Next Article in Journal
Sustainable and Eco-Friendly Packaging Films Based on Poly (Vinyl Alcohol) and Glass Flakes
Next Article in Special Issue
Thin-Film Composite Membranes with a Carbon Nanotube Interlayer for Organic Solvent Nanofiltration
Previous Article in Journal
Cobalt-Based Cathode Catalysts for Oxygen-Reduction Reaction in an Anion Exchange Membrane Fuel Cell
Previous Article in Special Issue
Comparison of Effects of Different Sacrificial Hydrogen Bonds on Performance of Polyurethane/Graphene Oxide Membrane
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Membranes
Volume 12
Issue 7
10.3390/membranes12070700
membranes-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Large-Scale Screening and Machine Learning for Metal–Organic Framework Membranes to Capture CO2 from Flue Gas

by
Yizhen Situ
1,
Xueying Yuan
1,
Xiangning Bai
1,
Shuhua Li
1,
Hong Liang
1,
Xin Zhu
1,*,
Bangfen Wang
1,* and
Zhiwei Qiao
1,2,*
1
Guangzhou Key Laboratory for New Energy and Green Catalysis, School of Chemistry and Chemical Engineering, Guangzhou University, Guangzhou 510006, China
2
Joint Institute of Guangzhou University & Institute of Corrosion Science and Technology, Guangzhou University, Guangzhou 510006, China
*
Authors to whom correspondence should be addressed.
Membranes 2022, 12(7), 700; https://doi.org/10.3390/membranes12070700
Submission received: 28 June 2022 / Revised: 7 July 2022 / Accepted: 8 July 2022 / Published: 11 July 2022
(This article belongs to the Special Issue Preparation of Organic-Inorganic Hybrid Membranes and Their Membrane Performance)
Browse Figures
Versions Notes

Abstract

:
To combat global warming, as an energy-saving technology, membrane separation can be applied to capture CO2 from flue gas. Metal–organic frameworks (MOFs) with characteristics like high porosity have great potential as membrane materials for gas mixture separation. In this work, through a combination of grand canonical Monte Carlo and molecular dynamics simulations, the permeability of three gases (CO2, N2, and O2) was calculated and estimated in 6013 computation–ready experimental MOF membranes (CoRE–MOFMs). Then, the relationship between structural descriptors and permeance performance, and the importance of available permeance area to permeance performance of gas molecules with smaller kinetic diameters were found by univariate analysis. Furthermore, comparing the prediction accuracy of seven classification machine learning algorithms, XGBoost was selected to analyze the order of importance of six structural descriptors to permeance performance, through which the conclusion of the univariate analysis was demonstrated one more time. Finally, seven promising CoRE-MOFMs were selected, and their structural characteristics were analyzed. This work provides explicit directions and powerful guidelines to experimenters to accelerate the research on membrane separation for the purification of flue gas.

1. Introduction

With the rapid development of industry, global warming has increased, with increasing emissions of CO2, which is very harmful to the life and development of humans [1]. Because it is impossible to completely develop industry without carbon in a short time, the capture of CO2 has become an essential way to reduce greenhouse gas emissions and improve the present condition. However, CO2 capture is an energy-intensive process for a series of reasons, among which is the gas composition [1]. Therefore, it is an urgent need to develop energy-saving technology to separate CO2 from gas mixtures. Comparing with a series of separation technologies, membrane separation has received much attention due to its efficiency, low–energy consumption, and requirement of relatively simple equipment [2]. However, the search is on for a membrane that shows good separation performance with wide application, especially when complex components are involved. The development of materials with good membrane separation performance has become a hot topic in recent years.
In the last twenty years, new materials metal–organic frameworks (MOFs), self-assembled by a wide range of organic links and metal nodes, have been considered to have potential for use in domains such as drug delivery [3], catalysis [4], gas storage [5,6,7], gas adsorption and separation [8,9,10,11,12] due to their excellent characteristics such as large surface area and high porosity. Commonly, MOFs can be applied as adsorbates and membranes for the separation of gas mixtures. For example, designing porous materials at the molecular level for adsorption-based application, PCN–88 was synthesized by Li et al. [13] based on a new concept, ‘single-molecule trap’, which showed excellent preferential adsorption of CO2 over N2 and CH4. With two new empirical equations for the prediction of hydrogen adsorption capacity by pore volume by Zhang et al. [14], NPF–200 was predicted and demonstrated as a promising MOF for H2 storage. Further, an MOF, CAU–10–NH2, with excellent water stability/reusability was synthesized by their team [15], regarded as a promising material for C2H2/CO2 separation. Boyd et al. [16] synthesized two MOFs containing the most hydrophobic adsorbaphore found by the evaluation of the CO2/N2 selectivity of MOFs in wet flue gas through computational screening. They found that both of their CO2/N2 separation performances were not affected by water. Nugent et al. [17] demonstrated the feasibility of a crystal engineering or reticular chemistry strategy, which controls pore functionality and size for the improvement of CO2 separation performance, through the synthesis of SIFSIX–2–Cu, SIFSIX–2–Cu–i, and SIFSIX–3–Zn. Although adsorption technology can be used to separate gas mixtures, membrane separation is an attractive option for its lower energy consumption [2]. In the field of membrane separation, MOF membranes (MOFMs) have been demonstrated as a kind of material with great potential to separate gas mixtures [2]. For example, Yin et al. [18] synthesized a thin tubular CAU–1 membrane exhibiting a high permeance of up to 1.34 × 10−6 mol·m−2·s−1·Pa−1 for CO2 and excellent selectivity of 17.4–22.8 for CO2/N2 mixture. Chang et al. [19] found that with the coating of a Pebax®1657 layer on the surface, the H2/CO2 separation performance of a ZIF–7–NH2 membrane can be improved. Kang et al. [20] found that the 1,2–bi–(4–pyridyl) ethylene (BPE) molecule distributed in channels can improve the H2/CO2 separation performance of the [Ni2(L–asp)2(BPE)]·(G) membrane. Under a two-step coating process, a new MOF-based membrane (PAN–γ–CD–MOF–PU membrane) was fabricated by Fan et al. [21], with a permeability to CO2 of over 70 barrer and selectivity to CO2/N2 and CO2/O2 of 253.46 and 154.28, respectively. Yan et al. [22] synthesized a UiO–66 membrane through tertiary growth at room temperature, which exhibited an optimal selectivity of 37.8 for CO2/N2. Chen et al. [23] fabricated a ZIF–8 membrane under different reaction conditions, and found that the optimal temperature for synthesis of ZIF–8 membrane is 80 °C. Further, they also found that the separation factor of CO2/N2 was 5.49 and the permeance of CO2 was 0.47 × 10–7 mol·m–2·s–1·Pa–1 under optimal conditions. However, with the increasing number of MOFs synthesized by experiments and built by computer technology, it is not practical to select potential MOFMs only through experiments due to the high costs and long time periods involved. Further, a series of chemical reagents are harmful to experimenters and the environment. There is an urgent need to develop a rapid method to select MOFMs with excellent performance.
Recently, high-throughput computational screening based on molecular simulation technology has been demonstrated as a useful way to accelerate the research on MOFMs by previous studies. For example, Qiao et al. [24] calculated the performance of 137,953 MOFMs by grand canonical Monte Carlo (GCMC) and molecular dynamics (MD) simulations and finally screened 24 optimal MOFMs for CO2/N2/CH4 separation. Appling GCMC and equilibrium MD simulations, Glover et al. [25] studied the separation performance of MOFMs for CO2/CH4 and H2S/CH4. They screened eight top-performing MOFMs superior than polymer membranes, zeolite membranes, and mixed matrix membranes (MMMs). Azar et al. [26] analyzed the H2/N2 separation performance of more than 3000 different types of MOF membranes and examined their separation potential in MMMs by molecular simulation. They found the characteristics of most promising MOFMs and the great advantage of incorporating MOFs into polymers. In the same way, Daglar et al. [27] screened optimal membranes for CO2/N2/H2O separation and explored the structure–performance relationship. They found that MOFMs with narrow pores, low surface areas, and monoclinic and lanthanide-containing structures are the best candidates for CO2/N2 membrane separation. Altintas et al. [28] used molecular simulation to explore the relationship between 175 different structures of MOFMs and the separation performance of C2H6/C2H4 and C2H6/CH4. They found that MOFMs with high C2H6 selectivity are those with cavity diameters between 6 and 9 Å, porosities lower than 0.5 and surface areas between 500 and 1000 m2g−1. Wang et al. [29] studied H2/CH4 separation using an IRMOF-1 membrane through a dual-force zone nonequilibrium molecular dynamics simulation. They reached the conclusion that both structural and chemical features of functionalized MOFMs determine gas separation performance. Bai et al. [30] screened MOFMs for the separation of gas pairs including H2 by molecular simulation and machine learning, and found 15 MOFMs with excellent separation performance.
Commonly, there are several compositions of industrial flue gas in addition to CO2, such as N2 and O2 [31,32,33]. In this work, the volume ratio of CO2, N2, and O2 is considered as 1:1:1. For the screening of MOFMs with great potential for ternary CO2/N2/O2 separation, the permeability (P) of pure CO2, N2, and O2 in MOFMs was calculated by GCMC and MD simulations. Then, through univariate analysis and machine learning, the relationship between structural descriptors and permeance performance was explored. Finally, seven promising MOFMs for ternary gas pair separation were screened.

