Quantum computers reflect experimental reality

Quantum computers are gaining a reputation among chemists and physicists as promising tools that can tackle problems that are too complex for ordinary computers and inaccessible experimentally. In a study that takes an important step toward that promise, researchers used a quantum computer to simulate an experiment and report on Accurately forecasted actual data (PDF). The study, which has not yet been peer-reviewed, was a collaborative effort between two national laboratories, several universities, and IBM Quantum.

“For the first time, we can directly compare the results of a quantum computer to a complex experiment,” says Alessandro CurioniVice President of IBM Europe and Africa and Director of IBM’s Research Laboratory in Zurich.

The researchers tested the quantum simulation against reality using data recorded in a neutron-scattering experiment conducted at Oak Ridge National Laboratory. In this type of experiment, scientists fire neutrons at a material and track how they scatter. The data reveal the arrangements and motions of atoms and electrons in the target material, information that scientists use to understand the material’s properties, such as its conductivity, catalytic activity and energy storage.

Researchers focus on crystalline KCuF₃A well-studied magnetic substance. They simulated neutron scattering through quantum calculations and compared the results to data measured at Oak Ridge. “And it works.” Curioni says. “You can see that the difference is negligible.”

“As a quantum computing achievement, this seems to be a very good progress,” says steven whiteA physicist at the University of California, Irvine, who was not involved in the study. But he says there are reasons to examine the development.

Classical methods outperform quantum methods

The team chose KCuF₃ Because its properties can be calculated by a classical computer. White and his colleagues have been simulating neutron scattering for similar materials with high precision using classical methods since 2006. Furthermore, the classical simulations conducted in this study beat the quantum version on every metric the researchers tested in terms of accuracy and computing efficiency.

Regarding efficiency, Curioni says quantum computers are often misunderstood as simply faster processors. “But that approach is very limited,” he says. He argues that quantum computing represents information in a fundamentally different way, changing the types of problems that can be tackled.

The accuracy of quantum simulations is limited Hardware in quantum machines. To predict the neutron-scattering process, a quantum computer simulates KCuF₃ The crystal grows as the neutrons collide, one small step at a time. At each step, the machine applies operations to neighboring qubits to mimic the way atoms in the crystal influence each other. The more steps the machine can perform, the faster the final output will be.

But these operations fail about once every thousand attempts, and these errors grow as the simulation runs. Like a photo that gets blurry every time it is copied and reprinted, the quantum result becomes noisy and gradually loses detail and becomes blurry. In contrast, the classical simulation preserves the sharp features of the internal dynamics of the crystal.

Paradoxically, noise can make quantum simulations match experimental data more closely than classical data. The results of actual experiments are also blurred due to thermal motions of atoms and detector limitations. But the researchers say that in this case the apparent match is merely a coincidence, not proof that the quantum processor is capturing physics that the classical method misses.

The classical machinery behind the quantum result

IBM uses quantum-centric supercomputing label. “This should be interpreted carefully,” says Jacob Biamontequantum computing researchers at Montreal’s School of Higher Technology (ETS), who were not involved in the study. “The workflow is strongly mixed with substantial classical input.”

Before the quantum processor can begin calculations, it must load information about the crystal’s ground state – the state of the material before the neutron hits it. That initial state is determined using classical simulations and then transferred to a quantum computer with 80–85% fidelity.

Full development for today’s quantum hardware is still very demanding. So the researchers used classical computers to compress the early part of the simulation into smaller, more manageable steps.

This quantum simulation does not outperform classical methods. Nevertheless it establishes a workflow for evaluating quantum simulations. “Benchmarking against actual physical measurements, rather than just classical simulations, is an important direction for the field,” says Biamonte.

Real quantum testing: beyond classically solvable problems

There are some problems that classical computers cannot solve even with all the resources available today. In principle, quantum computers could be used to solve those problems, creating what is known as quantum advantage. This domain is where quantum computers will be really useful to scientists.

Neutron scattering from KCuF₃ There is a problem that can be solved in a classical way. But other materials are still not amenable to calculations done on classical computers. “This is where quantum computers can ultimately help,” says White. “But those systems tend to be less hardware friendly.” We just need better hardware.

Curioni agrees today’s hardware Systems beyond classical reach cannot yet be dealt with. The open question is whether getting there will require gradual hardware improvements or major breakthroughs.

Although Curioni emphasizes how close IBM is to quantum advantage, some researchers are skeptical. But Biamonte is not one of them. “Given the current trajectory, such experiments may begin to yield benefits soon.”