In Silico Tuning of the Pore Surface Functionality in Al-MOFs for Trace CH3I Capture.

Aluminum (Al)-based metal-organic frameworks (MOFs) have been shown to have good stability toward γ irradiation, making them promising candidates for durable adsorbents for capturing volatile radioactive nuclides. In this work, we studied a series of existing Al-MOFs to capture trace radioactive organic iodide (ROI) from a gas composition (100 ppm CH3I, 400 ppm CO2, 21% O2, and 78% N2) resembling the off-gas composition from reprocessing the used nuclear fuel using Grand canonical Monte Carlo (GCMC) simulations and density functional theory (DFT) calculations. Based on the results and understanding established from studying the existing Al-MOFs, we proceed by functionalizing the top-performing CAU-11 with different functional groups to propose better MOFs for ROI capture. Our study suggests that extraordinary ROI adsorption and separation capability could be realized by -SO3H functionalization in CAU-11. It was mainly owing to the joint effect of the enhanced pore surface polarity arising from -SO3H functionalization and the μ-OH group of CAU-11.

Introduction

Nuclear energy is promising to be one of the leading emission-free energy source to power the global economy.[1,2] During reprocessing the used nuclear fuel or tackling nuclear accidents,[3] the removal of the harmful volatile radioactive nuclides, 129I, 131I with their molecular forms iodine (I2), and radioactive organic iodides (ROI), is essential to ensure the safe use of the nuclear energy.[4] Compared with iodine, CH3I, which is the main kind of the ROI species, is known to be much more difficult to be separated and immobilized because of its comparatively inert chemical reactivity.[5] Existing methods for ROI retention and separation based on adsorption include immobilizing gaseous iodine onto AgZ,[6] activated charcoals,[7] and amine-impregnated activated carbon.[8,9] However, high regeneration costs and poor separation efficiency have limited these adsorbents from having wide applications.[10]

As a novel class of functional porous crystalline solids that are assembled using “Molecular Lego”-like organic/inorganic blocks, metal–organic frameworks (MOFs) have been receiving growing interest for ROI capture in the past few years. Unlike conventional porous adsorbents, MOFs feature their tunable pore geometry, precise surface functionality control, large surface areas, and the ease for postsynthetic functionalization, which render them as a new generation of functional materials for adsorption and separation applications. Recently, Chebbi et al.[11] investigated a series of MOFs for CH3I capture at 35 °C, in which HKUST-1(Cu) had the highest saturated CH3I uptake (425 mg/g). Li et al.[12,13] suggested that a total CH3I uptake of 80 wt % could be reached under 150 °C by N,N′-dimethyl ethylenediamine (DMEDA)-grafted MIL-101(Cr) due to the strong ionic coupling of CH3I with tertiary nitrogen of an amine group forming ionic (R3N–CH3)+I–. Lan et al.[14] computationally screened 187 covalent organic frameworks COFs (which are MOFs’ analogue but the main composition is light element linked by covalent bonds) for I2 and CH3I capture. They found that 3D-Py-COFs with a larger accessible surface area or a void fraction have the best I2 uptake and COF-103 with a pore size of 9 Å is identified as the best-performing material for CH3I capture, respectively. The durability of MOFs has always been a concern when MOFs are employed in practical applications.[15,16]

Volkringer et al. performed a study of the structural stability of a series of MOFs under γ irradiation, where Al-based MOFs were demonstrated to be able to remain intact under a γ irradiation with doses up to 1.75 mGy, whereas the other tested transition-metal-based MOF materials, such as HKUST-1(Cu) and UiO-66(Zr), were destructed under such irradiation. The radiation stability of the Al-based MOFs was attributed to the smaller γ-ray absorption cross section of aluminum (2.5 barns per atom), compared to those of the transition metals, copper, zinc, and zirconium (>5.3 barns per atom); a smaller cross section implies a lower γ ray absorption and hence higher robustness of the MOF under γ irradiation.[17] Inspired by Volkringer et al.’s study, we envisage that other Al-based MOFs might also exhibit a high resistance to γ irradiation, which is a desirable property of the adsorbent materials for capturing volatile ROIs in an irradiation environment. However, we did not attempt any MOF synthesis or stability test under γ irradiation in this study, but we offer our predictions as experimental targets for the future.

