AtomDensityMap: Simulation-informed Point Defect Analysis

In the world of materials science, the path to innovation often lies within imperfection. While we might imagine that the best materials are perfectly ordered crystals, it is actually the tiny atomic defects - a missing atom here, an extra one there, or a single atom replaced by another element - that dictate how a material behaves. These microscopic flaws are the secret architects behind how a device conducts electricity, how a catalytic material helps to produce green hydrogen, or how a battery stores energy. Despite their importance, identifying these defects and understanding the tiny electric charge variations surrounding them remains one of the greatest challenges in electron microscopy, as they are often too small or too fragile to be captured reliably with current techniques. This project seeks to overcome these challenges by reimagining the way we use scanning transmission electron microscopy. Instead of simply taking a traditional image, we will capture the full complexity of how electrons scatter from every single point in a material, creating a massive and intricate dataset. This approach provides a wealth of hidden information that cannot be interpreted directly by the human eye. To unlock these secrets, we are developing advanced reconstruction methods that can translate this data into precise maps. These maps will reveal not only where individual atoms are located but also how electric charge is redistributed around a defect, providing a level of detail that was previously out of reach. To this end, we will bridge the gap between experiment and theory by integrating realistic computer simulations and modern machine learning into the reconstruction. These physics-based rules act as a safeguard, guiding our algorithms to produce meaningful results while filtering out the noise and experimental drift that often lead to false conclusions. Maximising the level of sensitivity is particularly crucial for fragile materials, as it allows us to observe defects using a very low dose of electrons, preventing the microscope`s beam from damaging the very structures we are trying to study. By testing our methods on well-controlled materials like layers of graphene and molybdenum disulphide, we will create a validated, open- source toolset for the global scientific community. Ultimately, this work will provide the foundational knowledge needed to design more efficient electronics, faster processors, and the next generation of sustainable energy technologies.