Thermal Treatment Effect on CO and NO Adsorption on Fe(II) and Fe(III) Species in Fe3O-Based MIL-Type Metal–Organic Frameworks: A Density Functional Theory Study

The properties of metal–organic frameworks (MOFs) based on triiron oxo-centered (Fe3O) metal nodes are often related to the efficiency of the removal of the solvent molecules and the counteranion chemisorbed on the Fe3O unit by postsynthetic thermal treatment. Temperature, time, and the reaction environment play a significant role in modifying key features of the materials, that is, the number of open metal sites and the reduction of Fe(III) centers to Fe(II). IR spectroscopy allows the inspection of these postsynthetic modifications by using carbon monoxide (CO) and nitric oxide (NO) as probe molecules. However, the reference data sets are based on spectra recorded for iron zeolites and oxides, whose structures are different from the Fe3O one. We used density functional theory to study how the adsorption enthalpy and the vibrational bands of CO and NO are modified upon dehydration and reduction of Fe3O metal nodes. We obtained a set of theoretical spectra that can model the modification observed in previously reported experimental spectra. Several CO and NO bands were previously assigned to heterogeneous Fe(II) and Fe(III) sites, suggesting a large defectivity of the materials. On the basis of the calculations, we propose an alternative assignment of these bands by considering only crystallographic iron sites. These findings affect the common description of Fe3O-based MOFs as highly defective materials. We expect these results to be of interest to the large community of scientists working on Fe(II)- and Fe(III)-based MOFs and related materials.


INTRODUCTION
Metal−organic frameworks (MOFs) based on a triiron oxocentered cluster (Fe 3 O) 1,2 are important materials for several applications, including gas storage and separation, 3−5 heat pump applications, 6,7 catalysis, 8−10 and drug delivery. 1,11,12 The most representative of this class of MOFs are MIL-100(Fe) 2 and MIL-101(Fe) (MIL = Materials Institute Lavoisier). 13 In the assynthesized material, the metal node has the formula [Fe III 3 (μ 3 -O)(X)(Z) 2 ] 6+ (Figure 1a), where Z is a solvent molecule (e.g., water) and X is a counteranion. 5 The anion originates from the reagents used in the MOF synthesis (e.g., −OH from NaOH or KOH, −F from HF, and −Cl from iron chloride salts). More than one kind of X and Z can be present in the same material.
Upon heating of the material in a vacuum or in a flow of inert gas, both the X and Z species can be removed from the metal node, creating an open iron site that can coordinate an adsorbate. 5 The oxidation state of the iron centers does not change upon the removal of Z. The removal of X, instead, causes the reduction of one Fe(III) to Fe(II). This means that upon increasing the temperature we can go from the as-synthesized sample, having all of the nodes with the formula [Fe III 3 (μ 3  x Fe III y (μ 3 -O)(X) l (Z) n ] 6+ , with x = 0 and 1, y = 3 − x, l = 1 − x; 0 ≤ n ≤ 2 + x. The presence of open iron centers and, in particular Fe(II) sites, has been considered to be one of the reasons for the successful performance of Fe 3 O-based MOFs in catalysis 8,9 and gas separation. 3−5 It is then important to evaluate the efficacy of the possible postsynthetic protocols in order to tailor the number of open Fe(II) and Fe(III) species for the different purposes. Vibrational and Mossbauer spectroscopies 4,5,9,14 and microcalorimetry 4 are commonly used to optimize postsynthetic protocols involving the iron centers as the thermal treatment of these materials and the grafting of guest species. 15 IR spectroscopy using nitric oxide (NO) and carbon monoxide (CO) as molecular probes is particularly suitable for the characterization of iron-based materials because the stretching frequency of these molecules is shifted to different spectral regions if they are coordinated to Fe(II) or Fe(III) centers. 