Postsynthetic Transformation of Imine- into Nitrone-Linked Covalent Organic Frameworks for Atmospheric Water Harvesting at Decreased Humidity

Herein, we report a facile postsynthetic linkage conversion method giving synthetic access to nitrone-linked covalent organic frameworks (COFs) from imine- and amine-linked COFs. The new two-dimensional (2D) nitrone-linked covalent organic frameworks, NO-PI-3-COF and NO-TTI-COF, are obtained with high crystallinity and large surface areas. Nitrone-modified pore channels induce condensation of water vapor at 20% lower humidity compared to their amine- or imine-linked precursor COFs. Thus, the topochemical transformation to nitrone linkages constitutes an attractive approach to postsynthetically fine-tune water adsorption properties in framework materials.


Methods and Equipment
General methods: All reactions, unless otherwise noted, were performed with magnetic stirring under inert gas (N2 or Ar) atmosphere using standard Schlenk techniques. Reaction temperatures were electronically monitored as external heating block temperatures. Unless otherwise noted, reagents were purchased from different commercial sources and used without further purification. Commercial m-CPBA was purified according to a procedure described below.
Infrared spectroscopy: IR spectra were recorded on a Perkin Elmer UATR Two FT-IR spectrometer equipped with an attenuated total reflection (ATR) measuring unit. IR data are reported in wavenumbers (cm -1 ) of normalized absorption. The IR bands are characterized as w (weak), m (medium), s (strong), or br (broad).
Gas sorption measurements: Sorption measurements for COFs were performed on a Quantachrome Instruments Autosorb iQ MP with nitrogen at 77 K or CO2 at designated temperature. The samples were degassed for 12 h at 120 °C under vacuum prior to the gas adsorption studies. Pore size distributions were determined from nitrogen adsorption isotherms using the QSDFT cylindrical pores in carbon model for nitrogen at 77 K. For multipoint BET surface area calculations, pressure ranges were chosen with the help of the BET assistant in the ASiQwin software, which chooses BET tags in accordance with the ISO recommendations equal or below the maximum in grams per square meter.
Values of the adsorbed amount of CO2 in VSTP [cm 3 g -1 ] were converted to molar amount adsorbed per gram of material [mmol 1 g -1 ] = VSTP/22.414. Heats of adsorption at zero coverage (θ) were estimated from CO2 adsorption isotherms measured at 273 K, 288 K and 298 K using Henry's law (Eq. 1). The low pressure region (0 < p < 50 Torr) of the isotherms was fitted linearly to derive the Henry coefficient (kH) normalized to a proportionality factor (α) from the slope (=kHα -1 ) of the fit, according to Eq. 2. The normalized Henry coefficients at respective temperatures (T) were then plotted semi logarithmically vs. T -1 . Heats of adsorption (ΔQST) were then calculated from the slope of the linear fits (Eq. 3). X-ray powder diffraction (XRPD): X-ray powder diffraction experiments were performed on a Stoe Stadi P diffractometer (Co-/Cu-Kα1, Ge(111)) in Debye-Scherrer geometry. The samples were measured in sealed glass capillaries (OD = 0.7 mm) and spun for improved particle statistics.
Rietveld refinements: Rietveld refinements were performed using TOPAS v6. The background was corrected with Chebychev polynomials (Order 5). Simple axial and zero-error corrections were used together with additional corrections for Lorentzian crystallite size and/or strain broadening.

Supercritical CO2 activation:
Activation of the methanol-soaked COF samples with supercritical CO2 was performed on a Leica EM CPD300 critical point dryer.
Thermogravimetric analysis: Thermogravimetric analysis (TGA) was performed on a NETZSCH STA 449 F3 Jupiter. Measurements were carried out with 3-6 mg of sample in an Al2O3 crucible under Ar flow (60 mL/min) a temperature range between 30 and 800 °C and a heating rate of 5 K/min. Deviating buoyancy effects between the reference crucible and the sample-loaded crucible were compensated by a correction of y-offsets. Baseline correction was achieved by subtracting reference measurements with an empty crucible.
Quantum-chemical Calculations: Atom positions and lattices of all periodic structures were optimized on RI-PBE-D3/def2-TZVP 1, 2, 3, 4 level of theory using an acceleration scheme based on the resolution of the identity (RI) technique and the continuous fast multipole method (CFMM) 5, 6, 7 implemented 8,9 in Turbomole version V7.3. 10 The CFMM uses multipole moments of maximum order 20, together with a well-separateness value of 3 and a basis function extent threshold of 10 -9 a.u. Grid 7 was used for the numerical integration of the exchange-correlation term. The norm of the gradient was converged to 10 -4 a.u. and the total energy to 10 -8 Hartree within the structure optimization using the gamma point approximation.
Structures for all investigated molecular compounds were optimized on PBE0-D3/def2-TZVP 2, 3, 11, 12 level of theory. Subsequent frequency calculations were performed on the same level of theory to ensure all minima to be true minima on the potential energy hypersurface.
NMR chemical shifts were obtained on B97-2/pcS-2 13, 14 level of theory using the FermiONs++ program package. 13,15,16 Purification of commercial mCPBA: meta-chloroperbenzoic acid (mCPBA) was obtained from Merck KGaA, Darmstadt. Commercial grade mCPBA contains water and impurities of benzoic acid, which were removed following a previously reported procedure. 17 mCPBA (25 g) was dissolved in Et2O (200 mL) and washed with an aqueous buffer solution (pH 7, 3 x 100 mL). The organic phase was dried (MgSO4). The solvent was removed under reduced pressure to afford pure mCPBA as white crystals. To avoid decomposition, the purified material was stored at -30°C under inert atmosphere in the dark.

