In Situ Investigation of Multicomponent MOF Crystallization during Rapid Continuous Flow Synthesis.

Access to the potential applications of metal-organic frameworks (MOFs) depends on rapid fabrication. While there have been advances in the large-scale production of single-component MOFs, rapid synthesis of multicomponent MOFs presents greater challenges. Multicomponent systems subjected to rapid synthesis conditions have the opportunity to form separate kinetic phases that are each built up using just one linker. We sought to investigate whether continuous flow chemistry could be adapted to the rapid formation of multicomponent MOFs, exploring the UMCM-1 and MUF-77 series. Surprisingly, phase pure, highly crystalline multicomponent materials emerge under these conditions. To explore this, in situ WAXS was undertaken to gain an understanding of the formation mechanisms at play during flow synthesis. Key differences were found between the ternary UMCM-1 and the quaternary MUF-7, and key details about how the MOFs form were then uncovered. Counterintuitively, despite consisting of just two ligands UMCM-1 proceeds via MOF-5, whereas MUF-7 consists of three ligands but is generated directly from the reaction mixture. By taking advantage of the scalable high-quality materials produced, C6 separations were achieved in breakthrough settings.

Introduction

Metal–organic frameworks (MOFs) are highly porous materials and have accumulated great interest due to their versatility of potential and actual applications. Consisting of judiciously selected metal ions or clusters and organic linkers, these materials can be synthesized for specific purposes. The idea of incorporating multiple functionalities into one MOF has always been attractive due to the promise of increased MOF complexity and specificity for target applications.[1] Examples of these are multivariate MOFs, which are isoreticular analogues of a parent system created through employing multiple linkers with different substituents but identical lengths.[1,2] Since the linkers are not sufficiently distinct from one another, they become randomly distributed in the crystalline lattice causing disorder in the structure.[2−5] This contrasts with multicomponent MOFs which use multiple topologically distinct linkers to form a MOF. The linkers are arranged with regularity in the MOF structure, enabling a higher degree of control in functional group distribution. For example, in catalysis, fine control of the pore chemistry through the inclusion of catalytic and modulating functional groups, on separate linkers, facilitates control of the catalytic reaction rate, regioselectivity, and/or stereoselectivity of the product.[6,7] Furthermore, the tunable nature of the pore chemistry can be extended to applications in luminescence, selective gas separation, and gas storage.[6−16]

The synthesis of multicomponent MOFs to date has been limited to time-intensive, laboratory-scale solvothermal batch methods. This is because precise molar ratio combinations are needed to converge on these materials and avoid the formation of undesired competing phases, such as MOFs comprised of only one kind of linker. However, the growing interest in multicomponent MOFs provides motivation to investigate methods for large-scale synthesis, for example, continuous flow methods (Figure ). UMCM-1 {[Zn4O(bdc)(btb)4/3]} is a ternary MOF comprised of benzene-1,3,5-tribenzoic acid (H3btb) and benzene-1,4-dicarboxylic acid (H2bdc), combined together in a precise molar ratio.[14] When this molar ratio is disregarded, other frameworks can be simultaneously formed, for instance, MOF-5 {[Zn4O(bdc)3]} and MOF-177 {[Zn4O(btb)2]}. The formation of multiple products can also be seen in quaternary systems, such as MUF-7, {[Zn4O(btb)4/3(bpdc)1/2(bdc)1/2]}, comprised of H3btb, H2bdc, and biphenyl-4,4′-dicarboxylic acid (H2bpdc). The addition of a third linker increases the number of potential undesired phases that can be formed, including IRMOF-10 {[Zn4O(bpdc)3]} and SUMOF-4 {[Zn4O(bdc)2(bpdc)]}. In particular, the rotational flexibility of the terminal phenyl rings in H3btb contributes to a propensity to form these side phases. This can be circumvented through use of a planar, tritopic linker such as 5,5′,10,10′,15,15′-hexamethyltruxene-2,7,12-tricarboxylic acid (H3hmtt), as seen with the development of MUF-77 {[Zn4O(hmtt)4/3(bpdc)1/2(bdc)1/2]}.[11] The complexity of synthesizing multicomponent MOFs increases with the number of components present, leading to different potential phase combinations of organic linkers.[11]

Figure 1

Real time in situ X-ray diffraction monitoring of continuous flow MOF synthesis for UMCM-1 (above) and MUF-77 (below) showing the difference in formation mechanism between the two MOF systems.