2. Methods

2.1. Model

MOFMs studied in this work were MOFs from the 2019 computation–ready experimental metal–organic frameworks (CoRE–MOFs) database [34]. The structural parameters of CoRE–MOFs were derived from experimental data after free solvent molecules were removed [34,35]. The atomic structure of MOFs was described by Lennard–Jones (LJ) parameters and electrostatic potentials
u L J + elec ( r ) = ∑ 4 ε ij [ ( σ i j r i j ) 12 − ( σ i j r i j ) 6 ] + ∑ q i q j 4 π ε 0 r ij
where εij is the potential energy parameter, σij represents the equilibrium distance between atoms, and rij is the distance between atom pairs. qi, qj is the atomic charge of atoms i and atoms j, ε0 = 8.8542 × 10−12 C2∙N−1, which represents the permittivity of vacuum. The atomic charge of MOFs was quickly calculated by the MEPO-Qeq method [36]. All LJ potential energy parameters of MOFs are listed in Table S1, and they come from universal force field (UFF) [37]. The structural characteristics of MOFs are represented by six descriptors, in which the volumetric surface area (VSA) and fraction (ϕ) were calculated using N2 with a diameter of 3.64 Å and He with a diameter of 2.58 Å as a probe using the RASPA software package, and both the pore limited diameter (PLD) and largest cavity diameter (LCD) were estimated by the Zeo++ software package [38,39].
For CO2, O2, and N2 molecules, force field parameters were adopted from the transferable potentials for phase equilibria (TraPPE) force field [40], listed in Table S2. For CO2, the bond length of C–O is 1.16 Å, and the bond angle ∠OCO is 180°. For N2, which is considered as a three-site model, the N–N bond length is 1.10 Å. O2 is a three–site atom. A large number of studies have demonstrated that the application of UFF for MOFs and TraPPE for gases can accurately predict the gas adsorption and diffusion in various MOFs [41,42,43,44].

2.2. Molecular Simulation

In this work, the adsorption, diffusion, and permeability behaviors of pure CO2, O2, and N2 in MOFMs were simulated by GCMC and MD. Each GCMC or MD simulation was independently carried out and the interaction between MOFs and adsorbate molecules was calculated by the Lorentz–Berthelot rule. The periodic boundaries were applied in the three–dimensional system and the unit cell of each MOF was expanded to at least 24 Å in all the dimensions. To calculate LJ interactions, the spherical cutoff was set to 12 Å for long-range correction, and the framework–gas and gas–gas electrostatic interactions were calculated by Ewald summation [45]. Each GCMC simulation was run for 10,000 cycles, with the first 5000 cycles for the equilibration of simulation system and the last 5000 cycles for ensemble averages. Each cycle consisted of n trial moves (n: the number of adsorbate molecules), including translation, rotation, regrowth, and swap (insertion and deletion). The final simulation state of GCMC was used as the initial simulation state of MD. The MD duration in each MOF was 7 ns with the last 5 ns for production. All simulations of GCMC and MD were run under the RASPA software package [38]. After GCMC and MD simulations of each MOFM, the permeability of pure gases was calculated by
Pi = Ki × Di
where Di is the diffusivity of component i in MOFMs, and Ki is Henry’s constant of component i.

