1.Introduction

Due to the combustion of tremendous amount of fossil fuels,the rising concentration of atmospheric carbon dioxide(CO2)up to 409 ppm have raised chronic concerns on the overall impact on global environment[1-3],resulting in the significant research efforts on CO2capture.Current commercial technologies for pre-or post-combustion CO2capture using solvents are extremely energy intensive[4].In contrast,CO2capture on solid adsorbents,such as MOFs(metalorganic frameworks)[5],COFs(covalent organic frameworks)[6],PPNs(porous polymer networks)[7]and zeolites[8],have been widely recognized as an effective and energy-saving way for CO2removal from various gas streams or flue exhausts.The target of adsorbent-based CO2capture approach is to find promising nanoporous materials with high CO2working capacity in CO2storage,as well as with high performance for related CO2-mixed gas separations.

In the meantime,the excessive reliance on petroleum as transportation fuels leads to the concerns over the sustainability of oil reserves as well as the vast emissions of CO2and air-pollutions,which can be mitigated by switching to methane(CH4)-based energy sources.Natural gas(NG),which mainly consists of CH4,has abundant reserves in the world and releases much lower CO2and polluting emissions during combustion[9].Therefore,NG is an attractive midterm alternative to petroleum-based fuels.For the practical on board usage of NG in vehicular applications,the NG needs to be densified in a reasonably-sized fuel tank to provide enough energy power.Currently,the densification strategies,such as compressed natural gas(CNG)(250 bar)or liquefied natural gas(LNG)(111 K),are in high cost and face the problems of portability and safety,making them hardly to be widely applied[10,11].Adsorbed natural gas(ANG),in which NG is physically adsorbed in nanoporous materials,is considered to be a simpler,safer and more cost-effective strategy for storing NG[12,13].The US Department of Energy(DOE)target demands that one volume of adsorbent material should deliver 315 volumes of methane at standard temperature and pressure(STP)to the engine at ambient temperature and a storage pressure of 65 bar.To realize this target,a large number of porous materials have been synthesized and evaluated experimentally,like MOFs,COFs and PPNs.

Apart from enormous porous materials reported experimentally,the genomic characteristic of advanced porous materials particularly creates in finite variety of their structures by assembling different molecular building blocks[14].It is obviously impractical to synthesize all the materials and evaluate whether for CO2capture or CH4storage or other applications using experimental techniques.In recognition of this problem,theoretical researchers have paid a lot of efforts on developing high-throughput(HT)computational screening methods to identify promising candidates from thousands of materials as well as to reveal the useful structure-property relationships[15,16].Encouragingly,the US and lately China launched the Materials Genome Initiative(MGI),which demands a large-scale collaboration between materials scientists(both experimentalists and theorists)and computer scientists to speed up the discovery and marketization of advanced new materials[17].Consequently,it can be envisioned that HT computational techniques will play a key role in the development of many classes of materials.

In this review,we devote our attention to concisely highlighting the progresses that employ HT computational techniques to construct and screen porous materials,including MOFs,COFs,PPNs and zeolites,for CO2capture and CH4 storage.Accompanying with the HT-utilized sources of these material structures described in Section 2,Section 3 briefly introducesthe computational characterization tools and methods that are commonly used for structural features and application-property predictions.Section 4 presents the recent advance and research perspectives of HT screening and design of nannoporous materials related to the above two application fields.Finally,some challenges and outlooks were provided.

民国时期是图书馆事业发展壮大的重要时期，摆在人们面前的重要问题是如何发展图书馆事业。图书馆事业作为教育机构的一部分，其发展也不外如此，卡内基其人及其图书馆捐助活动随着美国图书馆思想的传播被全面介绍和宣传到中国。

2.Databases of porous materials

Besides the collections of experimentally reported materials,databases of porous materials utilized in HT computational studies have also been obtained from hypothetical ones.For the porous materials introduced in this work,a great variety of building units can be utilized to build their structures.The modular genomic characteristics of these materials make their structures theoretically unlimited.Thus,one of the advantages of building hypothetical databases is that more comprehensive structure-property relationships can be computationally explored for specific applications and allows subsequent experimental endeavors to be deployed on the identified promising candidates.Among the four types of porous materials mentioned above,zeolites are traditional materials which have been produced industrially on a large scale.MOFs,PPNs and COFs are new members that feature tunable pore sizes,pore geometries and ultrahigh surface area,which have experienced explosive growth of preparation and characterization in the past decade,especially for MOFs.

4.2.1.MOFs

COFs represent a novel class of crystalline nanoporous materials with periodic networks that are naturally assembled by organic building units.Because they are composed of lightweight elements linked by strong covalent bonds,COFs show features of low mass density,high thermal stability and perpetual porosity.At present,more than 200 COFs were reported,however,the majority of their structures were not deposited in any database.In 2017,Tong et al.collected 187 COFs and constructed a CoRE COF database,which served as a new branch of existing porous materials library[30].All the structure files are solvent-free and disorder-free,which can be directly used as input files for computational purposes(https://core-cof.github.io/CoRE-COF-Database/).Constructionsof hypothetical COFs have also been studied by using top-down approach[31-34],among which Martin et al.[34]constructed a set of 4147 3D-COFs using only established synthetic paths,previously utilized tetrahedral building blocks,and commercially available “linker”molecules.All the structures are publicly available online(http://www.nanoporousmaterials.org/databases/).

PPNs are porous materials composed predominantly of non-metallic elements connected through strong covalent bonds.PPNs share many common features with COFs except that they are essentially amorphous networks with disordered structures and wide pore size distributions,which can be ascribe to their formation through irreversible condensation reactions.Consequently,the structures of amorphous PPNs are difficult to determine and no available structure database of experimental PPNs was reported.In spite of this,Martin et.al.considered that the crystalline assumption of PPNs can be a standardized metric for reporting material properties[35].Hypothetically,they constructed a large data set of publicly available 18,000 PPN materials with crystalline structures by using top-down approach(http://www.nanoporousmaterials.org/databases/).

Zeolites are crystalline nanoporous,aluminosilicate materials that are widely used in the industry as adsorbents and catalysts.At the moment,roughly 240 zeolite framework types have been identified.Every new zeolite structure has to be approved by the International Zeolite Association(IZA)Structure Commission and will be designated with a code of three letters[36,37].Crystalline structure files of hypothetical zeolites can be acquired from the Predicted Crystallography Open Database(PCOD)(http://www.crystallography.net/pcod/)which contains over 330,000 structures within+30 kJ mol-1Si of a-quartz in which the known zeolites lie[38].