Continuous flow chemistry is typically performed in a plug flow reactor, which exhibits tightly controlled reaction parameters, a smaller footprint compared to batch reactors, and enhanced process safety from requiring lower solvent volumes.[17−19] MOF synthesis (Figure ) using continuous flow is typically performed by mixing the metal and linker solutions which lead to reduced reaction times as a result of leveraging high reactor surface area to volume ratios and improved heat and mass transfer.[20−22] The versatility of flow chemistry to synthesize a variety of MOFs with high space time yields (STYs) has been demonstrated by numerous groups.[23−30] STY is the mass of product per unit volume of reactor in a 24 h period and, along with production rates, a key metric in determining the viability of a process.

MOF formation is known to undergo a number of transitions between reaction intermediates before forming the thermodynamic product.[31−33] For example, the solvothermal synthesis of MIL-53(Al) undergoes phase transitions from reaction intermediates, MOF-235 and MIL-101, before transitioning into MIL-53(Al).[34] Continuous syntheses are often optimized to maximize product yield and throughput, which is a result of the reaction kinetics and can be an issue for MOFs that require phase transitions. Therefore, careful management of residence time is needed to tune continuous flow synthesis and achieve the desired pure product. Previous continuous synthetic experiments of MOFs indicate that the kinetic product in some cases can be the thermodynamic product (i.e., Cu-BTC, UiO-66, and MIL-53(Al)) obtained in solvothermal syntheses.[24,30] While this is ideal, it remains unknown whether this is true for the synthesis of multicomponent MOFs, where there is the potential for numerous, undesired single-component kinetic products to be formed.

In addition to monitoring phase transitions, the kinetics of nucleation and growth have been obtained for single-component MOFs such as Cu-BTC, ZIF-8, Zr-Fum, and MIL-53(Fe).[31,35−37] The growth of these MOFs has been modeled using Avrami–Erofe’ev as a base model and also either the Gualtieri or Finke–Watzky model.[38−41] These models describe the formation mechanisms, with previous findings showing that the formation mechanism of these MOFs typically varies between nucleation rate limited or phase boundary (surface reaction) rate limited.[37]

Herein, we report the use of a continuous flow chemistry approach for the synthetic optimization of two multicomponent MOFs, UMCM-1 and MUF-77-methyl (Figure ), with high phase purity. We demonstrate for the first time the suitability of continuous flow chemistry for the synthesis of multicomponent MOFs of this type. In situ synchrotron wide-angle X-ray scattering (WAXS) was employed to determine the mechanism of crystal growth and associated kinetic parameters for ternary and quaternary MOFs. Exploiting this new synthetic avenue toward bulk multicomponent MOF production, we demonstrate its industrial practicality. Here, 2,3-dimethylbutane (2,3-DMB), a C6 isomer, was separated under breakthrough conditions using magnetic induction swing adsorption (MISA).

Experimental Section

Materials

The reagents zinc acetate dihydrate (Zn(OAc)2.2H2O), terephthalic acid (H2bdc), 4,4′-biphenyldicarboxylic acid (H2bpdc), aluminum chloride, acetyl chloride, 1,3,5-triphenylbenzene, sodium hydroxide (NaOH), anhydrous magnesium sulfate (MgSO4), bromine, and 1-indanone were purchased from Sigma-Aldrich and used without further purification. The solvents dimethylformamide (DMF), dichloromethane (DCM), ethanol, and 1,4-dioxane were of analytical grade and were purchased from Sigma-Aldrich and used as received. 4,4′,4″-Benzene-1,3,5-triyl-tribenzoic acid (H3btb) and requisite precursors were synthesized as described in the literature.[42] 5,5′,10,10′,15,15′-Hexamethyltruxene-2,7,12-tricarboxylic acid (H3hmtt) and requisite precursors were synthesized as described in the literature.[9,11]