2.3. Evaluation of the Performance of MOFMs

Normally, evaluation of the performance of MOFMs includes gas permeance performance and gas diffusion performance. To select a series of MOFMs to separate ternary gas pairs, the separation performance of MOFMs for two binary gas pairs was firstly analyzed considering the complexity of ternary gas pairs. In this work, considering the adsorption–diffusion mechanism in the process of porous membrane separation, adsorption selectivity (Sads) has also been considered, which was calculated by
Sads (i/j) = Ki/Kj
Meanwhile, gas diffusion performance was evaluated by the Di and diffusion selectivity (Sdiff), where Sdiff was calculated by
Sdiff (i/j) = Di/Dj
and gas permeance performance was evaluated by Pi and permselectivity (Sperm), where Sperm was calculated by
Sperm (i/j) = Pi/Pj = Sdiff(i/j) Sads (i/j)

2.4. Machine Learning

To comprehensively analyze the relationship between structural descriptors and the permeance performance of MOFMs, seven classification machine learning algorithms were used to predict categories of MOFMs and calculate the relative importance (RI) of structural descriptors, which are support vector machine (SVM), k-nearest neighbor (KNN), decision tree (DT), random forest (RF), gradient boosting decision tree (GBDT), light gradient boosting machine (LGBM), and extreme gradient boosting (XGBoost).
At this stage, two categories were divided from the middle based on the permeance performance of MOFMs. Further, six structural descriptors (LCD, PLD, VSA, ϕ, the density (ρ), and the pore size distribution percentage between 2.5 and 3.5 Å (PSD%(2.5–3.5))) were applied, in which PSD%(2.5–3.5) was calculated by
PSD%(d1−d2) = A12/Atotal × 100%
where Atotal is the area under the entire PSD curve for a given MOF and A12 is the area between two pore sizes, d1 and d2. With a larger PSD%, there is a more significant proportion of uniform pores.
With five permeance performances of three components, in order to obtain a uniform calculation method for RI, after simply comparing the prediction accuracy of the algorithm, the optimal algorithm was selected. Then, after comparing the accuracy and stability of the optimal algorithm in detail with k-fold cross validation (k = 5, 10, and 15) five times, the best model with the best number fold cross validation was selected to calculate the RI of structural descriptors. In our work, the prediction accuracy of the machine learning model is evaluated by the accuracy (A), the sensitive (SEN), and the specificity (SPC). More detailed descriptions of the above algorithms are presented as Supplementary Materials.

3. Results and Discussion

3.1. Univariate Analysis

To efficiently screen top-performing MOFMs to separate ternary gas pairs, CO2/N2/O2, the relationship between the structural descriptors and permeance performance was analyzed. In Figure 1a and Figure 2a, with the increase in VSA, PLD, LCD, and ϕ, PCO2 increased rapidly when they were low and finally plateaued. However, the relationship between ρ and PCO2 was contrary. With a larger ρ, PCO2 become smaller. A similar tendency can also be found for N2 and O2 in Figures S1a,b and S2c,d. In Figure 1b, Figure 2a,b, Figures S1c and S2a,b, for Sperm (CO2/O2) and Sperm (CO2/N2), most of them were both large when VSA, PLD, LCD, and ϕ were small, which decreased dramatically at first and then plateaued with the increase in VSA, PLD, LCD, and ϕ. The above phenomenon can be attributed to the size of available permeance area. Usually, gas molecules with smaller kinetic diameters can permeate through MOFMs even if the pore size of MOFMs is small. However, with a larger pore size, gas molecules with larger kinetic diameters can also spread through MOFMs if the pore size is large enough. For VSA, with a larger VSA, the available permeance area for gas molecules becomes larger. This indicates that the permeance performance of gas molecules with smaller kinetic diameters places a greater demand on the channel and geometry area of MOFMs.
Normally, the permeance performance of MOFMs is comprehensively decided by their adsorption and diffusion performance. Specially, diffusion performance is significantly related to PLD of MOFMs. So the relationship between diffusion performance and PLD was studied. In Figure S2e, DN2 and DO2 are similar, which is attributed to their similar kinetic diameter. Surprisingly, for CO2, with the smallest kinetic diameter in three gas molecules, DCO2 is smaller than DN2 and DO2. This is because there are not only kinetic diameters but also other gas molecule properties impacting gas molecule diffusion. A similar phenomenon has also been found in previous research [24]. In Figure S2e, with the increase in PLD, D increases at first and then plateaus when PLD is larger than approximately 4 Å. For Sdiff, with a larger PLD, Sdiff shows a similar trend with Sperm. For two binary gas pairs, PLD ranges from approximately 2.5 to 3.5 Å, and PCO2 and Sdiff are both the largest. For better analysis of the relationship between structure and permeance performance, PSD%(2.5–3.5) was used for the further study. However, a decreasing trend was found in Figure 1e,f and Figure S3.
To better understand the separation mechanism, the relationship between P, Sperm, and Sads/Sdiff was analyzed. In Figure 3a, the permeance performance of a large number of MOFMs for CO2/N2 separation was found to exceed the 2008 Robeson upper bound [46]. Further, it is easy to find that with the increase in PCO2, Sperm (CO2/O2) increases, as shown in Figure 3b. The above phenomenon indicates that there is a great possibility of finding a series of MOFMs for ternary gas pair separation. Moreover, the Sads/Sdiff becomes larger when PCO2 and Sperm increase, which indicates that it is the adsorption mechanism and not the diffusion mechanism that plays a dominant role in membrane separation for CO2/N2 and CO2/O2 separation. This is attributed to the large quadrupole moment of CO2 impacting adsorption performance, which has been demonstrated previously [47,48,49,50].

