hiphive has been developed by Fredrik Eriksson, Erik Fransson, and Paul Erhart at the Department of Physics of Chalmers University of Technology in Gothenburg, Sweden with funding from the Knut och Alice Wallenbergs Foundation, the Swedish Research Council, the Swedish Foundation for Strategic Research, and the Swedish National Infrastructure for Computing.
When using hiphive in your research please cite the following paper:
The Hiphive Package for the Extraction of High-Order Force Constants by Machine Learning
Fredrik Eriksson, Erik Fransson, and Paul Erhart
Advanced Theory and Simulations 2, 1800184 (2019)
You might also find the following paper useful, which discusses in detail the advantages and disadvantages of different regression schemes in various different application scenarios:
Efficient construction of linear models in materials modeling and applications to force constant expansions
Erik Fransson, Fredrik Eriksson, and Paul Erhart
npj Computational Materials 6, 135 (2020)
If you want to run large-scale molecular dynamics simulations with force constant potentials constructed via hiphive, you should find the GPU implementation in the GPUMD code useful, which is described in the following paper:
Efficient calculation of the lattice thermal conductivity by atomistic simulations with ab-initio accuracy
Joakim Brorsson, Arsalan Hashemi, Zheyong Fan, Erik Fransson, Fredrik Eriksson, Tapio Ala-Nissila, Arkady V. Krasheninnikov, Hannu-Pekka Komsa, and Paul Erhart
Advanced Theory and Simulations 4, 2100217 (2021)
hiphive implements methods that have evolved in the field over many years including work by, e.g.,
Parlinski, Li, and Kawazoe [ParLiKaw97]
Esfarjani and Stokes [EsfSto08]
Hellman, Abrikosov, and Simak [HelAbrSim11]
Tadano, Gohda, and Tsuneyuki [TadGohTsu14]
Zhou, Nielson, Xia, and Ozolins [ZhoNieXia14]
Togo and Tanaka [TogTan15]
Please cite these original papers as appropriate for your work.
For a general overview of the vibrational properties of materials see e.g., [Ful10].