Qedma Reports 30–50x Accuracy Improvement in Quantum Chemistry

Say a quantum computer was attempting to calculate the energy of a single water molecule, and getting the answer significantly wrong. This isn’t a minor deviation; it’s the kind of inaccuracy where each gate operation introduces more static, rendering a complex chemistry calculation nearly illegible. This scenario reflects the current reality of quantum hardware. Now, envision running the same calculation on the same chip, but with a software layer that treats noise as a characterized adversary rather than a random disturbance. The result: an answer 30 to 50 times closer to the truth.

This remarkable achievement recently came from Qedma and the HQC2 team.

The work is a collaboration between Qedma Quantum Computing, a software company founded by Dorit Aharonov, Nathaniel Lindner, and Asif Sinay, and researchers from the University of Copenhagen, the Technical University of Denmark, and the University of Southern Denmark. The findings came out as a preprint on arXiv and saw a presentation at Q2B Copenhagen. It’s important to note that this is a proof of concept, not a peer-reviewed result, and does not claim quantum advantage.

Understanding Error Mitigation

Fault-tolerant quantum computing, the ideal scenario, involves building in extensive redundancy to correct errors as they occur. However, this requires significantly more qubits than currently available, and while the field has been understanding quantum error correction for years, hardware development is still catching up.

Error mitigation offers a more immediate solution. Instead of correcting errors during computation, it runs the circuit, measures how noise corrupts the output, and then estimates what the correct answer would have been. It’s analogous to knowing your bathroom scale consistently reads five pounds heavy; you don’t fix the scale, you simply subtract five.

Qedma’s approach, called QESEM, characterizes the specific noise profile of the chip in use and builds its correction strategy around this unique “fingerprint.” The noise isn’t treated as an abstract problem; it’s a measured and modeled phenomenon with distinct characteristics that QESEM learns before applying corrections.

Why Water and This Ansatz?

The team chose to study the potential energy surface of a water molecule. This surface maps how the molecule’s energy changes as its bonds stretch and bend, providing crucial information for understanding chemical reactions, molecular stability, and reaction pathways.

They employed an orbital-optimized variational ansatz, a sophisticated method that uses a flexible, tunable estimation for electron arrangement, which the algorithm then refines. This was executed on IBM’s Aachen processor as part of Q-CHEMION, a project funded by the Eureka open call for applied quantum technologies.

The team chose water because its properties are well-understood classically. When you calculate something for which classical chemistry provides an exact energy, you have a definitive benchmark. This allows precise measurement of how much the quantum machine errs and how effectively the mitigation software improves accuracy. Water, in this context, serves as an honest test case.

The Measured Improvement

The 30 to 50 times improvement refers to the accuracy increase over raw results without any mitigation. This comparison is critical: QESEM brought the noisy answer 30 to 50 times closer to the reference value than the unmitigated run.

It is crucial to understand what this number does not represent. It’s not a speedup over a classical computer, nor does it indicate that the quantum machine outperforms a laptop at chemistry tasks, as a laptop can accurately model a water molecule with ease. As Prof. Stephan P. A. Sauer of the University of Copenhagen emphasized:

“Accuracy is particularly important in quantum chemistry, where even relatively small errors can affect the reliability of calculated molecular properties and energies,” said Sauer. “This study allowed us to explore how error mitigation can improve the accuracy of chemistry calculations performed on today’s noisy quantum hardware.”

The value of this demonstration lies in showing that a real chemistry problem, run on genuinely noisy hardware, can be mitigated by software to achieve accuracy levels acceptable to chemists. This represents a potential bridge between current imperfect chips and the precise calculations required by future scientific endeavors.

What This Means and What It Doesn’t

To clarify the scope of this achievement:

This study focuses on a single, small molecule. Water has ten electrons, and quantum chemistry problems become exponentially more complex as molecules grow. The demonstrated 30 to 50 times improvement for water does not necessarily imply the same margin will hold for larger molecules where classical computers struggle – which is the ultimate goal of quantum chemistry.

Additionally, this is a preprint, meaning it has not yet undergone peer review. Other experts will scrutinize the numbers and methodologies.

Error mitigation also has inherent limitations. The overhead increases with larger and noisier circuits, meaning the technique is highly effective in the current regime but quickly becomes resource-intensive beyond it. Though Qedma has separately demonstrated quantum advantage with IBM in quantum materials dynamics (a stronger claim), it presents this chemistry result as a humbler proof of concept. CEO Asif Sinay himself described it as exactly that, a proof of concept, commendably avoiding overstatement.

In summary: a noisy quantum chip poorly performed a small chemistry problem, but sophisticated software impressively rectified the errors. However, it remains to be proven whether this mitigation scales to the complex molecules that would truly necessitate a quantum computer.