Quantum-enhanced supercomputers are starting to do chemistry

A quantum computer and a conventional supercomputer that work together could become an invaluable tool for understanding chemicals. A collaboration between IBM and the Japanese scientific institute Riken has now established a way to get there.

Predding what a molecule will do in a reaction – for example, in the context of medical treatment or an industrial catalyst – often depends on the understanding of the quantum states of its electrons. Quantum computers could accelerate the process of calculating these states, but in their current form, they are always subject to errors. Conventional supercomputers can take these mistakes before becoming a problem.

In a joint declaration at New scientistSeiji Yunoki and Mitsuhisa Sato in Riken said that quantum computers can push traditional computers to new capacities. Now, they and their colleagues have used IBM’s quantum computer Heron and Riken’s Fugaku SuperCalculator to model molecular nitrogen, as well as two different iron and sulfur molecules.

The researchers used up to 77 quantum bits, or qubits, and an algorithm called SQD to divide the calculation of the quantum states of the molecules between the machines. The quantum computer has made calculations while the supercomputer checked and corrected the errors. For example, if Heron produced a mathematical function describing more electrons than contained in the molecule to be accomplished, Fugaku would throw this part of the solution and would have the update of the heron and would repeat the calculation.

This hybrid method does not yet exceed the best case of what a supercomputer could do alone, but it is competitive with certain standard approaches, explains Jay Gambetta at IBM, which was not involved in the experience. “It is 1751640309 Rightly compare the calculation tools. »»

In the short term, this intervention is the “secret sauce” to obtain quantum computers subject to errors to make chemistry, explains Kenneth Merz at the Cleveland Clinic in Ohio. Using a different IBM quantum computer on a conventional computer, its team has developed a variation in the SQD algorithm which can model molecules in solutions, which is a more realistic representation of chemical experiences than previous models.

In the opinion of Merz, other SQD optimizations could help the combination of quantum and conventional computers to gain tangible advantages compared to the latter in the next year.

“The combination of Quantum and Supercalcuting is not only worth it – it is inevitable,” says Sam Stanwyck of the Computer Company Nvidia. A realistic use of quantum computer science is that where quantum processors are integrated into powerful conventional processors in a supercomputer center, he says. Nvidia has already developed a software platform that aims to support these hybrid approaches.

Aseem Datar at Microsoft says that his business has its objective on the “enormous potential in the combination of quantum computers, supercalculculculis and AI to speed up and transform chemistry and science of materials”.

But while stakeholders in the quantum IT industry defend the idea, many challenges remain. Markus Reiher at Eth Zurich in Switzerland says that the results of the Riken experience are encouraging, but it is not yet clear if this approach will become the preferred way to carry out quantum chemistry calculations. On the one hand, the accuracy of the final response of the quantum-supercutive computerly pair remains uncertain. On the other hand, there are already conventional methods well established for carrying out such calculations – and they work very well.

The promise to incorporate a quantum computer into the calculation process is that it could help model larger molecules or operate more quickly. But Reiher says that the scaling of the new approach can be difficult.

Gambetta says that a new version of IBM’s Heron Quantum computer was installed at Riken in June – and that is already fewer errors than previous models. It provides even greater material improvements in the near future.

Researchers also refine SQD algorithm and optimize the way Heron and Fugaku work in parallel to make the process more effective. Merz says that the situation is similar to that of conventional superordinators in the 1980s: there is no shortage of open problems, but the integration of new technologies could offer major yields.