3.Characterization tools and methods

Structural feature characterizations.Characterization of the textural properties for the obtained structures,such as the pore limiting diameter(PLD)[22,39],largest cavity diameter(LCD)[40],accessible surface area[41],void fraction[42]and free volume[43],can provide useful information about the diversity of the structures,which are important knowledge in establishing reliable structure-property relationships.Moreover,characterization of PLD and LCD in advance can help determine which materials are inaccessible to guest molecules and thus narrow down the large database to a small one for HT screening.The tools for characterization of the textural properties of various porous materials can be Zeo++[28],Poreblazer[44].MOFs and zeolites can also be characterized by MOFomics and ZEOMICS[45].In addition to the textural properties,information about the structural topology can be analyzed by using TOPOS software[46].

4.1.3.PPNs

Application property predictions.In HT computational studies,molecular simulation methods are widely adopted by researchers to examine materials performance in connection with specific applications due to the accuracy and efficiency.Grand canonical Monte Carlo(GCMC)can be used to predict the adsorption properties of guest molecule in porous materials,such as adsorption amounts,heat of adsorption and separation selectivity.Molecular dynamics(MD)simulations can provide diffusion properties of guest molecules.Force field is a basic input demanded in GCMC or MD simulations to describe the energetic interactions.Generally,the nonbonded interatomic interactions are often modeled as a combination of Lennard-Jones(LJ)and Coulomb potentials[47].For MOFs,COFs and PPNs,LJ parameters for framework atoms are often taken from generic force fields such as the Universal or DREIDING Force Field(UFF)[48,49].For zeolites,several developed(transferrable)force fields for adsorption of CO2and CH4have been used in screening studies[50-55].In simulations,the framework atoms are usually fixed at their crystallographic coordinates.For the guest molecules,force fields that can reproduce experimental vapor-liquid equilibrium properties are recommended to be used,such as the TraPPE force fields for CO2and CH4[56].After HT computation,quantum mechanical calculations can be used to study several or dozens of structures of interest to obtain knowledge on details of electronic structure,minimum energy configurations and binding energies.

4.HT screening and design

With the aforementioned characterization tools and methods,the breakthrough of HT computational techniques in recent years greatly accelerate the pace of discovering new materials with excellent performance,by means of quickly identifying promising candidates and revealing comprehensive structure-property relationships.At the moment,HT computational researches are mainly focused on the adsorption of small gas molecules,including the most widely concerned CH4storage and CO2capture.

4.1.CH4storage

4.1.1.MOFs

Using a bottom-up structure generation approach(as shown in Fig.2),Wilmer et al.[29]constructed 137,953 hypothetical MOFs from a library of 102 building blocks(derived from the structures of real MOFs)and screened them for CH4storage.Their approach was validated by comparing the atomic coordinates and the computationally predicted adsorption isotherms of the resembled hypothetical and experimental structures.Both the atomic coordinates and the adsorption isotherms match are very close to each other.After screening,over 300 hypothetical MOFs were identified with higher adsorption amount at 35 bar than PCN-14,which is the world record holder at that time[57].They selected one of the promising MOF and verified the predicted capacity experimentally.The revealed structure-property relationships show that,the top-performing materials have a common void fraction around 0.8,and the most frequent pore sizes of the best MOFs are 4 and 8 Å which are exactly big enough for containing one or two CH4molecules.

对给排水管道进行布置、设计时应结合整个矿山总图规划。在管线设计过程中，要充分考虑管线的可用性，在可用性的基础上再考虑安全性。此外，还需对管线布置的简洁性和美观性进行优化。给排水管道布置、设计包含以下内容：

Using the above hypothetical MOF database,Fernandez et al.[58]undertook the further large-scale quantitative structure-property relationship(QSPR)analysis of MOFs for CH4storage.Several models were developed:decision trees(DTs),multilinear regression(MLR)models,and nonlinear support vector machines(SVMs).While DT and MLR are simple linear regressors,the SVMs are a broadly applicable machine learning method to many types of pattern recognition problems.DTs models can produce the rules for optimal design.The results indicate that for methane storage at 35 bar,best MOFs have void fractions larger than 0.52 and framework densities greater than 0.43 g/cm3;for the storage at 100 bar,best MOFs have void fractions larger than 0.62 and densities greater than 0.33 g/cm3.Nonlinear SVM models were found to outperform DTs and MLR in prediction of methane storage performance,with the accuracies of 85%at 35 bar and 93%at 100 bar.Properties of the dominant pore diameter and void fraction were most strongly correlated with materials storage performance.Fig.3 presents the results derived from the SVM model.Interestingly,the predications by the SVM models also show that there are unexplored MOFs with even better methane storage performance than the 137,953 hypothetical ones.

The above generation methods are mainly focused on the self-assembly process to place the SBUs into candidate periodic networks.The number of MOF-type compounds grows exponentially with length and branching structure of the organic ligands,however,the vast majority of these potentials may have poor methane storage properties.Thus,it is preferable to efficiently sample the part of the MOF composition space with favorable material performance.Rather than HT screening,Bao et al.develop a de novo evolutionary algorithm to evolve the MOFs to improve the methane deliverable capacity[59].In this algorithm,the generation was firstly initiated from a population of 100 linker molecules.If the newly constructed MOF has a larger deliverable capacity than the lowest one in the current population,the linker is inserted into the population in rank order(otherwise,the linker will be discarded),which are then used to build next generation of promising MOFs.After the initialization,the linker molecules of population remains 100 at all times during the algorithm running.Illustration of the algorithm is shown in Fig.4.Therefore,the authors did not exhaustively screen an existing database to get top-performing MOFs,but evolving materials(i.e.targeted design of MOFs)with better storage performance in each generation step.Finally,48 predicted linkers in networks of cds,nbo,acs,and pcu were found to surpass MOF-5,a typical material with good methane deliverable capacity under 65-5.8 bar pressure swing condition at 298 K.