Herein, we investigated a series of Al-based MOFs: BasoliteA520,[18] CAU-3-BDC,[19] CAU-3-NDC,[19] CAU-11,[20] DUT-5-bpdc,[21] DUT-5-ndc,[21] NOTT-300,[22] and CAU-8-ODB[23] for capturing trace CH3I resembling the used nuclear fuel reprocessing conditions (Figure ). All of the selected Al-MOFs were previously reported to have good thermal and chemical stability. Six out of the total nine Al-MOFs structures feature 1D metal-oxide chain except for 2D MOF CAU-8-ODB and 3D MOF CAU-3-NDC and CAU-3-BDC, which possess a similar rhombohedral system while differ in secondary building edges (BDC2– and NDC2–, respectively). Unlike other Al-MOFs, which contain a μ-OH group, O-CH3I in CAU-3 series bridges the Al3+ vertices. GCMC simulations and DFT calculations were employed to study the adsorption and separation performance of CH3I in the MOFs. Our modeling results suggested that CAU-11 possessed both the top-performing CH3I adsorption capacity and selectivity. Based on the experimental fact that CAU-11 has been successfully functionalized through a postsynthetic approach without harming the parent framework,[25] we propose a series of hypothetic CAU-11 derivatives (namely, CAU-11-X) by grafting a series of functional groups on to the SDBA linker of CAU-11 to further promote the separation performance of trace CH3I under a realistic condition. This study of fine-tuning the surface functionalities and pore geometry will aid in the novel design of low-cost and stable materials for effective ROI capture for industrial applications. To our best knowledge, this is the first systematic investigation of trace CH3I capture using Al-based radiation-resistant MOFs.

Figure 1

Space-filling model of the selected Al-MOFs in this work (color scheme: Al, pink; C, brown; O, red; and H, white. The model of CAU-8-ODB can be found in the Supporting Information (Figure S1)).

Results and Discussion

Table presents the geometric properties together with separation performance of the selected Al-MOFs. We noted that the theoretically calculated surface area and porosity of Al-MOFs are slightly higher than those reported from experiments. This could be attributed to imperfect crystalline or incomplete activation during MOF synthesis.[25] With the abundant presence of CO2, O2, and N2, the adsorption of dilute 100 ppm CH3I in most of the selected MOFs has been difficult due to the competing nature of the other gas molecules against CH3I. Nevertheless, there are a few well-performing MOFs with high CH3I uptake, which possess relatively narrow pore channels. CAU-11, BasoliteA520, and NOTT-300, which possess relatively small largest cavity diameters (LCD) (5.89, 5.82, and 5.72 Å, respectively) enjoy high selective capture toward CH3I (11.18, 2.24, and 1.19 cm3/g, respectively). Such adsorption behavior that the separation performance could be enhanced by tuned pore size was also observed in studies of acid gas removal[26] and purification of natural gas using MOFs.[27]

Table 1

Selected Al-MOF Properties and Separation Performance under Simulated Conditions

MOF	ligand molecules	uptake (cm3/g)	selectivity	LCDa (Å)	SAsimb (m2/g)	SAexp (m2/g)	Vp-simc (cm3/g)	Vp-exp (cm3/g)
BasoliteA520[18]	FAd	2.24	8.18 × 103	5.82	1174.66	1025	0.56	0.47
CAU-3-BDC[19]	H2BDCe	0.07	4.06 × 102	11.07	2389.16	1920	0.88	0.64
MIL-53[24]	H2BDC	0.52	2.19 × 103	6.97	1410.96	1140	0.61	0.68
CAU-3-NDC[19]	H2NDCf	0.04	2.02 × 102	13.81	3129.4	2750	1.17	0.95
CAU-11[20]	H2SDBAg	11.18	6.64 × 104	5.89	469.78	350	0.31	0.17
DUT-5-bpdc[21]	H2bpdch	0.04	1.93 × 102	11.33	2376.25	2335	1.08	0.81
DUT-5-ndc[21]	H2NDC	0.09	4.41 × 102	8.90	2067.21	1996	0.83	0.68
NOTT-300[22]	H4Li	1.19	3.75 × 103	5.72	1326.7	1370	0.54	0.38
CAU-8-ODB[23]	H2ODBj	0.36	1.91 × 103	6.74	1401.78	1004	0.67	0.47

Largest cavity diameter.

