← Latest papers
🔬 materials science

quantEM: Quantitative Electron Microscopy Analysis from Differentiable Physics to Neural Priors

This paper introduces quantEM, an open-source Python package that unifies diverse quantitative electron microscopy analyses through a modular, differentiable physics framework built on PyTorch and Zarr, enabling seamless multimodal integration, GPU acceleration, and reusable physical models.

Original author: Colin Ophus, Arthur RC McCray, Georgios Varnavides, Stephanie M Ribet

Published 2026-11-18
📖 5 min read🧠 Deep dive

Original author: Colin Ophus, Arthur RC McCray, Georgios Varnavides, Stephanie M Ribet

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the author. For technical accuracy, refer to the original paper. Read full disclaimer

In the world of materials science, the quest to understand matter often begins with looking at it. Scientists use powerful microscopes to peer inside solids, revealing the arrangement of atoms, the flow of electrons, and the chemical makeup of everything from tiny computer chips to the proteins that sustain life. These instruments do not just take pictures; they capture a vast array of data, including how electrons bounce off a sample, how they lose energy, and how they diffract into complex patterns. The challenge for researchers has long been turning these raw, massive streams of information into clear, three-dimensional maps of the material's structure. To do this, they must solve difficult mathematical puzzles, working backward from the messy data they see to the hidden reality that created it. As microscopes have become faster and more sensitive, the amount of data they produce has exploded, often filling terabytes of storage in a single day, making the old ways of analyzing this information too slow and too fragmented to keep up.

A team of researchers has now introduced a new open-source software tool called quantEM, designed to bring order to this chaos. Instead of relying on a patchwork of different programs, each built to handle only one specific type of measurement, this new package unifies the entire process. It treats the analysis of electron microscopy data as a single, flexible system where the physical laws governing how electrons interact with matter are written into the software itself. By using a modern computing approach that allows the computer to automatically figure out the best way to solve these puzzles, the software can handle imaging, diffraction, spectroscopy, and 3D tomography all at once. This means a scientist can take a single dataset and ask it many different questions—such as where the atoms are, how much strain is in the crystal, and what elements are present—without ever having to convert the data into a different format or switch to a different program.

The core innovation of quantEM lies in how it handles the math behind the scenes. Traditionally, when scientists wanted to create a new way to analyze their data, they had to manually derive complex equations to tell the computer how to adjust its guesses. This was a slow, error-prone process that often meant new methods could not easily share the work of older ones. The new software replaces this manual labor with a system that learns the rules of physics and automatically calculates the necessary adjustments. It builds a digital model of the experiment, including the microscope's lens, the detector, and the sample, and then lets the computer refine that model until it perfectly matches the real data. Because the system is built on a single, unified foundation, it can run on standard computer processors as well as powerful graphics cards, making it fast enough to handle the massive datasets generated by modern instruments.

This approach allows researchers to combine different techniques in ways that were previously difficult. For example, the software can take a series of images taken from different angles and reconstruct a three-dimensional view of the material, or it can combine chemical data with structural data to see exactly where specific elements are located within a crystal. The tool is designed to be modular, meaning scientists can swap out different parts of the analysis, such as changing how the material is represented or how the noise in the data is handled, without rewriting the entire program. It also includes features to save and resume work, allowing a long-running analysis to be paused and continued later, or shared with a colleague who can pick up exactly where the first person left off.

The developers have already populated the software with a wide range of capabilities, including tools to map the orientation of crystals, measure the strain in a material, and reconstruct 3D volumes from tilt series. It can also simulate how electrons should behave in a perfect crystal, allowing researchers to compare their real-world observations against theoretical predictions. By keeping all these tools in one place and ensuring they all speak the same language, the software removes the barriers that often slow down discovery. It does not just make the analysis faster; it makes it more reliable, ensuring that the calibration of the microscope and the units of measurement are tracked correctly from the first step to the final result.

The impact of this work is seen in the ability to tackle problems that were previously too complex to solve. For instance, the software can now handle "ptychography," a technique that reconstructs images from overlapping diffraction patterns, with a level of detail that reaches the scale of individual atoms. It can also process data from experiments where the sample is heated, cooled, or stretched, capturing how materials change in real time. By making these advanced methods accessible and easy to combine, the software empowers scientists to ask deeper questions about the materials that shape our world. The code is freely available to anyone, accompanied by tutorials and example datasets, inviting the global community of researchers to build upon it and extend its reach.

In the end, quantEM represents a shift in how scientific software is built. Rather than creating isolated tools for specific tasks, it offers a comprehensive framework where the physics of the experiment is the central guide. This allows the software to adapt to new types of data and new ways of thinking without requiring a complete overhaul. As the field of electron microscopy continues to generate ever-larger and more complex datasets, tools like this will be essential for turning raw numbers into genuine understanding, revealing the hidden structures of matter with a clarity that was previously out of reach.

Drowning in papers in your field?

Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.

Try Digest →