3.2. Machine Learning

To comprehensively understand the order of impact importance of structural descriptors to the permeance performance of MOFMs, seven ML classification algorithms (SVM, KNN, DT, RF, GBDT, LGBM, and XGBoost) were applied to predict categories of MOFMs. Based on the performance of MOFMs, two categories were divided from the middle at first, in which P1 represents MOFMs with worse performance and P2 represents MOFMs with better performance. After comparison, XGBoost with optimal prediction performance was selected to predict categories of permeance performance.
From ML research, a series of conclusions were reached. (1) For PCO2, XGBoost with 10-fold cross validation showed the best prediction. In Figure 4c, there was an accuracy of 81% in general, and an accuracy of 81% for P1 and 82% for P2 in the confusion matrix. According to ML calculation, the order of RI is LCD > ϕ > PLD > ρ > VSA > PSD%(2.5–3.5). (2) For PO2 and PN2, the best prediction was by XGBoost with 15-fold cross validation. In Figure S14a,b, there was an accuracy of 90% in general for PO2 and there was an accuracy of 91% in general for PN2. In Figure 4e, the order of RI for PO2 and PN2 is similar—the order for PO2 is ϕ > PLD > LCD >VSA > ρ > PSD%(2.5–3.5) and for PN2 is PLD > ϕ > LCD >VSA > ρ > PSD%(2.5–3.5). (3) For Sperm (CO2/O2) and Sperm (CO2/N2), optimal prediction was by XGBoost with 10-fold and 5-fold cross validation. From ML calculation, the order of RI is shown in Figure 4f. For Sperm (CO2/O2), the order of RI is LCD > VSA > PLD > ϕ > ρ > PSD%(2.5–3.5); while, for Sperm (CO2/N2), the order of RI is PLD > VSA > ϕ > ρ > LCD > PSD%(2.5–3.5).
Based on the above conclusions from ML, we found that of six structural descriptors, LCD, PLD, and ϕ have a greater impact on gas molecule permeability. Further, the order of RI for LCD, VSA, and ρ for the permeability of three gas molecules is contrary to the order of RI of kinetic diameter. For Sperm, the order of RI of LCD and ρ for Sperm (CO2/O2) is larger than Sperm (CO2/N2). Both confirm the conclusion by the univariate analysis that the permeance performance of gas molecules with smaller kinetic diameters places a greater demand on the channel and geometry area of MOFMs. On the contrary, the order of RI of PLD for the permeability of three gas molecules is the same as the order of RI of kinetic diameter (Dia), DiaN2 > DiaO2 > DiaCO2, because PLD plays an important role in the confirmation of diffusion barrier and determines the gas molecule diffusion barrier in porous materials [51,52]. This is also explained by the phenomenon that the importance of PLD to Sperm (CO2/N2) is larger than Sperm (CO2/O2).
To better apply the ML result, the relationships between LCD and Sperm (CO2/O2) and between PLD and Sperm (CO2/N2) were analyzed in detail. From Figure 2a,b, the optimal LCD range for MOFMs applied to separate CO2/O2 gas pair mainly is approximately 2.5–7.5 Å and the optimal PLD range for MOFMs applied to separate CO2/N2 gas pair is approximately 2–5 Å.

3.3. Separation of CO2/N2/O2 Pairs

In this work, the vol% of CO2/N2/O2 pairs at 298 K and 1 bar is 1:1:1. For the purpose of CO2 capture instead of O2 and N2, it is essential use MOFMs with a large PCO2, a small PN2 and a small PO2 even under large permselectivity. As such, 44 MOFMs used to separate CO2/O2 and 38 MOFMs used to separate CO2/N2 were, respectively, screened under the condition of PCO2 ≥ 106 barrer and Sperm ≥ 106 at first, as shown in the red area of Figure 3a,b. Of the above MOFMs, there are approximately 91% LCD of MOFMs for CO2/O2 separation under the optimal LCD range and approximately 84% PLD of MOFMs for CO2/N2 separation under the optimal PLD range, which demonstrates the effectiveness of ML analysis. Furthermore, in the search for promising MOFMs for practical application, seven top-performing MOFMs with PN2 and PO2 less than 100 barrer were selected, with structural characteristics found to completely follow the optimal LCD and PLD range. The details of seven top-performing MOFMs are listed in Table S10. Due to the similar atomistic structures of three MOFMs, the seven top-performing MOFMs are shown divided in Figure 5 and Figure S15.
Moreover, the ranges of other structural descriptors were analyzed. The VSA of the majority of MOFMs is smaller than 140 m2/cm3 and approximately 60% have a VSA smaller than 30 m2/cm3. The ϕ of the MOFMs is not greater than 0.3. The ρ of 71% of MOFMS is not larger than 1600 kg/m3. Further, the PSD%(2.5–3.5) of 86% of MOFMs is less than 4%. The above conclusions on the range of optimal potential MOFMs once more demonstrates the importance of available permeance area for the permeance performance of gas molecules with smaller kinetic diameters.

4. Conclusions

In this work, the permeability of pure CO2, N2, and O2 in CoRE-MOFMs was calculated by GCMC and MD simulations to select MOFMs for the purification of flue gas. Univariate analysis showed the great impact of available permeance area to the permeance performance of gas molecules with smaller kinetic diameters. Further, the adsorption mechanism has a dominant role in the membrane separation mechanism for both CO2/N2 and CO2/O2 gas pairs. Furthermore, to comprehensively understand the order of importance of structural descriptors to permeance performance, seven classification algorithms were applied to predict categories of permeance performance, from which XGBoost was selected due to optimal prediction accuracy. Through ML calculations, the impact of available permeance area was demonstrated one more time. LCD and PLD were found to significantly impact the separation of CO2/O2 and CO2/N2, respectively. Finally, considering the purpose of CO2 capture, seven promising MOFMs with optimal permeance performance were screened. Their LCD and PLD completely conformed to the optimal LCD and PLD ranges by mining big data and ML, respectively. After the analysis of other structural descriptor ranges, the importance of available permeance area to permeance performance was illustrated for membrane separation by MOFMs. This work can provide explicit directions and powerful guidelines to study the capture of CO2 in flue gas by membrane separation.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/membranes12070700/s1, Figure S1: PN2/PO2/Sperm (CO2/O2)-PLD, φ, VSA, and ρ; Figure S2: (a) LCD-PCO2-Sperm (CO2/N2); (b) PLD-PCO2-Sperm (CO2/O2); (c) PN2-LCD; (d) PO2-LCD; (e) D-PLD; Figure S3: P/Sperm-PSD%(2.5–3.5). Figure S4: The calculation schematic diagram for the accuracy, sensitive, and specificity; Figure S5: The accuracy, sensitive, and specificity comparison of seven algorithms for Sperm (CO2/N2) and Sperm (CO2/O2); Figure S6: The diagram of k-fold cross validation; Figure S7: KNN algorithm model; Figure S8: SVM algorithm model; Figure S9: DT algorithm model; Figure S10: RF algorithm model; Figure S11: GBDT algorithm model; Figure S12: The leaf-wise tree growth schematic diagram of LGBM algorithm model; Figure S13: XGBoost algorithm model; Figure S14: Confusion matrix from best model; Figure S15: Atomistic structures of top-performance MOFMs. Table S1: Lennard–Jones parameters of metal–organic frameworks (MOFs); Table S2: Lennard–Jones parameters and charges of adsorbates; Table S3: Optimal hyperparameters for SVM, KNN, DT, RF, and GBDT; Table S4: Optimal hyperparameters for LGBM and XGBoost; Table S5: Evaluation of seven ML algorithm for P; Table S6: Evaluation of seven ML algorithm for Sperm: Table S7: Evaluation of XGBoost for PCO2 and PO2; Table S8: Evaluation of XGBoost for PN2; Table S9: Evaluation of XGBoost for Sperm (CO2/O2) and Sperm (CO2/N2); Table S10: Seven top-performance MOFMs for CO2/N2/O2 separation [9,37,53,54,55,56,57,58,59,60,61].

