Can IA and Super-Computing Reduce Lithium Usage ?

The problem with Lithium

However, as technological reliance on lithium-ion batteries continues to surge, the demand for lithium follows suit, exhibiting an exponential trajectory. The propulsion of EVs into the mainstream and the ubiquity of electronic devices underscore this upward surge. This exponential demand paints a landscape where projections by authoritative bodies, including the International Energy Agency, anticipate a tenfold increase in the demand for lithium-ion batteries by 2030.Yet, this surge in demand casts a shadow of concern—the looming risk of a lithium shortage. Projections indicate that this shortage could materialize as early as 2025, prompting the need for preemptive measures to maintain global energy transition.

Microsoft and the Pacific Northwest National Laboratory (PNNL) set out to revolutionize battery technology by leveraging the powerful synergy of Artificial Intelligence (AI) and High Performance Computing (HPC).

Source : IEA Report "The Role of Critical Minerals in Clean Energy Transitions"

In addition, the use of liquid electrolytes in lithium-ion batteries raises safety concerns. The combustible nature of these electrolytes, which is necessary to achieve high energy density, has safety implications, including the risk of overheating and fire incidents. At the same time, the industry is constantly looking for ways to increase energy density in order to allow for longer operational times and greater device performance

Complicating matters, traditional lithium mining methods carry a substantial environmental cost. The extraction process involves substantial water consumption and vast amounts of energy, primarily sourced from non-renewable fossil fuels. This leads to potential water scarcity concerns in already stressed regions and exacerbate the environmental footprint. Additionally, the disposal of mining byproducts, often containing hazardous materials, poses risks of soil and water contamination.

The combination of AI & High Performance Computing

Confronted by these challenges, the imperative to explore alternative materials capable of reducing or replacing lithium in batteries becomes more pressing. In this context, a significant breakthrough has emerged in battery technology as Microsoft and the Pacific Northwest National Laboratory (PNNL) collaborated to discover a new potential material for the battery industry that holds the potential to substantially decrease lithium usage. This has been achieved through the prowess of advanced AI and high-performance computing (HPC), a type of cloud-based computing that combines large numbers of computers to solve complex scientific and mathematical tasks.

In a constantly evolving environment of technological advancements, the pursuit of sustainable energy storage solutions remains a critical challenge. At the heart of this difficulty is the role of lithium, also known as "white gold". Lithium stands as the elemental backbone powering rechargeable batteries, notably lithium-ion technology. Its unique properties have positioned it as an indispensable component, driving the efficiency and longevity of batteries that fuel our electric vehicles (EVs) and myriad electronic devices.

As energy transitions gather pace, clean energy technologies are becoming the fastest-growing segment. According to the International Energy Agency report, their share of total demand is expected to rise significantly in the next two decades regardless of the projection scenario.

In a scenario that meets the Paris Agreement goals ( IEA Sustainable Development Scenario [SDS]) - climate stabilization at “well below 2°C global temperature rise”), Lithium demand may exceed over 90%.

The Stated Policies Scenario (STEPS) provides a more conservative benchmark for the future as it is based on a detailed review of the current policy landscape. However, it is still showing an important increase of Lithium (74%) demand by clean energy technologies.

The Azure Quantum Elements platform, played a central role in this revolutionary discovery. It served as a virtual laboratory allowing researchers to navigate a massive material database with unprecedented precision.

Concretely, the AI algorithms screened over 32 million materials and narrowed them down to 18 promising candidates, ultimately leading to the unearthing of a new substance codenamed N2116, that has the potential to transform battery safety and sustainability.

The process of narrowing down material candidates - Microsoft

Surprising results

What makes this candidate stand out is that it combines lithium and sodium, an abundant element and the primary component of salt. Microsoft claims the new material could reduce the amount of lithium used in a battery by up to 70%.Furthermore, it could be used to create a solid-state battery that is safer than today's lithium-ion batteries, which use liquid electrolytes and are more susceptible to overheating. Moreover, This accelerated screening process was achieved in just one week and represents a paradigm shift in the pace of scientific discovery compared to traditional laboratory research methods. To put this in perspective, traditional laboratory research methods could have taken over two decades to produce a comparable result.

It is worth to highlight that the new material is still under testing, the exact chemistry is subject to optimization and might not work out when evaluated at larger scale. However, the story here is not about this particular battery material, but rather the speed at which the discovery has been made.

Through platforms like Azure Quantum Elements, these breakthroughs are becoming accessible to all, accelerating innovation and compressing centuries of progress into decades. The combined force of scientific expertise and AI has the potential to revolutionize industries and unlock a new era of discovery, ultimately shaping a brighter future for all.

At first, 32.6 million candidate materials were filtered using a workflow that combined our AI models of materials with traditional HPC-based simulations. This first application identified 500,000 materials that are predicted to be stable. Then, AI models were used to screen this pool of materials for functional properties such as redox potential and band gap, reducing the number of possible candidates to around 800.

The second stage of screening combined both physics simulations and AI models. High-performance computing resources in Microsoft Azure were harnessed to conduct intricate Density Functional Theory (DFT) calculations. This step served as a critical cross-check on the properties identified by the initial AI screening, as AI models can sometimes have prediction errors. By performing these DFT calculations, researchers essentially obtained a higher-fidelity assessment of the predicted properties.

Microsoft's Azure Quantum team achieved these results by combining cloud-based HPC calculations with new AI models that estimate material characteristics such as energy, force, stress, electronic band gap, and mechanical properties. These models have been trained on millions of data points from materials simulations, allowing them to minimize HPC calculations while predicting material properties 1,500 times faster than traditional density functional theory (DFT) calculations.

Then, Microsoft Quantum researchers used AI-accelerated MD simulations to investigate dynamic properties like ionic diffusivity. These simulations used AI models for forces at each MD step, rather than the slower DFT-based method. This stage reduced the number of candidates to 150. Then, practical features such as novelty, mechanics, and element availability were taken into consideration to create the set of 18 top candidates.

Subsequent investigation of the candidate materials employed advanced AI-accelerated molecular dynamics (MD) simulations. This innovative approach substituted the established, computationally intensive DFT method with AI models to calculate interatomic forces at each MD step. This significantly enhanced computational efficiency while enabling profound insight into dynamic properties, such as ionic diffusivity. Consequently, this stage drastically reduced the candidate pool to 150 materials exhibiting promising characteristics. Finally, pragmatic considerations encompassing novelty, mechanical properties, and element availability were critically appraised to select a set of 18 top candidates exhibiting significant potential was carefully selected for further in-depth investigation.

After all the sifting and sorting, it was PNNL's expertise that took the final call. The laboratory researchers identified additional screening parameters, effectively refining the selection of final structural candidates. Subsequently, the synthesized the most promising candidate, meticulously characterized its structural properties, and meticulously quantified its conductivity.