UMCM-1 Synthesis

Synthesis of UMCM-1 was performed using a Vaportec R4 reactor with R2 pump modules. A Zn(OAc)2·2H2O solution (0.1987 g, 0.905 mmol, 1 equiv) was prepared in 5 mL of DMF. The ligand solution was prepared in DMF (5 mL) containing H2bdc (0.0450 g, 0.271 mmol, 0.3 equiv) and H3btb (0.1067 g, 0.243 mmol, 0.27 equiv). The two solutions were each pumped into a 10 mL reactor coil at a rate of 0.5 mL/min for a combined flow rate of 1 mL/min. The reaction was conducted in a Vaportec R4 reactor with R2 pump modules at 85 °C and 5 bar pressure. The reaction was cooled to room temperature, and the precipitate was collected through centrifugation. The collected product was washed twice with 30 mL of DMF and solvent exchanged thrice with 30 mL of DCM over three days. The samples were dried under flowing N2 to yield a white solid (average yield 0.1069 g, 57%). All manipulations of the product while performing solvent exchange with DCM were under nitrogen atmosphere.

MUF-77-Methyl Synthesis

Synthesis of MUF-77-methyl was performed using a Vaportec R4 reactor with R2 pump modules. A Zn(OAc)2·2H2O solution (0.2200 g, 1.002 mmol, 1 equiv) was prepared in 5 mL of DMF. The ligand solution was prepared in DMF (5 mL) containing H2bdc (0.0159 g, 0.096 mmol, 0.09 equiv), H2bpdc (0.0242 g, 0.100 mmol, 0.10 equiv), and 0.0252 M H3hmtt (0.1408 g, 0.252 mmol, 0.25 equiv). The two solutions were pumped into a 10 mL reactor coil at a rate of 0.5 mL/min for a combined flow rate of 1 mL/min. The reaction was conducted in a Vaportec R4 reactor with R2 pump modules at 85 °C and 5 bar pressure. The collected product was washed twice with 30 mL of DMF and solvent exchanged thrice with 30 mL of DCM over three days. The samples were dried under flowing N2 to yield a pale-yellow solid (average yield 0.1152 g, 50%).

Results and Discussion

Synthetic Optimization

UMCM-1, {[Zn4O(bdc)(btb)4/3]}, is a ternary framework that exhibits an muo topology and possesses both micropores and mesopores which do not collapse upon desolvation of the framework.[14] Based on the reported solvothermal batch procedure for the synthesis of UMCM-1, we developed a continuous flow process for the synthesis of this multicomponent MOF.[14,43] For flow synthesis, zinc acetate dihydrate was used in place of zinc nitrate hexahydrate to minimize any safety concerns related to nitrate buildup at scale. The reaction conditions were optimized for reaction temperature and residence time within the reactor as outlined in Table . For each varying condition, the resultant white suspension was collected from the reactor and washed with DMF three times followed by solvent exchange with DCM over three days, with fresh DCM being replaced after each day.

Table 1

UMCM-1 Synthesis Optimizationa

parameter	reaction 1	reaction 2	reaction 3	reaction 4	reaction 5	reaction 6
temperature (° C)	85	85	130	130	25	25
residence time (min)	10	5	10	5	10	5
BET surface area (m2/g)	3380	3460	3500	3500	3550	3690
yield (%)	57	51	44	42	48	67
STY (kg/m3·day)	1539	2759	1208	2281	1310	3600
production rate (g/h)	0.64	1.15	0.50	0.95	0.55	1.50

Yield % is based on H3BTB as the limiting reagent.

