Targeted classification of metal-organic frameworks in the Cambridge structural database (CSD)

Abstract

Large-scale targeted exploration of metal-organic frameworks (MOFs) with characteristics such as specific surface chemistry or metal-cluster family has not been investigated so far. These definitions are particularly important because they can define the way MOFs interact with specific molecules (e.g. their hydrophilic/phobic character) or their physicochemical stability. We report here the development of algorithms to break down the overarching family of MOFs into a number of subgroups according to some of their key chemical and physical features. Available within the Cambridge Crystallographic Data Centre's (CCDC) software, we introduce new approaches to allow researchers to browse and efficiently look for targeted MOF families based on some of the most well-known secondary building units. We then classify them in terms of their crystalline properties: metal-cluster, network and pore dimensionality, surface chemistry (i.e. functional groups) and chirality. This dynamic database and family of algorithms allow experimentalists and computational users to benefit from the developed criteria to look for specific classes of MOFs but also enable users - and encourage them - to develop additional MOF queries based on desired chemistries. These tools are backed-up by an interactive web-based data explorer containing all the data obtained. We also demonstrate the usefulness of these tools with a high-throughput screening for hydrogen storage at room temperature. This toolbox, integrated in the CCDC software, will guide future exploration of MOFs and similar materials, as well as their design and development for an ever-increasing range of potential applications.

Fig. 1. Histograms comparing geometric properties for all the porous MOFs in the CSD MOF subset from 1995 to 2015. (a) Largest cavity diameter (LCD), (b) pore limiting diameter (PLD), (c) void fraction, (d) density, (e) gravimetric surface area, (f) volumetric surface area. The animated version of these graphs can be found in the ESI. All family-property relationships of the 8253 porous MOFs presented in this work can be found online at ; http://aam.ceb.cam.ac.uk/mof-explorer/CSD_MOF_subset.

Fig. 2. Criteria developed for the identification of MOF families in the CSD MOF subset based on specific secondary building units and their connection to the organic linkers. The target MOF families are zirconium oxide, MOF-74/CPO-27-like, ZIF-like, zinc oxide and IRMOF-like, as well as Cu–Cu paddle-wheeled materials. (a–d) Diagrams used to look for structures containing Cu–Cu paddlewheels. The dotted box for (c) and (d) means the structures inside should be considered as one single query. The red diagrams are queries used to eliminate undesired structures. See ESI for more details on each MOF family.

Fig. 3. Criteria developed to identify MOFs with common functionalities in the CSD MOF subset. (a) Polar groups (–NH₂, –NO₂, –CN, –COOH and –OH). For the –CN case, the red box represents queries which target dicyanides that are chosen to be eliminated. This dicyanide search is obtained via a combination of one “must-have” query and two “must not have” queries. The green diagram is thus an overall negative and the red diagrams are double negatives; (b) alkoxys (methoxy, ethoxy, propyloxy); (c) alkyls (methyl, ethyl, propyl); (c′) alkyls (with more than 4 carbon atoms on the left) and (d) halogens (–F, –Cl, –Br), and structures with perfluoroalkane groups. The variable bonds are all the same type for queries within the grey dotted box: single, double, aromatic or delocalized. For the three queries outside of the grey dotted box, the variable bonds are either aromatic or delocalized. See ESI for more details on each functional group.

Fig. 4. Histograms of the geometric properties of 1911 chiral structures with non-zero gravimetric surface area in the CSD MOF subset. (a) Largest cavity diameter, (b) pore limiting diameter, (c) void fraction, (d) density, (e) gravimetric surface area, (f) volumetric surface area.

Fig. 5. Analysis of MOFs included in the CSD. (a) Histograms of framework and channel/pore dimensionalities characterized for the 52 787 structures. (b) Non-cumulative evolution of R factors of the MOF subset from 1960 to 2015. Blue: boxplots of R-factors per year. Percentiles used: 1% (lower dash symbol), 25% (lower cross symbol), 50% (dash in the box), 75% (upper cross symbol), 99% (upper dash symbol). A black line connects the means across all the boxes; the orange curve shows the percentage of structures added to the database per year. The orange area under the orange curve highlights the number of structures with an R-factor higher than 10%. (c) Distribution of R-factors and density across different MOF families and crystal systems of low or high symmetry.

Fig. 6. Characterization of the 3D MOFs screened for hydrogen storage. Volumetric uptake vs. absolute uptake wt% H₂ at room temperature at 200, 500 and 900 bar. Each circle represents a MOF structure. The sizes of the circles represent the LCD in all plots. The dashed line corresponds to the volumetric uptake obtained in an empty tank. (a–c) Families of the screened structures; structures that have not been assigned a family are colored in grey in the background. The highlighted structure BAZGAM is shown in the inset at 900 bar. (d–f) Percolation of the screened structures. Structures containing 1D, 2D and 3D pore channels are respectively represented in yellow, blue and purple. (g–i) Functional groups identified in the screened structures. Structures that have no particular functional groups identified are colored in grey in the background. Full hydrogen adsorption data can be found online at ; http://aam.ceb.cam.ac.uk/mof-explorer/H2_HTS.

Fig. 7. Quantitative characterization of the 3D MOFs screened for hydrogen storage boxplots of volumetric uptake of H₂ at room temperature at 200, 500 and 900 bars versus (a–c) families of the screened structures, (d–f) percolation of the screened structures and (g–i) functional groups identified in the screened structures. The jittered points in the background give an idea on the number of structures considered for each boxplot. The markers represent the minimum, first quartile, median, third quartile, and maximum values, respectively. Outliers are represented by black data points. The dashed line corresponds to the volumetric uptake obtained in an empty tank.

Fig. 8. (a) Evolution of the structure with the highest hydrogen volumetric uptake at room temperature and at 500 bars in the CSD over the years. Each circle represents a structure. The size indicates the LCD, the color the corresponding R-factor. Each new best performing structure is highlighted with a yellow circle and the yellow line tracks the best performing structure over the years. (b) Snapshot of a supercell of BOMCUB. The counter-ions and water molecules were removed from the snapshot for clarity. (c) Snapshot of a supercell of JARMEU.