehydrated DMF. The activated MOF 1a was then obtained by evacuating the solvent at 140 °C under reduced pressure, and stored over desiccating silica gel.
Synthesis of 1b: 1b was prepared according to literature. 7 3.36 mmol of Zn(NO3)2•6H2O, 3.36 mmol of 1,4-benzenedicarboxylic acid, 1.67 mmol of triethylenediamine were dissolved in 20 ml of DMF, then heated at 120 °C for 48 h in a steel autoclave. After reaction, the resulting white powder was collected by centrifugation and washed several times with dehydrated DMF. The activated MOF 1b was then obtained by evacuating the solvent at 140 °C under reduced pressure, and stored over desiccating silica gel.
Synthesis of 2: 2 was prepared according to the literature. 8 3.4 mmol of Co(NO3)2•6H2O, 3.4 mmol of 2,6-naphthalenedicarboxylic acid and 1.7 mmol of 4,4-bipyridine were dissolved in 600 ml DMF and heated at 120 °C for 24 h. After cooling down, the resulting powder was collected by filtration, and washed with dehydrated DMF and methanol to yield 2•MeOH as a green powder. MeOH was then evacuated at room temperature under reduced pressure (300 Pa then 3 Pa) to yield 2 without guest (purple powder). The evacuated 2 was stored at 4 °C over desiccating silica gel.

Synthesis of functionalized PEG
Note that in this article, the molecular weight indicated for functionalized PEGs refers to the molecular weight of the parent hydroxylated chain.
Synthesis of tritylated PEGs: Tr-PEG-Tr (2 kDa and 20 kDa) and Me-PEG-Tr (2 kDa) were prepared by coupling the corresponding hydroxylated PEGs with trityl chloride. 9 In a typical synthesis, PEG (ca. 5 g) was dissolved in anhydrous dichloromethane. Excess triethylamine and trityl chloride (2.5 equivalents per OH group in the case of PEG (2 kDa), and 10 equivalents in the case of PEG (20 kDa)) were added. Reactions were conducted in closed vials for 96 h at room temperature. After reaction, the organic phase was washed with aqueous NH4Cl and deionized water, dried over MgSO4, then evaporated under reduced pressure. Tritylated PEG was further purified by dissolution in a dichloromethane/diethyl ether mixture (1:99 vv) followed by evacuation at 300 kPa to cause the precipitation of PEG. The conversion was quantitative, as determined by comparison of the integrals of 1 H NMR peaks for Tr groups and main chain.
Synthesis of alkylated PEGs: Me-PEG-Me, Et-PEG-Et and Bu-PEG-Bu were prepared by Williamson ether synthesis from the corresponding hydroxylated PEGs. 10 In a typical synthesis, H-PEG-H (ca. 5 g) was dissolved in 100 ml anhydrous toluene, and maintained at 0 °C. Excess NaH (suspension in oil) was then added and left to react for 15 min. After this, the desired alkyl halide (MeI, EtBr, or n-BuBr, 5 equivalents per OH group) was added. The reaction temperature was then maintained at 0 °C for 1 h then progressively increased to 60 °C for 24 h. Afterwards, the reaction temperature was decreased to 0 °C and MeOH was added to neutralize the remaining NaH. The organic phase was then washed with aqueous NH4Cl and deionized water, then evaporated under reduced pressure. The resulting mixture was dissolved in water, then washed with diethyl ether and extracted with dichloromethane. The dichloromethane phase was then filtered over activated charcoal and celite, and evaporated under reduced pressure at 60 °C to yield the alkylated PEG. The conversion was quantitative, as determined by comparison of the integrals of 1 H-NMR peaks for alkyl groups and main chain.

DSC analysis to determine the enthalpy of PEG insertion
Guest-free 1a and solid PEG were placed together in a DSC crucible. A very slow heating ramp (1 °C min -1 ) was used to allow for the quantitative insertion of PEG during the first DSC cycle. Upon heating, the sample exhibited an endothermic peak corresponding to PEG melting, followed by an exothermic peak originating from PEG insertion. By contrast, neat PEG exhibited only the endothermic peak. The integral of the exothermic peak corresponds the heat released upon introduction of PEG. Because the two peaks cannot be integrated separately, the total integral is compared to that of melting PEG without MOF. The released heat can be determined by difference.