16,17 IR spectroscopy of CO and NO is an efficient diagnostic tool to determine the iron oxidation state in MOF materials. Previous spectroscopic studies on MIL-100(Fe) suggested the presence of heterogeneous Fe(II) and Fe(III) sites. 5,14 Nevertheless, no explanation of the origin of this heterogeneity was provided besides its possible correlation with the presence of an unreacted linker 18 in the material and the presence of defects. 5 In fact, the only data that can be used for the assignment of the NO and CO bands in IR spectra in Fe-based MOFs are based on studies on oxides and zeolites, 5,14 whose structures are different compared to that of the Fe 3 O cluster. The study of CO and NO adsorption on Fe 3 O-based MOFs is relevant to gas separation and therapeutics. NO removal from exhaust gases is an important process. 19 CO removal from H 2 and CH 4 is a mandatory step for their safe use in fuel cells. 4,20 [21][22][23] and CO 22 are also important gasotransmitters in physiological and biological functions. MIL-100(Fe) and MIL-101(Fe) are among the most studied MOFs for drug delivery because of their high capacity and low toxicity. 1,11,12 Understanding how the presence of water modifies the interaction of NO and CO with the drug carrier is pivotal for these applications because water triggers the gas release. 4 Moreover, the effect of water on NO and CO adsorption in MOFs is important because water is often present as a contaminant in many gas mixtures and can affect the separations.
In this study, we used Kohn−Sham density functional theory (DFT) methods to determine if the presence of water molecules and of X influences the adsorption enthalpy and stretching frequency of CO and NO in Fe 3 O-based materials. The effect of CO and NO coverage was also evaluated. We considered two X, −OH and −Cl, to assess the vibrational shift of the molecular probes. We adopted cluster models, namely, [Fe II x Fe III y (μ 3 -O)(X) l (Z) n ] 6+

COMPUTATIONAL METHODS
DFT calculations were performed using the M06-L functional 25 in its unrestricted formalism (U) in combination with the def2-TZVP basis sets, 26,27 as implemented in the Gaussian 16 program. 28 Previous investigations showed that this level of theory correctly reproduces the electronic properties of iron centers in MOFs, 29 in particular Fe 3 O, when benchmarked versus multireference calculations. 8 The Fe 3 O model has been previously employed to describe N 2 O reactivity on MIL-100. 9 Moreover, this model has been used by Mavrandonakis et al. 30 to predict the adsorption enthalpies and vibrational frequencies of different adsorbates in trimetal oxo-centered MOFs. This model has shown results similar to those reported for a cluster coordinated to benzoate instead of formate groups in reactivity 31 and in adsorption studies. 30 Geometry optimizations were carried out by means of the Berny optimization algorithm with an analytical gradient. A (99, 590) pruned grid was used (i.e., 99 radial points and 590 angular points per radial point). The Gaussian 16 default convergence thresholds were set for optimization. All of the energetic data were corrected for basis set superposition error (BSSE) following the a posteriori method proposed by Boys and Bernardi, 32  Unscaled harmonic frequencies were obtained analytically. Enthalpies and Gibbs free energies were calculated at 1 atm and 298 K using the scheme proposed by De Moor et al., 33 whereby low-lying frequency modes (<50 cm −1 ) were replaced by a cutoff value (50 cm −1 ) in the calculation of the vibrational partition functions. 34−38 Charge and spin densities were obtained using Charge Model 5 (CM5) 39 and Hirshfeld population analysis, 40 respectively. Spin densities are expressed as the difference between the α and β electron densities.

RESULTS AND DISCUSSION
The dependence on the temperature of the water and X [or Fe(II)] content in Fe 3 O-based MOFs was determined experimentally in previous studies. 4,5 We summarize prior results to guide the reader in a comparison between our computational results and the experimental ones. Leclerc et al. 5 treated MIL-100(Fe) (X = 81% F, OH, trimesate) in a dynamic vacuum in the 25−300°C range: they observed the complete removal of free and bonded water at 150°C, with the formation of only a small fraction of Fe(II). Above 150°C, the concentration of Fe(II) increases with the temperature: Yoon et al. 4 reported a maximal removal of 40% of the initial Fe−X sites at 260°C. Above 260°C, the sample starts to decompose. A slightly different behavior was reported by Wuttke et al. 14 using a helium flow: at 150°C, they observed only the removal of physisorbed water, while the water directly bonded to the Fe 3 O clusters was fully desorbed at 200°C. Moreover, the removal of X is far less effective in a helium flow than in dynamic vacuum. 4 Also X plays a role in the thermal behavior of Fe 3 O samples: the removal of only 4−5% of Fe−X was reported at 250°C for a MIL-100(Fe) (X = 20% Cl, OH, trimesate) sample. 9,18 Different  The ground-state electronic configuration for all of the systems has the three iron centers in high spin states. However, the most stable configuration for the Fe 3 O-Cl, Fe 3 O-OH, and Fe 3 O is not the highest possible spin state for the cluster (HS) but the "broken-symmetry" solution (BS) where two high-spin Fe(III) centers couple antiferromagnetically. Although the BS energetics would be more accurate, 9,41 the corresponding wave function is not a spin eigenfunction nor does it have the correct spin density. Moreover, the BS solution is strongly dependent on the initial guess, hindering both the reproducibility of the BS results and the comparison among different studies. Following a common strategy, 41−43 we modeled all of the clusters considering the Fe 3 O node in HS. In general, the difference in energy between the HS and BS is 20−30 kJ mol −1 . For details on this choice, see the discussion reported in previous studies. 8,9,31 Coordination of the adsorbates can cause a change in the ground spin state of the iron centers. This has been evaluated for a range of spin states starting from the HS value of the bare triiron oxocentered clusters to determine the most stable spin state.
The calculations indicate that each iron center can coordinate only one adsorbate molecule: when more than one molecule is adsorbed on one metal node, each molecule is coordinated to a different metal site of the metal node. The highest coverage corresponds to filling of the position left free by X and Z (marked with green spheres in Figure 1a). This agrees with the available IR experiments on CO and NO adsorption that show the formation of monocarbonyls and mononitrosyls only. 5,14 Only in the NO case, forcing the formation of a Fe···2NO complex brings displacement of the carboxylate groups. This complex is a local minimum, but it is less stable than the mononitrosyl complex by 10 kJ mol −1 . Such a displacement is possible in a cluster model where no geometrical constraints were used in the optimization, while it is unlikely to happen in the MOF structure because of the framework constraints and because of the large cluster distortion, at least at subatmospheric NO pressures considered in the experiments. 24 Accordingly, the experimental spectra do not show the formation of dinitrosyls. 24 H  Table 1, Figure 2, and Table S6.   has only a slight dependence on the iron oxidation state [it increases upon going from Fe(II) to Fe(III)], while its dependence on the counteranion is negligible (it increases upon going from −Cl to −OH; Table 1 and  Table 1 and Figure 1 agree with the experimental values and reproduce the small dependence on the coverage of the isosteric heat on MIL-100(Fe).
