Building a quantum computer that pays off

At a Tuesday afternoon plenary session at SPIE Advanced Lithography and Patterning, Erik Hosler discussed the path to a useful quantum computer. The topic is of more than academic interest to Hosler. He heads the Process Exploration for Photonics Department at quantum computing startup PsiQuantum. So, he’s one of the people tasked with making quantum computing provide a return on investment.

Early in his talk he discussed a defining characteristic of a useful quantum computer. “It must impact society at large,” he said.

In terms of money, he added this means “the value of the computations it performs exceeds the cost to build and operate the computer.”

There are a number of possible uses for a quantum computer, which like its classical counterpart has logic, memory, and moves data between the two. Because it stores and communicates information in the quantum states of matter or light and uses such quantum concepts as superposition and entanglement in its operations, a quantum computer can solve problems a classical computer cannot, like breaking encryption.

Quantum computer applications have already been identified in finance, healthcare, materials, security, and transportation. For example, a quantum computer might help re-engineer lithium-ion batteries so they perform better. Analysts have speculated the quantum computing market might approach a trillion dollars by 2035, Hosler said.

Standing in the way of this lucrative future is the fact that no one has built a quantum computer that meets the definition of being useful. In 2019, Google claimed quantum supremacy by demonstrating the ability to solve a problem a conventional computer cannot. That demonstration ran on a 53-qubit, or quantum bit, chip for 20 clock cycles to spit out a solution to a problem of no practical use beyond the demonstration itself. Although state-of-the-art, that chip and its performance are far short of what’s required.

“We need hundreds to thousands of usable qubits with the capability to do billions of sequential operations to really do useful work,” Hosler said.

Noise in current qubits means that many physical qubits are needed to make up a single useable one. The ratio today is about 1000:1, according to Hosler. That number varies according to the noise level of the physical cubits.

Feeding the output of many less-than-perfect qubits into an error correction system produces a single, logical qubit that is sufficiently noise free. It can then be part of a group of logical qubits, with these then enabling a quantum computer to run long enough to do useful work. So, what’s really needed is about a million qubits.

There’s a race today among companies to produce that first useful quantum computer, with different companies backing different technologies. Some are matter based, typically running at temperatures a few millidegrees above absolute zero. Others, like the technology used by PsiQuantum, use photons. These also run at cryogenic temperatures but could, in theory at least, run at room temperature, Hosler said. Being able to run at even a few degrees warmer than the matter-based competition could be a substantial advantage.

PsiQuantum picked an approach based on light and silicon photonics in part because the company believes that will be the easiest way to scale from a single qubit — with the proof-of-principle concept already demonstrated in a lab — to a million in a commercially useful form. From a manufacturing and lithography point of view, there are many shared challenges between semiconductors found in conventional computers and quantum computing done using silicon photonics. These include patterning related issues, such as a need for low line-edge roughness as well as low overlay and edge placement error.

“We’d like to see perfection in our patterning, but we’ll settle for striving for perfection,” Hosler said in discussing the situation.

There are some ways, though, that lithography for quantum computing differs from that for semiconductors. The critical feature sizes, for instance, in quantum computing devices are in the hundreds of nanometers and even microns, far larger than today’s leading-edge semiconductors.

Nonetheless, Hosler said the semiconductor industry and its technology are essential to building a useful quantum computer. The patterning techniques developed for advanced EUV, for instance, might be needed in a photon-based quantum computer to achieve the necessary nanoscale dimensional control.

Hosler said PsiQuantum aims to build a million-qubit system, with manufacturing already underway. He acknowledged, though, that at the moment no approach can be definitively ruled in or out. In part, this is because simply getting a quantum computer to work won’t be enough.

During the Q&A after his talk, he said that a useful quantum computer will have to meet not only performance but also economic thresholds.

As he said, “We need to build a quantum computer that doesn’t break the fab and doesn’t break the bank.”

Hank Hogan is a science writer based in Reno, Nevada.