To determine the phase purity of the UMCM-1 synthesized under continuous flow conditions, the product was digested and analyzed by 1H NMR spectroscopy to identify the linker ratio. A linker ratio of 4/3 btb to 1 bdc was expected for phase pure UMCM-1, and instead a ratio of 4/3 btb and 5/4 bdc was observed in the 1H NMR spectra for all samples (Figure S1). This indicated an excess of 1/4 bdc in the flow-synthesized UMCM-1 product. This could be a result of bdc remaining trapped in the pores; however, coformation of MOF-5 {[Zn4O(bdc)3]} in the reaction is more likely as the MOF was washed extensively. In this case, the ratio of linkers would suggest a product ratio of 89% UMCM-1 and 11% MOF-5. Powder X-ray diffraction (PXRD) (Figure A and Figure S3) of the samples shows the desired UMCM-1 reflections and, to a lesser extent, the MOF-5 reflection at a 2θ angle of 6.8° which confirms the coformation of MOF-5.[11] Thermal stability of the UMCM-1 frameworks was examined using thermogravimetric analysis (TGA) and corresponded to the reported UMCM-1 thermal degradation profile by Matzger and co-workers.[14] An initial mass loss was observed at 150 °C, attributed to DMF remaining in the pores. The following mass loss at 450 °C was associated with the linker degradation and subsequent collapse of the framework (Figure S5).

Figure 2

PXRD patterns of (A) UMCM-1 and (B) MUF-77 synthesized at various temperatures with 10 min residence time (* denotes MOF-5 reflection at 6.8°). Background peaks are observed in (A) due to the presence of petroleum jelly (X-alliance GMBH) used to mitigate moisture sensitivity during data collection.

Brunauer–Emmett–Teller (BET) surface areas of the UMCM-1 MOFs were calculated from nitrogen adsorption isotherms recorded at 77 K (Table , Figure S7). The UMCM-1 BET surface areas ranged from 3375 to 3690 m2/g, with the room temperature (25 °C) conditions providing the highest surface area. These surface areas were below that reported in the literature by Matzger and co-workers at 4160 m2/g.[14] This decrease could possibly be attributed to the rapid formation of MOF and due to the zinc acetate precursor employed.[44] Scanning electron microscopy (SEM, Figure C and Figure S9) was used to determine the particle size and morphology of the MOFs. Micrographs of the MOFs showed fibrous crystals with diameters of 100 nm, compared to that of UMCM-1 reported by Walton and co-workers with 30 μm diameter crystals.[45] This was a result of using the zinc acetate salt in the synthetic reaction, as opposed to zinc nitrate salts, which increased the nucleation rate and limited crystal growth.[46]

Figure 3

SEM micrographs of (A) UMCM-1 synthesized at 130 °C, (B) MUF-77 synthesized at 25 °C, (C) MUF-77 synthesized at 85 °C, and (D) MUF-77 synthesized at 130 °C.

With respect to synthetic optimization, the space–time yields (STY) and production rates were higher when a 5 min residence time was employed; however, this was the result of higher throughputs of reagents through the reactor. Further increasing the residence time to 10 min results in a higher percentage yield of MOF for reactions at 85 and 130 °C, which are comparable with the reported yields by Matzger and co-workers.[14]

To extend these investigations and include an additional linker dimensionality, flow synthesis of a quaternary multicomponent system, MUF-77-methyl {[Zn4O(hmtt)4/3(bpdc)1/2(bdc)1/2)]}, was explored. MUF-77 exhibits three distinct pore sizes due to its ith-d topology. The reaction conditions, temperature, and residence time were varied according to Table . A pale yellow suspension was collected from the reactor, which was washed with DMF three times followed by solvent exchange with DCM over three days with fresh DCM being replaced after each day.

Table 2

MUF-77-Methyl Synthesis Optimizationa

parameter	reaction 1	reaction 2	reaction 3	reaction 4	reaction 5
temperature (° C)	85	130	130	25	25
residence time (min)	10	10	5	5	10
BET surface area (m2/g)	3340	3530	3560	1850	1360
yield (%)	50	65	36	21	86
STY (kg/m3·day)	1659	1955	1187	1365	2855
production rate (g/h)	0.69	0.81	0.49	0.57	1.19

Yield % is based on H3hmtt as the limiting reagent.