The vibrational bending modes of water, δ(H 2 O), can be used to determine the oxidation state of the metal node: the calculations indicate that δ(H 2 O) is shifted to lower wavenumbers if the cluster is in its oxidized form, while it is shifted to higher wavenumbers if it is reduced ( Figure S3). CO Adsorption. Relevant energetic, geometric, and spectroscopy parameters for all of the CO complexes are reported in Table 2, while additional parameters are listed in Tables S2 and  S5. CO is adsorbed end-on, C-side, in all of the complexes on the iron centers, with a linear geometry (  Table 2, with all of the values being between 43 and 51 cm −1 , with one outlier: the complex of one CO molecule with  CO adsorption on MIL-100(Fe) has been investigated by means of IR spectroscopy in previous works. 4,5,14 The experimental studies have discussed only the changes observed in the CO stretching frequency region. Additional bands associated with vibrational modes involving CO (e.g., the bending mode of Fe···CO) are expected in the spectral region below 800 cm −1 . The description of theoretical spectra in this range is reported in Figure S1. Yoon et al. 4 and Wuttke et al. 14 have studied it at room temperature in a flow of 10% CO in helium (Table 3) to determine how the CO surface species formed at a certain CO partial pressure change with the treatment temperature. Leclerc et al. 5 have investigated how the CO bands change with the coverage up to CO condensation: this allowed them to characterize all of the adsorption sites present on the MIL surface. Leclerc et al. 5 have recorded the spectra on MIL-100(Fe)-150C and MIL-100(Fe)-250C at −173°C in static conditions and by increasing the pressure up to 0.53 mbar ( Table 3).
The CO spectra reported in refs 4 and 14 show three bands, whose intensity changes with the treatment temperature: a band at 2189 cm −1 is present also after treatments at T < 150°C, associated with CO on Fe(III) sites, and two bands at 2182 (or 2185) and 2173 cm −1 gain significant intensity only after the reduction of Fe(III) to Fe(II). The intensity of the signals is   (Table 3).

Inorganic Chemistry pubs.acs.org/IC
Forum Article the spectral regions typical of −OH stretching frequencies and of carboxylate absorption. 5 The set of spectra recorded at −173°C is used to follow CO adsorption up to the filling of all of the open iron sites (Table  3). 5 The description of the bands for the intermediate coverage is the same as that reported at room temperature: the main difference is associated with the position of the bands, shifted of about −6 cm −1 as an effect of the temperature. 44 ) and/or a water molecule (∼45 cm −1 ). These shifts are close to those observed for the spectra recorded at the lowest coverage on samples degassed at different T values (Table 3). In particular, the models are able to reproduce the ΔvC O values for both Fe(II)···CO and Fe(III)···CO complexes observed experimentally (35 and 50 cm −1 , respectively). 5 The results obtained for Fe 3 O-Cl and Fe 3 O-OH are fully comparable, suggesting that the IR spectra of CO cannot help to distinguish between clusters with different X. The treatment protocol and the temperature (T IR ) and pressure (P IR ) conditions used in the IR measurements are reported. For each frequency, the assignment reported in the original paper is also shown. Reference values for CO in the gas phase in a microporous matrix used for the calculation of the CO stretching frequency shift (  For intermediate coverage, the experimental spectra show the presence of a doublet, where only the relative intensity of the peaks is dependent on T, while the position of the peaks is independent. The two peaks have been assigned to Fe(II)···CO (ΔvC O = 31 cm −1 ) and Fe(III)···CO (41 cm −1 ). 5 The calculations suggest an alternative assignment for the higherfrequency band of the doublet. When two CO molecules are adsorbed on the same metal node, each vibrational mode in the CO spectral region is associated with the combination of the modes of the two CO molecules, that is, to the asymmetric or the symmetric stretching of the two CO molecules. For 3CO/Fe 3 O, the three modes are associated with the symmetric stretching of all of the CO molecules (ΔvC O = 46 cm −1 ), the asymmetric C− O stretching of the two Fe···CO moieties (47 cm −1 ), and the C− O stretching of Fe···CO (48 cm −1 ), respectively. The predicted shift for modes associated with the 2CO and 3CO complexes is 45 cm −1 , independent of the metal node. This value is very close to the higher-frequency peak of the doublet (41 cm −1 ) that we assign, based also on the discussion above, to the formation of 2CO complexes on [Fe II Fe III 2 (μ 3 Table 4). The calculated enthalpy of NO adsorption, ΔH NO c , is strongly   (Table 4 and Figure 4d), unlike CO, for which ΔH CO c is independent of the coverage and only slightly decreasing with the reduction of the metal node (Table 2 and Figure 3c). ΔH NO c for 1NO complexes spans a range from −108 to −94 kJ mol −1 on the reduced clusters, while for the oxidized cluster, it is halved (−64 kJ mol −1 for Fe 3 O-Cl and −51 kJ mol −1 for Fe 3 O-Cl· 1H 2 O). The adsorption of a second NO is by far less exothermic than the first one, with ΔH NO c being −50 kJ mol −1 for the reduced clusters and −20 kJ mol −1 for the oxidized ones. Although no experimental values are reported for the energetics of NO adsorption, Eubank et al. 24 have studied the competitive adsorption of water in an NO-loaded MIL-100(Fe)-250C(Fe) using IR spectroscopy. They verified by using IR spectroscopy that NO is not fully released at room temperature after switching from a dry to a wet helium flow. Our calculations agree with these observations: the calculated adsorption Gibbs free energy for Fe 3 O is lower than that calculated for water for the first NO (Tables 1 and 4), while it is higher than water for the second NO. The calculations also indicate that ΔG NO c for Fe 3 O-Cl is always higher than ΔG H 2 O c . These results are an example of the importance of the temperature used in the postsynthetic treatment. For drug delivery, an Fe 3 O-based degassed at ≤150°C before to be loaded with NO will fully release NO after contact with a water-rich medium, while a MOF treated at ≥150°C would allow a gradual release in time of NO, with a large amount of NO delivered immediately, followed by a slow desorption due to the gradual substitution of NO by water. MOFs showing a larger affinity for NO than for H 2 O have also been suggested for the environmental removal of NO. 47 The treatment temperature will have an important effect also on NO capture: Fe 3 O-based MOFs treated at ≤150°C have ΔG NO c > ΔG H2O c , and then they will not be able to capture NO in wet streams. The smaller ΔG NO c than ΔG H 2 O c of Fe 3 O-based MOFs treated at ≥150°C would allow one to maintain good selectivity for NO also in wet gas streams, although their capacity will be limited to 1NO molecule per reduced Fe 3 O node.
The calculated ΔvÑ O is strongly dependent on the oxidation state of iron (Table 4 and Figure 4b,d). For a description of the spectra in the region below 800 cm −1 , see the Supporting Information and Figure S2. For 1NO complexes, ΔvÑ O spans from −57 to −90 cm −1 if adsorbed on a reduced cluster, while it is +5 cm −1 on oxidized clusters. A vibrational shift smaller than −70 cm −1 is in general associated with a bent geometry, that is, with Fe−N−O angles (∠Fe−N−O) significantly different from 180°. 19,50 The calculations can model this behavior (Figure 4a and Table S3). When the loading (2NO and 3NO complexes) is increased, a different evolution of the bands is predicted for the oxidized and reduced clusters. For the reduced ones, two bands are predicted: one band almost corresponds with the gas-phase value, and a second one is shifted to lower wavenumbers with respect to the 1NO complex. For 2NO/Fe 3 O-Cl, two spin states are calculated to be isoenergetic: 2S + 1 = 16 and 12 (Table 4). These spin states have different vibrational shifts, a doublet at 5 and 24 cm −1 for 2S + 1 = 16 and a doublet at 5 and −32 cm −1 for 2S + 1 = 12. For the reduced clusters, each vibrational mode is associated mainly with the stretching of a single NO molecule in 2NO complexes (Table 4), unlike for the Fe 3 O-Cl-based   refs 14, 24, 48, and 49). Their results are summarized in Table 5. The experimental ΔvÑ O assigned to NO···Fe(II) complexes in MIL-100(Fe)-250C agrees with the value calculated for 1NO/Fe 3 O: −55 versus −57 cm −1 , respectively. The shift calculated for the adsorption of the first NO molecule on the oxidized cluster is very small (3−5 cm −1 ), that is, almost indistinguishable from the gas-phase value. The signal of Fe(III)···NO complexes is associated with bands at ∼1895 cm −1 (+20 cm −1 ), a shift larger than the one predicted in the calculations. Nevertheless, there is a contradiction between the experimental results reported for IR and volumetric experiments. IR experiments either failed to detect NO adsorption on fully oxidized materials [e.g., MIL-100(Fe)-150C] at room temperature or obtained signals with very small intensity at ∼1895 cm −1 , 14,24 a symptom of a small interaction energy of NO with Fe(III) sites. Volumetric measurements indicate a large NO adsorption on the same materials under the same conditions. 