Efficiently Exploring Adsorption Space to Identify Privileged Adsorbents for Chemical Separations of a Diverse Set of Molecules

The Open DAC 2023 Dataset and Challenges for Sorbent Discovery in Direct Air Capture

Direct air capture (DAC) of CO2 with porous adsorbents such as metal-organic frameworks (MOFs) has the potential to aid large-scale decarbonization. Previous screening of MOFs for DAC relied on empirical force fields and ignored adsorbed H2O and MOF deformation. We performed quantum chemistry calculations overcoming these restrictions for thousands of MOFs. The resulting data enable efficient descriptions using machine learning.

Efficient Exploration of Adsorption Space for Separations in Metal-Organic Frameworks Combining the Use of Molecular Simulations, Machine Learning, and Ideal Adsorbed Solution Theory

Adsorption-based separations using metal-organic frameworks (MOFs) are promising candidates for replacing common energy-intensive separation processes. The so-called adsorption space formed by the combination of billions of possible molecules and thousands of reported MOFs is vast. It is very challenging to comprehensively evaluate the performance of MOFs for chemical separation through experiments. Molecular simulations and machine learning (ML) have been widely applied to make predictions for adsorption-based separations. Previous ML approaches to these issues were typically limited to smaller molecules and often had poor accuracy in the dilute limit. To enable exploration of a wider adsorption space, we carefully selected a diverse set of 45 molecules and 335 MOFs and generated single-component isotherms of 15,075 MOF-molecule pairs by grand canonical Monte Carlo. Using this database, we successfully developed accurate (r2 > 0.9) machine learning models predicting adsorption isotherms of diverse molecules in large libraries of MOFs. With this approach, we can efficiently make predictions of large collections of MOFs for arbitrary mixture separations. By combining molecular simulation data and ML predictions with Ideal Adsorbed Solution Theory, we tested the ability of these approaches to make predictions of adsorption selectivity and loading for challenging near-azeotropic mixtures.

Quantitative Simulations of Siloxane Adsorption in Metal-Organic Frameworks

We present a transferable force field (FF) for simulating the bulk properties of linear and cyclic siloxanes and the adsorption of these species in metal-organic frameworks (MOFs). Unlike previous FFs for siloxanes, our FF accurately reproduces the vapor-liquid equilibria of each species in the bulk phase. The quality of our FF combined with the Universal Force Field using standard Lorentz-Berthelot combining rules for MOF atoms was assessed in a wide range of MOFs without open metal sites, showing good agreement with dispersion-corrected density functional theory calculations. Predictions with this FF show good agreement with the limited experimental data for siloxane adsorption in MOFs that is available. As an example of using the FF to predict adsorption properties in MOFs, we present simulations examining entropy effects in binary linear and cyclic siloxane mixture coadsorption in the large-pore MOF with structure code FOTNIN.

Curated Collection of More than 20,000 Experimentally Reported One-Dimensional Metal-Organic Frameworks

A collection of more than 20,000 experimentally derived crystal structures for metal-organic frameworks (MOFs) that do not have two- or three-dimensional covalently bonded networks has been developed from the materials available at the Cambridge Crystallographic Data Centre. Of these 20,000 1D MOFs, more than 12,000 structures have been verified to be solvent-free and in exact agreement with the stoichiometry of the synthesized materials. More than 10% of the complete data set comprise materials including two or more distinct metals. The band gaps of more than 12,000 1D MOFs have been computed at the density functional theory-generalized gradient approximation level, finding more than 2000 materials that have a zero band gap. Molecular simulations of CH₄ adsorption in a small number of 1D MOFs indicated that adsorbate-induced deformation plays a significant role in determining adsorption isotherms in these materials. As a result, methods that have been used previously for high-throughput predictions of molecular adsorption in 3D MOFs are not suitable for 1D MOFs.

Gaussian approximation of dispersion potentials for efficient featurization and machine-learning predictions of metal–organic frameworks

Energy-related descriptors in machine learning are a promising strategy to predict adsorption properties of metal–organic frameworks (MOFs) in the low-pressure regime. Interactions between hosts and guests in these systems are typically expressed as a sum of dispersion and electrostatic potentials. The energy landscape of dispersion potentials plays a crucial role in defining Henry’s constants for simple probe molecules in MOFs. To incorporate more information about this energy landscape, we introduce the Gaussian-approximated Lennard-Jones (GALJ) potential, which fits pairwise Lennard-Jones potentials with multiple Gaussians by varying their heights and widths. The GALJ approach is capable of replicating information that can be obtained from the original LJ potentials and enables efficient development of Gaussian integral (GI) descriptors that account for spatial correlations in the dispersion energy environment. GI descriptors would be computationally inconvenient to compute using the usual direct evaluation of the dispersion potential energy surface. We show that these new GI descriptors lead to improvement in ML predictions of Henry’s constants for a diverse set of adsorbates in MOFs compared to previous approaches to this task.

