Selective Adsorption of Oxygen from Humid Air in a Metal–Organic Framework with Trigonal Pyramidal Copper(I) Sites

High or enriched-purity O2 is used in numerous industries and is predominantly produced from the cryogenic distillation of air, an extremely capital- and energy-intensive process. There is significant interest in the development of new approaches for O2-selective air separations, including the use of metal–organic frameworks featuring coordinatively unsaturated metal sites that can selectively bind O2 over N2via electron transfer. However, most of these materials exhibit appreciable and/or reversible O2 uptake only at low temperatures, and their open metal sites are also potential strong binding sites for the water present in air. Here, we study the framework CuI-MFU-4l (CuxZn5–xCl4–x(btdd)3; H2btdd = bis(1H-1,2,3-triazolo[4,5-b],[4′,5′-i])dibenzo[1,4]dioxin), which binds O2 reversibly at ambient temperature. We develop an optimized synthesis for the material to access a high density of trigonal pyramidal CuI sites, and we show that this material reversibly captures O2 from air at 25 °C, even in the presence of water. When exposed to air up to 100% relative humidity, CuI-MFU-4l retains a constant O2 capacity over the course of repeated cycling under dynamic breakthrough conditions. While this material simultaneously adsorbs N2, differences in O2 and N2 desorption kinetics allow for the isolation of high-purity O2 (>99%) under relatively mild regeneration conditions. Spectroscopic, magnetic, and computational analyses reveal that O2 binds to the copper(I) sites to form copper(II)–superoxide moieties that exhibit temperature-dependent side-on and end-on binding modes. Overall, these results suggest that CuI-MFU-4l is a promising material for the separation of O2 from ambient air, even without dehumidification.


■ INTRODUCTION
Enriched-or high-purity O 2 is a critical commodity in the medical, manufacturing, and aerospace industries and for the production of feedstock chemicals such as ethylene oxide and phthalic anhydride. 1,2The vast majority of O 2 is produced from the cryogenic distillation of air. 3,4This energy-intensive, multistep process involves the compression and pretreatment of air to remove volatile organic compounds, water, and CO 2 , and then the resulting gaseous mixture�predominantly O 2 , N 2 , and Ar is expanded and cooled to cryogenic temperatures upon passing through a series of heat exchangers, before being fed into distillation columns where O 2 is separated from N 2 and Ar.A simplified illustration of the basic steps required for the cryogenic distillation of air is shown in Figure 1a.Ultimately, while cryogenic distillation is the most mature and widely used technology for air separations, there is significant interest in identifying more energy-efficient and scalable methods for isolating O 2 from air.
The prospect of using O 2 -selective adsorbents for energyefficient air separations has garnered research attention for decades, 4,5 beginning with early studies of O 2 binding in molecular cobalt(II) complexes. 6A renaissance in this area has occurred within the previous decade with the discovery that certain porous, microcrystalline metal−organic frameworks (MOFs) featuring coordinatively unsaturated iron(II), 7−9 cobalt(II), 10−13 and chromium(II) 14,15 sites can bind O 2 via electron transfer mechanisms that give rise to excellent selectivities for O 2 over typically redox-inactive N 2 .The reader is referred to two recent perspective articles published on this topic. 4,16Importantly, air separations using cation-exchanged zeolites that selectively adsorb N 2 over O 2 (and Ar) are already used in industry to supplement cryogenic distillation for applications where O 2 purities <95% are sufficient (e.g., for medicinal use).As such, in particular for medium to small-scale applications, infrastructure is in place that could in principle be adapted to implement separations technology using O 2selective adsorbents 3,4 A porous adsorbent capable of selectively capturing O 2 over N 2 and the other components of air, such as water, could be used to produce high purity O 2 from air in a process that requires no pretreatment 17 (other than removal of particulate matter) and is thus in principle operationally simpler than cryogenic distillation or N 2 -selective adsorptive separations. 18,19An illustration of such a hypothetical process is given in Figure 1b, although we note that this is a conceptual diagram only, intended to highlight an idealized process flow for such an adsorbent.In principle, far less adsorbent would be required to treat a given quantity of air in such a process than would be needed for an equivalent air separation using an N 2selective zeolite, since the concentration of O 2 (21%) in air is much less than the concentration of N 2 (78%). 4Consequently, the capital and energy expenditures required for air separations using an O 2 -selective adsorbent could be significantly less than what is required for cryogenic distillation and current adsorptive separations. 4However, the majority of O 2 -selective MOFs studied to date adsorb appreciable O 2 only at subambient temperatures [7][8][9][10][11][12][13]20 or exhibit poor stability to repeated cycling.14,15,21 Additionally, these materials feature coordinatively unsaturated, Lewis acidic metal sites that can also serve as strong binding sites for water. 22,23Importantly, none of the corresponding studies has examined adsorbent O 2 selectivity and capacity in the presence of water vapor, which is a non-negligible component of air.
A noteworthy framework in the context of air separations is Cu I -MFU-4l (Cu x Zn 5−x (Cl/OOCH) 4−x (btdd) 3 ; H 2 btdd = bis(1H-1,2,3-triazolo [4,5-b],[4′,5′-i])dibenzo [1,4]dioxin), which features pentanuclear cluster nodes consisting of a central octahedral zinc(II) ion coordinated to four peripheral metal ions either pyramidal copper(I) or tetrahedral zinc(II) (Figure 2a,b).Under ambient conditions, the copper(I) sites in the framework have been shown to strongly and reversibly bind O 2 , 24 and the favorable calculated ΔG°of O 2 binding at 298 K in this material suggests that it may be promising candidate for O 2 -selective adsorptive air separations. 4Additionally, in the context of hard−soft acid−base theory, we hypothesized that the intrinsic mismatch between the soft copper(I) ion and hard water molecule may render Cu I -MFU-4l selective for O 2 even in the presence of water. 25erein, we disclose that Cu I -MFU-4l, synthesized with new optimized procedures that result in higher Cu I loadings, is able to reversibly adsorb O 2 from air at ambient temperatures with excellent cyclability, even in the presence of water vapor.Variable-temperature in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) and magnetic susceptibility  data indicate that O 2 binds at the copper(I) sites to form copper(II)−superoxide motifs and that side-and end-on superoxide binding modes are in equilibrium over a range of temperatures.We find that differences in the kinetics of desorption of O 2 and N 2 from the framework allow for the isolation of high-purity O 2 , providing a new approach to separate O 2 directly from ambient air.

■ RESULTS AND DISCUSSION
Materials Synthesis and Characterization.The framework Cu I -MFU-4l was initially synthesized following the literature protocol. 26In brief, Zn 5 Cl 4 (btdd) 3 (MFU-4l) 24 was treated with excess CuCl 2 in N,N-dimethylacetamide under an inert atmosphere at 60 °C to give Cu II S38).−29 The copper(I) sites in this material are known to strongly bind H 2 , 24,26 and therefore as a means of qualitatively estimating the number of these sites in Cu I -MFU-4l, we collected H 2 adsorption isotherms at 77 K and pressures ranging from 0 to 1.2 bar (Figure S7).The material exhibits steep H 2 uptake at low pressures and achieves a capacity of 1.3 mmol/g at 1 mbar, followed by more gradual uptake at higher pressures indicative of H 2 physisorption.If all of the copper sites in the material were trigonal pyramidal copper(I), and assuming a 1:1 stoichiometry for H 2 binding, 24 we would expect a lowpressure uptake of approximately 1.9 mmol/g (based on the copper site stoichiometry determined for the Cu II -MFU-4l precursor from ICP-OES).From the measured uptake of 1.3 mmol/g at 1 mbar, we then estimate that ∼68% of the copper ions are exposed copper(I) sites. 30While estimates of copper(I) loading achieved in this way are qualitative (see Figure S7), we propose that the H 2 uptake at 1 mbar may be a useful means of estimating and comparing copper(I) loading in Cu I -MFU-4l materials [see Table S2 for a comparison of reported copper(I) loadings in various Cu I -MFU-4l samples prepared in the literature and other qualitative approaches used to evaluate loadings].
With the goal of accessing a form of Cu I -MFU-4l featuring a greater number of copper(I) sites per node and therefore higher gas adsorption capacities, we sought to optimize the synthesis of this material.For simplicity, we denote Cu I -MFU-4l materials prepared via different routes simply as Cu x -MFU-4l, where x specifies the total number Cu sites per node as quantified by ICP-OES analysis of the copper(II) precursor (e.g., the shorthand for Cu I -MFU-4l prepared via the literature route 24,26 is Cu 2.2 -MFU-4l).Following extensive optimization, we found that treatment of Zn 5 Cl 4 (btdd) 3 with 40 equiv of CuCl 2 in anhydrous dimethyl sulfoxide at 60 °C yields a material with 2.7 Cu II ions per pentanuclear node, based on energy-dispersive X-ray spectroscopy and ICP-OES (Figures S29 and S30).Two sequential additions of lithium formate monohydrate followed by a thermolysis sequence ending with heating at 250 °C afforded a material with the formula Cu 2.7 Zn 2.3 H 0.9 Cl 0.7 (btdd) 3 (hereafter, Cu 2.7 -MFU-4l; see Section S2 of the Supporting Information for synthesis details and Figures S15, S29, S34, S37, and S38).
Analysis of H 2 adsorption data collected at 77 K revealed that Cu 2.7 -MFU-4l adsorbs 2.1 mmol/g at 1 mbar of H 2 , nearly double the capacity measured for Cu 2.2 -MFU-4l under the same conditions (Figure S8; Table S2).From this uptake, we estimate that approximately 89% of the copper sites in the material are copper(I), which is one of the highest levels of copper(I) incorporation reported for Cu I -MFU-4l to date.As validation of this approach, we also conducted an experiment where we dosed a tared sample of activated Cu 2.7 -MFU-4l with 50 mbar of CO at 298 K to saturate the copper(I) sites 24 and then evacuated the sample to remove physisorbed CO.The resulting sample mass increased by 5.8(1) wt %, corresponding to a copper(I) loading of 2.4(1) per node, consistent with the copper(I) loading estimated from the H 2 adsorption data (see Table S14).−34 Thermal decomposition profiles collected under dry N 2 and O 2 revealed that the material is stable until approximately 400 and 280 °C, respectively, under these gases (Figure S16).We found it is also possible to prepare Cu I -MFU-4l via the abovementioned route using hydrated CuCl 2 in dimethyl sulfoxide without the exclusion of air or water (see Section S2 of the Supporting Information for details).The resulting material exhibits a similarly high H 2 capacity of 1.9 mmol/g at 77 K and 1 mbar (Figure S9).
