High-throughput Molecular Simulation Study Of MOF Adsorbents For Chemical Warfare Agent Captur

Chemical warfare agents（CWAs）are highly toxic compounds often used in terror attacks and acts of war.Since its emergence,it has continuously caused injury and threat to human health and life safety.Therefore,it is very important to find suitable adsorption materials to capture these toxic gases in the environment.Our research uses metal-organic framework（MOF）as an adsorbent,high throughput computational screening（HTCS）and machine learning（ML）were employed to study the performance of MOF-four systems（adsorption and separation of mustard gas（HD）,Soman and simulants 2-Chloroethyl ethyl sulfide（2-CEES）,Dimethyl Methyl Phosphonate（DMMP）in the air）,respectively.The molecular simulation method adopts Grand Canonical Monte Carlo（GCMC）,the machine learning algorithm is Decision Tree（DT）,Random Forest（RF）and e Xtreme Gradient Boosting（XGB）.The MOFs material used in this study is from a Computational-Ready Experimental MOFs（Co RE-MOFs）.In this work,we first summarize the application of machine learning algorithm assisted high-throughput computing technology in multiple fields of MOFs materials.Then,on this basis,mainly carried out the following two aspects of research:1.The removal of chemical warfare agents from air is particularly important for battlefield and chemical warfare agents leakage.Firstly,the adsorption properties of Co RE-MOFs on low concentration chemical warfare agents HD and Soman,as well as simulants 2-CEES and DMMP in 6013 MOFs were calculated by GCMC simulation.Then machine learning algorithm DT,RF and XGB were used to predict the adsorption selectivity of MOF.The prediction effect of XGB was the best.The R value of the above four target gases were 0.90,0.88,0.90 and 0.88,respectively.Further applying machine learning to quantitatively analyze the relative importance of the MOF descriptor,it is found that the energy descriptor Henry Coefficient K_H,heat of adsorption Q_st⁰,and porosity ф are strongly correlated with the capturing power of the CWAs.Finally,20（two CWAs and two simulants,a total of 80）MOFs with the best performance were successfully screened,and it was found that MOFs that could adsorb a variety of CWAs all had quadrilateral pore structure,and the combination MOF of Zn with a metal center and organic linker 1,3,5-benzenetricarboxylic acid plays an important role in adsorbing four CWAs.The part of this research is helpful to promote the application of high-throughput computing combined with machine learning in the field of MOFs adsorption of chemical warfare agents.2.Based on the previous work,we screened 606 hydrophobic Co RE-MOFs by using the boundary of the Henry Coefficient of water to avoid the influence of wet environment which may lead to the collapse of MOFs material structure and the competitive adsorption on the removal of the four poisons.In this work,the capture effect of the hydrophobic Co RE-MOFs for four warfare agents（HD,2-CEES,Soman,and DMMP）in the air was obtained by combining HTCS and ML.Firstly,after HTCS,the relationship between the structure-performance（adsorption capacity（N）and selectivity（S）and comprehensive trade-off value（TSN）between S and N of MOF）was established.After comparison,there was a strong correlative among the porosity and K_H and the capture performance.Then,three machine learning algorithms,DT,RF and XGB,were used to train,learn and predict the performance of MOF.Compared with N and S,TSN had a good prediction effect（average R>0.95）for four warfare agents and the XGB model was suitable for the prediction of MOF-CWAs system.and K_H was the most relative importance by the big data analysis of this method.Finally,based on TSN,10 MOFs with high performance were screened for the adsorption of four warfare agents,showing that MOF materials with open metal sites had excellent adsorption performance of four warfare agents.High-throughput computational screening and machine learning are used to elaborate the laws between MOF and four warfare agents from different aspects,which provides a new direction for efficient screening MOFs with target performance to capture four warfare agents in the air.