Scientists at the Freie Universität Berlin have an AI-based solution for calculating the ground state of the Schrödinger equation in quantum chemistry.
The Schrödinger equation is mainly used to predict the chemical and physical properties of a molecule based on the arrangement of the atoms. The equation helps to determine where the electrons and nuclei of a molecule are and under a given set of conditions what their energy is.
The equation has the same central importance as Newton’s law movement, which can predict the position of an object at a given moment, but in quantum mechanics – that is, in atoms or subatomic particles.
The article describes how the neural network developed by the scientists at the Freie Universität Berlin provides more accuracy in solving the Schrödinger equation and what it means for the future.
AI brings more accuracy to the comparison
In principle, the Schrödinger equation can be solved to predict the exact location of atoms or subatomic particles in a molecule, but in practice this is extremely difficult as it is very approximate.
Central to the equation is a mathematical object, a wave function that specifies the behavior of electrons in a molecule. But the high dimensionality of the wave function makes it extremely difficult to find out how electrons affect each other. Most of what you get from the mathematical representations is therefore a probable version of it and not exact answers.
This limits the accuracy with which we can find properties of a molecule, such as the configuration, conformation, size and shape, that can help define the wave function. The process becomes so complex that it becomes impossible to implement the equation beyond single atoms.
Subscribe to our newsletter
Get the latest updates and relevant offers by sharing your email address.
As a replacement for the mathematical building blocks, the scientists at the Freie Universität Berlin devised a deep neural network that can teach the intricate patterns of how electrons are located around the nuclei.
The scientists developed a Deep Neural Networks (DNN) model, PauliNet, which has several advantages over conventional methods of studying quantum systems such as the Quantum Monte Carlo or other classical quantum chemistry methods.
The DNN model developed by these scientists is very flexible and provides for a varied approach that can help the accurate calculation of electronic properties outside the electronic energy.
Second, it also helps the easy calculation of multi-body and more complex correlation with fewer determinants, which reduces the need for higher computational power. The model mainly helped solve a major problem between accuracy and computational cost, which often faced the solution of the Schrodinger equation.
The model can also calculate the local energy of heavy nuclei such as heavy metals without using pseudo-potentials or approximations.
Finally, the model developed in the study has antisymmetry functions and other principles essential for electronic wave functions integrated into the DNN model, rather than teaching the model. The construction of fundamental physics in the model thus helped to make meaningful and accurate predictions.
AI help science
In recent years, artificial intelligence has helped solve many scientific problems that would otherwise have seemed impossible by using traditional methods.
AI has become instrumental in anticipating the results of experiments or simulations of quantum systems, especially because of their complex nature. In 2018, reinforcement learning was used to automatically design new quantum experiments in automated laboratories.
Recent efforts by the University of Warwick and another IBM and DeepMind have also attempted to solve the Schrödinger equation. With its greater accuracy in solving the equation now, PauliNet offers us the potential to use it in many real-world applications.
Understanding the composition of molecules can help accelerate the discovery of drugs, which used to be difficult due to the approaches to understanding their properties.
Similarly, it can also help to discover various other elements or metamaterials such as new catalysts, industrial chemical applications, new pesticides, among others. It can be used to characterize molecules synthesized in laboratories.
Several academic and commercial software use Schrödinger’s equation at its core, but are based on applications. The accuracy of this software will improve. Quantum calculation in itself is based on quantum phenomena of superposition and consists of qubits that use the principle. Quantum computer performance will improve as quits can be measured faster.
Finish
While the current study has come up with a faster, cheaper and accurate solution, there are many challenges to overcome before it is ready for the industry.
Once ready, however, the world will see many applications due to greater accuracy in solving Schrödinger’s equation.
If you liked this story, join our Telegram community.
You can also write to us and be one of the 500+ experts who have contributed stories to AIM. Share your nominations here.
