Computational Exploration of IRMOFs for Xenon Separation from Air.

Metal-organic frameworks (MOFs) found their well-deserved position in the field of gas adsorption and separation because of their unique properties. The separation of xenon from different gas mixtures containing this valuable and essential noble gas is also benefited from the exciting nature of MOFs. In this research, we chose a series of isoreticular MOFs as our study models to apply advanced molecular simulation techniques in the context of xenon separation from air. We investigated the separation performance of our model set through simulation of ternary gas adsorption isotherms and consequent calculation of separation performance descriptors, finding out that IRMOF-7 shows better recovering capabilities compared to the other studied MOFs. We benefited from visualization of xenon energy landscape within MOFs to obtain valuable information on possible reasoning behind our observations. We also examined temperature-based separation performance boosting strategy. Additionally, we noted that although promising candidates are present among the studied MOFs for xenon recovery from air, they are not suitable for xenon recovery from exhaled anesthetic gas mixture.

Introduction

Since their emergence,[1,2] metal–organic frameworks (MOFs), an exciting class of advanced materials, have attracted enormous attention among academic and industrial sections because of their unique properties such as high surface area, modularity in synthesis, and tunable pore structure,[3] which consequently make them promising for a wide range of applications including not only in gas storage/separation[4−6] but also in catalysis,[7,8] drug delivery,[9,10] hazardous compound elimination,[11] and electronic devices.[12] Although reticular chemistry behind the design of MOFs provides their interesting aspect, which is the possibility of constructing thousands of MOFs, it makes their study for a specific application challenging. In this regard, one could either benefit from in silico high-throughput screening to find suitable MOFs for desired applications such as methane[13]/hydrogen storage[14] and carbon dioxide separation[15−19] or select an established set of MOFs sharing common features and investigate them to obtain information on the molecular level to use them as a starting point for future computational and experimental studies.[20−28] In this research, we benefit from the latter approach to accomplish our objectives.

The xenon (Xe) separation from gas mixtures containing this expensive gas using MOFs is one of the exciting and relevant topics that was the subject of several accomplished studies[29] to provide Xe for several medical[30−32] and characterization[33−36] usages. However, as stated in an earlier publication by our group,[20] the worldwide usage of xenon is limited because of the high prices arising from energy-intensive cryogenic distillation, which is the dominant procedure for xenon production. An excellent alternative for the development of the xenon production units based on pressure swing adsorption (PSA) or membrane-based separation is MOFs, which have been covered by interesting results published by several research groups that mainly deal with the xenon separation from the Xe/Kr mixture.[37−42] For instance, Anderson et al. recently published an interesting paper[43] on the application of molecular simulations to the membrane-based Kr/Xe separation, where they quantified the effect of temperature on the separation process and revealed that a lower temperature improves the Kr/Xe separation performance of studied membranes. Thallapally and his co-workers recently published a comprehensive review paper[29] that covers nearly all related publications dealing with this issue. However, another alternative resource for xenon can be xenon/air mixture, which could be very beneficial from both the economic and energy sustainability point of view.[29,44]

In this research, we have benefited from advanced molecular simulation techniques to explore a set of isoreticular MOFs for recovering xenon at room temperature. This set of IRMOFs that was first reported[45] in 2002 represents a well-established class of MOFs that fits very well to serve as a model to investigate the effect of different organic linkers on specific separation performance. The gas adsorption characteristics of IRMOF-1 to 16 are assessed by simulating ternary gas mixtures and using the results to calculate corresponding separation performance parameters. We also investigated the process by more detailed atomistic molecular simulation data including the heat of adsorption and energy surface landscapes to get deeper insight into our separation process of interest, and finally, we explored the possible boosting by an affordable alteration of the operation condition.

Results and Discussion

We calculated the pore limiting diameter (PLD) and the largest cavity diameter (LCD) of MOFs (Table S1) to select the ones that could accommodate gas mixture components on the basis of their kinetic diameters (Xe: 3.96 Å, N2: 3.64 Å, and O2: 3.46 Å), resulting in the exclusion of IRMOF-4 (PLD: 3.35 Å) and IRMOF-5 (PLD: 1.72 Å) from further investigations. To assess the separation performance, we initially simulated the gas uptake of each MOF in contact with a ternary mixture of Xe (10%), N2 (70%), and O2 (20%); then, we calculated various parameters reported in Table to quantify the xenon separation capability of studied MOFs.

Table 1

Calculated Adsorption Properties and Separation Performance Parameters at 298 K

	NXeads a	ΔNXeb	αXe/N2c	αXe/O2	ΔHadsd	APIN2 e	APIO2
IRMOF-7	4.84	4.37	10.26	9.23	18.08	2.24	1.99
IRMOF-6	3.20	2.88	8.22	7.91	16.00	1.30	1.24
IRMOF-2	2.40	2.16	6.91	6.67	15.41	0.83	0.79
IRMOF-3	2.36	2.12	6.82	6.54	14.97	0.82	0.78
IRMOF-12	1.78	1.61	6.12	5.79	14.54	0.57	0.53
IRMOF-1	1.68	1.51	5.56	5.33	13.75	0.50	0.48
IRMOF-14	1.48	1.33	5.31	5.00	13.80	0.42	0.39
IRMOF-8	1.42	1.28	5.19	4.93	13.52	0.40	0.37
IRMOF-10	0.88	0.79	3.99	3.82	12.13	0.19	0.18
IRMOF-16	0.56	0.51	3.26	3.12	11.29	0.10	0.10

Xe uptake at 1.0 bar (cm3(STP) cm–3).

