摘要：To comply with the Kigali amendment to the Montreal Protocol in 2016, development of new refrigerants with low global warming potential is urgently required in addition to satisfying the conventional requirements of cooling performance, safety, and non-destructiveness to the ozone layer. Because these requirements closely correlated, the proper control of various chemical properties is necessary to fulfill the requirements. However, simultaneous satisfaction of all the requirements is extremely difficult because of the tradeoffs among the chemical properties. Hence, we must correctly recognize how chemical properties behave when the composition of molecules is changed. We performed an in-silico screening that combines quantum chemical calculations, machine learning, and database search, where 10,163 molecules were investigated exhaustively within the properly imposed constraints; subsequently we found a few candidates.
