Quantum Computing for Materials Discovery

When a research team evaluates a new battery chemistry, the work starts on a computer, before anyone builds a physical cell. The software predicts how a candidate will behave, including how much voltage and capacity it delivers, how it handles heat, and how long it lasts. Those predictions come from first-principles methods, which calculate material behavior directly from quantum mechanics rather than from fitted experimental data. For the bulk of a material, they are accurate predictions. The trouble starts at the interfaces—the thin regions where one material meets another.

Those interfaces are exactly where the models take shortcuts, and exactly where batteries fail. A model might promise that a cell holds its charge for a thousand cycles, only for the real cell to wear out after a few hundred. While the model captured the bulk behavior correctly, the reaction that caused the failure was simply never in the calculation. Capturing the interface chemistry in full is possible in principle, but the calculation grows so fast with each added atom that an exact solution would outrun any classical computer and any realistic deadline. So, researchers leave out details that sometimes prove to be decisive.

For engineers designing products, the gap between the model and reality becomes a practical design constraint. A battery’s chemistry sets how fast it can charge, how much heat it produces, what safety circuitry it needs, and how long the product can last. When the chemistry is modeled accurately up front, fewer failures emerge later—the kind that force teams to redesign charging, thermal, and safety systems after the fact.

Quantum computing could narrow that gap. A quantum processor stores and manipulates information using quantum mechanics, so it can represent the electronic structure of a material in ways a classical computer cannot replicate.   The calculations that overwhelm classical computers are often the same interface problems that decide whether a cell lasts in service.

Today, researchers discover materials through an established workflow, often called the stack: classical simulation, high-performance computing, machine learning, and lab validation. Mouser Electronics sat down with Phalgun Lolur, Head Scientist at Capgemini's Quantum Lab, to talk about where quantum computing fits into that stack. He puts it simply: “Quantum won't replace the stack. It will slot into it.” If quantum hardware matures on the timelines its developers project, which is a milestone that depends on error correction at scale, quantum methods could start joining that workflow within the decade.
 

Dr. Phalgun Lolur, Head Scientist, Capgemini’s Quantum Lab, leads the research and scientific direction within Capgemini’s Quantum Lab and Capgemini Engineering R&D, focusing on projects that bring together chemistry, physics, materials science, data science, and quantum computing. He is a Royal Society-endorsed expert in theoretical and computational chemistry, quantum mechanics, and quantum computing. His work centers on integrating quantum computing with established scientific methods and building practical tools and workflows for solving real-world materials and chemistry challenges.

Where Today’s Models Lose Accuracy

Battery researchers do not lack tools. First-principles methods have run on classical computers for decades, and the underlying physics and chemistry are well established. The key question is a practical one: can these methods run with the speed and reliability a design decision needs, rather than only confirm a result afterward?

“We are not short on models,” Lolur says. “What we struggle with is describing real materials accurately and quickly enough to influence design.”

The obstacle is fundamental to the mathematics, and adding more computing hardware does not remove it. Every atom adds electrons, interactions, and possible configurations the calculation has to track, and the number of possible arrangements grows exponentially as the system grows. Even a slightly larger material can become a dramatically harder calculation—far harder than researchers could handle by adding more processors. So, they simplify the problem from the start.

Lolur uses caffeine to illustrate the scale. It is a small, everyday molecule, yet modeling every one of its electrons exactly is already beyond a classical computer. Researchers get useful predictions for molecules that size only by approximating. The standard approximation, density functional theory (DFT), keeps the interactions most likely to matter and drops the rest. DFT works for many systems yet fails for others. It loses reliability when electrons interact too strongly to be treated one at a time, or when charge transfer depends on the exact local arrangement of atoms. Both conditions are rare inside a single molecule; however, both are common at a battery interface.

The DFT modeling method buys useful answers by leaving things out, and it pays for them in lost accuracy. Batteries make that cost hard to manage, because a working cell never stops changing. For example, electrolytes decompose, boundaries shift with every charge, and impurities build up. A chemistry that performs well at room temperature can falter in the cold or under fast charging. These conditions promote lithium plating, which is a buildup of metallic lithium on the electrode surface that permanently reduces capacity and poses a safety risk in lithium-ion cells. As a result, cycle 1,000 behaves very differently from cycle 1.

Most of these failures start at interfaces, where the chemistry is hardest to simplify. Protective layers form, side reactions begin, and small defects grow into reliability problems. “People get by with approximations,” Lolur says, “but those approximations can break down where the interface behavior determines performance.” Early in a cell’s life, a protective film called the solid electrolyte interphase (SEI) forms on the electrode. Whether that film grows evenly, whether lithium plating occurs during fast charging, and how side reactions spread across an electrode surface all depend on fine-grained local chemistry. Capturing it accurately in predictions is difficult, because the approximations that make the calculation manageable tend to discard the very details that decide how the cell performs.

The result is wasted development time and lab work. A promising electrolyte candidate might pass early tests, then degrade in production. A new chemistry might look stable in simulation, then fail after a few hundred cycles. In each case, the model captured the material’s interior correctly, while the failure traced back to an interface detail it left out.

Why Quantum Hardware Fits the Problem

Quantum hardware suits the complexity of battery materials discovery for one reason: the electrons that drive battery chemistry already obey quantum mechanics. To capture their collective behavior, a classical computer must translate it into ordinary bits, and that translation demands exponentially more memory and processing time as the system gets larger and more varied.