In our optimization of the synthesis of Cu I -MFU-4l, we also found that treatment of Zn 5 Cl 4 (btdd) 3 with copper(I) chloride dimethylsulfide in acetonitrile at 25 °C and activation of the resulting framework at 300 °C affords Cu 2.4 Zn 2.6 Cl 1.6 (btdd) 3 (Cu 2.4 -MFU-4l; see Figures 2a,b, S31, S35, and S36).The H 2 adsorption capacity measured for this material at 77 K and 1 mbar is 2.0 mmol/g, consistent with the theoretical capacity assuming binding of one molecule of H 2 per Cu(I) site in the material (2.0 mmol/g) (Figure S7; a consistent loading was also determined from analysis of CO uptake in the material, see Table S14).Advantageously, this approach affords access to Cu I -MFU-4l with a more well-defined formula and in fewer synthetic steps than the materials accessed via copper(II) substitution and autoreduction.The BET surface area of Cu 2.4 -MFU-4l is 3820(30) m 2 /g (Figure S1), which is slightly lower than the surface area measured for Cu 2.7 -MFU-4l and consistent with the presence of only chloride capping ligands, in contrast to the mixture of chloride and smaller hydride ligands in Cu 2.7 -MFU-4l.For all O 2 , N 2 , and Ar isotherm data collection (see below), we employed Cu 2.7 -MFU-4l based on its slightly higher estimated copper(I) loading, while Cu 2.4 -MFU-4l was used for spectroscopic analyses due to its greater homogeneity of copper ions and coordinated anions. 35tructural, Spectroscopic, and Computational Characterization of O 2 Binding.As noted above, Cu I -MFU-4l is known to reversibly bind O 2 at room temperature, 24 and a recent investigation of O 2 binding in Cu I -MFU-4l using in situ Cu L 2,3 -edge near-edge X-ray absorption fine structure spectroscopy and time-dependent density functional theory revealed that O 2 adsorption is accompanied by significant electron transfer from copper(I) to O 2 with partial oxidation of the copper ion. 27We sought to better understand the nature of the binding of the binding of the O 2 species in this material using a suite of structural, spectroscopic, and computational analyses.Dosing microcrystalline Cu 2.4 -MFU-4l with 8 mbar of O 2 at 195 K resulted in a rapid color change from off-white to pink, indicative of a change in the copper oxidation state.Analysis of powder X-ray diffraction data collected for Cu 2.4 -

Journal of the American Chemical Society
MFU-4l at 195 K before and after dosing with 8 mbar of O 2 revealed a shift in the peak positions to higher 2θ values with gas dosing, while dosing with higher O 2 pressures of 109 and 1005 mbar did not lead to further changes in the peak positions (Figure S45).
Pawley fits against the diffraction data for activated Cu 2.4 -MFU-4l collected under vacuum and the sample dosed with 8 mbar of O 2 revealed a unit cell contraction upon O 2 binding from 31.2090 (14) to 31.0044(3)Å (cubic space group Fm3̅ m, see Figures S42 and S43), while a smaller unit cell contraction was characterized upon dosing the material with 9 mbar of N 2 at 195 K [from 31.2090(14) to 31.0997(11)Å, Figure S44].These results suggest a greater perturbation of the local electronic structure around the copper(I) ions upon O 2 binding versus N 2 binding, and the decrease in unit cell parameter upon O 2 dosing is consistent with the shortening of the metal−ligand bonds due to copper oxidation.Rietveld refinement of the diffraction data collected for activated Cu 2.4 -MFU-4l revealed a structure consistent with that reported previously based on neutron powder diffraction data (Figure 2a,b). 26Rietveld refinement of the diffraction data for O 2dosed Cu 2.4 -MFU-4l using the structure of the activated framework as a starting model revealed electron density above the copper sites that was refined as O 2 , yielding an occupancy of 2.8(4) molecules per node (Figures S47 and S48).However, the structural disorder of the O 2 motif about the C 3 axis as well as the crystallographic superposition of the zinc and copper ions preclude meaningful commentary on structural metrics or the nature of O 2 binding.
We turned to variable-temperature in situ DRIFTS to further investigate the nature of the binding of O 2 in Cu 2.4 -MFU-4l.Dosing the material with up to 1 bar of O 2 at 300 K resulted in clear changes in the fingerprint region (600 to 1400 cm −1 ) (Figure S35).To elucidate the features arising from bound O 2 , we conducted identical experiments in which Cu 2.4 -MFU-4l was dosed at 263 K with 8 mbar of either natural abundance O 2 (99.8% 16 O 2 ) or 18 O 2 (97 atom % 18 O).Spectra were then collected at 5 K intervals as the sample was warmed from 263 to 298 K.A set of difference spectra generated by subtracting the 18 O 2 -dosed spectra from the O 2 -dosed spectra are plotted in Figure 3a and clearly show positive and negative peaks corresponding to 16 O 2 and 18 O 2 vibrations, respectively.Intriguingly, two sets of features were generated upon O 2 dosing: one pair at 1131 and 1051 cm −1 and another pair at 1073 and 993 cm −1 for the O 2 -and 18 O 2 -dosed samples, respectively.The isotopic shifts for both features are consistent with predictions based on the simple harmonic oscillator model for O 2 (peaks for 18 O 2 -dosed material are predicted to be at 1066 and 990 cm −1 ).As such, we assign both sets of features to the O 2 species.