24 Moreover, the volumetric measurements showed that NO can only be partially desorbed, indicating a strong interaction with the material that cannot be explained only by the presence of 2% Fe(II) sites. 24 The results reported in Table 4 explain this apparent contradiction. The Q band of NO in the gas phase is present in the experimental IR spectrum. The detection of bands slightly shifted from the gas-phase value is then difficult because these bands can be hidden behind the gasphase absorption. The signals associated with 1NO/Fe 3 O-Cl complexes, if not too intense, can be confused with the gas-phase The treatment protocol and the temperature (T IR ) and pressure (P IR ) conditions at which the band appears during the IR measurement are reported. For each frequency, the assignment reported in the original paper is also shown. Reference value for NO in the gas phase: 1876 cm −1 . Frequencies (cm −1 ), temperatures (°C), and pressures (mbar). b (w) = weak, (s) = shoulder, (l) = large. c 1% NO in a helium flow. No experimental spectra for the full coverage of iron sites by NO molecules have been reported, and they cannot thus be used to benchmark the evolution of the spectra predicted by the calculations. These results suggest that the NO spectra on MOFs treated at ≥150°C are composed of three families of bands associated with 3NO/Fe 3 O and 2NO/Fe 3 O-Cl complexes ( Figure 4b). Unlike CO, characterized by a single broad band at higher CO/Fe coverage, NO can differentiate the different Fe 3 O nodes even at NO/Fe ∼ 1 and is thus a more suitable molecular probe to verify the efficacy of thermal treatments of Fe 3 O-based MOFs.

CONCLUSIONS
Postsynthesis thermal treatments are effective ways to modify the performance of Fe 3 O-based MOFs in most applications. We have used DFT to study the CO and NO adsorption on metal nodes of Fe 3 O-based MOFs, subject to thermal treatments of different efficacy. The calculations allowed us to characterize how the adsorption of small molecules on Fe 3 O-based clusters evolves with the coverage and how the desorption of chemisorbed species (water molecules and counteranions) affects the interaction of the clusters with adsorbates. We compared the simulated IR bands of CO and NO with experimental spectra reported in the literature. The calculations reproduce the changes observed in the spectra. NO showed a larger sensitivity to the presence of adsorbed species than CO at all coverages, and it is then a more suitable molecular probe for quick quality control checks. On the basis of the calculations, we propose to reassign some of the bands previously inaccurately assigned because of the absence of reference data on systems with a structure close to the Fe 3 O structure. Several experimental bands were formerly associated with a large concentration of defects. These bands are here reassigned by considering only crystallographic sites. These findings help in changing the common belief that MILs are highly defective materials and are useful for a more precise assignment of the CO and NO bands in iron-based MOFs, confirming the importance of the synergy between theory and experiments.
The calculated enthalpy of adsorption for both CO and NO was also assessed using the experimental data and compared with the enthalpy of adsorption of water, present often in the different applications of Fe 3 O-based MOFs, as a contaminant or as a solvent. The importance of the interaction of CO and NO with Fe 3 O-based MOFs plays a role in important MOF applications like drug delivery 1,4,11,12,21−23 and gas mixture purification. 19,4,20 Future studies should be aimed at enlarging the set of theoretical IR spectra of adsorbates on Fe 3 O-based MOFs, including common probe molecules such as pyridine. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.