Deep neural network for fitting analytical potential energy curve of diatomic molecules from ro-vibrational spectra

ABSTRACT

We present a new approach which employs a deep neural network to obtain parameters of analytical representation of potential energy curve of diatomic molecule. We test the approach to find spectroscopic characteristics for the ground $X^2{\mathrm{\Sigma }}^{+}$ electronic state of MgF molecule based on the experimental energies of ro-vibrational transitions. The result shows that a deep neural network can be applied in characterisation of interatomic potential of diatomic molecule. Our approach is competitive with those obtained using other methods tested, i.e. shallow neural network and the so-called brute force method.

KEYWORDS:

• Diatomic molecule
• deep neural network
• potential energy curve
• magnesium fluoride
• expanded morse oscillator

Acknowledgments

We are grateful to Dr Tomasz Tylec from IF Research Poland for helpful suggestions. This work was supported by the National Science Centre Poland (Narodowe Centrum Nauki) [grant number UMO2015/17/B/ST4/04016]. We are very grateful to one of the Reviewers for the suggestion to include the spectroscopic constants determined by the tested methods.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

To get the supplementary materials, please visit: https://gitlab.com/dominikhorwat/dnnet_potentials.

Additional information

Funding

This work was supported by the National Science Centre Poland (Narodowe Centrum Nauki) [grant number UMO2015/17/B/ST4/04016].