To determine the origins of these resonances, we performed another set of experiments in which the activated framework was dosed with either 45 mbar O 2 or 18 O 2 at 300 K, cooled to 100 K to saturate the copper(I) sites, and then spectra were collected as the sample was incrementally warmed to room temperature (Figure 3b).Interestingly, a reversible color change from pink to gray-brown was observed upon warming the sample from 100 K to room temperature (Figure S61).At 100 K, only the peaks at 1051 and 993 cm −1 were present, while above 200 K, the peaks at 1131 and 1073 cm −1 were also apparent.−41 Interestingly, this is to our knowledge the first example of a copper−O 2 adduct where the coordinated O 2 species exhibits a temperature-dependent equilibrium between binding modes.Dosing Cu 2.4 -MFU-4l with air at room temperature results in an additional resonance at 2242 cm −1 associated with N 2 adsorption. 24This resonance is red-shifted from the Raman active mode of free N 2 (2331 cm −1 ), 42 indicating the copper(I) sites function as a weak π-donor for N 2 . 27he spin state of a copper−O 2 adduct depends on the O 2 binding mode.In particular, η 2 -O 2 adducts have been found to possess a singlet ground state (S = 0), 38,43 while η 1 -O 2 adducts possess a triplet ground state (S = 1). 38,44As such, to support the assignments made based on the in situ DRIFTS data, variable-temperature dc magnetic susceptibility data were collected at 1 T for a sample of Cu 2.4 -MFU-4l dosed with O 2 at 195 K (see the Supporting Information for details and Figures S49 and S50).Below 195 K, the magnitude of the molar magnetic susceptibility−temperature product (χ M T) is close to zero.However, as the material is warmed above 200 K, χ M T steadily increases to 0.25 emu•K/mol.Although the values of χ M T should be treated qualitatively due to challenges with the sample diamagnetic correction and desorption of O 2 at higher temperatures, these data are consistent with an equilibrium between the S = 0 and S = 1 species, in which the higher spin state is at least partially accessed upon warming.Density functional theory (DFT) calculations carried out on a pentanuclear cluster model of Cu I -MFU-4l further support an equilibrium between the O 2 bound side-on and end-on to the copper sites in the framework (see Section S10 of the Supporting Information and Figure 2c).In particular, the η 2 -O 2 binding mode was found to be favored based on electronic energies over the η 1  Single-component O 2 , N 2 , and Ar isotherms were collected for Cu 2.7 -MFU-4l at 298 K and pressures ranging from 0 to 1 bar (Figure 4a).Consistent with binding of the O 2 at the open copper sites, the material exhibits relatively steep uptake of the O 2 at low pressures and achieves a capacity of 1.5 mmol/g at 210 mbar, the partial pressure of the O 2 in air.This is the second highest O 2 capacity reported for a MOF under these conditions (see Table S3), exceeded only by Cr 3 [(Cr 4 Cl) 3 (BTT) 8 ] 2 (Cr-BTT; BTT 3− = 1,3,5-benzenetristetrazolate), which exhibits a capacity of 2.2 mmol/g. 15owever, Cr-BTT is not stable for repeated cycling with O 2 under conditions relevant to uptake from ambient air, unlike Cu 2.7 -MFU-4l (see below).Oxygen uptake in Cu 2.7 -MFU-4l begins to level off at higher pressures, reaching a value of 2.0 mmol/g at 1 bar O 2 .Assuming that one O 2 molecule binds at every copper(I) site, the theoretical capacity (excluding adsorption at secondary sites in the material) is 2.1 mmol/g.Considering that the capacity at 1 bar represents both chemisorption and physisorption, the experimental uptake suggests that not all copper(I) sites in this framework are saturated at this pressure.Nitrogen uptake in Cu 2.7 -MFU-4l is more gradual at low pressures, and the material adsorbs less N 2 than does O 2 over the entire pressure range (1.5 mmol/g at 1 bar).However, at the partial pressure of N 2 in air (780 mbar), the material exhibits a N 2 capacity of 1.3 mmol/g N 2 that is only slightly less than the O 2 capacity at 210 mbar.Finally, Cu 2.7 -MFU-4l adsorbs very little Ar at 298 K between 0 and 1 bar, and at 9 mbar, the partial pressure of Ar in air, the material adsorbs <0.01 mmol/g.
Variable-temperature Ar, O 2 , and N 2 adsorption isotherms were collected to determine the enthalpies of adsorption in Cu 2.7 -MFU-4l (Figures S5 and S6).Low-temperature Ar isotherms were collected at 170, 180, and 190 K, and these data could be simultaneously modeled using the single-site Langmuir−Freundlich equation (Figure S6).In contrast, a dual-site Langmuir−Freundlich equation was needed to model O 2 and N 2 isotherms collected at 288, 298, and 308 K, consistent with primary gas binding at the copper sites and secondary interactions with the framework (Table S4).Using the fits to these data and the Clausius−Clapeyron equation, we determined O 2 , N 2 , and Ar adsorption enthalpies as a function of loading (Figure S13).The isosteric enthalpy of adsorption (ΔH ads ) at low loadings of O 2 is −56.8(1)kJ/mol, higher than the values determined for N 2 [−38.9(4)kJ/mol] and Ar [−10.9(1)kJ/mol].The enthalpy of Ar adsorption is consistent with weak adsorbate−framework interactions 46 and remains essentially constant with loading, while the heats of adsorption for O 2 and N 2 gradually decline as the copper sites become saturated and secondary adsorption sites within the framework are occupied.The O 2 and N 2 adsorption enthalpies are consistent with previously reported isosteric enthalpy (heat) of adsorption in Cu I -MFU-4l [ΔH ads = −Q st = −52.6(6)and −41.6(6) kJ/mol, respectively; see Table S3 for heats of O 2 and N 2 adsorption reported for other relevant frameworks] 24 and indicative of strong interactions with the open copper(I) sites.This result is consistent with the DRIFTS data and the electron transfer from copper(I) to O 2 .