Geometric surface area.

Pore volume.

Fumaric acid.

Terephthalic acid.

1,4-Naphthalenedicarboxylic acid.

4,4′-Sulfonyldibenzoic acid.

4,4′-Biphenyldicarboxylic acid.

Biphenyl-3,3′,5,5′-tetracarboxylic acid.

4,4′-Oxydibenzoic acid.

To further reveal the correlation between MOF characteristics and the CH3I separation capability, structure–property relationships were mapped for the selected nine Al-MOFs; as shown in Figure a, the CH3I gravimetric uptake increases as density increases. However, a nonlinear correlation could be observed. Figure b–d shows that there is a sharp drop of CH3I uptake when LCD, the surface area, and the pore volume of the MOFs are increased to around 6.5 Å, 1000 m2/g, and 0.5 cm3/g, respectively. Above all, extraordinary CH3I capture performance (11.18 cm3/g uptake and 6.64 × 104 selectivity) could be found in CAU-11 owing to compact interaction with CH3I provided by a constricted porosity (surface area: 469.78 m2/g and pore volume: 0.31 cm3/g), followed by BasoliteA520 (surface area: 469.78 m2/g and pore volume: 0.31 cm3/g) and NOTT-300 (surface area: 469.78 m2/g and pore volume: 0.31 cm3/g).

Figure 2

Competitive CH3I uptake structure–property relationship of the selected Al-MOFs as a function of (a) framework density, (b) largest cavity diameter Å, (c) surface area m2/g, and (d) pore volume cm3/g (dashed line represents the fitting curve of the scatter trend).

Figure a presents the asymmetric edge unit of CAU-11, which forms the lozenge-shaped narrow channel geometry. Interconnected by an SDBA linker, Al3+ vertices were bridged through μ-OH groups.[25] Compared with NOTT-300 (Figure b) where only the μ-OH group contributes to an electrostatic field, the −SO2 vertex of the SDBA ligand bridging two phenyl rings along with the μ-OH group also provides a polarized channel surface enhancing the electrostatic potential field. As indicated from the DDEC calculated atomic charges on the asymmetric unit of CAU-11 (Figure a), the sulfur atom formulating the V shape of the SDBA linker possess an atomic charge of 0.978e while the two neighboring oxygen atoms are charged around −0.631e. The hydrogen and oxygen atoms of the μ-OH group possess point charges of 0.465e and −1.122e, respectively. With the aid of 1-D narrow pore channels, the polarized pore surface provides the overlapped electrostatic interaction with CH3I to enhance its uptake in CAU-11. As shown in Figure c, the isosteric heat of CH3I in CAU-11 outperforms all other selected MOFs. Such tuned polarity of the channel surface in MOFs was studied to play vital roles in CO2 storage and capture.[28,29]

Figure 3

(a, b) Schematic illustrations of the asymmetric unit of CAU-11 and NOTT-300 with the atomic point charge (color scheme: C, gray; H, white; O, red; Al, pink; and S, gold). (c) Isosteric heat of adsorption of the adsorbates at infinite dilution in the Al-MOFs.

As shown in Figure a, the phenol rings of SDBA ligands in CAU-11 have good coverage of the pore surface in CAU-11, which are the least polarized part of the ligands. We further grafted eight different types of functional groups in silico varying between polarities and size onto a phenol ring of the SDBA ligand to study to effect of pore size and tuned functionality onto the selective adsorption nature toward trace CH3I. Potential synthetic routes for the grafted ligands (Figure b) are proposed in Section S10 of the Supporting Information. The −SO3H-functionalized ligand, diphenylsulfone-3,3′-disulfo-4,4′-dicarboxylate, has been synthesized and used to construct MOFs before.[30] More generally, diarylsulfones show good versatility in incorporating chemical functionalities.[31] However, we envisage that some syntheses might be elaborate and challenging.

Figure 4

(a) Space-filling model of the SDBA linker. (b) Asymmetric unit of SDBA grafted by different functional group designed in this work (SO2R refers to the other symmetric part of the SDBA-X ligand, X = functional group; CH3I-inaccessible CAU-11-X were excluded).

