The study, published in 'Science', details how this AI-driven method can solve fundamental equations that describe complex molecular systems.
Excited States of Molecules
The research focused on the challenge of understanding how molecules shift between 'excited states.' When molecules absorb significant energy-through light exposure or high temperatures-electrons may shift into a temporary, energized configuration, known as an excited state.
These transitions between states involve specific energy absorption and release patterns, which create unique molecular 'fingerprints.' These fingerprints are crucial for understanding and optimizing the performance of technologies such as solar panels, LEDs, semiconductors, and photocatalysts. Additionally, they are vital for biological processes involving light, such as photosynthesis and vision.
However, accurately modeling these excited states is notoriously difficult due to the quantum nature of electrons, whose positions can only be described probabilistically.
Dr. David Pfau, the lead researcher from Google DeepMind and Imperial's Department of Physics, explained: "Representing the state of a quantum system is extremely challenging. A probability has to be assigned to every possible configuration of electron positions.
"The space of all possible configurations is enormous - if you tried to represent it as a grid with 100 points along each dimension, then the number of possible electron configurations for the silicon atom would be larger than the number of atoms in the universe. This is exactly where we thought deep neural networks could help."
Neural Network Approach
The research team developed a novel mathematical technique and applied it using a neural network called FermiNet (Fermionic Neural Network). FermiNet is notable for being the first deep learning model capable of accurately computing the energy of atoms and molecules based on fundamental principles.
Testing this approach on a variety of cases, the researchers achieved promising outcomes. For instance, on the complex carbon dimer molecule, they reached a mean absolute error (MAE) of 4 millielectronvolts (meV)-five times closer to experimental results compared to previous methods, which had an MAE of 20 meV.
Dr. Pfau added: "We tested our method on some of the most challenging systems in computational chemistry, where two electrons are excited simultaneously, and found we were within around 0.1 eV of the most demanding, complex calculations done to date.
"Today, we're making our latest work open source, and hope the research community will build upon our methods to explore the unexpected ways matter interacts with light."
Research Report:Accurate Computation of Quantum Excited States with Neural Networks
Related Links
Imperial College London
Understanding Time and Space
| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