Ideal adsorption solution theory (IAST) 47 was used to predict the equilibrium adsorption behavior of Cu 2.7 -MFU-4l exposed to a binary O 2 /N 2 mixture and extract O 2 /N 2 adsorption selectivities as a function of O 2 concentration (see the Supporting Information for details).For a binary mixture containing 21% O 2 , the O 2 /N 2 selectivity is 10 at 298 K, which would correspond to 72% adsorbed phase purity (74% at 288 K; Figure 4b).However, it should be noted that one of the assumptions of IAST is that there is a homogeneous distribution of guests within the pores of the material, 48 an assumption that is not likely to hold for Cu I -MFU-4l, where electron transfer is a driving force in O 2 binding but not in the case of N 2 binding.In a realistic scenario where O 2 preferentially clusters around the copper(I) sites, IAST is expected to overestimate actual selectivity values. 48However, breakthrough analysis using dry and humid compressed air streams indicates that O 2 does indeed bind selectively over N 2 under dynamic conditions (see below).Interestingly, predicted N 2 /Ar and O 2 /Ar selectivities suggest that Cu 2.7 -MFU-4l may also be appropriate for removing N 2 and O 2 impurities in Ar purification (Figure S12).
In addition to a high capacity and selectivity for O 2 over the other components of air, an ideal O 2 -selective adsorbent would exhibit robustness to humidity and a low affinity for water, thus potentially enabling air separations without the need for pretreatment to remove moisture.The Cu I -MFU-4l framework was previously reported to be air-stable, and in our hands, a sample of Cu 2.7 -MFU-4l was found to be robust to ambient air for 3 months, based on powder X-ray diffraction analysis (Figure S40).Furthermore, DRIFTS data collected for a sample of Cu 2.7 -MFU-4l exposed to the atmosphere at 300 K revealed the stable coordination of both N 2 and O 2 over the course of at least 10 h without any additional changes in color (Figure S36).A water isotherm collected for Cu 2.7 -MFU-4l at 298 K at relative humidity levels ranging from 2 to 80% revealed that the material has a low affinity for water (Figure S14), in contrast to the parent framework MFU-4l. 49onsistent with this result, DFT calculations predict an electronic energy of −27 kJ/mol for water adsorption at the copper(I) sites in the framework, indicating a weaker metal− adsorbate interaction compared to the binding of O 2 or N 2 to the exposed copper(I) site. 50dditionally, while the O 2 , N 2 , and Ar adsorption isotherms of Cu 2.7 -MFU-4l were completed within several hours, the completion of the water adsorption isotherm required approximately 50 h, indicative of sluggish water uptake kinetics.A small amount of hysteresis upon water desorption may be due to the formation of water clusters within the pores. 49,51Powder X-ray diffraction analysis of Cu 2.7 -MFU-4l following water adsorption/desorption isotherms collected at 298 and 308 K revealed that the material remains crystalline (Figure S38).These observations contrast the behavior of Cu 2.2 -MFU-4l, which exhibits substantial loss of crystallinity following water adsorption/desorption isotherms at 298 K. Finally, to probe the stability of Cu 2.7 -MFU-4l at 100% relative humidity under relevant conditions, we dosed a sample of the material with air presaturated with water for 30 min at 25 °C.The material was then reactivated (see Section S1.16 of the Supporting Information for details) and powder X-ray diffraction and O 2 adsorption and desorption data were collected.The framework retained crystallinity, and the isothermal adsorption data are indistinguishable from that collected for the pristine material (see Figures S41 and S10, respectively), indicating excellent material stability.
Adsorption/Desorption Cycling Performance under Dry Air.To gain initial insight into the cycling stability of Cu 2.7 -MFU-4l under more realistic O 2 capture conditions, we performed thermogravimetric analysis adsorption/desorption cycling experiments by exposing a sample of the material to flowing dry air (10 min at 30 °C), followed by desorption under a simulated vacuum (Ar purge, 10 min at 30 °C).Remarkably, the material retained >99.9% of its total capacity over the course of 40 cycles, although we note that the weight change measured under these conditions reflects both adsorbed O 2 and N 2 (Figure S18).Adsorption/desorption cycling data were also collected under simulated temperature swing conditions by exposing the material to dry air (5 min at 30 °C), followed by desorption under an O 2 purge at higher temperature (30 s at 100 °C).Under these conditions, incomplete desorption was observed, and a capacity of ∼95% (O 2 and N 2 ) was retained over 40 cycles (Figure S19).The slightly lower capacity relative to that measured under simulated pressure swing conditions is attributed to the highly oxidizing conditions used for desorption.