The hypothetical structures with grafting modification were optimized using DFT calculations as presented in the methodology section. To ensure the accessibility of CH3I into the hypothetical functionalized CAU-11-X series, MOFs whose pore limiting diameter (PLD) are smaller than 4.23 Å (the smallest Van der Waals diameter of CH3I) are treated as inaccessible MOFs and excluded; a resulting eight CAU-11-X MOFs were obtained for further investigation (Figure b). Theoretical characterized porosity of the hypothetically designed MOFs and models of CH3I-inaccessible CAU-11-X series are provided in the Section S7 of the Supporting Information. The adsorption isotherms of the four gases (CH3I, CO2, O2, and N2) up to 1 bar in the CAU-11-X MOFs were simulated by GCMC simulations and are given in Figure S22. Compared with CO2, O2, and N2, CH3I features a relatively larger molecular size, as shown in Figure S19, which contributes to an enhanced overlapped interaction between the gas molecules and the pore surface. As shown in Figure S22a, CH3I adsorption isotherms of functionalized CAU-11-X series feature type I micropore filling adsorption mechanism, where the uptake can be saturated at relatively low pressure.

To elucidate the trace CH3I separation performance of CAU-11-X series MOFs, we further studied the trace CH3I separation performance under resembled nuclear industrial off-gas composition as the same conditions in the investigations of preliminary Al-MOFs. It appears that merely judging the CH3I separation performance from adsorption isotherms up to 1 bar has provided false impression that postsynthetic modification is not working. As illustrated in Table , CAU-11-SO3H gives the best performance in selectively capturing CH3I by GCMC simulations (19.7 cm3/g and 5.14 × 105), which is much higher than that of pristine CAU-11. The functionalization of CAU-11 has provided higher isosteric heat of adsorption (Qst) of CH3I with the MOFs with the exception of −CH3 functionalization, which enabled higher CH3I uptake, as shown in Figure a. The overall Qst ranking is −SO3H > −CN > −NH2 > −NO2 > −Cl > −Br > −F > −H > −CH3, which fits well the CH3I uptake ranking. The Qst ranking from the GCMC simulations agrees well with the binding affinities obtained from DFT calculations, which confirms the validity of the GCMC simulation results. Moreover, as shown in Figure b, CH3I selectivity ranking in the MOFs is positively correlated with the binding affinity differences between CH3I and CO2, N2, and O2. In general, CO2 poses stronger adsorption competitiveness than N2 and O2 during the separation of CH3I.

Table 2

CH3I Separation Performance of the Functionalized CAU-11-X and Adsorption Affinity of the Adsorbates

CAU-11-X	CH3I uptake (cm3/g)	selectivity	ECH3Ia (kJ/mol)	ECO2 (kJ/mol)	EN2 (kJ/mol)	EO2 (kJ/mol)
–SO3H	19.7	5.14 × 105	91.29	47.09	28.35	26.91
–Br	9.84	9.37 × 104	77.1	35.91	24.23	23.5
–CH3	7.75	7.18 × 104	72.44	33.45	21.92	21.82
–CN	14.6	1.35 × 105	79.75	40.61	25.61	24.3
–NO2	12.11	9.23 × 104	78.44	37.59	24.8	24.04
–NH2	12.7	1.03 × 105	79.27	36.87	24.72	23.89
–Cl	11.95	9.58 × 104	77.34	36.03	24.12	23.38
–F	8.54	5.81 × 104	75.32	38.31	24.86	24.32
–H	11.18	6.64 × 104	71.9	32.85	21.87	21.04

DFT-derived binding enthalpy.

Figure 5

(a) Correlations between adsorption affinity and adsorption capacity of CH3I in CAU-11-X series (blue bars stand for isosteric heat of adsorption calculated from GCMC simulation; orange bars stand for DFT-calculated adsorption enthalpy; and black dots stand for selectively adsorption capacity of CH3I). (b) Correlations between adsorption enthalpy difference of CH3I against other adsorbates and adsorption selectivity in CAU-11-X series (blue bars stand for adsorption enthalpy difference between CH3I and CO2; orange bars stand for adsorption enthalpy difference between CH3I and N2; gray bars stand for adsorption enthalpy difference between CH3I and O2; DFT-calculated adsorption enthalpy; and black dots stand for selectively adsorption selectivity of CH3I).