To determine the phase purity of MUF-77-methyl, the product:linker ratios were determined through 1H NMR spectroscopy of samples digested in base (Figure S2). This confirmed the expected ratio of 4/3 hmtt, 1/2 bpdc, and 1/2 bdc in MUF-77-methyl. PXRD of the products further confirmed the MOF pure phase formation, matching simulated MUF-77-methyl patterns reported by Telfer and co-workers (Figure B and Figure S4).[6] The absence of XRD diffraction peaks from undesired MOF phases contrasted with the flow synthesis of UMCM-1. TGA of MUF-77-methyl showed structural degradation occurring at 435 °C (Figure S6).[6] Nitrogen adsorption isotherms (Figure S8) recorded at 77 K for MUF-77-methyl exhibited a broad range of BET surface areas (Table ). For MUF-77-methyl synthesized at high temperatures, a BET range of 3340–3558 was observed, consistent with literature surface areas reported at 3600 m2/g.[11] Comparatively, MUF-77-methyl synthesized at 25 °C resulted in poorer quality products. This was evident in the PXRD patterns with respect to peak broadening as a result of smaller crystallites and lower BET surface areas that are less than 2000 m2/g. SEM micrographs of MUF-77-methyl synthesized at 25 °C (Figure B) showed spherical particulates, with a decrease in size to 50 nm. The smaller crystallite size indicates rapid nucleation with insufficient growth which necessitated the elevated temperatures to promote the crystal formation process.[47]

With respect to synthetic optimization (Table ), the room temperature synthesis gave good yields at 10 min residence time but poor surface areas. Increasing the reaction temperature to 85 °C resulted in an increase in the surface area up to 3340 m2/g and a yield of 50%, which is comparable with reported yields by Telfer and Liu.[11] Further increasing the reaction temperature to 130 °C resulted in higher surface areas and yields with longer residence times. This indicates that a temperature over 85 °C is required for high surface area MUF-77-methyl formation, and residence time influences the yield.

In Situ X-ray Analysis of Flow Product Formation

In situ synchrotron wide-angle X-ray scattering (WAXS) analysis was performed to observe the formation of multicomponent MOFs during continuous flow. A variety of in situ analytical techniques such as XRD and small-angle X-ray scattering (SAXS) have been performed on MOF syntheses to investigate particle and crystal growth formation mechanisms.[32,35,36,48,49] However, they have been typically performed under static solvothermal conditions with few reports of in situ flow synthesis analysis.[50−52]

To investigate the in situ flow syntheses of multicomponent MOF systems, UMCM-1 and MUF-7 {[Zn4O(btb)4/3(bpdc)1/2(bdc)1/2)]} were selected as candidate materials. MUF-7 is a variant of MUF-77, where H3hmtt was substituted by H3btb and was selected as the quaternary representative for practical purposes.[10] Under steady state conditions, the length along a flow reactor is equivalent to reaction time progression, and as such, diffraction data were obtained at various positions along the reactor.[53] The steady state conditions also allowed for syntheses to be studied where the progression is on the order of seconds to minutes, which can be difficult to achieve under static solvothermal conditions.

The apparatus shown in Figure S11 was used to perform the experiment where the sample cell (Figure S12) was mounted with brackets on an aluminum post and aligned to the center of the beam and the detector positioned 742 mm past the sample cell. The sample cell was fed by the continuous flow reactor positioned adjacent to the beamline which was connected via Touhy Borst adapters which enabled connection to the reactor setup and allowed for manipulation of the reactor length. This allowed for the sample cell to be fixed and enabled the control of the residence time through the substitution of tubing with known lengths (i.e., 505 mm = 30 s, 1142 mm = 60 s). The minimum reaction time measurable was 6.2 s due to the swept volume of the fittings and sample cell (quartz capillary) used to obtain the measurements. The maximum time point studied was 10 min and equivalent to the maximum residence time used in the flow synthesis optimization. For each data collection, Bragg peak intensities were obtained, plotted, and fitted to the Avrami–Erofe’ev (AE) model and Finke–Watzky (FW) model to determine crystal growth kinetic parameters (see the Supporting Information for additional information).[39,41,54−56]