Kinetics Measurements.Adsorption and desorption kinetics are also critical factors to consider in assessing the utility of a candidate adsorbent.We investigated the kinetics of O 2 and N 2 adsorption and desorption in Cu 2.7 -MFU-4l at temperatures of 288, 298, and 308 K after dosing with initial quantities of each adsorbate (0.5, 1.0, 5.0, or 10.0 mmol/g) corresponding to equilibrated pressures ranging from approximately 3 to 330 mbar (see Table S5; these pressures reflecting the steep region of the single-component isotherms for each gas).Adsorption of O 2 and N 2 occurred rapidly in Cu 2.7 -MFU-4l, with pressure equilibration occurring within approximately 50 s or less in both cases and for all temperatures and dosing conditions (Figure S20).For both gases, the rate of adsorption increased with an increase in temperature, although this effect is more pronounced for N 2 (Figure 5a).At the lowest two dosing concentrations and all three temperatures, N 2 uptake in Cu 2.7 -MFU-4l equilibrated more rapidly than that in O 2 (∼20 versus 50 s).This difference was minimized at the highest two dosing concentrations of 5.0 and 10.0 mmol/g (Figures 5b,  S22, and S23), and the saturation times for both gases appeared to approach the diffusion-controlled time scales measured for Ar adsorption kinetics (∼20 s, see Figure S21).The adsorption data collected following dosing with 1.0 mmol/g of O 2 or N 2 could be satisfactorily fit (R 2 > 0.99) with a pseudo-first order rate law model using the Lagergren equation. 52Activation energies calculated using the Arrhenius equation are similar for O 2 adsorption [E a = 10(1) kJ/mol] and N 2 adsorption [E a = 12(1) kJ/mol] (Table S6 and Figure S26).Identical activation barriers were calculated using data collected under the most dilute dosing conditions (0.5 mmol/ g; see Figure S27).Diffusion time constants calculated from the kinetics data (see the Supporting Information for details) indicate that diffusion of N 2 is more rapid than O 2 , which we attribute to a weaker N 2 binding within the framework (Table S7 and Figure S28). 53The relatively large diffusion time constants are indicative of rapid diffusion kinetics facilitated by large framework pores.
Following adsorption analysis, variable-temperature O 2 and N 2 desorption kinetics data were collected under reduced pressure (Figures 5c, S24, and S25), which revealed that O 2 desorption from the material is more sluggish than N 2 desorption.For example, after dosing with 1.0 mmol/g of each adsorbate at 298 K, half-life (t 1/2 ) values extracted for O 2 and N 2 desorption were 161 and 16 s, respectively (Figure 5c), and this difference becomes even more pronounced upon lowering the temperature to 288 K (t 1/2 values of 260 and 53 s, respectively; see Figures S24 and S25).The desorption curves for both gases were fit using a pseudo-first order rate law, and the resulting data were used to calculate activation barriers for O 2 and N 2 desorption of E a = 45(1) and 30(1) kJ/mol, respectively (Figure S26 and Table S6).Importantly, although the kinetics of O 2 and N 2 adsorption in Cu 2.7 -MFU-4l are similar and both gases are expected to be adsorb under conditions relevant to air capture, these results suggest that it may be possible to tailor the desorption conditions to isolate high-purity O 2 .
Breakthrough Analysis.Breakthrough measurements were conducted at 25 °C using pelletized Cu 2.7 -MFU-4l and compressed air inlet streams (2 mL/min) with varying levels of humidity to assess the selectivity of Cu 2.7 -MFU-4l for O 2 and N 2 under more realistic conditions (see Section S1.13 of the Supporting Information for details).We note that negligible uptake of CO 2 in anticipated under these conditions based on single-component CO 2 adsorption data collected for isotherm Cu 2.7 -MFU-4l at 25 °C (Figure S11).When the material was exposed to dry air, a sharp breakthrough of N 2 occurred after 10 min, followed by breakthrough of O 2 after 25 min.This result highlights the selective nature of the binding of O 2 in Cu 2.7 -MFU-4l under these conditions (Figure 6a).Additionally, while not the focus of this work, the separation of the N 2 and O 2 breakthrough products under these conditions suggests that Cu 2.7 -MFU-4l may also be a viable adsorbent for purifying N 2 from air, although we note that in this case pretreatment of the air, or post-treatment of the recovered N 2 , may be needed to separate other trace air contaminants, depending on the intended use and required N 2 purity.Following breakthrough of N 2 from the column, the normalized outlet flow rate (F/F 0 ) for N 2 temporarily exceeded the inlet flow rate, indicative of roll-up 54 from displacement of bound N 2 by O 2 due to competitive adsorption at the same bind sites.After this initial breakthrough run, the material was regenerated with heating at 150 °C under a He purge (until O 2 , N 2 , and H 2 O were no longer detected in the outlet stream), completing the first breakthrough "cycle".The material was then cooled to ambient temperature, and two more adsorption/desorption cycles were performed under the same conditions.The breakthrough times and capacities for O 2 and N 2 did not change over the course of these three cycles (Figures 6a and S51 and Table S12).
Using the same sample, three breakthrough cycles were subsequently carried out in succession, involving adsorption of compressed air streams with relative humidity levels of 25, 50, 75, and 100% and regeneration with heating at 150 °C under He (Figures S52−S55).There was no significant change in the measured breakthrough times for O 2 over the course of the 15 adsorption runs (Figure 6a), and there was no apparent change in the corresponding material capacity (Figures S56 and S57 and Table S12).Indeed, the average O 2 capacity from triplicate measurements under dry conditions was the same as that measured at the highest humidity (Figure S56 and Table S12), indicating that water neither hinders the material performance nor competes with O 2 for binding at the copper(I) sites.Finally, after the third breakthrough run at 100% relative humidity, two additional breakthrough cycles at the same relative humidity were performed with more mild regeneration under flowing He at only 50 °C (Figure S55).Although some  S12).Small quantities of O 2 detected after 10 min before breakthrough are attributed to displacement of a small amount of O 2 by N 2 .A small amount of both gases detected above the baseline at t = 0 is attributed to trace air present within the connection between the breakthrough column and the gas chromatograph that remains after flushing the system prior to the start of the measurement (see Section S1.13 of the Supporting Information and Figure S62 for details).(b) Oxygen and N 2 desorption breakthrough profiles for Cu 2.7 -MFU-4l.Nitrogen was desorbed first by purging the material with He gas at 25 °C, and O 2 was subsequently desorbed under He gas at 50 °C.The O 2 and N 2 concentrations were quantified using gas chromatography, which led to a relatively low signal-to-noise ratio.
water remains adsorbed in the material following desorption under these conditions (Figure S60), there was no apparent change in the material capacity.
The O 2 and N 2 capacities determined from averaging over all 17 breakthrough runs are 1.2 and 1.7 mmol/g, respectively.These capacities differ slightly from those determined from single-component isotherm data, namely, 1.5 and 1.3 mmol/g, respectively.A lower O 2 capacity from breakthrough analysis is consistent with competition between O 2 and N 2 binding in the material, highlighting the fact that single-component adsorption data may not accurately reflect the adsorption behavior of a MOF when exposed to a mixed-gas stream.Interestingly, the N 2 capacities determined from the humid breakthrough data are overall higher than the single-component adsorption capacity.This phenomenon is currently not understood, and while the capacity values should be interpreted with caution in the absence of statistical errors, this result may indicate that the presence of humidity enhances N 2 uptake in Cu 2.7 -MFU-4l.Importantly, the breakthrough data reveal that the framework retains selectivity for O 2 over N 2 when exposed to air streams with varying humidity levels, and also that there is an enhancement in the quantity of O 2 in the adsorbed phase in Cu 2.7 -MFU-4l relative to ambient air.
Finally, we sought to exploit the differences in N 2 and O 2 desorption kinetics discussed above and identify conditions, under which it would be possible to separately isolate adsorbed N 2 and O 2 .We found that it is indeed possible to desorb the majority of the bound N 2 from the material at 25 °C under simulated vacuum with a He purge, after which point highpurity O 2 can be isolated from the material (containing <0.5% N based on analysis of the stream composition using gas chromatography) by ramping the temperature to 50 °C (Figure 6b).Desorption of both N 2 and O 2 could also be conducted entirely at 25 °C, albeit with relatively sluggish kinetics for complete O 2 desorption, indicating a trade-off between desorption rates and thermal input for regeneration (Figure S59).

■ CONCLUSIONS
We have optimized the synthesis of the well-known metal− organic framework Cu I -MFU-4l and studied its O 2 binding properties under various conditions relevant to O 2 capture from air in the presence of water vapor.Spectroscopic, magnetic, and computational analyses revealed that the copper(I) sites bind to O 2 via electron transfer to form copper(II)−superoxo species.Interestingly, the superoxo moieties bind in both side-and end-on modes at the copper(II) sites, and these modes are in equilibrium over a range of temperatures.Breakthrough cycling experiments indicate the material is stable to extended cycling under dry and humid air streams and reversibly captures O 2 from ambient air in the presence of water.While both O 2 and N 2 rapidly adsorb in the material under these conditions, the activation barrier for O 2 desorption is higher than that for N 2 desorption, and this feature can be exploited to access highpurity O 2 (>99%) after initial N 2 desorption.Breakthrough analyses further indicate that the O 2 capacity of the material is unaffected by the presence of humidity, suggesting coadsorbed water does not bind to the exposed Cu I sites.These results highlight the advantages of using soft copper(I) sites in MOFs for selective O 2 capture in the presence of water.Further, while high adsorption selectivity for O 2 over N 2 has traditionally been sought in candidate MOFs for O 2 -selective air separations, our results reveal that differences in the desorption kinetics can be used to access high purity O 2 even when adsorption behavior suggests relatively low selectivity for O 2 .This discovery suggests that it may be worthwhile to reinvestigate existing materials that have been overlooked based on preliminary analysis of their single-component O 2 and N 2 adsorption behavior.Research is ongoing in our laboratory to further advance Cu I -MFU-4l for practical air separations, including the scaleup of the materials developed here, the impact of trace air contaminants on long-term stability, and the study of hydrophobic polymer 55 coatings to minimize water uptake.

Figure 1 .
Figure1.(a) Simplified process flow diagram of current industrial air separation, entailing air pretreatment to remove condensable volatiles, followed by cryogenic distillation to isolate O 2 .Note that the initial pretreatment step often entails adsorption along with its associated unit operations, which are not shown in detail here.A subsequent secondary distillation is used to separate N 2 and Ar.VOC denotes volatile organic compounds, which present safety issues due to potential uncontrolled oxidations with liquid oxygen in the distillation "cold box".18,19(b) More desirable adsorbent-based air separation would entail the direct removal of O 2 from untreated air at ambient temperatures, followed by secondary separation to purify N 2 and Ar.

Figure 2 .
Figure 2. (a) Solid-state structure of Cu 2.4 -MFU-4l determined from Rietveld refinement of synchrotron powder X-ray diffraction data (Figure S50).The framework Cu 2.4 -MFU-4l was prepared via an optimized synthesis route involving the direct reaction of Zn 5 Cl 4 (btdd) 3 with copper(I) chloride dimethylsulfide.Note that the Cu and Zn sites are disordered such that the aggregate ratio of Cu to Zn is 2.4 to 2.6 for the framework.(b) (Upper) Expanded view of a pentanuclear node in Cu 2.4 -MFU-4l.(Lower) Structure of the H 2 btdd linker.(c) Illustration of calculated structures for superoxide bound in a side-on and end-on fashion to copper(II) sites in the model pentanuclear cluster Cu 2 Zn 3 Cl 2 (bta) 6 (ta − = 1,2,3-benzotriazolate; see the Supporting Information for details).Vibrational spectroscopy analysis supports that O 2 adsorbs in Cu 2.4 -MFU-4l to generate both side-on and end-on superoxide bound to copper(II), which are in temperature-dependent equilibrium (Figure 3).Brown, light blue, green, red, dark blue, and gray spheres represent Cu, Zn, O, N, Cl, and C atoms, respectively.

Figure 3 .
Figure 3. (a) Difference spectra obtained by subtracting DRIFTS data collected upon warming (263 to 298 K) a sample of Cu 2.4 -MFU-4l dosed with O 2 from the corresponding data for Cu 2.4 -MFU-4l dosed with 18 O 2 .(b) Difference spectra generated by subtracting DRIFTS data collected upon warming (100 to 300 K) a sample of Cu 2.4 -MFU-4l dosed with18 O 2 (negative features) from the corresponding data for Cu 2.4 -MFU-4l dosed with O 2 (positive features).Peaks corresponding to a secondary, less activated superoxide species appear at 200 K.The lower and higher energy features in both sets of spectra are assigned to superoxide species bound to copper(II) in a side-on and end-on fashion, respectively.Insets depict superoxide overtones.
-O 2 binding mode (ΔE = −65 versus −46 kJ mol −1 , respectively).Additionally, the calculated O−O bond lengths for O 2 bound in an η 2 and η 1 fashion are slightly longer than the O−O bond length for gaseous O 2 (calculated 1.31 and 1.26 Å for O 2 versus 1.21 Å, respectively), consistent with electron transfer from copper(I) to O 2 to form a copper(II)− superoxo (O 2 •− ) moiety.Taken together, the structural, spectroscopic, and computational results support the formation of Cu II −O 2 •− moieties upon O 2 binding in Cu I -MFU-4l. 45Investigation of O 2 , N 2 , Ar, and H 2 O Adsorption.

Figure 4 .
Figure 4. (a) Single-component O 2 , N 2 , and Ar adsorption (filled circles) and desorption (open circles) isotherm data collected at 298 K for Cu 2.7 -MFU-4l.Colored lines represent calculated curves obtained from simultaneous fitting of three single-component isotherms at different temperatures with either a dual-site Langmuir−Freundlich equation (N 2 , O 2 : 288, 298, and 308 K) or a single Langmuir−Freundlich equation (Ar: 170, 180, and 190 K).(b) Variable-temperature IAST selectivities calculated for a binary O 2 /N 2 mixture.The O 2 concentration in air (21%) is denoted with a vertical gray line.

Figure 5 .
Figure 5. (a) Normalized adsorption kinetic traces for Cu 2.7 -MFU-4l dosed with 5.0 mmol/g N 2 and O 2 at 288, 298, and 308 K. Nitrogen adsorption and equilibrium in the material occurs slightly more rapidly than for O 2 at all temperatures.(b) Higher initial dosing quantity for N 2 and O 2 minimizes differences in equilibration times seen at lower dosing concentrations.(c) Normalized O 2 and N 2 desorption kinetic traces collected for Cu 2.7 -MFU-4l following dosing at variable temperatures.Oxygen desorption is more gradual than N 2 desorption, and this difference is enhanced at a lower temperature.

Figure 6 .
Figure 6.(a) Breakthrough profiles collected for Cu 2.7 -MFU-4l exposed to compressed air streams with the indicated humidity levels.Sharp breakthrough of N 2 occurs before O 2 , highlighting the selectivity of the framework for O 2 over N 2 .Symbols correspond to averages of data from triplicate runs, and solid lines are guides for the eye (see Figures S51−S55 for individual data sets and TableS12).Small quantities of O 2 detected after 10 min before breakthrough are attributed to displacement of a small amount of O 2 by N 2 .A small amount of both gases detected above the baseline at t = 0 is attributed to trace air present within the connection between the breakthrough column and the gas chromatograph that remains after flushing the system prior to the start of the measurement (see Section S1.13 of the Supporting Information and FigureS62for details).(b) Oxygen and N 2 desorption breakthrough profiles for Cu 2.7 -MFU-4l.Nitrogen was desorbed first by purging the material with He gas at 25 °C, and O 2 was subsequently desorbed under He gas at 50 °C.The O 2 and N 2 concentrations were quantified using gas chromatography, which led to a relatively low signal-to-noise ratio.