Periodic Reporting for period 3 - AI-DEMON (Artificial intelligence design of molecular nano-magnets and molecular qubits)
Período documentado: 2024-01-01 hasta 2025-06-30
The magnetic moment associated with electrons, namely the electron's spin, is the key ingredient of any magnetic device, such as those ubiquitously found around us in hard magnets and sensors. Spin is also among the frontrunners among the possible physical implementations of quantum devices, which hold the promise for unprecedented computer resources and ultra-sensitive sensors. However, harnessing the features of electrons at the atomic scale without at the same time degrading them is far from a simple task to achieve. In particular, as temperature is increased from sub-Kelvin toward ambient values, the motion of atoms inside molecules and crystals is found to lead to a drastic decrease in spin lifetime, i.e. the time it takes spin to relax to thermal equilibrium and compromise its ability to store or process information. So far, we only have a very limited understanding of the microscopic mechanisms leading to spin relaxation, with the dramatic consequence of making it extremely hard to design new compounds with improved properties.
AI-DEMONS aims at developing an unprecedented computational framework able to tackle all the most important steps towards the design of novel magnetic molecules able to retain their properties at temperatures approaching the ambient one. In order to achieve this grand objective, AI-DEMON avails of an interdisciplinary theoretical and computational approach that brings together electronic structure theory, machine learning, open quantum systems theory, high-performance computing, and coordination chemistry. Building on such a large cohort of disciplines, AI-DEMON tackles the development of a quantitative ab initio theory of spin decoherence. Once combined with machine learning tools, the knowledge of how physics dictates the relaxation of electrons' spin in solid-state environments will make it possible to explore the chemical space of coordination compounds in search of ideal candidates to be exploited for technological applications such as quantum sensing and high-density information storage.
AI-DEMONS aims at developing an unprecedented computational framework able to tackle all the most important steps towards the design of novel magnetic molecules able to retain their properties at temperatures approaching the ambient one. In order to achieve this grand objective, AI-DEMON avails of an interdisciplinary theoretical and computational approach that brings together electronic structure theory, machine learning, open quantum systems theory, high-performance computing, and coordination chemistry. Building on such a large cohort of disciplines, AI-DEMON tackles the development of a quantitative ab initio theory of spin decoherence. Once combined with machine learning tools, the knowledge of how physics dictates the relaxation of electrons' spin in solid-state environments will make it possible to explore the chemical space of coordination compounds in search of ideal candidates to be exploited for technological applications such as quantum sensing and high-density information storage.
The team recruited to work toward AI-DEMON's goals has developed an unprecedented theory of spin relaxation and decoherence and has combined it with advanced electronic structure simulations in order to achieve a truly predictive computational method. In this approach, we have exploited the theory of open quantum systems, where the dynamics of the electron's spin is dictated by its interaction with a large environment of lattice vibrations at the thermal equilibrium. According to this theoretical framework, one or two lattice vibrations are able to interact with the spin in order to change its state and eventually drive it to thermal equilibrium. This theoretical framework has then been combined with advanced electronic structure methods, which make it possible to harness high-performance computing facilities to approximately solve the Schroedinger equation for real materials. The combination of these two approaches makes it possible to obtain a parameter-free description of spin relaxation and the application of this novel methodology has shown to be able to quantitatively reproduce available experimental results for crystals of magnetic molecules. These results have provided unprecedented insights into the mechanism of spin relaxation in magnetic molecules and have made it possible to compile a list of synthetic guidelines to help chemists realize novel chemical systems with improved properties.
Although ab initio spin dynamics theory allows us to obtain unprecedented insights into the relaxation of molecular spin moments, it only makes it possible to address a few molecules at a time due to its enormous computational cost. In order to tackle this issue and extend the application of ab initio spin dynamics to a large portion of the chemical space, AI-DEMON has integrated machine learning methods into its computational framework. Thanks to the development of dedicated machine-learning algorithms, AI-DEMON has enabled the computationally inexpensive prediction of both the potential energy surface and the spin properties of magnetic molecules, paving the way to the computational design of novel magnetic molecules with long spin lifetime.
Although ab initio spin dynamics theory allows us to obtain unprecedented insights into the relaxation of molecular spin moments, it only makes it possible to address a few molecules at a time due to its enormous computational cost. In order to tackle this issue and extend the application of ab initio spin dynamics to a large portion of the chemical space, AI-DEMON has integrated machine learning methods into its computational framework. Thanks to the development of dedicated machine-learning algorithms, AI-DEMON has enabled the computationally inexpensive prediction of both the potential energy surface and the spin properties of magnetic molecules, paving the way to the computational design of novel magnetic molecules with long spin lifetime.
The achievement of a quantitative ab initio theory of spin relaxation is a ground-breaking result that goes well beyond the state-of-the-art. Until recently, spin relaxation theory has been constrained to a phenomenological ground due to the impossibility of determining the many parameters that populate the equations describing spin dynamics. The combination of advanced electronic structure methods with a robust theoretical framework has made it possible to interpret experimental results that have baffled the scientific community for decades. Moreover, this result represents the stepping stone for a systematic exploration of the links between the chemical nature of magnetic molecules and their spin relaxation properties, eventually leading to a ground-breaking new paradigm for the design of quantum memories and sensors.
In order to achieve this second ground-breaking result, the AI-DEMON's team is now developing a machine-learning-driven high-throughput computational strategy to explore the chemical space of magnetic molecules. Machine learning will assist the automatic assembly of novel molecules and at the same time will provide a prediction of their magnetic and vibrational properties, making it possible to effectively explore the virtually infinite chemical space of magnetic coordination compounds in search of ideal molecular magnets and molecular qubits.
In order to achieve this second ground-breaking result, the AI-DEMON's team is now developing a machine-learning-driven high-throughput computational strategy to explore the chemical space of magnetic molecules. Machine learning will assist the automatic assembly of novel molecules and at the same time will provide a prediction of their magnetic and vibrational properties, making it possible to effectively explore the virtually infinite chemical space of magnetic coordination compounds in search of ideal molecular magnets and molecular qubits.