Xe working capacity (cm3(STP) cm–3).

Adsorption selectivity.

Xe enthalpy of adsorption (kJ mol–1).

Adsorbent performance indicator.

The xenon uptake at adsorption pressure (NXeads) reflects the ability of each MOF to adsorb xenon from the mixture. IRMOF-7 provides the highest uptake (4.84 cm3(STP) cm–3), which is ∼10 times greater than that of IRMOF-16 (0.56 cm3(STP) cm–3). Because solid adsorbents are used in a cyclic process of adsorption and desorption, the investigation of the delivery characteristics of MOFs in the process is also needed. Therefore, we calculated the working capacity of xenon (ΔNXe) for each MOF, which is equal to (NXeads – NXedes). IRMOF-7 also provides a promising value of 4.37 cm3(STP) cm–3, which shows that its performance is almost 35% better than that of the second-ranked IRMOF-6 and almost 90% better than that of IRMOF-16. The next important parameter is the adsorption selectivity (α), which is calculated using eq from the amount of mixture components in adsorbed (x) and gas (y) phases.where i and j denote Xe and (N2 or O2), respectively. We observe that in spite of having a range of different values among studied MOFs, all are xenon-selective at an operation condition and no selectivity inversion occurs. In passing, IRMOF-7 outperforms other MOFs by showing better Xe selectivity toward both N2 and O2. It is also noteworthy that although the major component of the gas mixture is N2, all studied adsorbents show a slightly higher αXe/N compared to αXe/O. However, to have a conclusive ranking of studied MOFs for the separation process of interest, we considered using the multiparameter adsorbent performance indictor (API) developed by Wiersum et al.[46] calculated using eq .where ΔHads (kJ mol–1) is the xenon enthalpy of adsorption along with the other parameters explained earlier in the text. As we are dealing with a purification case, we assumed A, B, and C exponents from ref (46) to be the same and equal to 1. RASPA developers have implemented the fluctuation method[47] to calculate ΔHads using eq .where R, T, U, N, and brackets are the gas constant, temperature, energy, number of particles, and ensemble average, respectively. Table also gives the calculated values for ΔHads,Xe and API. We observe that IRMOF-7 is conclusively the top-performing MOF above IRMOF-6 with APIN of 2.24 and 1.99, respectively; even the former suffers from the negative effect of higher enthalpy of adsorption. We subsequently studied structures in more detail to have better insight.

In this regard, Figures and S1–S9 illustrate simulated ternary adsorption isotherms up to 1 bar. We note that IRMOF-7 shows higher xenon uptake in comparison to the nitrogen and oxygen uptake in the whole pressure range, which is consistent with the above discussion; despite having more N2 in the gas mixture, IRMOF-7 uptakes more xenon. It also reveals that between N2 and O2, the former is adsorbed more actively in all studied MOFs, suggesting that Xe/N2 separation can be considered as the challenging part rather than the Xe/O2 one.

Figure 1

Simulated adsorption isotherms of Xe (0.1), N2 (0.7), and O2 (0.2) ternary mixture on IRMOF-7 at 298 K.

Afterward, we studied the isosteric heat of adsorption (Qst = −ΔHads) of each component of the ternary mixture to obtain an overview on the observed separation performances from a different point of view (Figures and S10–S12). These plots generally exhibit a plateau trend for all studied gaseous species. This reveals that gas molecules are facing energetically homogenous adsorption surfaces within the structures, which can be related to the relatively large pores within the studied MOFs. This is also in good agreement with the ternary adsorption isotherms showing an increasing linear trend that implies that all MOFs are far from being saturated by adsorbates.

Figure 2

Calculated (a) isosteric heat of adsorption (Qst) for Xe within Xe/N2/O2 ternary mixture and (b) ΔQst (Qst(Xe) – Qst(N2)) at 298 K.

We also benefited from another descriptor, ΔQst, of Xe and N2 (Figure b) or O2 (Figure S12) to investigate its effect on the obtained trends. We observed that IRMOF-7 is completely separated from the rest of MOFs by having ΔQst of ∼7 kJ mol–1. We can suggest that having a large ΔQst can be a beneficial factor for high separation performance.