Author Contributions

Conceptualization, Z.Q., X.Z. and B.W.; methodology, Z.Q., B.W. and Y.S.; software, Z.Q., H.L. and S.L.; validation, Y.S., X.Y. and X.B.; formal analysis, X.Y. and X.B.; investigation, X.Z.; resources, Z.Q.; data curation, Z.Q.; writing—original draft preparation, Y.S., B.W. and Z.Q.; writing—review and editing, X.Z., B.W. and Z.Q.; visualization, Y.S.; supervision, X.Z.; project administration, B.W.; funding acquisition, Z.Q. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (21978058 and 21676094), the Pearl River Talent Recruitment Program (2019QN01L255), the Natural Science Foundation of Guangdong Province (2022A1515011446 and 2020A1515010800), the Guangzhou Municipal Science and Technology Project (202102020875) and the R & D Program of the Joint Institute of GZHU and ICoST (GI202102).

Institutional Review Board Statement

“Not applicable” for studies not involving humans or animals.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data is contained within the article or Supplementary Material.

Acknowledgments

We gratefully thank the National Natural Science Foundation of China (21978058 and 21676094), the Pearl River Talent Recruitment Program (2019QN01L255), the Natural Science Foundation of Guangdong Province (2022A1515011446 and 2020A1515010800), the Guangzhou Municipal Science and Technology Project (202102020875) and the R & D Program of the Joint Institute of GZHU and ICoST (GI202102) for financial support.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Yagihara, K.; Ohno, H.; Guzman-Urbina, A.; Ni, J.; Fukushima, Y. Analyzing flue gas properties emitted from power and industrial sectors toward heat-integrated carbon capture. Energy 2022, 250, 123775. [Google Scholar] [CrossRef]
  2. Qian, Q.; Asinger, P.A.; Lee, M.J.; Han, G.; Mizrahi Rodriguez, K.; Lin, S.; Benedetti, F.M.; Wu, A.X.; Chi, W.S.; Smith, Z.P. MOF-Based Membranes for Gas Separations. Chem. Rev. 2020, 120, 8161–8266. [Google Scholar] [CrossRef]
  3. Liu, X.; Liang, T.; Zhang, R.; Ding, Q.; Wu, S.; Li, C.; Lin, Y.; Ye, Y.; Zhong, Z.; Zhou, M. Iron-Based Metal-Organic Frameworks in Drug Delivery and Biomedicine. ACS Appl. Mater. Interfaces 2021, 13, 9643–9655. [Google Scholar] [CrossRef]
  4. Guo, M.; Zhang, M.; Liu, R.; Zhang, X.; Li, G. State-of-the-Art Advancements in Photocatalytic Hydrogenation: Reaction Mechanism and Recent Progress in Metal-Organic Framework (MOF)-Based Catalysts. Adv. Sci. 2022, 9, e2103361. [Google Scholar] [CrossRef]
  5. Lim, D.W.; Chyun, S.A.; Suh, M.P. Hydrogen storage in a potassium-ion-bound metal-organic framework incorporating crown ether struts as specific cation binding sites. Angew. Chem.-Int. Edit. 2014, 53, 7819–7822. [Google Scholar] [CrossRef] [PubMed]
  6. Mason, J.A.; Oktawiec, J.; Taylor, M.K.; Hudson, M.R.; Rodriguez, J.; Bachman, J.E.; Gonzalez, M.I.; Cervellino, A.; Guagliardi, A.; Brown, C.M.; et al. Methane storage in flexible metal-organic frameworks with intrinsic thermal management. Nature 2015, 527, 357–361. [Google Scholar] [CrossRef] [PubMed]
  7. Prasetyo, N.; Pambudi, F.I. Toward hydrogen storage material in fluorinated zirconium metal-organic framework (MOF-801): A periodic density functional theory (DFT) study of fluorination and adsorption. Int. J. Hydrog. Energy 2021, 46, 4222–4228. [Google Scholar] [CrossRef]
  8. Shi, Z.; Yang, W.; Deng, X.; Cai, C.; Yan, Y.; Liang, H.; Liu, Z.; Qiao, Z. Machine-learning-assisted high-throughput computational screening of high performance metal–organic frameworks. Mol. Syst. Des. Eng. 2020, 5, 725–742. [Google Scholar] [CrossRef]
  9. Qiao, Z.; Li, L.; Li, S.; Liang, H.; Zhou, J.; Snurr, R.Q. Molecular fingerprint and machine learning to accelerate design of high-performance homochiral metal–organic frameworks. AlChE J. 2021, 67, e17352. [Google Scholar] [CrossRef]
  10. Yuan, X.; Deng, X.; Cai, C.; Shi, Z.; Liang, H.; Li, S.; Qiao, Z. Machine learning and high-throughput computational screening of hydrophobic metal–organic frameworks for capture of formaldehyde from air. Green Energy Environ. 2021, 6, 759–770. [Google Scholar] [CrossRef]
  11. Yuan, X.; Li, L.; Shi, Z.; Liang, H.; Li, S.; Qiao, Z. Molecular-fingerprint machine-learning-assisted design and prediction for high-performance MOFs for capture of NMHCs from air. Adv. Powder Mater. 2022, 1, 100026. [Google Scholar] [CrossRef]
  12. McDonald, T.M.; Mason, J.A.; Kong, X.; Bloch, E.D.; Gygi, D.; Dani, A.; Crocella, V.; Giordanino, F.; Odoh, S.O.; Drisdell, W.S.; et al. Cooperative insertion of CO2 in diamine-appended metal-organic frameworks. Nature 2015, 519, 303–308. [Google Scholar] [CrossRef]
  13. Li, J.R.; Yu, J.; Lu, W.; Sun, L.B.; Sculley, J.; Balbuena, P.B.; Zhou, H.C. Porous materials with pre-designed single-molecule traps for CO2 selective adsorption. Nat. Commun. 2013, 4, 1538. [Google Scholar] [CrossRef]