The adsorption enthalpy of CH3I in CAU-11-SO3H, at a close-to-saturation loading of 23.38 cc/g, was 85 kJ/mol, which is high for physisorption and could indicate difficulty in desorption. To evaluate the possibility of releasing CH3I from CAU-11-SO3H, we compared the zero-coverage isosteric heats of adsorption (Qst) of CH3I in CAU-11-SO3H and HKUST-1(Cu), using the Widom particle insertion method. The zero-coverage Qst is a direct measure of the adsorbent–adsorbate interaction strengths, which was calculated to be 77 and 67 kJ/mol for CH3I in CAU-11-SO3H and HKUST-1(Cu), respectively. In our calculations, any strong adsorbate interaction with the open metal sites of HKUST-1(Cu) was not captured beyond what was described by the used force field model for HKUST-1(Cu), which was a combination of UFF and DDEC charges. Therefore, the more accurate Qst for HKUST-1(Cu) would be much higher than 67 kJ/mol should the strong adsorbate interactions with the open metal sites be included. Indeed, an appreciable percentage (5%) of adsorbed CH3I in HKUST-1(Cu) was found to be from chemisorption; nonetheless, it was also shown that HKUST-1(Cu) exhibited a good adsorption/desorption cyclability.[11] Based on the relative adsorption strengths of CH3I in CAU-11-SO3H and HKUST-1(Cu), we envisage that the release of CH3I from CAU-11-SO3H might be possible.

Overall, we would like to propose CAU-11-SO3H to be the best-performing trace CH3I separation material, compared to other functionalization modifications, as it possesses the best CH3I uptake and selectivity against (N2, O2, and CO2) (19.7 cm3/g and 5.14 × 105). However, the predicted adsorption capacity is not as high as the capacity measured for HKUST-1(Cu) by Chebbi et al.,[11] where breakthrough experiments were employed to investigate the dynamic adsorption properties of CH3I in MOFs; see Table S19 for a comparison between CAU-11-SO3H and selected MOFs from the literature. Different from breakthrough experiments, our GCMC simulations predict adsorption uptakes for a perfect crystalline structure, at a certain temperature and pressure, under thermodynamic equilibrium. By contrast, the dynamic adsorption in breakthrough experiments is additionally influenced by mesoscale factors, such as the height to the diameter ratio of the adsorbent bed and the flow conditions of the feed. It is also worth noting that the breakthrough experiments for HKUST-1(Cu) were conducted for 1333 ppm CH3I with argon as a carrier gas, whereas we simulated for trace (100 ppm) CH3I capture from a gas mixture also containing CO2, N2, and O2.

To gain a molecular level of understanding of the competitive CH3I adsorption behaviors in CAU-11-SO3H, which proved to be the best adsorbent candidate in this study, a detailed investigation of adsorption sites and geometric positions of the adsorbates were carried out. Figure a presents the GCMC-simulated adsorption density plots of CH3I molecules in CAU-11-SO3H. CH3I molecules are mainly located at the center of the pore channel. Due to the uniformly packed 1-D channel chain build up by SDBA-SO3H and the induced joint effect by two polar centers provided by μ-OH groups and −SO2, the density contour was lozenge shaped and seated parallel against the backbones of the framework. The DFT-optimized CH3I binding position agrees well with the GCMC simulations results. It further reveals the sitting orientation of CH3I in the channel. As shown in Figure b, the DFT-optimized CH3I molecule is located in the center of the channel, where close distances of I···O and I···S are found to be 4.62 and 4.91 Å, respectively, indicating the favorable electrostatic interaction between the atom pairs. As with a less electron-dense site of the C center, the distances of C···S and C···O are 5.11 and 6.35 Å, respectively. The radial distribution analysis reveals that these atom pairs (I···O, I···S, C···S, and I···S) show clear regular peaks, indicating the regular sitting of CH3I molecules along the pore channel of CAU-11-SO3H. Such clear well-defined peaks are missing for N2, O2, and CO2 adsorption in CAU-11-SO3H, indicating their random sitting in the MOF.