HT screening studies have also been performed on large number of experimental MOFs.For example,Chung et al.carried out the methane storage study based on 4764 experimental MOFs in the CoRE MOF database[21].Although the CoRE MOF database is much smaller than the hypothetical MOF database,the former contains over 350 unique topologies while the latter only contains 6 topological nets,where over 90%of structures have the 6-coordinated pcu topology.The calculations predicted that MIL-53(Al)would be the top-performing material for methane storage capacity.They also investigated the structural properties of the CoRE MOFs that dominate the methane storage capacity and found that the relationships agree well with those obtained from a large database of hypothetical MOFs[29].Notably,both studies using the CoRE MOF and hypothetical MOF databases predicted that methane storage capacity is optimized at a helium void fraction of around 0.8 and that the deliverable capacity is maximized at the heats of adsorption between 10 and 15 kJ/mol.The above studies illustrate the potential of HT computational techniques to identify promising candidates for synthesis,together with the disclosed useful structureproperty relationships.

4.1.2.COFs

To date,only hypothetical COFs have been explored for methane storage using HT screening approaches.Martin et al.[34]constructed a hypothetical 3D-COF database which contains 4147 structures using top-down structure generation approach.The structure enumeration was restricted to established synthetic routes and compatible substrates.All the structures were generated according to the templates of dia,ctn,or bor topology.They screened the hypothetical COFs for methane storage under the conditions of pressure swing adsorption(PSA)process.The adsorption and desorption pressure are respectively 65 bar(or 35 bar)and 5.8 bar(or 1 bar)according to the Advanced Research Projects Agency-Energy(ARPA-E)of the DOE.Fig.5 shows that the highest deliverable capacity of the COF can achieve 181.67 v(STP)/v for the 65-5.8 bar operation.While the record high deliverable capacity(65-5.8 bar)at that time is held by the MOF-5 and HKUST-1,with deliverable capacities of～185 v(STP)/v[60,61].In addition,several structure-property relationships were revealed.For example,the highest deliverable capacities appear in a range of pore sizes from 8.5 to 13 Å,the deliverable capacities of the COF materials increase with surface area,and materials with low crystal density,high void fraction,small surface area,and a large included sphere approach an empty tank limit of 62 v(STP)/v.No clear relationships were found between the number of interpenetrated nets and performance.The regions that are most favorable for methane adsorption show the potential energy contours between-10 and-12 kJ/mol.

Black women have gradually internalized the subordinating position and remained to be the“Other”in the hegemonic America.

目前给高铁系统划分的应用频点是在5.8 GHz,如果系统对滤波器远端抑制要求较高,则高次谐波与寄生通带会严重影响带外的滤波性能(如图4)。通过图8 与图5及图4中的协同仿真结果对比,发现在通带指标影响不大的前提下,远端(8.7 GHz～10 GHz)谐波抑制改善了30 dB。说明带通与低通滤波器在级联时,其滤波性能具有一定的叠加性,特别是对通带远端谐波与寄生通带的抑制有着显著的改善作用。所以在设计有着远端抑制要求的高频段滤波器时,一般采用级联低通滤波器的方式去实现。

Using the top-down structure generation approach,Martin et al.[35]introduced a data set of hypothetical PPN materials,which comprises 17,846 predicted materials with dia topology[62],on the basis of commercially chemical monomers and experimentally established synthetic routes.In screening PPNs for methane storage,they found that pore diameter plays a vital role in governing the deliverable capacity as the PPNs are consisted of dia topology with vast space of pore sizes.For DC(65,5.8)and DC(35,1),respectively referring to deliverable capacities between 65(35)and 5.8(1)bar,the optimal pore diameters are approximately 10 Å and 7 Å.Moreover,for materials with suitable pore size,higher surface area generally benefits the deliverable capacity.For PPNs with large pore diameters,interpenetrated materials generally show higher deliverable capacities than non-interpenetrated ones due to the additionalattractive forcesfrom the framework atoms.Regarding to the heat of adsorption,structures with the highest deliverable capacities exhibit an intermediate heat of adsorption(color-coding in Fig.6).Furthermore,for the topperforming PPNs,the heat of adsorption increases with pressure(and thus loading)due to the favorable methane-methane interactions,which in turn attract more methane at the adsorption pressure while not at the desorption pressure.The highest DC(35,1)was 194 cm3v(STP)/v,and the highest DC(65,5.8)was 178 v(STP)/v.Totally,only three out of 17,846 materials surpassed the previous DOE target of 180 v(STP)/v,and only for DC(35,1),illustrating the degree of difficulty in designing materials to surpass this target,and let alone to achieve the more recent DOE target of 315 v(STP)/v.

在高等教育的今天，就业压力异常大，因此从新生入学开始就对他们进行职业生涯规划教育，让学生早期就对自己有个目标和规划，让同学朝着今后的目标奋斗。

最高长官的临阵脱逃，直接动摇了守城官兵的信心，这让临危受命，誓死决战的陈颐磊非常寒心。现在，唯一能掌控守城军心的就靠他了。好在六十七师是自己的老部队，从上到下都是自己培植的亲信。

4.1.4.Limit of porous materials for CH4storage

The MGI aims to enhance the understanding of the fundamentals of materials science by providing useful information needed to accelerate the development of new materials.Simon et al.[14]collected over 650,000 materials and compared their methane uptakes on a consistent basis,and employed the computational results to search for high performing adsorbents to store natural gas in vehicular fuel tanks.The screened materials include hypothetical MOFs,PPNs,ZIFs(a type of zeolite-like MOFs),zeolites and experimental MOFs.Firstly,the geometric properties characterizations depicted in Fig.7a shows that PPNs generally have the lowest crystal densities while zeolites have the highest,in contrast,MOFs have a broad density distribution.Both the geometric void fraction and pore size correlates inversely with the crystal density.Among the several types of materials,MOFs and PPNs inclined to achieve the highest surface areas.Then,the authors calculated the DC(65,5.8)to assess thermodynamic limits to methane storage performance and identify promising candidate materials.