We calculated and visualized the energy surface landscapes associated with xenon and plotted them at different isocontour levels in Figures and S13–S21 to complement the previous discussion. The gray surface (0 K energy) illustrates the pore walls for xenon adsorption, and we observe that it is almost apparent within a whole unit cell of IRMOF-7 (Figure a), whereas IRMOF-16 (Figure b) lacks it because of having large pores. Two other energy surface landscapes at lower isocontour levels (green: ∼−12.5 kJ mol–1 and violet: −16.5 kJ mol–1; the latter is close to the lowest energy) reveal the possible xenon adsorption sites within the MOFs. It is noteworthy that the surface is distorted in IRMOF-7 because of the presence of 1,4-naphthalenedicarboxylic acid moieties oriented toward the pores, which results in the formation of an inhomogeneous energy surface at the corners, providing more favorable interaction sites for xenon atoms, and it can be the possible reason behind observing better performance for this particular MOF. In the other MOFs, the energy surfaces are mostly homogenous surrounding the metal nodes and corners (Figures S13–S21).

Figure 3

Xenon energy surfaces of (a) IRMOF-7 and (b) IRMOF-16 with isocontour values of 0 (gray), −12.5 kJ mol–1 (green), and −16.5 kJ mol–1 (violet).

In the final part of our study, we have considered the effect of a possible performance-boosting strategy, i.e., temperature decrease to 273 K. We plotted API-related correlation plots in Figure . We observe that we could enhance the API by a factor of ∼2.5 by reducing the temperature to 273 K (Figure a). This also reveals that the temperature-assisted boosting had more influence on working capacity (Figure c, enhanced by ∼1.9 times), followed by selectivity (Figure b, enhanced by ∼1.3 times). It is noteworthy that temperature decrease does not show a significant effect on ΔHads and therefore regeneration of occupied adsorption sites should not become an issue.

Figure 4

Correlation plots for (a) API, (b) selectivity, (c) working capacity, and (d) enthalpy of adsorption at 273 and 298 K.

Conclusions and Prospects

We successfully implemented state-of-the-art molecular simulation techniques to investigate IRMOF-1 to 16 for the room temperature xenon separation from air. Our study showed that IRMOF-7 is the top-performing one among the studied MOFs. We examined the systems in more detail by visualizing the xenon energy landscape within the MOFs, and it revealed that IRMOF-7 benefits from the presence of 1,4-naphthalenedicarboxylic acid linker, which is oriented toward the pores. It consequently generated an inhomogeneous energy surface for xenon. We, then, examined the effect of temperature lowering on the separation performance and found that not only the API can be enhanced 2.5 times but also while the working capacity, followed by selectivity, is influenced, the heat of adsorption remains unchanged. Our findings within the context of current research can be beneficial for designing future MOFs based on the studied model compounds to deliver advanced MOFs, which can provide xenon at a lower price and help the energy sustainability by lowering the energy consumption.

Simulation Details

We adopted the study set from 20 MOF structures reported by Eddaoudi et al.[45] In their original paper, four of MOFs (IRMOF-9, 11, 13, and 15) are doubly interpenetrated structures associated with IRMOF-8, 10, 14, and 16, respectively, and we excluded the former ones from our investigation. We calculated the PLD and LCD of structures using Zeo++ (version 0.3)[48] with a high-accuracy flag using implemented CSD atomic radius.

The gas adsorption isotherms were simulated using RASPA code[49] (version 2.0.35) and via Grand Canonical Monte Carlo[50,51] (GCMC) approach. All of the structures were considered rigid during simulations and taken from the shipped library along with RASPA. Nonbonded interactions among adsorbates and framework atoms were described using 12–6 Lennard–Jones and Coulomb potentials by applying eq (50,51)where i and j represent interacting atoms and r corresponds to their distance. σ and ε are L–J well depth and diameter, respectively. q denotes the partial charge of atoms, and ε0 is the dielectric constant. Partial atomic charges were calculated using extended charge equilibration (EQeq) method using default values.[52,53] A 12.8 Å cutoff was used to truncate the energies without being shifted and tail correction. Cross-term L–J parameters were calculated through Lorentz–Berthelot mixing rules.[50] The Ewald summation technique with a precision of 1 × 10–6 was used to model long-range electrostatic interactions. The force field parameters for framework atoms were all taken from universal force field.[54] Gas molecules were modeled using TraPPE force field parameters[55,56] except xenon, which was taken from Potoff et al.[57] The simulations included 5 × 105 MC cycles at each pressure point with the first half for equilibration and the remaining for the production run. An MC cycle consists of n steps, which is the highest number between 20 and the number of molecules at the beginning of each given simulation point. The Peng–Robinson equation of state was used to convert pressure to fugacity during the calculations.[49] The insertion, deletion, translation, rotation, and identity-change MC trial moves were used in all GCMC calculations. We have validated our simulation setup in our previous publications.[20,21] The energy surface landscapes were calculated and visualized using iRASPA.[58] Considering the purpose of this research, we adopted the separation condition from a recent publication[44] in which 1:9 molar ratio was used for the Xe/air mixture. Because we were interested in exploring our set of model MOFs for vacuum PSA, we slightly modified the adsorption and desorption pressures to 1.0 and 0.1 bar, respectively. We performed the simulations at 298 and 273 K.