Synthesis of NO-TTI-COF:
To a cooled suspension of rTTI-COF (30.0 mg, 42.9 µmol, 1.0 equiv.) in acetone (3.0 mL), a solution of mCPBA (45.6 mg, 0.26 mmol, 6.0 equiv.) in acetone (2.0 mL) was added dropwise at 0°C. The reaction mixture was stirred for 24 h, during which the mixture was allowed to warm to room temperature. The crude product was filtered and washed with acetone (3 x 10 mL) and methanol (3 x 10 mL). Soxhlet extraction with methanol for 24 h and supercritical drying with CO2 afforded NO-TTI-COF (23.0 mg, 72%) as a yellow-orange powder. Wavenumber (cm -1 ) -NH

Figure S 30: N2 sorption isotherm (a), pore-size distribution (b), and BET plot (c) for PI-3-COF and NO-PI-3-COF obtained via direct oxidation of PI-3-COF (reproduction experiment). NO-PI-3-COF was treated with scCO2 and residual adsorbed water was removed by storing the sample in a desiccator over CaCl2. Compared to a previous sample (Figure S 19, heated under dynamic vacuum) the loss in porosity after oxidation of PI-3-COF was reduced and allowed to obtain a sample of NO-PI-3 COF with larger
surface area of SBET = 1186 m²g -1 . Relative Pressure (P/P sat )

Figure S 32: Water vapor sorption isotherms of three consecutive cycles measured at T = 25°C for PI-3-COF and NO-PI-3-COF.
Within the experimental error, no reduction in uptake capacity within these cycles is observed.       In situ X-ray powder diffraction: XRPD patterns under humidity controlled atmosphere were collected using a Bruker D8 advanced diffractometer using Cu-Kα1 radiation from a Johann-type Ge111 monochromator and a Lynx Eye detector (Bruker) equipped with a humidity chamber (Anton Paar). The humidity within the chamber was adjusted by mixing a dry and a water vapor saturated nitrogen stream. A total flow rate of 500 mL/min was applied and the temperature of the chamber was constantly kept at 25.0 ± 0.2 °C. NO-PI-3-COF was exposed to 0.1 and 40 % r.H for switching between the dehydrated and hydrated state. For measurements on the hydration and dehydration kinetics a total scan time of 10 minutes was used. The cycling behavior was monitored by exposing NO-PI-3-COF to the "dry" (0.1 % R.H.) and "wet" state (40 % R.H.) for one hour per step. A delay time of 40 minutes, prior to a measurement using a scan time of 20 minutes was applied. The XRPD data analyses was performed using TOPAS v6. 22