Generally,MOFs and PPNs have significantly greater deliverable capacities than zeolites.The predicted highest deliverable capacity in the 650,000 materials is 196 v(STP)/v,as shown in Fig.7b.The materials with the best experimentally measured deliverable capacities to date are MOF-519(208 v(STP)/v),UTSA-76(194 v(STP)/v),HKUST-1(185 v(STP)/v)and MOF-5(185 v(STP)/v),the limiting deliverable capacity was found around 200 v(STP)/v from both the large screening data and experimental results,quite far from the ARPA-E target of 315 v(STP)/v.The authors therefore suggested some other sets of operating pressures and temperatures that might make the natural gas tanks more appealing.In addition,the huge amounts of data show that the crystal densities of materials with the highest deliverable capacities appear in a narrow range of 300-800 kg/m3.The optimal pore diameter is 11 Å,which provides a useful guidedance for synthesis work since the size of the building blocks used in the experiment directly governs the pore size.The comprehensive structure/performance relationships open up the possibility for machine-learning techniques and datamining to rapidly search the structural space and predict material performance from simple,easily computed structural descriptors.This work is also an example illustration of how the MGI approach can drastically reduce the discovering period of novel materials by eliminating unproductive tasks,allowing experimental efforts to be deployed on the most promising candidates that demonstrate the highest performance.

4.2.CO2capture

To date,the HT computational screening studies related to CO2capture are mainly focused on MOFs and zeolites.Different from CH4molecules,which are usually modeled as neutral species,framework charges should be considered for CO2capture simulations since coulombic interactions are important for CO2.Fast methods[63-66]for estimating MOF partial charges,such as the connectivity-based atom contribution(CBAC)[63]method and extended charge equilibration methods(EQeq)[65],are the two approaches that are frequently adopted in the HT computations rather than using expensive quantum mechanical(QM)calculations.These works greatly facilitate the efforts on revealing clear structure-property relationships,applying novel algorithms to quickly identify or construct high-performance materials,and developing optimal processes to select cost-effective materials for CO2capture.

MOFs are crystalline materials containing metal clusters connected by organic linkers.The Cambridge Structural Database(CSD)includes the most complete crystal structures of experimental MOFs[18].The latest study by Moghadam and coworkers show that the number of MOFs in CSD has increased dramatically in the past decade to an estimated number of ca.70,000 materials,as shown in Fig.1[19].To build a database which is exclusively devoted to MOFs,several groups put efforts on extracting MOFs from CSD[19-22].Among them,Chung et al.created the publicly available “Computation-Ready,Experimental MOFs”(CoRE MOF)database which contains over 5109 3D MOFs with atomic coordinates online(https://doi.org/10.11578/1118280)[23].Moghadam et al.[19]report the most complete collection of 69,666 MOFs maintained and updated by the Cambridge Crystallographic Data Centre(CCDC),in which the“CSD non-disordered MOF subset”containing 54,808 nondisordered MOF structures can be directly used in HT computations.However,it was not mentioned whether the missing atoms like H atoms have been added in these structures,which often occurs in the deposited structures.For constructing hypothetical MOFs,several groups have used top-down or bottom-up approaches to generate the structures[24-29].The bottom-up approach works in the way of sequentially connecting secondary building units(SBUs)until forming a periodic crystal structure.The top-down approach begins with a given topology or net,and the appropriate building blocks are then placed onto the net to construct the structure,as illustrated in Fig.2.At present,an accessible hypothetical MOFs database on large scale was provided by Wilmer et al.[29]which contains 137,953 hypothetical MOFs(see the website:hmofs.northwestern.edu).

Using the large amount of 130,000 hypothetical MOFs,Wilmer et al.[67]established clear correlations between the CO2separation ability and purely structural characteristics of MOFs,which were previously impossible to discern due to the small sample size available.In this work,CO2/CH4and CO2/N2mixtures with different compositions were separated using PSA or VSA(vacuum-swing adsorption)conditions.Besides the metric of selectivity,other four adsorbent evaluation criteria from the Engineering literature shown in Table 1 were used to comprehensively assess the potentials of MOFs in CO2 separations.The five metrics of different cases were correlated with structural characteristics(i.e.pore volume,pore size and surface area),heat of adsorption of CO2(Qst),and chemical features(i.e.,functional groups).The best-performed materials in each correlation correspond to an optimum region of the structural characteristics and Qst.For example,it was found that for CO2/CH4(50/50)separation using PSA process(adsorption at 5 bar,desorption at 1 bar),high-workingcapacity materials are those with Qstvalues of～21 kJ/mol and void fractions of 0.8.Chemical functional groups,especially those with fluorine or chlorine atoms,are usually ranked in the best performers in all separation cases,and approximately 50%of those with the highest selectivity(S)contain fluorine groups.The resulting qualitative structure-property relationships provide useful guidance for experimental synthesis going forward.

根据规范，管网漏失水量和未预见水量之和，按上述用水量之和的10%取值，则上述W1和W2加上未预见水量后，分别为489 m3/d、3.1 m3/d，共492.1 m3/d，最终确定本次工程日用水总量为500 m3/d。为了控制工程规模同时结合调蓄水池的调节作用，确定供水时长为10 h。同时日变化系数取1.2，最终计算得设计流量为60 m3/h。供水衔接工程为向三交镇供水，日用水量为500 m3/d，取时变化系数为3，供水时长16 h，则供水衔接工程设计流量为93.8 m3/h。

Besides the hypothetical MOFs,CO2capture capability of experimental MOFs under humidity environment were also evaluated by Li et al.[23].Frequently the appearance of H2O molecules may compete for adsorption sites,while dehumidification before CO2capture increases process complicacy and cost.Therefore,they carried out a large-scale screening to search for MOFs with high selectivity toward CO2over H2O from CoRE database which contains 5109 MOFs.The authors first computationally identified MOFs with high CO2/H2O selectivities based on the ratio of Henry's law constants of CO2 and H2O.These top MOFs were subsequently demonstrated to perform well for CO2/N2separation.Study on the effects of framework charges shows that,for CO2and N2interactions with MOFs,the van der Waals interaction plays a more important role than electrostatic interactions.For polar H2O molecules,electrostatic interactions are more important than the van der Waals ones between H2O and MOF atoms,therefore,accurate charge calculation methods should be used to simulate H2O con fined in MOFs.

