Computational Modelling — Metallic Metal-Organic Frameworks Predicted by the Combination of Machine Learning Methods and Ab Initio Calculations

13 measurement groups · 137 results

Reported values remain attached to the sample, method, conditions, extraction quality and source location that produced them.

DFT+HSE electronic band-structure calculation

Density functional theory

DFT model of CdC4 (Cd2C8 model) · Model

Hybrid-functional check of the Cd2C8 CdC4 model beyond DFT+PBE to assess whether semilocal self-interaction caused fictitious metallisation.

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 4 · Results and discussion · Figure S6

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
HSE band-structure figure support	band structure computed at high-symmetry points	—	—	Caption Qualitative	13 · Ab Initio Calculations · Figure S6
DFT+HSE metallicity checkMarked as a best value within this paper	Cd2C8 remains metallic	—	—	Text Qualitative	4 · Results and discussion · Figure S6

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of CdC4 (Cd2C8 model) · Model

Kinetic energy cutoff 450 eV; Monkhorst-Pack grid 8x8x8; structures relaxed in volume and atomic positions until forces were below 0.01 eV/angstrom before self-consistent density and band-structure calculations.

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
large band dispersion / bandwidth	up to a few hundred meV, particularly for Cd2C8 and Mn4Re12Te16C12N12	—	—	Text Approximate	5 · Results and discussion · Figure 4
DFT-optimised crystal structure type	face centered cubic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
calculated DOS near Fermi level	DOS around the Fermi level calculated	—	—	Caption Qualitative	12 · Ab Initio Calculations · Figure S5
Monkhorst-Pack k-point grid	8x8x8	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	8.9364	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	8.9364	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	8.9364	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity statusMarked as a best value within this paper	DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=metallic; SVC=not reported metallic; NN=not reported metallic; RF=metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
estimated Mott constant C	0.0838	—	—	Table Exact Reported	4 · Results and discussion · Table 2
DFT+PBE band gapMarked as a best value within this paper	0	—	—	Table Exact Reported	4 · Results and discussion · Table 2

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of CdC4 (Cd8C32 model) · Model

Kinetic energy cutoff 450 eV; Monkhorst-Pack grid 4x4x2; structures relaxed in volume and atomic positions until forces were below 0.01 eV/angstrom before self-consistent density and band-structure calculations.

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
DFT-optimised crystal structure type	simple monoclinic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
Monkhorst-Pack k-point grid	4x4x2	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	9.003	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	12.046	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	18.87	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity status	not DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=metallic; SVC=not reported metallic; NN=not reported metallic; RF=metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
DFT+PBE band gap	2.19	—	—	Table Exact Reported	4 · Results and discussion · Table 2

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of Hg[SCN]4Co[NCS]4 · Model

Kinetic energy cutoff 450 eV; Monkhorst-Pack grid 8x4x4; structures relaxed in volume and atomic positions until forces were below 0.01 eV/angstrom before self-consistent density and band-structure calculations.

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
DFT-optimised crystal structure type	simple orthorhombic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
calculated DOS near Fermi level	DOS around the Fermi level calculated	—	—	Caption Qualitative	12 · Ab Initio Calculations · Figure S5
Monkhorst-Pack k-point grid	8x4x4	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	7.6237	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	14.2872	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	15.4499	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity statusMarked as a best value within this paper	DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=not reported metallic; SVC=metallic; NN=metallic; RF=metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
estimated Mott constant C	0.089	—	—	Table Exact Reported	4 · Results and discussion · Table 2
DFT+PBE band gapMarked as a best value within this paper	0	—	—	Table Exact Reported	4 · Results and discussion · Table 2

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of KNd[Re4Te4(CN)12] · Model

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
DFT-optimised crystal structure type	base centered orthorhombic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
Monkhorst-Pack k-point grid	4x4x2	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	12.2367	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	12.2367	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	18.9469	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity status	not DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=metallic; SVC=not reported metallic; NN=metallic; RF=not reported metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
DFT+PBE band gap	1.49	—	—	Table Exact Reported	4 · Results and discussion · Table 2

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of Mn[Re3Te4(CN)3] · Model

Kinetic energy cutoff 450 eV; Monkhorst-Pack grid 6x6x4; structures relaxed in volume and atomic positions until forces were below 0.01 eV/angstrom before self-consistent density and band-structure calculations.

