The Vision
Understand and predict the properties and behaviors of materials from the fundamental laws of physics, using the tools of atomistic simulations.
To predict the behavior of matters, the bottleneck is with computational cost, according to Paul Dirac.
"The fundamental laws necessary for the mathematical treatment of a large part of physics and the whole of chemistry are thus completely known, and the difficulty lies only in the fact that application of these laws leads to equations that are too complex to be solved. "
The challenge in computational chemistry is to bring down the cost.
“...approximate practical methods of applying quantum mechanics should be developed, which can lead to an explanation of the main features of complex atomic systems without too much computation.” Paul Dirac (6 April 1929)
Is machine learning a free lunch?
To combine the accuracy of the first principles methods and efficiency of the empirical force fields, we can exploit machine learning (ML) techniques to “learn” the atomic interactions from quantum mechanics.
In essence, a ML interatomic potential “remembers” a pre-computed database of reference structure-property pairs, and makes predictions about a new structure by comparing it to the members of the reference set.
Statistical mechanics meets machine learning
By marrying advanced statistical mechanics methods with data-driven machine learning interatomic potentials, we want to develop and apply a method to predict the behavior of materials at finite temperatures using first principles methods based on quantum chemistry.
Can we predict material properties, including thermodynamic stabilities, from first principles?
From the stone age, through the bronze and iron ages to the silicon age, technological breakthroughs in human history have been heralded by the discovery and utilization of new materials. In the past, novel materials were exclusively invented and characterized experimentally in brick-and-mortar lab facilities, often relying on the chemical intuition of personnel. Nowadays, ever-increasing computational power has enabled us to predict material properties using first principles methods based on the fundamental laws of quantum mechanics. This paradigmatic change allows us to predict a wide variety of properties of a novel material before it is even experimentally synthesized, to gain physical understanding at the atomic scale, and to screen hundreds or thousands of new materials for certain desired properties.
In particular, we are interested in predicting the thermodynamic stabilities of novel materials, which is crucial for in silico material design – a novel material cannot be useful if it is unstable at its operating conditions, and indeed, if it cannot be synthesized at all.