Besides CO2adsorption separation,membrane-based separation using zeolites has also been investigatedviaHTscreening approach.Kim et al.[73]conducted a large-scale screening of 87,000 zeolites including IZA structures and hypothetical pure silica zeolite structures to evaluate their CO2/N2and CO2/CH4 membrane separation capabilities.Compared with the calculations in the scope of adsorption processes,screening of membranes demands not only the information on the adsorption properties but also the diffusion coefficients,which requires expensive MD simulations.To avoid performing MD simulations for thousands of structures,they take full advantage of the information obtained from the free energy landscape of the crystal structure and apply the transition-state theory(TST)to calculate the diffusion properties.Moreover,the algorithm was mapped to the HT processing power of the graphics processing units(GPUs),enabling an accurate and fast characterization of the adsorption and the diffusion properties of 87,000 zeolites.The screening also showed that the top structures in the predicted zeolite database outperform the best known zeolite by a factor of 4-7.Mechanism analysis revealed that the topper forming structures are usually featured by channels with uniformly spread adsorption sites across the entire channel so that they can well balance the CO 2adsorption and the diffusion properties.They provided a simple experimental signature to discern such a material forCO2/CH4separation:arelativelylow heat of adsorption(i.e.,-30to-20kJ/mol)and anintermediate range of CO2Henry coefficient(i.e.,10-5＜＜10-4mol/(kg⋅Pa)).Finally,an optimal set of zeolite structures was identified for an inverse process,in which CO2is intercepted while N2or CH4can go through a membrane.

做到这一点又难又不难。说难，那是需要高智商和好体力的；说不难，只要准确理解不同的身份要求，并扮演好自己的角色就行了。当然，身份多了，难免会很累。但要在这个世界上出人头地，就不能怕累。张仲平是一个谋定而后动的人，已经习惯了在做任何事情时都权衡利弊，他觉得，只有这样，才称得上一个真正的商人和一个真正成功的男人。

To rapidly recognize high-performance MOFs for CO2 capture rather than exhaustive screening all the hundreds and thousands of MOFs,Fernandez et al.[68]developed QSPR models using advanced machine learning algorithms to make the virtual screening of large searching space feasible.They introduced material geometrical features(void fraction,pore size and surface area),and more importantly,the descriptor of atomic property-weighted radial distribution function(APRDF)that capture the chemical features of a periodic material,from which accurate QSPR models were developed to predict the absolute CO2uptakes of MOFs.In this work,the database contains 324,500 predicted MOFs generated by connecting 66 SBUs and 19 functional groups.In the beginning,10%of the database(32,450 MOFs)was randomly selected to form the calibration set which was used to train the QSPR models,while the remaining MOFs constituted the test set for validating the models.Machine learning techniques of SVMs were used to accurately distinguish the complex RDF pro files.The QSPR classifier could discover 945 of the top 1000 MOFs in the test set while flagging only 10%of the large library for exhaustively screening.Thus,the methodology applied in this work would result in reduction of computational time in an order of magnitude and thus allow the sreening of intractably large structure database and searching space.

Similar to the idea of Bao et al.[59]mentioned previously,Chung et al.[69]report the in silico discovery of highperformance adsorbentsfor precombustion CO2capture(CO2/H2separation),where a genetic algorithm(GA)was applied to efficiently design top candidates based on a large database of MOFs.Three independent GA runs were carried out to separately optimize three different evaluation criteria,namely,the CO2/H2selectivity(),the CO2working capacity(ΔN1),and the adsorbent performance score(APS),which is the product of the former two quantities.As the GA running,the top-performing MOF evolved after each generation,for example,the CO2/H2selectivity improved from ca.700 to 2600,the CO2working capacity improved from ca.7 mol/kg to 8 mol/kg,and the APS improved from ca.1000 mol/kg to 1200 mol/kg.The GA can reduce the computational time by at least two orders of magnitude compared with a brute force search.NOTT-101/OEt,one of the top-performing MOFs,was synthesized and tested.The IAST-predicted mixture isotherms show that the CO2working capacity and CO2/H2selectivity of NOTT-101/OEt are 3.8 mol/kg and 60,respectively;the former quantity is the highest among all the MOFs reported publicly under the same operating conditions at that time.In addition,the obtainedstructure-property relationships in this work could be used to search for top-performing MOFs in different databases without the need of additional large-scale simulations.

To develop cost-effective materials and processes for CO2 capture,HT screening methodology has also been applied to evaluate both IZA structures and large number of hypothetical zeolites to attain comprehensive information on zeolites family.Lin et al.[72]determined an optimization process by minimizing the electric load imposed on a power plant through a temperature-pressure swing process when using a material.The minimum load,called parasitic energy(PE),was used as a evaluator to make a comparison of different materials.They set a target of seeking the materials with a PE significantly lower than 1060 kJ(kg CO2)-1,which is a reference value of the state-of-the-art amine capture process.The screening showed a large amount of zeolite structures that have a PE below the current technology 1060 kJ(kg CO2)-1.Investigation of these cost-effective structures highlights their diversity:there are one-,two-,and three-dimensional channel structures,cage-like topologies,and even more complex geometries.The calculations also identified many predicted structures with lower PE values than that of the known structures.Fig.9 shows the correlation that optimal regions of CO2Henry coefficient and CO2binding energy for getting ideal parasitic energy.A clear conclusion of their study is that a highperforming carbon capture material should have a plenty number of adsorption sites with medium binding energy,which is adequately large to be selective,but not too large to bring difficulty for the desorption.

在Matlab中调用其模糊逻辑工具箱，根据之前的分析和控制策略设计机械臂拾取系统的模糊控制器[6]。设计流程为：在Matlab命令窗口输入fuzzy调出模糊逻辑控制器，建立模糊控制器的输入输出量和控制规则之间的逻辑关系，确定各自适用的隶属度函数以及之前确定的各个变量所对应的论域，根据模糊规则表输入模糊控制规则，完成模糊控制器的设计。最终得到能够全面反映控制器输入位移偏差和位移偏差变化率以及输出压力之间关系的三维曲面视图如图5，所得到的三维曲面显示越光滑平整其系统的控制效果会越好。

传统的假日购物季通常从黑色星期五开始，即感恩节第二天，然后一直持续至圣诞节。在这期间，零售商会打折，购物者则去寻找礼物、在商店排队、匆忙跑过通道以获得最佳优惠。这是以前的常态，当然在一些地区现在仍是常态，不过由于百货店和在线零售商不断更改折扣日期，在很多地方，购物季早在感恩节之前便已经开始，而且成为动态事件，没有固定的日期。

ILME技术是采用离子液体作为萃取剂，取代了传统的有机试剂，其萃取操作过程与传统萃取相似，加入所有试剂定容后可采用手摇或者超声辅助的方式完成离子液体的分散和萃取。