A quantum computer skips that translation, because it works on the same principle as the electrons it models: superposition, the ability of a quantum system to occupy many states at once. Electrons in a material exist in superpositions of possible configurations, and a quantum computer stores information the same way. It does this with qubits. A qubit is the quantum counterpart of the classical bit, with one defining difference—where a bit must be either 0 or 1, a qubit can hold many states at once. Because qubits and electrons follow the same underlying rules, a quantum processor can represent all the interacting electrons in a material directly instead of approximating them. The advantage here is accuracy. Quantum processors could reduce the approximation error exactly where DFT is weakest, i.e., in the strongly interacting electrons and the interface charge transfer, even when it runs no faster than a classical machine. This ability would let researchers model more of the real material.

The hardware has not reached that point yet. Today’s machines run up to a thousand physical qubits, although they are error prone. ,  Useful chemistry will likely need hundreds of logical qubits, each one stable because it is built from many physical qubits that work together to correct each other’s errors. Reaching that point could require millions of physical qubits in total. ,  Major vendors have published roadmaps toward machines at that scale, though those dates are projections that the hardware has not yet met.  Building a larger machine is not sufficient on its own. What matters is if it returns trustworthy answers inside real workflows.

Lolur frames the transition from today’s error-prone machines to future error-corrected ones as an opening, a chance to begin work on problems that are out of reach today.

“Once quantum hardware becomes reliable enough, we can start designing solutions for problems we cannot even represent on today’s computers,” he says. “Will we solve them right away? Not necessarily. But it gives us room to start studying the chemistry at material interfaces, and to consider new materials we have no way to design today.”

“The key,” he adds, “is not which material quantum computing solves first. It is which modeling errors it helps remove.”

The same lesson may apply well beyond batteries. Battery electrodes, corrosion-resistant alloys, carbon-capture materials, and catalysts may look like separate problems, yet many of them share the same obstacle: chemistry at the boundary between two different materials, where the usual approximations stop being reliable. If quantum methods improve the calculation for one kind of interface, the gains may transfer to the others.

A New Layer in the Workflow

In battery materials discovery, quantum computing adds a new layer to the existing stack of classical simulation that includes high-performance computing (HPC), machine learning (ML), and lab validation. Rather than replacing these methods, quantum integrates into the workflow as a specialized accelerator for a narrow set of particularly difficult calculations—similar to how a graphics processing unit (GPU) handles particular tasks inside a larger computer. The overall process remains largely classical. Quantum methods are introduced only where classical approaches are least reliable, extending the toolkit without changing the foundation of the workflow.

ML is already valuable because it can quickly search large sets of candidate materials. But the goal in materials discovery is often to design something no one has built, so the training data may be sparse or missing. When data is sparse, the search depends on the physics used to evaluate each candidate, and quantum simulation could supply high-accuracy reference points to check or train the models where experimental data does not exist.

Lolur does not regard AI and quantum computing as competitors. “Artificial intelligence and quantum computing are going to work very closely together,” he says. “As quantum computing matures, it will give the AI models much more accurate starting points, which reduces their workload.”

At Capgemini’s Quantum Lab, the team builds these hybrid workflows with industry partners, combining ML, classical simulation, and quantum methods with experimental validation, so that better modeling reaches the design decisions that depend on it.

How the Benefits Reach Design Engineers

For most engineers, the benefits of quantum-assisted materials work arrive as more dependable chemistries: higher energy density, longer cycle life, and wider operating windows. Those properties shape charging strategy, thermal design, safety systems, maintenance assumptions, and the tradeoffs in any product that stores energy. No one is going to write quantum programs to choose a cell chemistry. The results simply arrive inside the components.

Those gains come from better modeling upstream. If materials teams can model how an electrolyte breaks down, they can design cells that last longer. If they can predict interface chemistry with more confidence, they can screen designs before building them. If they can rule out weak candidates by calculation, development moves faster and lab time goes to the chemistries worth pursuing.

The challenge of modeling chemistry at interfaces is not unique to batteries. Corrosion alone costs the world about $2.5 trillion a year, according to the Association for Materials Protection and Performance (AMPP, formerly NACE International) IMPACT study,  and at its core it is another interface-chemistry problem. Catalysts, alloys, and carbon-capture materials face the same difficulty—chemistry at a boundary, where simplified models miss the behavior that decides performance.

Conclusion

The physics of battery chemistry is well understood. The barrier to discovering new chemistries is computational. Predicting how a candidate material will behave takes more calculation than classical computers can deliver in full. To make that physics solvable, today’s models rely on approximations, which fail most often at the interfaces that decide how long a cell lasts. Quantum computing targets that weakness directly. By representing electrons in their natural quantum form, it could reduce the approximation error in the calculations where classical methods are weakest.

If quantum computing hardware matures on the timeline its developers project, materials discovery would run as a layered process: ML proposes candidates, classical simulation screens the bulk of them, quantum simulation resolves the interface cases classical methods cannot, and the lab confirms the result. The models will still simplify the problem, but they may reduce the need for simplifying assumptions in the most challenging parts of the calculation.

That future depends on progress that has not yet arrived. It needs error-corrected hardware at logical-qubit scale, quantum methods that integrate with classical workflows, and a demonstrated accuracy advantage on a genuine interface problem. For a design engineer, the takeaway is to calibrate rather than wait. Track those milestones as they are demonstrated, weigh near-term vendor performance claims against them, and treat the gap between a roadmap and a validated result as the measure of how far the technology really is.