Thermal Management for Hydrogen Charging and Discharging in a Screened Metal-Organic Framework Particle Tank

Thermal management of H₂ gas storage in a tank is crucial for determining the H₂ gas deliverable capacity. In this study, a strategy for the design of an excellent comprehensive performance fuel storage tank from the screening of microscopic materials to the design of macroscopic particle adsorption tank performance is proposed. The best metal-organic framework (MOF) for H₂ deliverable capacity in a computation-ready experimental MOF database is first screened using a grand canonical Monte Carlo (GCMC) method. An upscale model that combines the finite volume method with GCMC is then established to investigate the H₂ charging and discharging processes in a screened best MOF-filled adsorption particle tank that is integrated with a phase-change material (PCM) jacket. The process of the heat and mass transfer in the screened best MOF particle adsorption tank with and without the PCM jacket-inserted metal foam is studied. The results show that the prescreened XAWVUN has the highest gravimetric and considerable volumetric deliverable capacity among 503 MOFs, which can reach up to 23.1 mol·kg^-1 and 20.8 kg·m^-3 at 298 K and pressures between 35 000 kPa (adsorption pressure) and 160 kPa (desorption pressure), respectively. The H₂ deliverable capacity can be maximized by 3.2 and 12.1% for PCM jackets inserted with metal foam in the H₂ charging and discharging processes when it is compared with the case without the PCM jacket, respectively. The above study will facilitate the development of new equipment for hydrogen storage.

Incorporating Flexibility Effects into Metal-Organic Framework Adsorption Simulations Using Different Models

High-throughput calculations based on molecular simulations to predict the adsorption of molecules inside metal-organic frameworks (MOFs) have become a useful complement to experimental efforts to identify promising adsorbents for chemical separations and storage. For computational convenience, all existing efforts of this kind have relied on simulations in which the MOF is approximated as rigid. In this paper, we use extensive adsorption-relaxation simulations that fully include MOF flexibility effects to explore the validity of the rigid framework approximation. We also examine the accuracy of several approximate methods to incorporate framework flexibility that are more computationally efficient than adsorption-relaxation calculations. We first benchmark various models of MOF flexibility for four MOFs with well-established CO₂ experimental consensus isotherms. We then consider a range of adsorption properties, including Henry's constants, nondilute loadings, and adsorption selectivity, for seven adsorbates in 15 MOFs randomly selected from the CoRE MOF database. Our results indicate that in many MOFs adsorption-relaxation simulations are necessary to make quantitative predictions of adsorption, particularly for adsorption at dilute concentrations, although more standard calculations based on rigid structures can provide useful information. Finally, we investigate whether a correlation exists between the elastic properties of empty MOFs and the importance of including framework flexibility in making accurate predictions of molecular adsorption. Our results did not identify a simple correlation of this type.

Deep learning combined with IAST to screen thermodynamically feasible MOFs for adsorption-based separation of multiple binary mixtures