Aiming at the experimental zeolites in IZA database which now contains approximately 240 framework types,many groups[38,55,70,71]have carried out simulation studies to evaluate CO2capture performance of the IZA zeolite structures.One of the main obstacles to deploy carbon dioxide capture and storage(CCS)in a large-scale in power plants is the energy cost required to separate the CO2from flue gas.To searching for novel materials for cost-effective CO2capture,Faruque Hasan et al.[70]conducted an adsorbent screening by combining material characterization and process optimization methodologies,as illustrated in Fig.8.195 IZA zeolites are screened based on metrics of size,shape,and adsorption selectivities.They optimized an adsorption process for the selected zeolites to prepare a final rank-ordered list based on the total cost of capture and compression.During optimization,the optimal process conditions that can satisfy recovery,purity,and other process constraints were obtained.The top ten zeolites(NAB,AHT,AWO,ABW,TON,MVY,VNI,WEI,OFF and ITW)can capture and compress CO2to 150 bar from a CO2/N2(14/86)mixture with cost less than$30 per ton of CO2captured.

4.2.2.Zeolites

车地无线通信采用成熟的LTE技术。该技术具备高可靠的抗干扰能力，可满足互联互通CBTC系统车地之间数据在高速移动环境下的稳定传输[7]。同时，针对空口消息的伪装风险，可采用安全加密技术防护，将其直接部署在TAU(车载终端)和BBU(轨旁基带单元)上来实现鉴权和加密机制，保障车地无线通信的信息安全。安全加密技术采用满足LTE国际加密标准的国密算法——祖冲之(ZUC)算法。

5.Remarks and outlooks

During the past few years,HT computational screening techniques have been extensively used in areas like batteries[74],solar cells[75],drug discovery[76],energy storage[77],thermal conductivity[78]and catalysis[79].HT tools expand our horizons on rapidly evaluating hundreds and thousands of materials,identifying bestcandidates and establishing comprehensive structure-property relationships for specific application fields.It has been widely recognized that the knowledge obtained from HT computational screening can indicate the direction of synthesis efforts,while the experimental validation results can in turn help the rectification of the deviations in HT computational predictions.We anticipate that the computational approaches along with closely coupled experimentations will play a significant role in the discovery of advance materials in various research fields.Regarding to the domains of CH4storage and CO2capture,we think the following are the challenges should be devoted to in the future:

1.For hypothetical database,high efficient self-assembly technique plays a key role in generating numerous structures.A major problem of present self-assembly technique is that the generated topologies are very limited.For example,experimental MOFs contains more than 350 topologies,while the 137,953 hypothetical MOFs generated by Snurr and co-workers are only in 6 topologies.Therefore,the topology-diversity needs to be enriched in future development of high efficient self-assembly techniques.The success of self-assembly techniques in porous materials will stimulate their applications in other kinds of advanced materials.

2.Development of simulation methods with higher efficiency in large-scale screening.As illustrated above,QSPR models[58,68]and machine learning techniques can contribute a lot to data-mining,and GPUs[73]have been used to speed up the simulations.Other methods[80,81]reported in early studies that can accelerate calculations may also be considered in future HT screenings.

3.Development of algorithms for purposefully designing high-performance materials.Bao et al.[59]initiated the in silico GA to evolve and optimize MOFs in methane storage.Chung et al.[69]made further efforts on developing GA and realized the design and synthesis of promising MOFs through cooperation with experimental researchers.In this way,the computational investigations should not try to screen large database of materials since the majority may perform poorly for a certain application,but rather automatically designing materials according to certain metrics like high selectivity or high deliverable capacity.

4.Improved force fields that can accurately describe the interactions involved in special materials.Introducing open metal sites(OMSs)or novel functional groups into MOF structures are widely adopted as efficient means to achieve the materials with excellent application performance.If the force fields can be generated from high-level first principle calculations automatically,it will greatly benefit HT computations.

5.Employment of more comprehensive assessment metrics.In addition to the common metrics that describe the adsorption/separation performance,metrics related to practical use,such as stability,effectiveness and energy cost,are also useful information that deserve to be investigated in HT computations.

6.Efforts with the incorporation of experimentalist considerations.To date,only a limited number of generated structures have been experimentally examined from the inspiration of the HT screening and design[29,69,82].As a typical example,NOTT-101/OEt,which was computationally identified as a top-performing material for CO2/H2,was synthesized and tested by Chung et al.[69].This material exhibits a higher CO2working capacity than any MOF reported in the literature under the same operating conditions.However,record-breaking validation has never been an easy job.The reasons may be:(i)performance of the predicted structures are not superior enough that deserve the experimentalists to have a try;(ii)the predicted structures are hard to be realized by experiments.For the first question,with the obtained useful structure-property relationships,computational chemists can make bold assumptions and manage to design novel materials that may have significantly improved performance than present materials used industrially.These materials may be difficult to be synthesized at the moment,but can provide new insights for future study.For the second one,computational chemists need to perform a deep and thorough research on the type of structures that experimental chemists expert at (i.e., establishing customized structure-property relationships for these structures)and endeavor to generate new structures with improved performance.

Conflict of interest

The authors declare no conflict of interest.

Acknowledgment

This work was supported by the Natural Science Foundation of China(Nos.21706106,21536001 and 21322603),the National Key Basic Research Program of China(“973”)(No.2013CB733503),the Natural Science Foundation of Jiangsu Normal University(16XLR011)and Priority Academic Program Development of Jiangsu Higher Education Institutions.

References

[1]Earth's CO2.http://co2now.org/(Accessed 7 May 2017).

[2]S.Chu,Science 325(2009)1599.

[3]R.S.Haszeldine,Science 325(2009)1647-1652.

[4]H.Yang,Z.Xu,M.Fan,R.Gupta,R.B.Slimane,A.E.Bland,I.Wright,J.Environ.Sci.20(2008)14-27.

[5]K.Sumida,D.L.Rogow,J.A.Mason,T.M.McDonald,E.D.Bloch,Z.R.Herm,T.-H.Bae,J.R.Long,Chem.Rev.112(2012)724-781.

[6]H.Furukawa,O.M.Yaghi,J.Am.Chem.Soc.131(2009)8875-8883.