The obtained time-resolved diffraction patterns (Figure A) show that the (002) reflection from MOF-5 appears initially. This is followed by the respective reflections from UMCM-1, with (111) and (010) being the prominent Bragg peaks, which increase in intensity over time. Comparing the Scherrer crystallite size between UMCM-1 and MOF-5 (Figure S15), the size of MOF-5 remains constant after 100 s of reaction time, whereas there is an increase in crystallite size for UMCM-1 until 200 s reaction time. The kinetic parameters of the AE and FW fitted curves (Figure D and Figure S13) were obtained and highlight a difference in growth limitation during flow synthesis. Here the AE growth exponent shows that growth in the [010] direction is limited by nucleation, compared to [111] being phase boundary limited (see Table S3 and the Supporting Information for discussion). The FW kinetic parameters for [010] also support this, suggesting a higher rate of growth compared to nucleation. Considering the crystallite sizes calculated for the two Bragg peaks at the maximum residence time of 10 min, the average calculated (010) crystallite size was 770 nm as compared with the average (111) crystallite size of 680 nm. The larger overall crystallite sizes observed in the [010] direction can be a result of the faster rate of growth as determined by the FW kinetic model. Upon further investigating the difference in the growth kinetics between the two planes, the ratio between the peak intensity of the (010) and (111) planes changed over time (Figure S14). Overall, the growth kinetics and mechanism indicate a preferential growth in the [010] direction.

Figure 4

Diffraction patterns at various points of time within the flow reactor for (A) UMCM-1 at 26 °C, (B) MUF-7 at 26 °C, and (C) MUF-7 at 80 °C (red dash line denotes (002) reflection from MOF-5). Extent of crystallization over time, AE and FW model fits for (D) UMCM-1 based on peak height at (010) reflection at 26 °C, (E) MUF-7 based on peak height at (042) reflection at 26 °C, and (F) MUF-7 based on peak height at (042) reflection at 80 °C.

MUF-7 time-resolved diffraction patterns (Figure B and C) were obtained at 26 and 80 °C, at a maximum reaction time of 10 min. The minimum reaction time attainable for the 80 °C experiment was 30 s due to the minimum required length that can be heated and plumbed within the reactor housing. The progression of the prominent Bragg peaks, (022), (042), and (422), for MUF-7 and (002) for MOF-5 was tracked. The reflections from the possible impurities (MOF-177, IRMOF-10, UMCM-1, and SUMOF-4) (Figure S16) were not observed in the data collected and therefore were not tracked. In this case, the formation of MUF-7 occurs without the formation of an intermediate phase, and the observed pathway was simply escalating crystal growth. The kinetic parameters (Table S4) obtained for the various MUF-7 reflections (022), (042), and (422) at 26 and 80 °C were consistent with each other. This indicates a nonpreferential direction of growth for the MOF. Increasing the reaction temperature to 80 °C resulted in doubling the 26 °C reaction AE rate constant for each of the Bragg peaks. With respect to the FW model parameters, the 26 °C reaction exhibited a nucleation rate lower than the growth rate, indicating a nucleation-limited reaction. The 80 °C reaction, conversely, exhibited a higher nucleation rate than the growth rate, indicating a phase boundary-limited reaction. In absolute terms, the intensities observed for the (042) Bragg peak (Figure S17) between 26 and 80 °C showd an increase in the overall intensity at similar time points with the higher reaction temperature. The Scherrer crystallite size (Figure S18) was obtained using the (042) Bragg reflection and was at a maximum size of 900 and 800 Å for temperatures of 80 and 26 °C respectively.

Overall, the synthesis of UMCM-1 and MUF-7 show two different pathways. The UMCM-1 initially formed an intermediary phase in MOF-5 before the formation of the multicomponent phase, whereas a direct formation of the multicomponent MOF was observed for MUF-7.

