Three Frosty Innovations for Better Quantum Computers

Putting these in the cryogenic freezer could make quantum computers more powerful and compact

3 min read
Chalmers University of Technology is working on cryo-LNAs. Their circuit uses high-electron-mobility transistors (HEMTs).
A close-up picture of the wire bonded InP HEMT device in the three-stage 4-8 GHz LNA for quantum computing.
Photo: Chalmers University of Technology

For most quantum computers, heat is the enemy. Heat creates error in the qubits that make a quantum computer tick, scuttling the operations the computer is carrying out. So quantum computers need to be kept very cold, just a tad above absolute zero.

“But to operate a computer, you need some interface with the non-quantum world,” says Jan Cranickx, a research scientist at imec. Today, that means a lot of bulky backend electronics that sit at room temperature. To make better quantum computers, scientists and engineers are looking to bring more of those electronics into the dilution refrigerator that houses the qubits themselves.

At December’s IEEE International Electron Devices Meeting (IEDM), researchers from than a half dozen companies and universities presented new ways to run circuits at cryogenic temperatures. Here are three such efforts: 

Google’s cryogenic control circuit could start shrinking quantum computers

Google\u2019s first generation cryogenic-CMOS single-qubit controller (center and zoomed on the right) packaged and ready to be deployed inside our cryostat. The controller measures 1mm by 1.6mm. Google’s cryo-CMOS integrated circuit, ready to control a single qubit. Photo: Google

At Google, researchers have developed a cryogenic integrated circuit for controlling the qubits, connecting them with other electronics. The Google team actually first unveiled their work back in 2019, but they’re continuing to scale up the technology, with an eye for building larger quantum computers.

This cryo-CMOS circuit isn’t much different from its room-temperature counterparts, says Joseph Bardin, a research scientist with Google Quantum AI and a professor at the University of Massachusetts, Amherst. But designing it isn’t so straightforward. Existing simulations and models of components aren’t tailored for cryogenic operation. Much of the researchers’ challenge comes in adapting those models for cold temperatures.

Google’s device operates at 4 kelvins inside the refrigerator, just slightly warmer than the qubits that are about 50 centimeters away. That could drastically shrink what are now room-sized racks of electronics. Bardin claims that their cryo-IC approach “could also eventually bring the cost of the control electronics way down.” Efficiently controlling quantum computers, he says, is crucial as they reach 100 qubits or more.

Cryogenic low-noise amplifiers make reading qubits easier

A key part of a quantum computer are the electronics to read out the qubits. On their own, those qubits emit weak RF signals. Enter the low-noise amplifier (LNA), which can boost those signals and make the qubits far easier to read. It’s not just quantum computers that benefit from cryogenic LNAs; radio telescopes and deep-space communications networks use them, too.

Researchers at Chalmers University of Technology in Gothenburg, Sweden, are among those trying to make cryo-LNAs. Their circuit uses high-electron-mobility transistors (HEMTs), which are especially useful for rapidly switching and amplifying current. The Chalmers researchers use transistors made from indium phosphide (InP), a familiar material for LNAs, though gallium arsenide is more common commercially. Jan Grahn, a professor at Chalmers University of Technology, states that InP HEMTs are ideal for the deep freeze, because the material does an even better job of conducting electrons at low temperatures than at room temperature.

Researchers have tinkered with InP HEMTs in LNAs for some time, but the Chalmers group are pushing their circuits to run at lower temperatures and to use less power than ever. Their devices operate as low as 4 kelvins, a temperature which makes them at home in the upper reaches of a quantum computer’s dilution refrigerator.

imec researchers are pruning those cables

Any image of a quantum computer is dominated by the byzantine cabling. Those cables connect the qubits to their control electronics, reading out of the states of the qubits and feeding back inputs. Some of those cables can be weeded out by an RF multiplexer (RF MUX), a circuit which can control the signals to and from multiple qubits. And researchers at imec have developed an RF MUX that can join the qubits in the fridge.

imec's RF MUX Photo: imec

Unlike many experimental cryogenic circuits, which work at 4 kelvins, imec’s RF MUX can operate down to millikelvins. Jan Cranickx says that getting an RF MUX to work that temperature meant entering a world where the researchers and device physicists had no models to work from. He describes fabricating the device as a process of “trial and error,” of cooling components down to millikelvins and seeing how well they still work. “It’s totally unknown territory,” he says. “Nobody’s ever done that.”

This circuit sits right next to the qubits, deep in the cold heart of the dilution refrigerator. Further up and away, researchers can connect other devices, such as LNAs, and other control circuits. This setup could make it less necessary for each individual qubit to have its own complex readout circuit, and make it much easier to build complex quantum computers with much larger numbers of qubits—perhaps even thousands.

The Conversation (0)

The Future of Deep Learning Is Photonic

Computing with light could slash the energy needs of neural networks

10 min read

This computer rendering depicts the pattern on a photonic chip that the author and his colleagues have devised for performing neural-network calculations using light.

Alexander Sludds
DarkBlue1

Think of the many tasks to which computers are being applied that in the not-so-distant past required human intuition. Computers routinely identify objects in images, transcribe speech, translate between languages, diagnose medical conditions, play complex games, and drive cars.

The technique that has empowered these stunning developments is called deep learning, a term that refers to mathematical models known as artificial neural networks. Deep learning is a subfield of machine learning, a branch of computer science based on fitting complex models to data.

Keep Reading ↓ Show less