Figure 6

(a) Adsorption density plot pictures of CH3I adsorbed in CAU-11-SO3H during competitive adsorption. (b) DFT-optimized geometric positions of CH3I in the CAU-11-SO3H pore channel (color scheme: C, gray (brown in the density picture); H, white; O, red; Al, pink; and S, yellow).

Figure 7

(a) Radial distribution functions of the O atoms of the SO3H groups of CAU-11-SO3H pairing various atoms of the adsorbates. (b) Radial distribution functions of the S atoms of the SO3H groups of CAU-11-SO3H pairing various atoms of the adsorbates (the C and I atoms were chosen for CH3I).

Conclusions

To sum up, we have identified a promising Al-based MOF, CAU-11, for trace CH3I capture, based on the combined GCMC- and DFT-calculated results. We demonstrated that the trace CH3I separation performance could be enhanced by tuning the functionalization of the pore surface of CAU-11. The best-performing hypothetical CAU-11-SO3H was proposed to possess both high selectivity and high adsorption capacity (19.7 cm3/g and 5.14 × 105) in trace CH3I capture, owing to the synergistic effect of the highly polarized functional group and good pore confinement. Since Al-based MOFs have been demonstrated to be highly radiation-resistant compared with MOF structures based on other metal types,[17] our work will provide good guidance on the design of materials that could be well applied in nuclear waste management.

Computational Details

Grand Canonical Monte Carlo Simulations

The RASPA package[32] was used to carry out the GCMC simulations of gas molecules’ adsorption and separation in the MOFs. A typical GCMC simulation consists of an equilibrium run of 2 × 106 Monte Carlo (MC) cycles and a production run of 8 × 106 MC cycles. All of the MOF structures were kept rigid with periodic boundary conditions applied where the unit cell numbers were adjusted to ensure the perpendicular cell width to be longer than 24.0 Å (twice as the cutoff distance of 12.0 Å). The simulated gas composition was set to 100 ppm CH3I, 400 ppm CO2, 78% N2, and 21% O2 under 423 K, 1 bar. Such a system is intended to simulate a gaseous mixture released from the reprocessing of the spent nuclear fuel.[4,11] The selectivity of CH3I in the simulated system is calculated by the following equation, where XA and XB are the mole fractions of targeted gas CH3I and other gas components in the adsorbed phase, and YA and YB are the mole fractions of targeted gas CH3I and other gas components in the bulk phase.

The structural properties, such as surface areas and porosity of the MOFs were characterized using the Zeo++ package.[33] To accurately address the electrostatic interaction, atomic charges of the structures were computed using the high-quality density-derived electrostatic charges (DDECs) method.[34] Lennard-Jones parameters of the adsorbents were taken from the universal force field.[35] Lorentz–Berthelot mixing rules were implemented to address the nonbonded interaction (Table ).

Table 3

Computational Model Parameters Governing Nonbonded Interactions

	atom label	ε/kb (K)	σ (Å)
MOFs	Al	254.09	4.01
	C	52.83	3.43
	O	30.19	3.12
	H	51.22	3.40
	S	22.14	2.57
	N	34.72	3.26
	F	25.16	3
	Br	126.29	3.73
	Cl	114.21	3.52
adsorbates[36−38]	nitrogen_N	36.0	3.31
	methyl iodide_H	10.01	2.2.0
	methyl iodide_C	51.22	3.40
	methyl iodide_I	324.06	4.12
	carbon dioxide_O	79.0	3.05
	carbon dioxide_C	27.0	2.80
	oxygen_O	49.0	3.02

Density Function Theory Calculations

DFT calculations were carried out using VASP[39−41] (Vienna Ab initio Simulation Package) with the plane-wave pseudopotential formalism. Geometric optimization of the MOFs and their binding complexes with the gas molecules were performed with the Perdew–Burke–Ernzerhof exchange–correlation functional.[42] DFT-D3(BJ) dispersion[43,44] correction was implemented to address the noncovalent bond interaction and the plane-wave cutoff energy was set to 500 eV.

The gas–MOF binding energies were calculated by the following equationwhere Ebind is the zero-point energy of the optimized binding complex, Ehost is the zero-point energy of the MOF cell, and Eguest is the zero-point energy of the guest gas molecules. During the DFT calculations, multiple initial positions of gas molecules were attempted to search for the global minimum interaction position of the gas molecules in the MOFs.