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
DFT-optimised crystal structure type	simple monoclinic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
calculated DOS near Fermi level	DOS around the Fermi level calculated	—	—	Caption Qualitative	12 · Ab Initio Calculations · Figure S5
Monkhorst-Pack k-point grid	6x6x4	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	9.6777	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	10.8246	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	16.0906	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity statusMarked as a best value within this paper	DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=metallic; SVC=not reported metallic; NN=metallic; RF=not reported metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
estimated Mott constant C	0.1152	—	—	Table Exact Reported	4 · Results and discussion · Table 2
DFT+PBE band gapMarked as a best value within this paper	0	—	—	Table Exact Reported	4 · Results and discussion · Table 2

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of Mn2[Re6S8(CN)6]4 · Model

Kinetic energy cutoff 450 eV; Monkhorst-Pack grid 4x4x4; structures relaxed in volume and atomic positions until forces were below 0.01 eV/angstrom before self-consistent density and band-structure calculations.

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
DFT-optimised crystal structure type	simple cubic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
calculated DOS near Fermi level	DOS around the Fermi level calculated	—	—	Caption Qualitative	12 · Ab Initio Calculations · Figure S5
Monkhorst-Pack k-point grid	4x4x4	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	16.3842	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	16.3842	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	16.3842	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity statusMarked as a best value within this paper	DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=metallic; SVC=not reported metallic; NN=metallic; RF=not reported metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
estimated Mott constant C	0.1025	—	—	Table Exact Reported	4 · Results and discussion · Table 2
DFT+PBE band gapMarked as a best value within this paper	0	—	—	Table Exact Reported	4 · Results and discussion · Table 2

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of Mn2[Re6Se8(CN)6]4 · Model

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
DFT-optimised crystal structure type	simple orthorhombic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
calculated DOS near Fermi level	DOS around the Fermi level calculated	—	—	Caption Qualitative	12 · Ab Initio Calculations · Figure S5
Monkhorst-Pack k-point grid	4x4x4	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	16.3862	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	16.0077	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	15.9653	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity statusMarked as a best value within this paper	DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=metallic; SVC=not reported metallic; NN=metallic; RF=not reported metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
estimated Mott constant C	0.1042	—	—	Table Exact Reported	4 · Results and discussion · Table 2
DFT+PBE band gapMarked as a best value within this paper	0	—	—	Table Exact Reported	4 · Results and discussion · Table 2

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of Mn2[Re6Te8(CN)6]4 · Model

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
DFT-optimised crystal structure type	simple orthorhombic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
calculated DOS near Fermi level	DOS around the Fermi level calculated	—	—	Caption Qualitative	12 · Ab Initio Calculations · Figure S5
Monkhorst-Pack k-point grid	4x4x4	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	16.7127	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	16.1979	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	15.8171	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity statusMarked as a best value within this paper	DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=metallic; SVC=not reported metallic; NN=metallic; RF=not reported metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
estimated Mott constant C	0.1034	—	—	Table Exact Reported	4 · Results and discussion · Table 2
DFT+PBE band gapMarked as a best value within this paper	0	—	—	Table Exact Reported	4 · Results and discussion · Table 2

VASP DFT with GGA-PBE functional and PAW pseudopotentials

Density functional theory

DFT model of Na13Fe4Sb2W18(C4O43)2 · Model

Geometry: periodic crystal model
Context: pristine computational framework model
Measurement source: 9-10 · Ab Initio Calculations · Table SIV

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
DFT-optimised crystal structure type	triclinic	—	—	SI Table Qualitative	10 · Ab Initio Calculations · Table SIV
Monkhorst-Pack k-point grid	4x4x4	—	—	Text Exact Reported	9-10 · Ab Initio Calculations · Methods paragraph
DFT-optimised lattice parameter a	12.9005	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter b	14.399	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT-optimised lattice parameter c	19.9229	—	—	SI Table Exact Reported	10 · Ab Initio Calculations · Table SIV
DFT+PBE metallicity status	not DFT-confirmed metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
classifier metallic votes	LR=metallic; SVC=not reported metallic; NN=metallic; RF=not reported metallic	—	—	Table Qualitative	4 · Results and discussion · Table 2
DFT+PBE band gap	0.52	—	—	Table Exact Reported	4 · Results and discussion · Table 2

t-SNE/MDS reduction of 45-dimensional feature spaces

Paper-level ML/DFT screening workflow · Model

Inorganic and MOF feature spaces reduced to two dimensions to assess transfer-learning overlap. SI Figure S3 evaluates overlap ratio versus number of inorganic materials.

Context: computational model screening
Measurement source: 3 · Feature-space analysis · Figure 3

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
MOF/inorganic feature-space overlap	large regions of overlap	—	—	Caption Qualitative	3 · Feature-space analysis · Figure 3
feature selection outcome	all 45 generated material descriptors were used	—	—	Text Exact Reported	4 · Feature Selection · Figure S2

Classifier performance evaluation on inorganic training/test sets

Paper-level ML/DFT screening workflow · Model

52,300 inorganic data split 70% training and 30% test; 5-fold cross-validation for model optimisation; metrics reported as accuracy, precision, recall and classifier-specific candidate counts.

Context: computational model screening
Measurement source: 3 · Model evaluation · Table 1

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
testing DC accuracy	50%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing DC precision	50%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing DC recall	50%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing LR accuracy	88%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing LR precision	85%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing LR recall	92%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing NN accuracy	94%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing NN precision	92%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing NN recall	96%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing RF accuracy	97%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing RF precision	96%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing RF recall	97%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing SVC_rbf accuracy	95%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing SVC_rbf precision	94%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
testing SVC_rbf recall	96%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training DC accuracy	50%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training DC precision	50%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training DC recall	50%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training LR accuracy	87%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training LR precision	84%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training LR recall	92%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training NN accuracy	95%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training NN precision	94%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training NN recall	97%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training RF accuracy	98%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training RF precision	98%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training RF recall	99%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training SVC_rbf accuracy	97%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training SVC_rbf precision	96%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
training SVC_rbf recall	98%	—	—	Table Exact Reported	3 · Model evaluation · Table 1
LR dft confirmed	5	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
LR ml predicted	13	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
LR mv predicted	8	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
NN dft confirmed	5	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
NN ml predicted	26	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
NN mv predicted	7	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
RF dft confirmed	2	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
RF ml predicted	3	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
RF mv predicted	3	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
SVC_rbf dft confirmed	1	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
SVC_rbf ml predicted	40	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII
SVC_rbf mv predicted	1	—	—	SI Table Exact Reported	7 · Machine Learning Methods · Table SIII

Transfer learning with logistic regression, SVC with rbf kernel, neural network, random forest, and statistical multivoting

Paper-level ML/DFT screening workflow · Model

Models trained on about 52,300 OQMD inorganic compounds using 45 material descriptors; 2932 CoRE-MOF structures screened; candidates predicted metallic by multivoting were validated with DFT.

Context: computational model screening
Measurement source: 2 · Methods/results narrative · Figure 1

Property	Reported value	Normalised value	Uncertainty	Origin and quality	Source
dft confirmed metallic count	six intrinsic metallic MOFs	—	—	Text Exact Reported	2-6 · Methods/results narrative · Table 2
generated descriptor count	45 material descriptors	—	—	Text Exact Reported	2-6 · Methods/results narrative · Table 2
multivoting accuracy	67%	—	—	Text Rounded Reported	2-6 · Methods/results narrative · Table 2
multivoting candidate count	nine predicted metallic MOFs	—	—	Text Exact Reported	2-6 · Methods/results narrative · Table 2
oqmd training pool count	about 52,300 inorganic compounds	—	—	Text Approximate	2-6 · Methods/results narrative · Table 2
screened mofs count	2932 MOF structures screened	—	—	Text Exact Reported	2-6 · Methods/results narrative · Table 2