Phase transitions such as melting and boiling characterize a material's interaction with the environment. Further, properties such as transition temperature can influence whether a material is suitable for a given application. Examples of phase-driven applications include drug delivery, melt-casting explosives, or energy havesting using phase-change materials. However, a well-known difficulty of materials discovery is that chemical synthesis is costly and time consuming. Therefore, identifying potential candidates may proceed a long and arduous synthesis task. A priori knowledge of transition temperatures would limit the chemical space of candidate compounds, thus expediting discovery.<br/><br/>Predicting physicochemical properties has long been the work of linear regression-based quantitative structure-property relationships. These models have had success but are often limited by their inability to find nonlinear mappings from easy-to-derive descriptors. Recently, there has been a strong interest in applying the tools of data science to chemistry. Nonlinear machine learning (ML) algorithms have made their presence from the atomic scale to the continuum. This work builds on these modern-day predictive tools for transition temperatures. Specifically, we propose a set of features made up of group-constitutive and geometrical descriptors, previously shown to map to enthalpy and entropy--two thermodynamic quantities that drive phase transitions. A notable advantage of the method is that descriptors are derived purely from a compound's SMILES string. Thus, besides relatively simple structural characteristics such as connectivity and hybridization, there are no high-level or numerically-intensive calculations necessary. While it is generally a challenge to train models to limited experimental data, we find that a concise set of domain-specific descriptors, combined with robust and efficient nonlinear ML algorithms, provide an appealing framework for predicting transition temperatures in a diverse set of compounds.

Machine learning transition temperatures from 2D structure

Macromolecular Chemistry: The Second Century

[Author Benefits](https://scimeetings.acs.org/?utm_source=pubs_content_marketing&utm_medium=website&utm_campaign=0320_MCF_NPI_Launch_Spring_Homepage&ref=pubs_content_marketing)

[How to use SciMeetings](https://storage.googleapis.com/dl.morressier.com/acs/SciMeetings-ACSSpring2020-HowToGuide_M3020-Final.pdf)

[Frequently Asked Questions](https://storage.googleapis.com/dl.morressier.com/acs/SciMtng_ACSNMSp2020_FAQ_M3120_Final.pdf)

ACS Spring 2020 National Meeting & Expo

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.