DSC analysis to determine the kinetics of PEG insertion
A known amount of MOF (mMOF) and PEG (mPEG) was introduced into a DSC crucible. The amount of PEG added was slightly above the maximal capacity of the MOF, to allow for a complete introduction. A very short heating treatment was performed to guarantee the homogeneity of PEG and MOF. The sample was then submitted to successive heating plateaus to allow some insertion to proceed (1a: 70 °C for 10 min, 2: 45 °C for 10 min). Between each plateau, the sample was cooled down to 10 °C, then heated again, so a melting of PEG could be observed after each plateau. The enthalpy of fusion (H(N th plateau)) corresponds to free PEG remaining outside of the MOF, and the quantity of PEG inserted after each plateau could be deduced by difference. The loading L, expressed in w% of the MOF (without guest), was determined using the formula ( ) = 0 -( ) For H-PEG-H, the first cycle presents a peak with a lower intensity than that of neat H-PEG-H, and an early prepeak. This was attributed to the partial insertion (ca. 30%) of PEG in the MOF that occurred during the solvent evacuation at room temperature. The PEG melting peak disappears during the second cycle and onward. This was because of the DSC analysis cycle that can be seen as a thermal annealing process. For Tr-PEG-Tr and Me-PEG-Tr, the presence of bulky terminal groups prevents the insertion, even at high temperature. The melting peak of PEG remained unchanged on all cycles during the analysis. Mn of PEG used in these works was 2 kDa. In this study, DMF was used as a solvent because of characteristic peaks of 2 containing PEG and DMF, so that the individual phases can be clearly identified. Supplementary  Fig. 17a presents the evolution of 2 with DMF under reduced pressure (<0.3 kPa). The initial phase has a composition of 2 and DMF. As DMF is evacuated, the peak at 21.1° is progressively shifted toward the larger angles and reaches its final position at 21.8° after 6 h. This peak does not move even after additional 8 hours in vacuum, revealing the existence of a stable adduct of 2 and DMF. This continuous shift of a unit cell in a flexible MOF is analogous to a recent report by Carrington et al.. 11 Interestingly, the peak at 20.4° remains unchanged upon DMF evacuation, and can be thus considered as a marker of the phase of 2-DMF composite. Supplementary Fig. 17b presents the evolution of 2 maintained in contact with a solution of H-PEG-H in DMF. Formation of 2 including PEG was not observed, showing that the PEG insertion proceeds only if solvent is removed. Supplementary Fig. 17c presents the evolution of 2 in contact with a solution of H-PEG-H in DMF, as vacuum (0.3 kPa) was applied. Initially, 2 adopts the configuration of the DMF adduct. However, this phase disappears completely within 10 min to be converted into 2 containing PEG. This indicates that the encapsulation of PEG is caused by the solvent removal. Furthermore, DMF evacuation is significantly faster in the presence of PEG, revealing that PEG can displace the solvent to cause its own insertion. Figure 18. PXRD patterns of 2 before and after immersion in various solvents. In apolar (e.g., hexane), highly polar (e.g., H2O, ethylene glycol), or bulky solvent (e.g., 1,3,5trimethylbenzene, methyl tert-butyl ether), 2 presents the pattern of a closed-pore phase, denoting the absence of insertion. In contrast, 2 presents an opened-pore phase pattern with diffraction peaks below 8° when immersed in solvents of intermediate polarity and small size, such as ethanol, acetonitrile, dichloromethane, and 1,2-dimethoxyethane. As observed for PEG of higher molecular weight, 2 presents a closed-pore phase pattern in tetraethylene glycol, but an open-pore phase pattern in tetraethylene glycol monomethyl ether and dimethyl ether. With this method, encapsulation of all R 1 -PEG-R 2 (with R 1 ; R 2 :H, Me, Et, and Bu) could be achieved readily. This is in sharp contrast with the direct introduction of molten PEG that did not proceed for H-PEG-H. Mn of PEG used in these experiments was 1 kDa.