  14. Zhang, X.; Lin, R.B.; Wang, J.; Wang, B.; Liang, B.; Yildirim, T.; Zhang, J.; Zhou, W.; Chen, B. Optimization of the Pore Structures of MOFs for Record High Hydrogen Volumetric Working Capacity. Adv. Mater. 2020, 32, e1907995. [Google Scholar] [CrossRef]
  15. Zhang, X.; Lin, R.-B.; Wu, H.; Huang, Y.; Ye, Y.; Duan, J.; Zhou, W.; Li, J.-R.; Chen, B. Maximizing acetylene packing density for highly efficient C2H2/CO2 separation through immobilization of amine sites within a prototype MOF. Chem. Eng. J. 2022, 431, 134184. [Google Scholar] [CrossRef]
  16. Boyd, P.G.; Chidambaram, A.; Garcia-Diez, E.; Ireland, C.P.; Daff, T.D.; Bounds, R.; Gladysiak, A.; Schouwink, P.; Moosavi, S.M.; Maroto-Valer, M.M.; et al. Data-driven design of metal-organic frameworks for wet flue gas CO2 capture. Nature 2019, 576, 253–256. [Google Scholar] [CrossRef] [Green Version]
  17. Nugent, P.; Belmabkhout, Y.; Burd, S.D.; Cairns, A.J.; Luebke, R.; Forrest, K.; Pham, T.; Ma, S.; Space, B.; Wojtas, L.; et al. Porous materials with optimal adsorption thermodynamics and kinetics for CO2 separation. Nature 2013, 495, 80–84. [Google Scholar] [CrossRef]
  18. Yin, H.; Wang, J.; Xie, Z.; Yang, J.; Bai, J.; Lu, J.; Zhang, Y.; Yin, D.; Lin, J.Y. A highly permeable and selective amino-functionalized MOF CAU-1 membrane for CO2-N2 separation. Chem. Commun. 2014, 50, 3699–3701. [Google Scholar] [CrossRef]
  19. Chang, H.; Wang, Y.; Xiang, L.; Liu, D.; Wang, C.; Pan, Y. Improved H2/CO2 separation performance on mixed-linker ZIF-7 polycrystalline membranes. Chem. Eng. Sci. 2018, 192, 85–93. [Google Scholar] [CrossRef]
  20. Kang, Z.; Fan, L.; Wang, S.; Sun, D.; Xue, M.; Qiu, S. In situ confinement of free linkers within a stable MOF membrane for highly improved gas separation properties. CrystEngComm 2017, 19, 1601–1606. [Google Scholar] [CrossRef]
  21. Fan, S.T.; Qiu, Z.J.; Xu, R.Y.; Zhang, S.X.; Chen, Z.H.; Nie, Z.J.; Shu, H.R.; Guo, K.; Zhang, S.; Li, B.J. Ultrahigh Carbon Dioxide-Selective Composite Membrane Containing a gamma-CD-MOF Layer. ACS Appl. Mater. Interfaces 2021, 13, 13034–13043. [Google Scholar] [CrossRef]
  22. Yan, J.; Sun, Y.; Ji, T.; Zhang, C.; Liu, L.; Liu, Y. Room-temperature synthesis of defect-engineered Zirconium-MOF membrane enabling superior CO2/N2 selectivity with zirconium-oxo cluster source. J. Membr. Sci. 2022, 653, 120496. [Google Scholar] [CrossRef]
  23. Chen, S.S.; Yang, Z.-J.; Chang, C.-H.; Koh, H.-U.; Al-Saeedi, S.I.; Tung, K.-L.; Wu, K.C.W. Interfacial nanoarchitectonics for ZIF-8 membranes with enhanced gas separation. Beilstein J. Nanotechnol. 2022, 13, 313–324. [Google Scholar] [CrossRef]
  24. Qiao, Z.; Peng, C.; Zhou, J.; Jiang, J. High-throughput computational screening of 137953 metal–organic frameworks for membrane separation of a CO2/N2/CH4 mixture. J. Mater. Chem. A 2016, 4, 15904–15912. [Google Scholar] [CrossRef]
  25. Glover, J.; Besley, E. A high-throughput screening of metal-organic framework based membranes for biogas upgrading. Faraday Discuss. 2021, 231, 235–257. [Google Scholar] [CrossRef]
  26. Azar, A.N.V.; Velioglu, S.; Keskin, S. Large-Scale Computational Screening of Metal Organic Framework (MOF) Membranes and MOF-Based Polymer Membranes for H2/N2 Separations. ACS Sustain. Chem. Eng. 2019, 7, 9525–9536. [Google Scholar] [CrossRef] [Green Version]
  27. Daglar, H.; Keskin, S. Computational Screening of Metal-Organic Frameworks for Membrane-Based CO2/N2/H2O Separations: Best Materials for Flue Gas Separation. J. Phys. Chem. C 2018, 122, 17347–17357. [Google Scholar] [CrossRef] [Green Version]
  28. Altintas, C.; Keskin, S. Molecular simulations of MOF membranes for separation of ethane/ethene and ethane/methane mixtures. RSC Adv. 2017, 7, 52283–52295. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  29. Wang, F.; Sose, A.T.; Singh, S.K.; Deshmukh, S.A. Dual-Force Zone Nonequilibrium Molecular Dynamics Simulations on Nanoporous Metal–Organic Framework Membranes for Separation of H2/CH4 Mixtures. ACS Appl. Nano Mater. 2022, 5, 4048–4061. [Google Scholar] [CrossRef]
  30. Bai, X.; Shi, Z.; Xia, H.; Li, S.; Liu, Z.; Liang, H.; Liu, Z.; Wang, B.; Qiao, Z. Machine-Learning-Assisted High-Throughput computational screening of Metal–Organic framework membranes for hydrogen separation. Chem. Eng. J. 2022, 446, 136783. [Google Scholar] [CrossRef]
  31. Su, Y.; Zhang, S.; Liu, L.; Xu, D.; Xiong, Y. Investigation of representative components of flue gas used as torrefaction pretreatment atmosphere and its effects on fast pyrolysis behaviors. Bioresour. Technol. 2018, 267, 584–590. [Google Scholar] [CrossRef]
  32. He, X.; Hägg, M.-B. Structural, kinetic and performance characterization of hollow fiber carbon membranes. J. Membr. Sci. 2012, 390, 23–31. [Google Scholar] [CrossRef]
  33. Mabuza, M.; Premlall, K.; Onyango, M.; Daramola, M.O. Low-high temperature flue gas direct injection in south African bituminous and anthracite coals: Sorption capacity assessment. Curr. Sci. 2018, 115, 682–691. [Google Scholar] [CrossRef]
