porous materials AI gym

open data sets for machine learning pertaining to porous materials.

• MOF = metal-organic framework
• COF = covalent organic framework

crystal structures

experimental

• MOFs: CoRE MOFs (Paper, Database), CSD MOF Subset (Paper, Database), CSD MOF Collection (Paper, Database)
• COFs: CURATED COFs (Paper, Database), CoRE COFs (Paper, Database)
• zeolites: IZA database (Database)

hypothetical

• MOFs: B&W (Paper, Database), ToBaCCo (Paper, Database, Code), hMOFs (Paper, Database), Anderson et al. (Paper, Database), MOF-5 analogues (Paper, Database), PORMAKE (Paper, Code)
• COFs: Mercado et al. (Paper, Database), Haranczyk's 3D COF database (Paper, Database)

labeled porous materials for supervised learning

material class	target y	features x provided?	Reference	size of data set
MOFs (hypothetical)	CO2, N2 adsorption (sim)	yes	Paper, Database	ca. 325,000
MOFs (experimental and hypothetical)	Band gaps, density of states, charge densities (sim)	yes	Paper, Database	ca. 18,000
MOFs (experimental)	Color (exp)	yes	Paper, Database	?
COFs (hypothetical)	CH4 deliverable capacity (sim)	yes, hand-crafted features provided.	Paper, Database	ca. 70,000
COFs (experimental)	CH4, H2, O2, Xe, Kr, H2S adsorption (sim)	?	Paper	ca. 500
MOFs (hypothetical)	H2 adsorption (sim)	yes	Paper / Database	ca. 100K
MOFs (experimental)	thermal stability, solvent removal stability	yes (RAC & geometric features)	Paper / Database	ca. 2-3K (extracted from experimental lit)
MOFs (experimental)	CO2, H2O DFT-calculated adsorption energy	no	Paper, DataBase	ca. 8400 MOFs, but 38M DFT calcs

labeled nodes for supervised learning

material class	target y	Reference	size of data set (# materials)
MOFs (experimental)	DDEC6 charges on atoms (sim)	Paper, Database	ca. 3,000
MOFs (experimental and hypothetical)	DDEC6/CM5/Bader charges on atoms (sim)	Paper, Database	ca. 18,000 (DDEC6/CM5), ca. 5,000 (Bader)
MOFs (experimental and hypothetical)	Effective bond orders on atoms (sim)	Paper, Database	ca. 18,000
MOFs (experimental)	Formal oxidation states on atoms (exp)	Paper, Database	ca. 49,000

other data sets, pertaining to materials, for machine learning

see matminer here. pointed out by Jack Evans.

construct your own crystal structures!

here is a list of open-source codes for building your own crystal structure models.

name of code	link to code	link to associated paper
tobacco	link	link
pormake	link	link
ToBasCCo	link	link
Zeo++	link	link
stk	link	link
PoreMatMod.jl (only modifies)	link	link
pyCOFBuilder	link	link