Breakthrough Experiments

The observations made using in situ X-ray diffraction revealed that the quaternary multicomponent MOFs do not undergo phase transitions throughout synthesis. In addition, the kinetic information allows further optimization of the reactions for future scale-up of multicomponent MOF synthesis. To test that the measured structural parameters were amenable to practical use, the continuous-flow-synthesized MUF-77 vapor adsorption experiments were investigated, as these are typical processes which require large quantities of adsorbent material. Adsorption processes with zeolites, and more recently MOFs, have been proposed as energy efficient alternatives for gas separation processes.[57,58] These adsorptive processes have been applied in a variety of separations with varying feedstocks for the purpose of separating different gas mixtures. Recently, Macreadie et al. demonstrated the remarkable ability for a multicomponent MOF, CUB-30, to separate benzene and cyclohexane through in silico breakthrough measurements.[9] This study highlighted the industrial potential of multicomponent MOFs, further supporting our investigation to discover high-throughput strategies for their synthesis. To extend this investigation, we explored the separation of 2,3-dimethylbutane (2,3-DMB) from nitrogen due to the pressing need to adopt alternative separation processes to distillation methods, particularly in the case of C6 isomers.[59] The adsorption of 2,3-DMB has previously been performed at elevated temperatures (343–473 K) using zeolites and MOFs (UiO-66, Fe2(BDP)3).[57,58,60] In the two analogous MOF systems, MOF-5 and CUB-5, the selective adsorption of 2,3-DMB, among other hexane isomers, was observed at 298 K at low pressures.[61] Using these vapor sorption investigations as a model, the effectiveness of flow-synthesized MUF-77-methyl to adsorb 2,3-DMB was verified using breakthrough experiments. A comparison of 2,3-DMB adsorption capacity for MUF-77-methyl at vapor pressures of 3 and 25 kPa was made with the previously reported materials (Tables S8 and S9).

Initially, low-pressure vapor sorption and nitrogen sorption isotherms were obtained at 298 K (Figure A). The 2,3-DMB isotherm exhibited a high capacity, as expected, while MUF-77-methyl maintained a low affinity for nitrogen at 298 K.[9] For the breakthrough experiments, magnesium ferrite (MgFe2O4) nanoparticles were incorporated with the flow-synthesized MUF-77-methyl at 10 wt % loading to form a MgFe2O4@MUF-77-methyl composite. The magnetic nanoparticles within the composite cause localized heating within the adsorption bed and enable the use of magnetic induction swing adsorption (MISA).[22,62−65] This minimizes heat transfer losses due to the thermally insulating nature of MOFs. The MgFe2O4@MUF-77-methyl was characterized with SEM (backscattered imaging and energy dispersive X-ray spectroscopy (EDX)) and vibrating sample magnetometry (VSM). EDX mapping (Figure S20) shows a good distribution of the magnetic nanoparticles among the MOF powder, while VSM (Figure S22) of the composite powder reveals a hysteretic curve with a magnetization strength of 8 emu/g. A temperature rise profile of the MgFe2O4@MUF-77-methyl powder was obtained by measuring the bed temperature while being subjected to an external magnetic field with a field strength of 31 mT. When compared to the bare magnetic nanoparticles, a significant decrease in the maximum achievable temperature was observed in the composite, which can be attributed to the insulating nature of the large pore size and consequent pore volume of the MOF.[66,67] This further highlights the need for localized heat generation to reduce any heat transfer effects to which conventional temperature swing adsorption may be susceptible.

Figure 5

(A) Low-pressure adsorption isotherms for MUF-77-methyl at 298 K. Note: the saturation pressure of 2,3-DMB at 298 K is 31.07 kPa. (B) Simplified flow diagram of a breakthrough apparatus.

2,3-DMB breakthrough experiments were performed with an experimental setup outlined in Figure B (see Figure S23 for a detailed diagram). 2,3-DMB vapor and nitrogen were delivered to the MgFe2O4@MUF-77-methyl packed bed. The outlet gas of the bed was analyzed with a mass spectrometer. After saturation, the bed was then regenerated by applying an alternating magnetic field, causing the MgFe2O4 to generate heat and expel the 2,3-DMB. The breakthrough curves (Figure ) obtained from the experiments show an immediate breakthrough of nitrogen when in the adsorption phase for both 3% and 25% feed concentration experiments. The breakthrough (Figure A) for the 25% 2,3-DMB occurred after 100 s, with saturation of the bed occurring after 300 s. The 3% feed concentration (Figure C), as expected, showed a longer residence time before breakthrough at 300 s and saturation at 850 s.

Figure 6

(A and C) Breakthrough plots normalized to the maximum component feed concentration. (B and D) Concentration profile of 2,3-DMB and N2 overlaid with the adsorption bed temperature.

