Enlarge / A single logical qubit is built from a large series of hardware qubits.
Certainly one of many more hanging issues about quantum computing is that the self-discipline, regardless of now not having proven itself especially invaluable, has already spawned a series of startups that are targeted on constructing one thing utterly different than qubits. It may very well be easy to brush aside this as opportunism—attempting to cash in on the hype surrounding quantum computing. Then again it can be invaluable to see at the issues these startups are targeting, because they can be an indication of hard concerns in quantum computing that haven’t yet been solved by any certainly one of many substantial companies focused on that space—companies savor Amazon, Google, IBM, or Intel.
In the case of a UK-based company called Riverlane, the unsolved fragment that is being addressed is the grand amount of classical computations that are going to be necessary to make the quantum hardware work. Specifically, or now not it is targeting the grand amount of data processing that will be needed for a key part of quantum error correction: recognizing when an error has took place.
Error detection vs. the data
All qubits are fragile, tending to lose their state for the duration of operations, or simply over time. No matter what the abilities—chilly atoms, superconducting transmons, whatever—these error rates establish a hard restrict on the amount of computation that can be accomplished earlier than an error is inevitable. That rules out doing almost every invaluable computation operating immediately on new hardware qubits.
The generally accepted resolution to this is to work with what are called logical qubits. These contain linking a number of hardware qubits together and spreading the quantum information among them. Additional hardware qubits are linked in so that they can be measured to video display errors affecting the data, allowing them to be corrected. It can take dozens of hardware qubits to make a single logical qubit, meaning even the largest new systems can easiest enhance about 50 tough logical qubits.
Riverlane’s founder and CEO, Steve Brierley, instructed Ars that error correction doesn’t easiest stress the qubit hardware; it stresses the classical share of the system as well. Each of the measurements of the qubits outdated for monitoring the system needs to be processed to detect and elaborate any errors. We are going to need roughly 100 logical qubits to enact a few of the greatest fascinating calculations, meaning monitoring thousands of hardware qubits. Doing more sophisticated calculations may mean thousands of logical qubits.
That error-correction data (termed syndrome data in the self-discipline) needs to be read between each operation, which makes for a lot of data. “At scale, we’re talking a hundred terabytes per 2d,” said Brierley. “At a million physical qubits, we’ll be processing about a hundred terabytes per 2d, which is Netflix global streaming.”
It also has to be processed in real time, in any other case computations will glean held up waiting for error correction to happen. To avoid that, errors have to be detected in real time. For transmon-based qubits, syndrome data is generated roughly every microsecond, so real time means ending the processing of the data—presumably Terabytes of it—with a frequency of around a Megahertz. And Riverlane was based to make hardware that’s capable of handling it.
Handling the data
The system the company has developed is described in a paper that it has posted on the arXiv. Or now not it is designed to handle syndrome data after utterly different hardware has already transformed the analog signals into digital originate. This allows Riverlane’s hardware to sit down initiate air any low-temperature hardware that’s needed for some varieties of physical qubits.
That data is accelerate via an algorithm the paper phrases a “Collision Clustering decoder,” which handles the error detection. To demonstrate its effectiveness, they implement it based on a typical Area Programmable Gate Array from Xilinx, the place it occupies easiest about 5 p.c of the chip however can handle a logical qubit built from nearly 900 hardware qubits (simulated, in this case).
The company also demonstrated a personalized chip that handled an even larger logical qubit, whereas easiest occupying a miniature fraction of a square millimeter and spellbinding accurate 8 milliwatts of vitality.
Each of these variations are extremely specialized; they simply feed the error information for various parts of the system to act on. So, it is far a extremely targeted resolution. Then again or now not it is also fairly flexible in that it works with various error-correction codes. Critically, it also integrates with systems designed to manipulate a qubit based on very utterly different physics, including chilly atoms, trapped ions, and transmons.
“I assume early on it was a little bit of a puzzle,” Brierley said. “You may have obtained all these utterly different varieties of physics; how are we going to enact this?” It became out now not to be a major challenge.