The structures of metal-organic frameworks (MOFs) can be tuned to reproducibly create adsorption properties that enable the use of these materials in fixed-adsorption beds for non-thermal separations. However, with millions of possible MOF structures, the challenge is to find the MOF with the best adsorption properties to separate a given mixture. Thus, computational, rather than experimental, screening is necessary to identify promising MOF structures that merit further examination, a process traditionally done using molecular simulation. However, even molecular simulation can become intractable when screening an expansive MOF database for their separation properties at more than a few composition, temperature, and pressure combinations. Here, we illustrate progress toward an alternative computational framework that can efficiently identify the highest-performing MOFs for separating various gas mixtures at a variety of conditions and at a fraction of the computational cost of molecular simulation. This framework uses a "multipurpose" multilayer perceptron (MLP) model that can predict single component adsorption of various small adsorbates, which, upon coupling with ideal adsorbed solution theory (IAST), can predict binary adsorption for mixtures such as Xe/Kr, CH₄/CH₆, N₂/CH₄, and Ar/Kr at multiple compositions and pressures. For this MLP+IAST framework to work with sufficient accuracy, we found it critical for the MLP to make accurate predictions at low pressures (0.01-0.1 bar). After training a model with this capability, we found that MOFs in the 95th and 90th percentiles of separation performance determined from MLP+IAST calculations were 65% and 87%, respectively, the same as MOFs in the simulation-predicted 95th percentile across several mixtures at diverse conditions (on average). After validating our MLP+IAST framework, we used a clustering algorithm to identify "privileged" MOFs that are high performing for multiple separations at multiple conditions. As an example, we focused on MOFs that were high performing for the industrially relevant separations 80/20 Xe/Kr at 1 bar and 80/20 N₂/CH₄ at 5 bars. Finally, we used the MOF free energies (calculated on our entire database) to identify privileged MOFs that were also likely synthetically accessible, at least from a thermodynamic perspective.

Construction of an Anion-Pillared MOF Database and the Screening of MOFs Suitable for Xe/Kr Separation

The separation of xenon/krypton (Xe/Kr) mixtures is a challenging process. Many porous materials allow the adsorption of both Xe and Kr but only with low selectivity. Anion-pillared metal-organic frameworks (MOFs), featuring the anion groups as structural pillars, show potential in gas separations, but only a limited number of them have been synthesized. Here, we describe a collection of 936 anion-pillared MOFs based on 22 experimentally available structures. We performed density functional theory (DFT) optimization and then assigned density-derived electrostatic and chemical (DDEC) charges for each MOF to make them well suited to molecular simulations. The structural properties of the MOFs vary more strongly with the choice of the organic ligand than with other aspects like fluorine groups and metal centers. We then screened the entire collection of MOFs in the context of Xe/Kr separation at room temperature. Compared with previously reported MOFs, the interpenetrated MOF SIFSIX-6-Cd-i is predicted to perform better for Xe/Kr separations, with a good balance between working capacity (1.62 mmol/g) and separation selectivity (16.4) at 298 K and 100 kPa. We also found that the heterogeneity of fluorine groups within a MOF can help to enhance Xe working capacity without reducing the Xe/Kr selectivity, suggesting that synthesis of anion-pillared MOFs with mixed fluorine groups may lead to improved Xe/Kr separation performance.

Adsorption-Based Separation of Near-Azeotropic Mixtures-A Challenging Example for High-Throughput Development of Adsorbents

Adsorption of gas mixtures is central to adsorption-based gas separations, and the number of adsorbate mixture/adsorbent systems that exist is staggering. Because examples of machine learning (ML) models predicting single-component adsorption of arbitrary molecules in large libraries of crystalline adsorbents have been developed, it is interesting to determine whether these models can accurately predict mixture adsorption. Here, we use molecular simulations to generate mixture adsorption data with a set of 12 near-azeotropic molecules in a diverse set of MOFs. These data provide a challenging example for any method to rapidly predict mixture adsorption in MOFs. We combine a previous ML single-component isotherm model with ideal adsorbed solution theory (IAST) to make predictions that can be compared directly with molecular simulation data for these adsorbed mixtures. This combination of ML and IAST illustrates the scope that is available with these methods, but the accuracy of the resulting predictions is disappointing. By examining the same examples with IAST based on minimal molecular simulation data for single-component isotherms, we show that having an accurate description of adsorption in the dilute loading limit is critical to being able to accurately predict mixture adsorption. This observation points to a useful direction for future work developing robust ML models of adsorption isotherms for diverse collections of molecules and adsorbents.

High-throughput screening of metal-organic frameworks for kinetic separation of propane and propene

We apply molecular simulations to screen a database of reported metal-organic framework structures from the computation-ready, experimental (CoRE) MOF database to identify materials potentially capable of separating propane and propene by diffusion. We report a screening workflow that uses descriptor analysis, conventional molecular dynamics (MD), and Nudged Elastic Band (NEB) energy barrier calculations at both classical force field and Density Functional Theory (DFT) levels. For the first time, the effects of framework flexibility on guest transport properties were fully considered in a screening process and led to the identification of candidate MOFs. The hits identified by this proof-of-concept workflow include ZIF-8 and ZIF-67 previously shown to have large differences in propane and propene diffusivities as well as two other materials that have not been tested experimentally yet. This work emphasises the importance of taking into account framework flexibility when studying guest transport in porous materials, demonstrates the potential of the data-driven identification of high-performance materials and highlights the ways of improving the predictive power of the screening workflow.

Adsorption Isotherm Predictions for Multiple Molecules in MOFs Using the Same Deep Learning Model

Tailoring the structure and chemistry of metal-organic frameworks (MOFs) enables the manipulation of their adsorption properties to suit specific energy and environmental applications. As there are millions of possible MOFs (with tens of thousands already synthesized), molecular simulation has frequently been used to rapidly evaluate the adsorption performance of a large set of MOFs. This allows subsequent experiments to focus only on a small subset of the most promising MOFs. In many instances, however, even molecular simulation becomes prohibitively time-consuming, underscoring the need for alternative screening methods, such as machine learning, to precede molecular simulation efforts. In this study, as a proof of concept, we trained a neural network-specifically, a multilayer perceptron (MLP)-as the first example of a machine learning model capable of predicting full adsorption isotherms of different molecules not included in the training of the model. To achieve this, we trained our MLP on "alchemical" species, represented only by variables derived from their force-field parameters, to predict the loadings of real adsorbates. Alchemical species used for training were small, near-spherical, and nonpolar, enabling the prediction of analogous real molecules relevant for chemical separations such as argon, krypton, xenon, methane, ethane, and nitrogen. MOFs were also represented by simple descriptors (e.g., geometric properties and chemical moieties). The trained model was shown to make accurate adsorption predictions for these six adsorbates in both hypothetical and existing MOFs. The MLP presented here is not expected to be applied "as is" to more complex adsorbates with properties not considered during its training. However, our results illustrate a new philosophy of training that can be built upon with the goal of predicting adsorption isotherms in not only a database of MOFs but also a database of adsorbates and over a range of relevant operating conditions.

A Strong Test of Atomically Detailed Models of Molecular Adsorption in Zeolites Using Multilaboratory Experimental Data for CO2 Adsorption in Ammonium ZSM-5

A recent international interlaboratory study led by the U.S. National Institute of Standards (NIST) reported CO₂ adsorption isotherms measured independently by 11 groups on reference material RM 8852, an ammonium ZSM-5 zeolite. Good reproducibility and high reliability of this experimental data provide a strong test for the ability of atomically detailed models to predict adsorption of CO₂ in zeolites. We developed force fields for CO₂ in ammonium zeolites based on first-principles calculations and also independently performed experiments with RM 8852 by microcalorimetry. At low pressures good agreement was obtained between predictions and experiments. At high pressures, however, deviations were observed. We show that the charge-balancing cations in the experimental material are the predominant source of the discrepancy between simulation and experiment at high pressures; the experimental sample treatment causes deammoniation. In addition, accounting for a small amount of noncrystalline mesoporosity in the zeolite brings predictions into much better agreement with experiments.

Impact of intrinsic framework flexibility for selective adsorption of sarin in non-aqueous solvents using metal–organic frameworks

We assess the nontrivial deviation in predicting the adsorption selectivity from bulk mixtures of complex molecules using nanoporous adsorbents approximated as rigid and intrinsically flexible.

Effects of Intrinsic Flexibility on Adsorption Properties of Metal-Organic Frameworks at Dilute and Nondilute Loadings

Molecular simulation of adsorption in nanoporous materials has become a valuable complement to experimental studies of these materials. In almost all cases, these simulations treat the adsorbing material as rigid. We use molecular simulations to examine the validity of this approximation for the adsorption in metal-organic frameworks (MOFs) that have framework flexibility without change in their unit cells because of thermal vibrations. All nanoporous materials are subject to this kind of framework flexibility. We examine the adsorption of nine molecules (CO₂, CH₄, ethane, ethene, propane, propene, butane, Xe, and Kr) and four molecular mixtures (CO₂/CH₄, ethane/ethene, propane/propene/butane, and Xe/Kr) in 100 MOFs at dilute and nondilute adsorption conditions. Our results show that single-component adsorption uptakes at nondilute conditions are only weakly affected by framework flexibility, but adsorption selectivities at both dilute and nondilute conditions can be significantly affected by flexibility. The most dramatic impacts of framework flexibility occur for adsorption uptake in the limit of dilute adsorption. These results suggest that the importance of including framework flexibility when attempting to make quantitative predictions of adsorption selectivity in MOFs and similar materials may have been underestimated in the past.