Enter a composition, or a list of up to 2000 compositions separated by commas, to return a prediction of the thermal conductivity of each formulation.The prediction was found to have a R² of 0.71, and root mean squared error of 0.55 log₁₀(W m^-1 K^-1) on a test set of thermoelectric materials comparing the true logarithm of thermal conductivity to the predicted logarithm of thermal conductivity. This application was reported in The Liverpool Materials Discovery Server: A suite of tools for the collaborative discovery of materials., and is an improved version of the model first used in Discovery of a Low Thermal Conductivity Oxide Guided by Probe Structure Prediction and Machine Learning. Please consider citing these papers if you use this in your work.
Click here for information as to how to use this model with an API