  34. Chung, Y.G.; Haldoupis, E.; Bucior, B.J.; Haranczyk, M.; Lee, S.; Zhang, H.; Vogiatzis, K.D.; Milisavljevic, M.; Ling, S.; Camp, J.S.; et al. Advances, Updates, and Analytics for the Computation-Ready, Experimental Metal–Organic Framework Database: CoRE MOF 2019. J. Chem. Eng. Data 2019, 64, 5985–5998. [Google Scholar] [CrossRef]
  35. Chung, Y.G.; Camp, J.; Haranczyk, M.; Sikora, B.J.; Bury, W.; Krungleviciute, V.; Yildirim, T.; Farha, O.K.; Sholl, D.S.; Snurr, R.Q. Computation-Ready, Experimental Metal–Organic Frameworks: A Tool to Enable High-Throughput Screening of Nanoporous Crystals. Chem. Mater. 2014, 26, 6185–6192. [Google Scholar] [CrossRef]
  36. Kadantsev, E.S.; Boyd, P.G.; Daff, T.D.; Woo, T.K. Fast and Accurate Electrostatics in Metal Organic Frameworks with a Robust Charge Equilibration Parameterization for High-Throughput Virtual Screening of Gas Adsorption. J. Phys. Chem. Lett. 2013, 4, 3056–3061. [Google Scholar] [CrossRef]
  37. Rappe, A.K.; Casewit, C.J.; Colwell, K.S.; Goddard, W.A.; Skiff, W.M. UFF A Full Periodic Table Force Field for Molecular Mechanics and Molecular Dynamics Simulation. J. Am. Chem. Soc. 1992, 114, 10024–10035. [Google Scholar] [CrossRef]
  38. Dubbeldam, D.; Calero, S.; Ellis, D.E.; Snurr, R.Q. RASPA: Molecular simulation software for adsorption and diffusion in flexible nanoporous materials. Mol. Simul. 2016, 42, 81–101. [Google Scholar] [CrossRef] [Green Version]
  39. Willems, T.F.; Rycroft, C.H.; Kazi, M.; Meza, J.C.; Haranczyk, M. Algorithms and tools for high-throughput geometry-based analysis of crystalline porous materials. Microporous Mesoporous Mater. 2012, 149, 134–141. [Google Scholar] [CrossRef]
  40. Potoff, J.J.; Siepmann, J.I. Vapor-liquid equilibria of mixtures containing alkanes, carbon dioxide, and nitrogen. AlChE J. 2001, 47, 1676–1682. [Google Scholar] [CrossRef]
  41. Shi, Z.; Liang, H.; Yang, W.; Liu, J.; Liu, Z.; Qiao, Z. Machine learning and in silico discovery of metal-organic frameworks: Methanol as a working fluid in adsorption-driven heat pumps and chillers. Chem. Eng. Sci. 2020, 214, 115430. [Google Scholar] [CrossRef]
  42. Shi, Z.; Yuan, X.; Yan, Y.; Tang, Y.; Li, J.; Liang, H.; Tong, L.; Qiao, Z. Techno-economic analysis of metal–organic frameworks for adsorption heat pumps/chillers: From directional computational screening, machine learning to experiment. J. Mater. Chem. A 2021, 9, 7656–7666. [Google Scholar] [CrossRef]
  43. Park, J.; Suh, B.L.; Kim, J. Computational Design of a Photoresponsive Metal–Organic Framework for Post Combustion Carbon Capture. J. Phys. Chem. C 2020, 124, 13162–13167. [Google Scholar] [CrossRef]
  44. Yan, Y.; Shi, Z.; Li, H.; Li, L.; Yang, X.; Li, S.; Liang, H.; Qiao, Z. Machine learning and in-silico screening of metal–organic frameworks for O2/N2 dynamic adsorption and separation. Chem. Eng. J. 2022, 427, 131604. [Google Scholar] [CrossRef]
  45. Ewald, P.P. Die Berechnung optischer und elektrostatischer Gitterpotentiale. Ann. Phys. 1921, 369, 253–287. [Google Scholar] [CrossRef] [Green Version]
  46. Robeson, L.M. The upper bound revisited. J. Membr. Sci. 2008, 320, 390–400. [Google Scholar] [CrossRef]
  47. Watanabe, T.; Manz, T.A.; Sholl, D.S. Accurate Treatment of Electrostatics during Molecular Adsorption in Nanoporous Crystals without Assigning Point Charges to Framework Atoms. J. Phys. Chem. C 2011, 115, 4824–4836. [Google Scholar] [CrossRef]
  48. Haldoupis, E.; Nair, S.; Sholl, D.S. Finding MOFs for highly selective CO2/N2 adsorption using materials screening based on efficient assignment of atomic point charges. J. Am. Chem. Soc. 2012, 134, 4313–4323. [Google Scholar] [CrossRef]
  49. Watanabe, T.; Sholl, D.S. Accelerating applications of metal-organic frameworks for gas adsorption and separation by computational screening of materials. Langmuir 2012, 28, 14114–14128. [Google Scholar] [CrossRef]
  50. Altintas, C.; Keskin, S. Role of partial charge assignment methods in high-throughput screening of MOF adsorbents and membranes for CO2/CH4 separation. Mol. Syst. Des. Eng. 2020, 5, 532–543. [Google Scholar] [CrossRef] [Green Version]
  51. Wang, J.; Zhou, M.; Lu, D.; Fei, W.; Wu, J. Computational screening and design of nanoporous membranes for efficient carbon isotope separation. Green Energy Environ. 2020, 5, 364–373. [Google Scholar] [CrossRef]
  52. Zhou, M.; Vassallo, A.; Wu, J. Toward the inverse design of MOF membranes for efficient D2/H2 separation by combination of physics-based and data-driven modeling. J. Membr. Sci. 2020, 598, 117675. [Google Scholar] [CrossRef]
  53. Shah, M.S.; Tsapatsis, M.; Siepmann, J.I. Development of the Transferable Potentials for Phase Equilibria Model for Hydrogen Sulfide. J. Phys. Chem. B 2015, 119, 7041–7052. [Google Scholar] [CrossRef] [PubMed]
  54. Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.; Prettenhofer, P.; Weiss, R.; Dubourg, V.; et al. Scikit-learn: Machine Learning in Python. J. Mach. Learn. Res. 2011, 12, 2825–2830. [Google Scholar]