The calculated capacities based on the break point times for complete adsorption (usable bed) of 2,3-DMB do not allow for the mass transfer zone (period between initial detection and saturation) adsorption capacities (Figure ). A decrease in the calculated breakthrough capacity of 176 cm3/g for the 25/75 feed mixture is seen compared with the equilibrium isotherm capacity at a vapor pressure of 25 kPa of 200 cm3/g. This decrease can be attributed to discounting the mass transfer zone capacity from the overall breakthrough capacity zone to avoid including the volume of potential vapor condensation within the bed from the saturated feed mixture. A wide mass transfer zone of 600 s was observed for the 3/97 feed mixture experiments, which resulted in a substantial decrease in usable bed capacity. This broad mass transfer zone can be attributed to the lower concentration of vapor in the feed stream, which decreases the diffusion kinetics based on a smaller concentration gradient. After the mass transfer zone capacity (i.e., total adsorption capacity) was included, the amount adsorbed was approximately 17 cm3, equating to a capacity of 142 cm3 of vapor per gram of MOF (Table S7), which is closer to the equilibrium capacity of 184 cm3/g. This decrease in capacity can be attributed to a need for further optimization of the mass transfer zone through adjustment of the bed dimensions and adsorbent geometry to improve the diffusivity of the vapors to the pores.

Figure 7

MUF-77-methyl capacity for 2,3-dimethybutane from breakthrough experiments. (Green) 25:75 2,3-DMB:N2 feed concentration. (Purple) 3:97 2,3-DMB:N2 feed concentration. (Square) Usable bed capacity (cm3/g). (Circle) Regeneration capacity (cm3/g).

Regeneration was performed by applying an alternating magnetic field, strength of 31 mT, with a constant flow of 50 mL/min of helium gas. The alternating magnetic field generates heat through hysteretic losses and rises in temperature, following the temperature rise profile. The released gas composition was analyzed with the mass spectrometer to determine the 2,3-DMB content being regenerated. While nitrogen was detected in this outlet stream, the negligible increase in nitrogen during the regeneration phase demonstrates that solely 2,3-DMB is adsorbed to MOF (Figure B and 6D). The flow rate exiting the adsorption bed was assumed to be 50 mL/min for regeneration calculations. The average amount was 18.54 cm3 for the 25/75 2,3-DMB/N2 feed mixture over five cycles and 18.18 cm3 for the 3/97 2,3-DMB/N2 feed mixture over three cycles. This equates to a regeneration capacity of around 150 cm3/g for both feed compositions or a production rate of 180 and 169 cm3 2,3-DMB vapor per gram of MOF per hour of operation for 25/75 and 3/97 feed compositions, respectively.

Conclusion

This research highlights the use of continuous flow methodology for the synthesis of phase pure multicomponent MOFs on scale. Remarkably, in the time of 10 min for continuous flow synthesis compared to 72 and 12 h for UMCM-1 and MUF-77 solvothermal syntheses, respectively, comparable yields were achieved. Importantly, this result creates a new foundation for multicomponent MOFs to be rationally considered for industrial applications. The high phase purity product obtained for ternary and quaternary MOFs, UMCM-1, and MUF-77, respectively, exhibited similar physical and behavioral properties to material obtained by solvothermal syntheses. Counterintuitively, MOF-5 serves as an intermediate to the ternary UMCM-1 framework whereas the quaternary MUF-7 is generated directly from its dissolved components. This result is promising for the future use of multicomponent MOFs in large-scale applications with the respective MOF families potentially synthesizable under similar conditions. In situ X-ray analysis revealed directed growth for UMCM-1 with nucleation and phase boundary limited reactions in the two reflections studied. In contrast, MUF-7 exhibited nonpreferential growth, with formation being highly temperature dependent. This affects the nucleation rate of the reaction and, in turn, affects the main mechanism of formation from being either nucleation limited or phase boundary limited in the 26 and 80 °C reaction, respectively. Finally, the industrial importance of multicomponent MOFs was demonstrated through employing a composite material, MgFe2O4@MUF-77, to separate 2,3-dimethylbutane from nitrogen. The breakthrough results obtained from the study showed a regenerable material that exhibits similar capacities to that obtained from equilibrium-based studies.