Microsoft and Pacific Northwest National Laboratory bring AI to quantum chemistry research

Microsoft and the Pacific Northwest National Laboratory are banking on artificial intelligence and high-performance computing’s potential to advance research breakthroughs in chemistry and materials science, unveiling an updated software suite for widespread access across scientific communities. 

The new offering, a series of software tools moved onto Microsoft’s Azure Quantum Elements cloud computing platform, is specifically engineered to provide high-performance AI tools for chemistry and materials sciences researchers. 

“Our focus with PNNL is taking the AI capabilities that already exist on the platform and augmenting them with best in class tools for doing scientific discovery using physics based methods on chemistry problems,” Nathan Baker, an Azure Quantum product lead at Microsoft, told Nextgov/FCW.

The tools — which rely on advanced algorithms to process complex research data and draw sophisticated conclusions — are publicly available on Github. 

“We provide for the cloud computing [an] element of resources that is really needed. And if you have access to the high performance cloud resources, you can perform pretty large simulations,” said PNNL computational chemist Karol Kowalski in an interview with Nextgov/FCW. 

Physics-based AI models are central to the updated Azure suite. Long developed and utilized by the scientific community, these algorithms are trained on laws of physics combined with specific data reflective of different scientific disciplines. These AI models are then leveraged to solve specific scientific problems, such as how to break down “forever chemicals,” discover new drugs and support other outstanding scientific challenges. 

These hyper-specific algorithms are often the basis for much of the training data used in other large-scale, accelerated AI models, Baker said. 

“The purpose of [Azure Quantum Elements] is not to replace traditional laboratory experimentation, but to supplement it by offering a powerful computing platform,” Baker said. “We're building a more scalable framework where scientists everywhere can take advantage of those tools, not just the ones that have access to the supercomputers.”

Legacy software containing these physics-based algorithms now included in Microsoft’s updated AQE suite include NWChem — an open-source computational chemistry package originally developed at PNNL — NWChemEx,  FLOSIC, ExaChem and others. Leveraging these softwares across Azure Quantum’s cloud platform aims to enable complex chemistry simulations while also incorporating accuracy assessments. 

“It made sense to start developing an ecosystem around tools like high performance computing for physics based models; like artificial intelligence for accelerating the process of discovery; and like quantum computing for transforming accuracy,” Baker said.  

Both Baker and Kowalski emphasized that democratizing specialized research software for researchers with expertise in varying scientific fields is arguably the biggest achievement of this collaboration. 

“This opens the gates for the non-expert users,” Kowalski said, explaining that scientists who don’t have a strong basis in quantum chemistry should still be able to benefit from such tools.