  55. Available online: http://scikit-optimize.github.io/stable/modules/bayessearchcv.html (accessed on 8 June 2022).
  56. Cover, T.M.; Hart, P.E. Nearest Neighbor Pattern Classification. IEEE Trans. Inf. Theory 1967, 13, 21. [Google Scholar] [CrossRef] [Green Version]
  57. Scholkopf, B.; Sung, K.-K.; Burges, C.J.C.; Girosi, F.; Niyogi, P.; Poggio, T.; Vapnik, V.; Tran, I. Comparing Support Vector Machines with Gaussian Kernels to Radial Basis Function Classiers. IEEE Trans. Signal Process. 1997, 45, 2758–2765. [Google Scholar] [CrossRef] [Green Version]
  58. Yin, F.; Shao, X.; Zhao, L.; Li, X.; Zhou, J.; Cheng, Y.; He, X.; Lei, S.; Li, J.; Wang, J. Predicting prognosis of endometrioid endometrial adenocarcinoma on the basis of gene expression and clinical features using Random Forest. Oncol. Lett. 2019, 18, 1597–1606. [Google Scholar] [CrossRef]
  59. Friedman, J.H. Greedy function approximation: A gradient boosting machine. Ann. Stat. 2001, 29, 1189–1232. [Google Scholar] [CrossRef]
  60. Available online: https://lightgbm.readthedocs.io/en/latest/Features.html (accessed on 8 June 2022).
  61. Available online: https://xgboost.readthedocs.io/en/stable/ (accessed on 8 June 2022).
Membranes 12 00700 g001 550
Figure 1. Structures–performance relationship studied by univariate analysis. (a) PCO2–VSA, ϕ, PLD, and ρ; (b) Sperm (CO2/N2)–VSA, ϕ, PLD, and ρ; (c) Sdiff (CO2/N2)–PLD; (d) Sdiff (CO2/O2)–PLD; (e) PCO2–PSD%(2.5–3.5); (f) Sperm (CO2/N2) –PSD%(2.5–3.5). In (a), the colors of balls represent PLD and the sizes of ball represent ρ of MOFMs. In (b), the colors of balls represent the VSA and the sizes of ball represent ρ of MOFMs.
Figure 1. Structures–performance relationship studied by univariate analysis. (a) PCO2–VSA, ϕ, PLD, and ρ; (b) Sperm (CO2/N2)–VSA, ϕ, PLD, and ρ; (c) Sdiff (CO2/N2)–PLD; (d) Sdiff (CO2/O2)–PLD; (e) PCO2–PSD%(2.5–3.5); (f) Sperm (CO2/N2) –PSD%(2.5–3.5). In (a), the colors of balls represent PLD and the sizes of ball represent ρ of MOFMs. In (b), the colors of balls represent the VSA and the sizes of ball represent ρ of MOFMs.
Membranes 12 00700 g001
Membranes 12 00700 g002 550
Figure 2. Relationship between LCD/PLD and permeance performance. (a) LCD–PCO2–Sperm (CO2/O2); (b) PLD–PCO2–Sperm (CO2/N2).
Figure 2. Relationship between LCD/PLD and permeance performance. (a) LCD–PCO2–Sperm (CO2/O2); (b) PLD–PCO2–Sperm (CO2/N2).
Membranes 12 00700 g002
Membranes 12 00700 g003 550
Figure 3. Relationship between P, Sperm, and Sads/Sdiff. (a) PCO2–Sperm (CO2/N2)–Sads (CO2/N2)/Sdiff (CO2/N2); (b) PCO2–Sperm (CO2/O2)–Sads (CO2/O2)/Sdiff (CO2/O2). In (a), the red line represents the 2008 Robeson upper bound.
Figure 3. Relationship between P, Sperm, and Sads/Sdiff. (a) PCO2–Sperm (CO2/N2)–Sads (CO2/N2)/Sdiff (CO2/N2); (b) PCO2–Sperm (CO2/O2)–Sads (CO2/O2)/Sdiff (CO2/O2). In (a), the red line represents the 2008 Robeson upper bound.
Membranes 12 00700 g003
Membranes 12 00700 g004 550
Figure 4. (a,b) Prediction accuracy comparison of seven classification algorithms; (c) confusion matrix for PCO2; (d) confusion matrix for Sperm (CO2/N2); (e) RI comparison of PCO2, PO2, and PN2; (f) RI comparison of Sperm (CO2/N2) and Sperm (CO2/O2).
Figure 4. (a,b) Prediction accuracy comparison of seven classification algorithms; (c) confusion matrix for PCO2; (d) confusion matrix for Sperm (CO2/N2); (e) RI comparison of PCO2, PO2, and PN2; (f) RI comparison of Sperm (CO2/N2) and Sperm (CO2/O2).
Membranes 12 00700 g004
Membranes 12 00700 g005 550
Figure 5. Atomistic structures of top-performing MOFMs. (a) CARGEI; (b) YUJWAD; (c) RIPWEU; (d) VEHNED; (e) WOCJII.
Figure 5. Atomistic structures of top-performing MOFMs. (a) CARGEI; (b) YUJWAD; (c) RIPWEU; (d) VEHNED; (e) WOCJII.
Membranes 12 00700 g005
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Situ, Y.; Yuan, X.; Bai, X.; Li, S.; Liang, H.; Zhu, X.; Wang, B.; Qiao, Z. Large-Scale Screening and Machine Learning for Metal–Organic Framework Membranes to Capture CO2 from Flue Gas. Membranes 2022, 12, 700. https://doi.org/10.3390/membranes12070700

AMA Style

Situ Y, Yuan X, Bai X, Li S, Liang H, Zhu X, Wang B, Qiao Z. Large-Scale Screening and Machine Learning for Metal–Organic Framework Membranes to Capture CO2 from Flue Gas. Membranes. 2022; 12(7):700. https://doi.org/10.3390/membranes12070700

Chicago/Turabian Style

Situ, Yizhen, Xueying Yuan, Xiangning Bai, Shuhua Li, Hong Liang, Xin Zhu, Bangfen Wang, and Zhiwei Qiao. 2022. "Large-Scale Screening and Machine Learning for Metal–Organic Framework Membranes to Capture CO2 from Flue Gas" Membranes 12, no. 7: 700. https://doi.org/10.3390/membranes12070700

APA Style

Situ, Y., Yuan, X., Bai, X., Li, S., Liang, H., Zhu, X., Wang, B., & Qiao, Z. (2022). Large-Scale Screening and Machine Learning for Metal–Organic Framework Membranes to Capture CO2 from Flue Gas. Membranes, 12(7), 700. https://doi.org/10.3390/membranes12070700

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop