text

Single-Point Calculations

Rowan supports a variety of methods to calculate the energy and properties of molecular structures.

Many Levels Of Theory

Rowan supports a wide variety of density functionals, including leading modern functionals like ωB97X-V and ωB97M–V that have been demonstrated to give high-accuracy results across many domains. For cases in which maximum accuracy is needed, Rowan also permits calculations with the double-hybrid functional DSD-BLYP-D3BJ, which a 2017 study found to be the most accurate among 217 different levels of theory studied.

Rowan supports multiple forms of non-local corrections, including "D3BJ" dispersion corrections and the "VV10" non-local correction.

Supported Electronic Structure Methods and ML Potentials

AIMNet2
GFN-FF
GFN0-xTB
GFN2-xTB
HF-3c
B97-3c
r²SCAN-3c
ωB97X-3c
B3LYP
M06-2X
ωB97X-D3
ωB97M-D3BJ
PBE
PBE0
r²SCAN
TPSS
TPSSh
M06-L
M06
CAM-B3LYP
ωB97X-V
ωB97M-V
DSD-BLYP-D3BJ
HF

Accurate Energies, Faster

Rowan enables calculations to be run with modern "composite"/"3c" methods, which feature custom basis sets and corrections tailored to provide maximum accuracy at a minimal cost. Relative to "standard" DFT methods like B3LYP/6-31G(d), composite methods offer significantly improved performance without introducing the same performance penalties as e.g. adding a massive basis set would. Composite methods like r2SCAN-3c have been adopted as the default choice for routine calculations by leading computational chemistry researchers, and Rowan makes it possible to use these methods in your own research.

For situations where speed is at a premium, Rowan also offers low-cost semiempirical quantum mechanical methods through xTB. xTB methods run many times faster than DFT calculations, work over the entire periodic table, and frequently offer energies and charges that are good enough for a first-pass.

Finally, Rowan allows for calculations to be run with machine-learned interatomic potentials. These methods use machine learning to produce high-quality energy and properties at a fraction of the cost of conventional DFT calculations, fundamentally changing the relationship between cost and accuracy provided by physics-based methods.

Property Prediction

Rowan supports prediction of a variety of single-point properties. Dipole moments can be computed, which give information about the distribution and asymmetry of charge in a molecule and also can be used to predict the strength and direction of intermolecular interactions. Similarly, atom-centered charges and spin densities can be computed by a variety of methods. These population parameters can be used to predict downsteam interactions and reactivity, and provide valuable information about the electronic distribution of a molecule.

text