[7]W.Lu,J.P.Sculley,D.Yuan,R.Krishna,Z.Wei,H.-C.Zhou,Angew.Chem.Int.Ed.51(2012)7480-7484.

[8]T.-H.Bae,M.R.Hudson,J.A.Mason,W.L.Queen,J.J.Dutton,K.Sumida,K.J.Micklash,S.S.Kaye,C.M.Brown,J.R.Long,Energy Environ.Sci.6(2013)128-138.

[9]J.Stephenson,International Association of NGVs,1993.

[10]R.A.Alvarez,S.W.Pacala,J.J.Winebrake,W.L.Chameides,S.P.Hamburg,Proc.Natl.Acad.Sci.U.S.A.109(2012)6435-6440.

[11]S.Yeh,Energy Policy 35(2007)5865-5875.

[12]V.C.Menon,S.Komarneni,J.Porous Mater.5(1998)43-58.

[13]T.A.Makal,J.-R.Li,W.Lua,H.-C.Zhou,Chem.Soc.Rev.41(2012)7761-7779.

[14]C.M.Simon,J.Kim,D.A.Gomez-Gualdron,J.S.Camp,Y.G.Chung,R.L.Martin,R.Mercado,M.W.Deem,D.Gunter,M.Haranczyk,D.S.Sholl,R.Q.Snurr,B.Smit,Energy Environ.Sci.8(2015)1190-1199.

[15]C.Mellot-Draznieks,J.M.Newsam,A.M.Gorman,C.M.Freeman,G.Férey,Angew.Chem.,Int.Ed.39(2000)2270-2275.

[16]Y.J.Col′on,R.Q.Snurr,Chem.Soc.Rev.43(2014)5735-5749.

[17]A.Jain,S.P.Ong,G.Hautier,W.Chen,W.D.Richards,S.Dacek,S.Cholia,D.Gunter,D.Skinner,G.Ceder,K.A.Persson,Apl.Mater.1(2013)011002.

[18]F.H.Allen,Acta Crystallogr.,Sect.B Struct.Sci.58(2002)380-388.

[19]P.Z.Moghadam,A.Li,S.B.Wiggin,A.Tao,A.G.P.Maloney,P.A.Wood,S.C.Ward,D.Fairen-Jimenez,Chem.Mater.29(2017)2618-2625.

[20]J.Goldsmith,A.G.Wong-Foy,M.J.Cafarella,D.J.Siegel,Chem.Mater.25(2013)3373-3382.

[21]Y.G.Chung,J.Camp,M.Haranczyk,B.J.Sikora,W.Bury,V.Krungleviciute,T.Yildirim,O.K.Farha,D.S.Sholl,R.Q.Snurr,Chem.Mater.26(2014)6185-6192.

[22]T.Watanabe,D.S.Sholl,Langmuir 28(2012)14114-14128.

[23]S.Li,Y.G.Chung,R.Q.Snurr,Langmuir 32(2016)10368-10376.

[24]I.A.Baburin,S.Leoni,CrystEngComm 12(2010)2809-2816.

[25]L.-C.Lin,A.H.Berger,R.L.Martin,J.Kim,J.A.Swisher,K.Jariwala,C.H.Rycroft,A.S.Bhown,M.W.Deem,M.Haranczyk,B.Smit,Nat.Mater.11(2012)633-641.

[26]H.Hayashi,A.P.Cote,H.Furukawa,M.O'Keeffe,O.M.Yaghi,Nat.Mater.6(2007)501-506.

[27]D.W.Lewis,A.R.Ruiz-Salvador,A.Gomez,L.M.RodriguezAlbelo,F.-X.Coudert,B.Slater,A.K.Cheetham,C.Mellot-Draznieks,CrystEngComm 11(2009)2272-2276.

[28]T.F.Willems,C.H.Rycroft,M.Kazi,J.C.Meza,M.Haranczyk,Microporous Mesoporous Mater.149(2012)134-141.

[29]C.E.Wilmer,M.Leaf,C.Y.Lee,O.K.Farha,B.G.Hauser,J.T.Hupp,R.Q.Snurr,Nat.Chem.4(2012)83-89.

[30]M.Tong,Y.Lan,Q.Yang,C.Zhong,Chem.Eng.Sci.168 456-464

[31]S.Bureekaew,R.Schmid,CrystEngComm 15(2013)1551-1562.

[32]B.Lukose,A.Kuc,J.Frenzel,T.Heine,Beilstein J.Nanotechnol.1(2010)60-70.

[33]B.Lukose,A.Kuc,T.Heine,J.Mol.Model.19(2013)2143-2148.

[34]R.L.Martin,C.M.Simon,B.Medasani,D.K.Britt,B.Smit,M.Haranczyk,J.Phys.Chem.C 118(2014)23790-23802.

[35]R.L.Martin,C.M.Simon,B.Smit,M.Haranczyk,J.Am.Chem.Soc.136(2014)5006-5022.

[36]C.Baerlocher,W.M.Meier,D.H.Olson,Atlas of Zeolite Framework Types,seventh ed.,Elsevier,Amsterdam,2007.

[37]International Zeolite Association.http://www.iza-online.org/.

[38]I. Matito-Martos, A. Martin-Calvo, J.J. Gutiérrez-Sevillano,M.Haranczyk,M.Doblare,J.B.Parra,C.O.Aniae,S.Calero,Phys.Chem.Chem.Phys.16(2014)19884-19893.

[39]E.Haldoupis,S.Nair,D.S.Sholl,J.Am.Chem.Soc.132(2010)7528-7539.

[40]T.Van Heest,S.L.Teich-McGoldrick,J.A.Greathouse,M.D.Allendorf,D.S.Sholl,J.Phys.Chem.C 116(2012)13183-13195.

[41]T.Düren,F.Millange,G.Ferey,K.S.Walton,R.Q.Snurr,J.Phys.Chem.C 111(2007)15350-15356.

[42]A.L.Myers,P.A.Monson,Langmuir 18(2002)10261-10273.

[43]E.M.Sevick,P.A.Monson,J.M.Ottino,J.Chem.Phys.88(1988)1198-1206.

[44]L.Sarkisov,A.Harrison,Mol.Simul.37(2011)1248-1257.

[45]E.L.First,C.E.Gounaris,J.Wei,C.A.Floudas,Phys.Chem.Chem.Phys.13(2011)17339-17358.

