Berkeley Lab Powers Worldwide Materials Research with Dash and Python

Introduction

Lawrence Berkeley National Laboratory (Berkeley Lab) is managed by the University of California for the U.S. Department of Energy’s Office of Science. Scientists and developers from multiple divisions at Berkeley Lab have built a core infrastructure to support worldwide collaboration on a suite of open-source tools that are accelerating materials science research around the world.

Challenge

Inventing innovative materials to address the world’s energy challenges is incredibly time-insensitive and costly. Solar cells, batteries, and lasers are some of the key technologies that utilize special crystalline materials. Identifying and making new crystalline materials in a lab setting is excruciatingly long. As an example, gallium nitride, the material used in LED lighting, is the product of several decades of research before commercialization was possible! Calculating even the basic properties of these materials requires extensive quantum physics calculations; without resources like a supercomputer, the right software, and broad access to data, progress is slow.

Solution

The Materials Project brings the required resources together in a single, free, open-source platform, where researchers who might not have access to all the necessities can benefit from pre-computed data. Over the last decade, a large, multi-institution collaboration of scientists and researchers across the world coalesced into the project as it stands today, where over 200,000 registered users have access to applications for exploring and analyzing nearly 150,000 different inorganic materials.

The core infrastructure staff consists of four individuals who administer the project’s website, API, and database infrastructure while also creating new tools for hundreds of thousands of users — no easy task. The small team leveraged the power of Python and Dash as an efficient and accessible way to tackle the work.

The project website itself is a Dash application written in Python. This approach eliminated the need for a full-stack web development team and capitalized on the Python skill sets of the team’s materials science researchers. A previous version of the website had been built with the Django framework, but with only a single web developer on the team, adding new front end features was slow. Switching to the Dash framework enabled the entire team to contribute to the website using Python and set the stage for end users to contribute their own applications as well.

Results

Fourteen applications for data exploration and analysis are available on the project website. Furthermore, the project team also open-sourced their own library of Dash components, giving researchers the ability to more easily create their own applications using analytics and user interface components specific to the domain of materials science.

The team has also used Dash to develop a user contribution framework for the website, enabling scientists around the world to upload their own data on new crystal structures they have discovered; an automated Python pipeline performs the requisite quantum physics calculations, identifies and corrects errors, and exports the data to a database that all users can freely download to further their own research.

Because Dash is a Python framework, the Materials Project benefits from leveraging the broader Python ecosystem in its tooling. This reduces the time costs of innovation and connects materials science domain experts more directly with tool development and data sharing. Since its creation, peer-reviewed publications associated with the project have been cited in research over 19,000 times and have helped foster collaborations across the public and private sectors. Applications in areas including carbon capture, better battery technology, and energy-efficient cooling and heating suggest an optimistic role for open-source science in addressing climate change and its impacts.