Quantum Chemistry Dataset · microsoft
msr-acc-tae25
A dataset of 73,040 coupled-cluster total atomization energies for evaluating and training quantum chemistry models.
What msr-acc-tae25 is
The Microsoft Research Accurate Chemistry Collection Total Atomization Energies dataset (MSR-ACC/TAE25) is an electronic-structure dataset developed by Microsoft Research. It contains 73,040 total atomization energies computed using high-accuracy coupled-cluster theory. The dataset was designed to train and benchmark machine-learning functionals, semi-empirical approaches, and density functional theory (DFT) methods across broad regions of chemical space.
Conventional benchmark collections for atomization energies are frequently small or biased towards organic compounds. MSR-ACC/TAE25 addresses this limitation by systematically covering closed-shell, charge-neutral, covalently bound equilibrium molecular structures containing up to five non-hydrogen atoms from elements up to argon that lack significant multireference character.
The dataset enables researchers to uncover systematic errors in approximate electronic structure algorithms, especially across diverse s- and p-block chemistries. It has also served as training data for deep-learning-based exchange-correlation functionals achieving chemical accuracy in atomization energies.
How it was made
Reference labels are calculated at the CCSD(T)/CBS level using the W1-F12 thermochemical protocol, with the primary reference energy provided in the tae@w1-f12 column. Molecules exhibiting significant multireference character are excluded.
Each record includes sufficient structural details to construct complete QCSchema documents, including symbols, coordinates, molecular charges, and multiplicities. The entries also store total atomization energies and their individual components across various theoretical levels, singlet-triplet energy gaps (S0-T1), perturbative triples energy fractions, full and non-hydrogen molecular graphs, and atomic number arrays.
Key facts
- Dataset size
- 73,040 molecular structures
- Level of theory
- CCSD(T)/CBS via W1-F12 thermochemical protocol
- Chemical scope
- Up to 5 non-hydrogen atoms from elements H through Ar
- Structural constraints
- Closed-shell, charge-neutral, covalently bound equilibrium structures
- Data schema
- QCSchema compatible
- File format
- Parquet
What you can use it for
- Machine-learning functional development
- Train and evaluate deep-learning exchange-correlation functionals and semi-empirical methods against coupled-cluster atomization energies.
- Graph neural network benchmarking
- Evaluate the ability of molecular graph architectures to generalize beyond organic chemical space using diverse s- and p-block compounds.
- Domain-specific functional selection
- Filter the dataset to benchmark density functional approximations on target chemical families, such as phosphorus sulfide systems.
Explore the data
The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from microsoft/msr-acc-tae25 at the moment you asked. Nothing is cached or stored — every row above came from that request.
How the columns are distributed
Over 71,871 rows of default/train
atom-count:all
int- min
- 2
- median
- 7
- mean
- 7.41
- max
- 17
atom-count:non-h
int- min
- 0
- median
- 4
- mean
- 4.10
- max
- 5
atomic_numbers
list- min
- 2
- median
- 7
- mean
- 7.41
- max
- 17
fragments
int- min
- 1
- median
- 1
- mean
- 1
- max
- 1
geometry
list- min
- 2
- median
- 7
- mean
- 7.41
- max
- 17
graph:all
list- min
- 1
- median
- 7
- mean
- 7.33
- max
- 18
graph:non-h
list- min
- 0
- median
- 4
- mean
- 3.93
- max
- 9
molecular_charge
float- min
- 0
- median
- 0
- mean
- 0
- max
- 0
molecular_multiplicity
int- min
- 1
- median
- 1
- mean
- 1
- max
- 1
name
string_text- min
- 2
- median
- 7
- mean
- 7.19
- max
- 12
sha1
string_text- min
- 40
- median
- 40
- mean
- 40
- max
- 40
singlet-triplet-gap-s0-t1@b3lyp/def2-tzvp
float- min
- 3.0e-5
- median
- 0.08
- mean
- 0.09
- max
- 0.37
0.00% empty
symbols
list- min
- 2
- median
- 7
- mean
- 7.41
- max
- 17
tae:frac[(T)]@ccsd(t)/6-31g*
float- min
- 0
- median
- 0.02
- mean
- 0.02
- max
- 0.29
0.35% empty
tae:frac[(T)]@w1-f12
float- min
- 0
- median
- 0.03
- mean
- 0.03
- max
- 0.06
0.00% empty
tae@b3lyp-d3/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.04% empty
tae@b3lyp-d3/def2-svp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.56
0.05% empty
tae@b3lyp-d3/def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.05% empty
tae@b3lyp-d3/ma-def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.07% empty
tae@b3lyp-d3/ma-def2-svp
float- min
- 0.03
- median
- 0.82
- mean
- 0.86
- max
- 2.55
0.05% empty
tae@b3lyp-d3/ma-def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.06% empty
tae@b3lyp/def2-qzvp
float- min
- 0.03
- median
- 0.82
- mean
- 0.85
- max
- 2.52
0.04% empty
tae@b3lyp/def2-svp
float- min
- 0.03
- median
- 0.81
- mean
- 0.85
- max
- 2.54
0.05% empty
tae@b3lyp/def2-tzvp
float- min
- 0.03
- median
- 0.82
- mean
- 0.85
- max
- 2.52
0.05% empty
tae@b3lyp/ma-def2-qzvp
float- min
- 0.03
- median
- 0.82
- mean
- 0.85
- max
- 2.52
0.07% empty
tae@b3lyp/ma-def2-svp
float- min
- 0.03
- median
- 0.81
- mean
- 0.84
- max
- 2.53
0.05% empty
tae@b3lyp/ma-def2-tzvp
float- min
- 0.03
- median
- 0.82
- mean
- 0.85
- max
- 2.52
0.06% empty
tae@b97-3c
float- min
- 6.1e-4
- median
- 0.79
- mean
- 0.82
- max
- 2.45
tae@b97m-v/def2-qzvp
float- min
- 0.03
- median
- 0.85
- mean
- 0.88
- max
- 2.55
0.03% empty
tae@b97m-v/def2-svp
float- min
- 0.03
- median
- 0.84
- mean
- 0.88
- max
- 2.58
0.04% empty
tae@b97m-v/def2-tzvp
float- min
- 0.03
- median
- 0.84
- mean
- 0.87
- max
- 2.54
0.03% empty
tae@b97m-v/ma-def2-qzvp
float- min
- 0.03
- median
- 0.85
- mean
- 0.88
- max
- 2.55
0.27% empty
tae@b97m-v/ma-def2-svp
float- min
- 0.06
- median
- 0.86
- mean
- 0.89
- max
- 2.58
14.7% empty
tae@b97m-v/ma-def2-tzvp
float- min
- 0.07
- median
- 0.86
- mean
- 0.89
- max
- 2.54
14.7% empty
tae@gfn1-xtb
float- min
- 0.15
- median
- 1.29
- mean
- 2.03
- max
- 259.74
0.33% empty
tae@gfn2-xtb
float- min
- 0.10
- median
- 1.13
- mean
- 1.17
- max
- 3.14
0.07% empty
tae@m062x-d3/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.04% empty
tae@m062x-d3/def2-svp
float- min
- 0.04
- median
- 0.82
- mean
- 0.86
- max
- 2.55
0.05% empty
tae@m062x-d3/def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.05% empty
tae@m062x-d3/ma-def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.13% empty
tae@m062x-d3/ma-def2-svp
float- min
- 0.04
- median
- 0.82
- mean
- 0.86
- max
- 2.55
0.05% empty
tae@m062x-d3/ma-def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.07% empty
tae@m062x/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.04% empty
tae@m062x/def2-svp
float- min
- 0.04
- median
- 0.82
- mean
- 0.86
- max
- 2.55
0.05% empty
tae@m062x/def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.05% empty
tae@m062x/ma-def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.13% empty
tae@m062x/ma-def2-svp
float- min
- 0.04
- median
- 0.82
- mean
- 0.86
- max
- 2.55
0.05% empty
tae@m062x/ma-def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.07% empty
tae@pbeh-3c
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.62
tae@r2scan-3c
float- min
- 0.02
- median
- 0.83
- mean
- 0.87
- max
- 2.57
0.00% empty
tae@r2scan-d3/def2-qzvp
float- min
- 0.03
- median
- 0.84
- mean
- 0.87
- max
- 2.55
0.04% empty
tae@r2scan-d3/def2-svp
float- min
- 0.03
- median
- 0.83
- mean
- 0.87
- max
- 2.56
0.05% empty
tae@r2scan-d3/def2-tzvp
float- min
- 0.03
- median
- 0.84
- mean
- 0.87
- max
- 2.55
0.05% empty
tae@r2scan-d3/ma-def2-qzvp
float- min
- 0.03
- median
- 0.84
- mean
- 0.87
- max
- 2.55
0.07% empty
tae@r2scan-d3/ma-def2-svp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.56
0.05% empty
tae@r2scan-d3/ma-def2-tzvp
float- min
- 0.03
- median
- 0.84
- mean
- 0.87
- max
- 2.54
0.05% empty
tae@r2scan/def2-qzvp
float- min
- 0.02
- median
- 0.84
- mean
- 0.87
- max
- 2.55
0.04% empty
tae@r2scan/def2-svp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.56
0.05% empty
tae@r2scan/def2-tzvp
float- min
- 0.02
- median
- 0.83
- mean
- 0.87
- max
- 2.54
0.05% empty
tae@r2scan/ma-def2-qzvp
float- min
- 0.03
- median
- 0.84
- mean
- 0.87
- max
- 2.55
0.07% empty
tae@r2scan/ma-def2-svp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.56
0.05% empty
tae@r2scan/ma-def2-tzvp
float- min
- 0.02
- median
- 0.83
- mean
- 0.87
- max
- 2.54
0.05% empty
tae@revdsd-pbep86-d3/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.87
- max
- 2.56
0.11% empty
tae@revdsd-pbep86-d3/def2-svp
float- min
- 0.03
- median
- 0.80
- mean
- 0.83
- max
- 2.52
0.10% empty
tae@revdsd-pbep86-d3/def2-tzvp
float- min
- 0.03
- median
- 0.82
- mean
- 0.85
- max
- 2.52
0.10% empty
tae@revdsd-pbep86-d3/ma-def2-qzvp
float- min
- 0.03
- median
- 0.84
- mean
- 0.87
- max
- 2.56
0.13% empty
tae@revdsd-pbep86-d3/ma-def2-svp
float- min
- 0.03
- median
- 0.80
- mean
- 0.83
- max
- 2.52
0.10% empty
tae@revdsd-pbep86-d3/ma-def2-tzvp
float- min
- 0.03
- median
- 0.82
- mean
- 0.85
- max
- 2.52
0.12% empty
tae@revdsd-pbep86/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.55
0.11% empty
tae@revdsd-pbep86/def2-svp
float- min
- 0.03
- median
- 0.79
- mean
- 0.82
- max
- 2.52
0.10% empty
tae@revdsd-pbep86/def2-tzvp
float- min
- 0.03
- median
- 0.81
- mean
- 0.84
- max
- 2.51
0.10% empty
tae@revdsd-pbep86/ma-def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.55
0.13% empty
tae@revdsd-pbep86/ma-def2-svp
float- min
- 0.03
- median
- 0.79
- mean
- 0.83
- max
- 2.51
0.10% empty
tae@revdsd-pbep86/ma-def2-tzvp
float- min
- 0.03
- median
- 0.81
- mean
- 0.84
- max
- 2.51
0.12% empty
tae@w1-f12
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.55
tae@wb97m-v/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.87
- max
- 2.55
0.03% empty
tae@wb97m-v/def2-svp
float- min
- 0.03
- median
- 0.84
- mean
- 0.87
- max
- 2.59
0.02% empty
tae@wb97m-v/def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.03% empty
tae@wb97m-v/ma-def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.87
- max
- 2.55
0.62% empty
tae@wb97m-v/ma-def2-svp
float- min
- 0.03
- median
- 0.83
- mean
- 0.87
- max
- 2.58
0.03% empty
tae@wb97m-v/ma-def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.05% empty
tae@wb97x-2/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.53
0.11% empty
tae@wb97x-2/def2-svp
float- min
- 0.03
- median
- 0.79
- mean
- 0.82
- max
- 2.50
0.10% empty
tae@wb97x-2/def2-tzvp
float- min
- 0.03
- median
- 0.81
- mean
- 0.84
- max
- 2.49
0.10% empty
tae@wb97x-2/ma-def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.53
0.13% empty
tae@wb97x-2/ma-def2-svp
float- min
- 0.03
- median
- 0.79
- mean
- 0.83
- max
- 2.49
0.10% empty
tae@wb97x-2/ma-def2-tzvp
float- min
- 0.03
- median
- 0.81
- mean
- 0.84
- max
- 2.49
0.13% empty
tae@wb97x-d3/def2-qzvp
float- min
- 0.04
- median
- 0.85
- mean
- 0.88
- max
- 2.56
0.03% empty
tae@wb97x-d3/def2-svp
float- min
- 0.04
- median
- 0.84
- mean
- 0.88
- max
- 2.58
0.05% empty
tae@wb97x-d3/def2-tzvp
float- min
- 0.04
- median
- 0.85
- mean
- 0.88
- max
- 2.56
0.05% empty
tae@wb97x-d3/ma-def2-qzvp
float- min
- 0.04
- median
- 0.85
- mean
- 0.88
- max
- 2.56
0.06% empty
tae@wb97x-d3/ma-def2-svp
float- min
- 0.04
- median
- 0.84
- mean
- 0.88
- max
- 2.58
0.05% empty
tae@wb97x-d3/ma-def2-tzvp
float- min
- 0.04
- median
- 0.85
- mean
- 0.88
- max
- 2.56
0.05% empty
tae@wb97x-v/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.87
- max
- 2.55
0.02% empty
tae@wb97x-v/def2-svp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.56
0.02% empty
tae@wb97x-v/def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.02% empty
tae@wb97x-v/ma-def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.87
- max
- 2.55
0.03% empty
tae@wb97x-v/ma-def2-svp
float- min
- 0.03
- median
- 0.82
- mean
- 0.86
- max
- 2.56
0.02% empty
tae@wb97x-v/ma-def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.02% empty
tae@wb97x/def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.03% empty
tae@wb97x/def2-svp
float- min
- 0.03
- median
- 0.82
- mean
- 0.86
- max
- 2.56
0.05% empty
tae@wb97x/def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.05% empty
tae@wb97x/ma-def2-qzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.06% empty
tae@wb97x/ma-def2-svp
float- min
- 0.03
- median
- 0.82
- mean
- 0.85
- max
- 2.55
0.05% empty
tae@wb97x/ma-def2-tzvp
float- min
- 0.03
- median
- 0.83
- mean
- 0.86
- max
- 2.54
0.05% empty
tae[HF]@w1-f12
float- min
- 2.1e-3
- median
- 0.57
- mean
- 0.60
- max
- 1.97
tae[∆(T)]@w1-f12
float- min
- 0
- median
- 0.02
- mean
- 0.02
- max
- 0.06
tae[∆CCSD]@w1-f12
float- min
- 0.03
- median
- 0.23
- mean
- 0.24
- max
- 0.57
tae[∆CV]@w1-f12
float- min
- -6.7e-3
- median
- 2.7e-3
- mean
- 2.8e-3
- max
- 0.01
Strengths and limits
Good at
- Delivers 73,040 reference total atomization energies calculated at the CCSD(T)/CBS level via the W1-F12 protocol.
- Exhaustively samples closed-shell equilibrium structures up to five heavy atoms across s- and p-block elements up to argon.
- Provides decomposed energy components, singlet-triplet gaps, and perturbative triples fractions alongside total energies.
- Structures are provided in formats compatible with the standardized QCSchema representation.
Watch out for
- Restricted to closed-shell, charge-neutral molecules in equilibrium geometries.
- Excludes molecules with significant multireference electronic character.
- Limited to molecules containing five or fewer non-hydrogen atoms with atomic numbers up to argon.
Questions
Can I look inside msr-acc-tae25 without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser, with full-text search and a SQL filter. Nothing is downloaded to your machine.
What licence is msr-acc-tae25 under?
Its authors declare cdla-permissive-2.0. Read the terms on the original page before reusing the data.
How is the total atomization energy (TAE) defined in this dataset?
The TAE is calculated as the sum of the ground-state isolated atomic energies for all constituent atoms minus the total molecular energy: TAE = n * E(A) - E(M).
What atomic ground-state multiplicities are assumed?
Multiplicities used are singlet for Be and Mg; doublet for H, Li, B, F, Na, Al, and Cl; triplet for C, O, Si, and S; and quartet for N and P.
Which column contains the highest-accuracy ground-truth energy?
The tae@w1-f12 column contains the most accurate reference atomization energy relative to experimental data.
How does MSR-ACC/TAE25 compare to datasets like GDB-9?
While GDB-9 focuses largely on organic molecules, MSR-ACC/TAE25 includes a wider variety of challenging s- and p-block compounds containing elements up to argon.
How to cite it
Ehlert, Sebastian and Hermann, Jan and Vogels, Thijs and Satorras, Victor Garcia and Lanius, Stephanie and Segler, Marwin and Giesbertz, Klaas JH and Kooi, Derk P and Takeda, Kenji and Huang, Chin-Wei and Luise, Giulia and van den Berg, Rianne and Gori-Giorgi, Paola and Karton, Amir. Accurate Chemistry Collection: Coupled cluster atomization energies for broad chemical space. arXiv:2506.14492 [physics.chem-ph], 2025.
Written from msr-acc-tae25's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.