Figure S 54: In situ XRPD measurements NO-PI-3-COF (a): XRPD patterns in dehydrated (black lines) and hydrated states R.H. (blue lines) including the diffraction signal attributed to the empty sample holder and the humidity chamber (grey line, asterisks) and selected reflection indices, (b) time dependent in situ XRPD patterns during hydration and dehydration, (c) quantitative analyses of the diffraction patterns during hydration and dehydration.
Upon hydration, the diffraction signals of the COF seem to decrease in intensity ( Figure S 54a, blue lines). A close inspection of XRPD patterns, however, reveals that only certain reflections like 100, 110 or 200 become weaker, whereas the intensity of other reflections like 120 stays constant (inset). As the peaks do not become broader, these diffraction effects are attributed to the incorporation of water into the crystal structure (leading to a reduced scattering contrast) rather than to a loss of the crystallinity of the COF (see below). Moreover, the incorporation of water into the structure leads to a significant upshift of the position of the 001 reflection, which corresponds to a decrease of the c-lattice parameter from 3.489(5) Å to 3.417(4) Å. When the relative humidity is subsequently decreased, the peak intensities of 100, 110 or 200 increase again (Figure S 54b, c), which shows that the hydration of the COFs is reversible. We conducted repeated measurements to gain insights into the de-and rehydration kinetics: the change in peak intensities upon de-and rehydration was observed to occur within 30-40 minutes ( Figure S 54b). At the end of the re-and dehydration cycle, the intensities do not fully revert to their initial state, which can be depicted best by monitoring the evolution of the 100 reflection intensity (Figure S 54c, black squares). After dehydration, the intensity of the 100 reflection only reaches 85 % of its original value, which suggests that after dehydration some water remains in the pores, even at lower relative humidity. The quality of the diffraction data of the NO-PI-3-COF allowed us to perform a more in depth analyses by fully weighted Rietveld 23 refinements ( Figure S 54c) for tracking the evolution of the lattice parameters. Hydration of the NO-PI-3-COF leads to a significant reduction of the unit cell volume by more than 2.5 % (brown circles), mainly driven by the contraction of the c-lattice parameter of more than 2 % (green circles), whereas the a-axis only slightly shortens by < 0.5 % (magenta circles). In conclusion, an uptake of water leads to a contraction of the mean interlayer distance and therefore to a contraction of the unit cell. This might appear counterintuitive, however by filling the pores with water molecules, the interlayer interactions can be increased mediated by hydrogen bonds among water molecules and neighboring COF layers. In addition, the water incorporation could trigger a conformational change of linker related groups and therefore lead to a more efficient packing of the COF layers. All XRPD data analyses were performed by fully weighted Rietveld 23 refinements using idealized models for the COF structures with planar layer and all possible torsion angles of the linker components fixed. The filling of the pores by disordered water molecules was simulated using oxygen atoms with artificially high thermal displacement parameters in a closed packed arrangement ( Figure S 55a, green balls). For the NO-PI-3-COF one pore was filled with seven water molecules per pore and layer. In order to visualize the impact of the pore filling on the diffraction patterns, we performed systematic simulations, where we calculated the XRPD patterns while incrementally increasing the occupancy of the pore water related oxygen sites from 0 (Figure S 55b, black lines) to 1 (blue lines), simulating a change from empty to water filled pores. An increasing pore content has significant impact on the XRPD pattern, especially on the 100 reflection, which shows a significant decrease in intensity. This decrease in peak intensity corresponds to the interaction of the increasing amount of diffuse electron density in the pores with the electron density of the COF-framework. All other reflection are not or only effected to lesser extent by this effect. It must be noted that an anisotropic occupation of the possible sites for pore water can change the magnitude of the decrease in the 100 peak intensity and can change the effect on other reflections. For the NO-PI-3-COF a significant decrease of the 110 and 200 peak intensities was noticed upon pore filling, whereas the intensity of the 120 stayed constant. Simulations with an isotropic pore filling, the intensity of the 110 and 200 diffraction lines hardly changes. This can point to an anisotropic filling of the pore in the real structure.
In the simulations, the magnitude of the decrease in peak intensity was also impacted by the size of the isothermal displacement parameter that we chose arbitrarily. Given this and the possible change of the stacking order, degree of stacking faulting and torsion of the linkers of the COF framework, we decided to not use this model for quantitative analyses of the pore filling, i.e. for tracking the occupancy of the pore water related oxygen sites. The reversible hydration and dehydration behavior of the COF material was monitored by in situ XRPD. Each hydration and dehydration cycle was started at 0.1 % R.H. Afterwards the humidity was ramped up to 40 % R.H. within one minute, kept constant, and subsequently the humidity was ramped down within one minute again. For every step the relative humidity was kept constant for one hour, which corresponds to a total cycle-length of two hours. The XRPD patterns reveal that the relative peak intensity (100), corresponding to the "dry" state of the COF, decreases during the first 3 cycles. This decrease is most pronounced after completion of the first hydration-cycle (Figure S 56b) and then the peak intensity stays constant upon continued hydration-dehydration cycles. The 100 reflection intensity in the dehydrated state stabilizes after four cycles at ≈ 75 % of its initial value (Figure S 56b, black dashed line), whereas the corresponding intensity in hydrated state stays constant at around 20 % throughout the entire experiment. This trend is also reflected by the evolution of the lattice parameters and of the unit cell volume (Figure S 56c). These data show, that after the completion of the hydration-dehydration cycles, water remains in the material that is not easily desorbed by a simple reduction of the relative humidity. On the other hand the constant intensity of the 100 reflection for the hydrated state signifies that the material reaches the same level of hydration in each adsorption cycle -and thus remains chemically and structurally intact throughout the cycles. This conclusion is also in agreement with the observed cycling stability during the presented volumetric water vapor adsorption experiments.