Welcome to our online tool kit for predicting properties of thermoelectric materials!
Our motivation
Thermoelectric materials are materials that exhibit a strong coupling between electric and thermal responses, producing an electric potential (thermopower) in the presence of a temperature gradient and vice-versa. Current use of these materials is limited to niche refrigeration/cooling applications and small electric generators due to the poor efficiency of energy conversion, quantified by a figure of merit, ZT. Based on current estimates that 59% of all primary energy produced in the United States is lost as waste heat, recovery of even 1% of this heat by a novel high-ZT thermoelectric material would amount to nine times the energy currently produced by photovoltaics. Thermoelectric energy conversion systems use the thermoelectric effect to convert heat, either from a primary source such as a burner, or waste heat, such as that in the exhaust stream of an automobile or power plant, to useful electricity. However, the disjointed nature of materials property databases and relative immaturity of the thermoelectrics multi-scale functional relationships make integrated optimization across numerous length and time scales particularly difficult in this case. A data-driven design of high-ZT thermoelectrics is possible with tools capable of integrating materials databases into a consistent framework and finding the multi-scale underlying processing/structure/property functional relationships responsible for thermoelectrics performance.
Our sponsors
We are currently funded under DARPA's Simplifying Complexity in Scientific Discovery (SIMPLEX) program through SPAWAR (Contract #N66001-15-C-4036). This program seeks to develop unified mathematical frameworks and tools for scientific data analysis. The ultimate goal of the program is to facilitate big hypothesis generation and accelerate discovery by correlating data across scientific domains.Description of currently available toolkit components
Currently the toolkit consists of two components that contribute to ZT. They are meant to help material scientists in their quest for discovery of better materials for various applications.Seebeck coefficient predictor
A temperature gradient applied to a material gives rise to a voltage difference across it. Such effect is a fundamental electronic transport property called the Seebeck effect. The Seebeck coefficient measures the entropy transported with a charge carrier as it moves, divided by the carrier's charge. Therefore, it could be represented as a sum of contributions arising from the presence of charge carriers and their transport. Both components of Seebeck coefficient are affected by charge carrier concentration, electron-phonon drag effects, carrier-induced softening, formation of polarons, and presence of impurities, magnetic ions, or specimen boundaries, and nanostructuring.Predictions of Seebeck coefficient are generated at four different temperatures: 300K, 400K, 700K, and 1000K. The models take into account the atomic composition of material. In order to capture the effects of other important features, the production method and crystallinity of material are also included as input parameters.
Bulk modulus predictor
The bulk modulus is an important physical parameter of crystals. It reflects bonding characters in crystals and, in many instances, is used as an indicator for crystal strength and hardness. An accurate determination of bulk modulus is a complicated issue involving a careful analysis of elastic parameters, plastic deformation, and even experiment measurement processes. Therefore, many experimental and theoretical attempts have been made to determine the bulk moduli of various crystals, and some relations between the bulk modulus and physical parameters have been obtained. For example, the correlation between the bulk modulus and the Debye temperature, the bulk modulus-volume-ionicity relationship, a simple relation between the bulk modulus and the bond length for some binary semiconductors, a linear relation between the bulk modulus and plasma energy, and so on. For the complex crystals, hovewer, it is well-known that quantum mechanics does not easily accommodate this picture.In this module, supervised machine learning is employed to find correlations between multiple properties of different types (from unary to quinary) of crystalline compounds and their bulk modulus, and subsequently predict the bulk modulus for new compounds unseen by the model.
Publications:
- A. Furmanchuk, J. Saal, J. W. Doak, G. B. Olson, A. Choudhary, A. Agrawal, "Prediction of Seebeck coefficient for compounds without restriction to fixed stoichiometry: A machine learning approach", Journal of Computational Chemistry, 2017, DOI: 10.1002/jcc.25067 [url]
- A. Furmanchuk, A. Agrawal, A. Choudhary, "Predictive analytics for crystalline materials: bulk modulus", RSC Advances, 2016, 6, 95246 - 95251. [url]