[46]E.V.Alexandrov,V.A.Blatov,A.V.Kochetkov,D.M.Proserpio,CrystEngComm 13(2011)3947-3958.

[47]Q.Yang,D.Liu,C.Zhong,J.-R.Li,Chem.Rev.113(2013)8261-8323.

[48]S.L.Mayo,B.D.Olafson,W.A.Goddard,J.Phys.Chem.94(1990)8897-8909.

[49]A.K.Rappé,C.J.Casewit,K.S.Colwell,W.A.G.III,W.M.Skiff,J.Am.Chem.Soc.114(1992)10024-10035.

[50]K.Makrodimitris,G.K.Papadopoulos,D.N.Theodorou,J.Phys.Chem.B 105(2001)777-788.

[51]A.I.Skoulidas,D.S.Sholl,J.Phys.Chem.B 106(2002)5058-5067.

[52]A.García-Sánchez,C.O.Ania,J.B.Parra,D.Dubbeldam,T.J.H.Vlugt,R.Krishna,S.Calero,J.Phys.Chem.C 113(2009)8814-8820.

[53]D.Dubbeldam,S.Calero,T.J.H.Vlugt,R.Krishna,T.L.M.Maesen,B.Smit,J.Phys.Chem.B 108(2004)12301-12313.

[54]D.Dubbeldam,S.Calero,T.J.H.Vlugt,R.Krishna,T.L.M.Maesen,E.Beerdsen,B.Smit,Phys.Rev.Lett.93(2004)088302.

[55]H.Fang,A.Kulkarni,P.Kamakoti,R.Awati,P.I.Ravikovitch,D.S.Sholl,Chem.Mater 28(2016)3887-3896.

[56]M.G.Martin,J.I.Siepmann,J.Phys.Chem.B 103(1999)4508-4517.

[57]S.Q.Ma,D.F.Sun,J.M.Simmons,C.D.Collier,D.Q.Yuan,H.C.Zhou,J.Am.Chem.Soc.130(2008)1012-1016.

[58]M.Fernandez,T.K.Woo,C.E.Wilmer,R.Q.Snurr,J.Phys.Chem.C 117(2013)7681-7689.

[59]Y.Bao,R.L.Martin,C.M.Simon,M.Haranczyk,B.Smit,M.W.Deem,J.Phys.Chem.C 119(2015)186-195.

[60]J.A.Mason,M.Veenstra,J.R.Long,Chem.Sci.5(2014)32-51.

[61]Y.Peng,V.Krungleviciute,I.Eryazici,J.T.Hupp,O.K.Farha,T.Yildirim,J.Am.Chem.Soc.135(2013)11887-11894.

[62]M.O'Keeffe,M.A.Peskov,S.J.Ramsden,O.M.Yaghi,Acc.Chem.Res.41(2008)1782-1789.

[63]Q.Xu,C.Zhong,J.Phys.Chem.C 114(2010)5035-5042.

[64]E.Haldoupis,S.Nair,D.S.Sholl,J.Am.Chem.Soc.134(2012)4313-4323.

[65]C.E.Wilmer,K.C.Kim,R.Q.Snurr,J.Phys.Chem.Lett.3(2012)2506-2511.

[66]E.S.Kadantsev,P.G.Boyd,T.D.Daff,T.K.Woo,J.Phys.Chem.Lett.4(2013)3056-3061.

[67]C.E.Wilmer,O.K.Farha,Y.-S.Bae,J.T.Hupp,R.Q.Snurr,Energy Environ.Sci.5(2012)9849-9856.

[68]M.Fernandez,P.G.Boyd,T.D.Daff,M.Z.Aghaji,T.K.Woo,J.Phys.Chem.Lett.5(2014)3056-3060.

[69]Y.G.Chung,D.A.G′omez-Gualdr′on,P.Li,K.T.Leperi,P.Deria,H.Zhang,N.A.Vermeulen,J.F.Stoddart,F.You,J.T.Hupp,O.K.Farha,R.Q.Snurr,Sci.Adv.2(2016)e1600909.

[70]M.M.Faruque Hasan,E.L.Firstz,C.A.Floudas,Phys.Chem.Chem.Phys.15(2013)17601-17618.

[71]R.Krishna,J.M.van Baten,J.Membr.Sci.360(2010)323-333.

[72]L.-C.Lin,A.Berger,R.L.Martin,J.Kim,J.A.Swisher,K.Jariwala,C.H.Rycroft,A.Bhown,M.W.Deem,M.Haranczyk,B.Smit,Nat.Mater.11(2012)633-641.

[73]J.Kim,M.Abouelnasr,L.-C.Lin,B.Smit,J.Am.Chem.Soc.135(2013)7545-7552.

[74]S.Kirklin,B.Meredig,C.Wolverton,Adv.Energy Mater.3(2013)252-262.

[75]A.Martsinovich,J.Phys.Chem.C 115(2011)11781-11792.

[76]N.Okimoto,N.Futatsugi,H.Fuji,A.Suenaga,G.Morimoto,R.Yanai,Y.Ohno,T.Narumi,M.Taiji,Plos.Comput.Biol.5(2009)e1000528.

[77]L.Cheng,R.S.Assary,X.Qu,A.Jain,S.P.Ong,N.N.Rajput,K.Persson,L.A.Curtiss,J.Phys.Chem.Lett.6(2015)283-291.

[78]C.Toher,J.J.Plata,O.Levy,M.Jong,M.Asta,M.B.Nardelli,S.Curtarolo,Phys.Rev.B 90(2014)174107.

[79]M.P.Andersson,T.Bligaard,A.Kustov,K.E.Larsen,J.Greeley,T.Johannessen,C.H.Christensen,J.K.Nørskov,J.Catal.239(2006)501-506.

[80]D.W.Siderius,V.K.Shen,D.W.Siderius,V.K.Shen,J.Phys.Chem.C 117(2013)5861-5872.

[81]K.Yu,J.G.McDaniel,J.R.Schmidt,J.Chem.Phys.137(2012)244102.

[82]D.A.Gomez-Gualdron,O.V.Gutov,V.Krungleviciute,B.Borah,J.E.Mondloch,J.T.Hupp,T.Yildirim,O.K.Farha,R.Q.Snurr,Chem.Mater.26(2014)5632-5639.