AI 4 Quantum

Overview

Within the AI 4 Quantum project, we develop and apply novel machine learning (ML) approaches to enable the computational study of complex quantum systems relevant to the green energy transition. The groups’ research foci are:

– Developing novel neural network quantum states that efficiently describe complex quantum matter via physics-inspired and symmetry-preserving methods.

– Using deep learning methodologies to aid in quantum computing algorithm design and initialise noise-resilient quantum circuit Ansätze.

– Extending the reach and enhancing the accuracy of existing quantum Monte Carlo methods for complex quantum systems through AI approaches.

Motivation

One of the main obstacles to advancing green technologies, such as bio-catalysis for efficient ammonia production, artificial photosynthesis or loss-less energy transmission via superconductors, is the complexity of their underlying quantum mechanical mechanisms. These processes, enabled by small quantum systems with strong electronic correlations, are not yet understood and pose significant challenges to experimental and theoretical investigations. Hence, there is a pressing need for sophisticated computational methods to elucidate these phenomena and support the development of green technologies. For example, artificial nitrogen fixation – converting molecular nitrogen into ammonia – at lower temperatures and pressures would offer a more sustainable alternative to the energy-intensive Haber-Bosch process, drastically reducing greenhouse gas emissions.

Fortunately, in quantum chemistry and solid state physics, the fundamental equation that describes these processes – the Schrödinger equation – is at hand. A solution of the correlated motion of electrons – given by its wavefunction \(\Psi(\mathbf{x})\) – would allow a description of groundbreaking physical and chemical phenomena, including nitrogen fixation, photosynthesis and superconductivity. An accurate fundamental understanding of these processes at the quantum scale would enable a bottom-up materials design approach to mimic them artificially.

However, due to the correlation between electrons, the Schrödinger equation is not analytically solvable, as the size of the exact wavefunction \(\Psi(\mathbf{x})\) grows exponentially with the problem size. Thus, we need the aid of approximations and computational approaches. There exists a wide variety of computational methods, ranging from affordable mean-field Hartree-Fock (HF) and density functional theory (DFT) to highly accurate but computationally expensive approaches like quantum Monte Carlo (QMC). Mean-field methods yield satisfactory results for so-called weakly correlated systems. However, HF and DFT are inadequate to describe the strongly correlated quantum systems that enable artificial nitrogen fixation or superconductivity. Advanced and costly computational methods are required to study these systems. However, the computational cost of these accurate methods grows steeply with the size of the investigated system – at worst exponentially. Thus, the application of accurate methods is currently limited to only the tiniest of problems.

The aim of AI 4 Quantum is to develop and apply novel deep machine learning, neural network and artificial intelligence approaches to extend the reach of accurate computational methods to study strongly correlated quantum matter relevant for the green energy transition.

To achieve this goal, my team and I will pursue three separate yet complementary research pillars:

(Under construction)

Hosted by:

Funded by: