QuamCore raises €9m for digital logic in the quantum age
Cette publication existe aussi en Français
QuamCore in Israel has emerged from stealth with €9m funding for a new type of digital logic for building quantum computers with 1m qubits.
Rather than using CMOS, QuamCom is using superconducting flux loops to provide digital logic with a new architecture that can implement a quantum computer with 1m qubits inside a single cryostat. Chief executive Alon Cohen talks to Nick Flaherty about the technology.
“We want to achieve a fully tolerant fault tolerant quantum computer with 1m qubits and the main bottleneck is scaling. So this is what we started two years ago to tackle,” said Cohen, as the company raises $9m to develop its architecture and a path to a quantum computer with 1m qubits that measures 55 x 52 x 28cm to fit into a single cryostat.
Cohen is previously the co-founder of the EyeC Radar Group at Mobileye and Intel where he was the chief architect and head of the algorithm group, and he holds over 40 patents in radar, communication, and signal processing.
The current challenge for scaling superconducting quantum computers is that cryostat. Around 5000qbits can fit into each one, so a 1m qubit machine would need 200 cryostats. This all needs interface electronics, which even in low power cryoCMOS would take 40kW of power.
“Instead we bring the controller inside the cryostat with digital communication to the outside world,” says Cohen. “We developed a controller with 300uW and a solution to 116uW and we can reduce that even further.”
This digital superconducting logic is based on Single Flux Quantum loops which are also used by D-Wave and SEEQC, along with a multiplexing protocol to manage all the signals and a new error correction scheme. The digital logic can work with any type if superconducting qubits, from the transmons used by IBM and Google to the Cat qubits from Alice and Bob or the silicon spin qubits from Quobly.
“The main point is that the controller uses superconducting elements and we did the digital logic for control rather than analog. Just like CMOS pmos and nmos, we are implementing current loops that hold a flux inside and you can create building blocks such as logic gates. This is working technology today.”
“But it’s not enough as it reduces the power by a factor of 1000 and you need a reduction of a billion for 1m qubits. So we changed the manufacturing process to be compatible with 10mK and this reduces the power further.”
Instead of transferring the data, it can be compressed the data in the digital domain to reduce the number of instruction used to control the qubits. “For example to implement an error correction it’s the same operations so you don’t need to send every instruction so you can implement an order of magnitude of compression.”
“We solved problem after problem and we have patents on all of them so we have full silicon design. We have not built this yet. Before we spend a lot of money on the fab and testing we want to see the entire design. It’s an architecture-first approach.”
“We are dealing with engineering. If something is not solid and proven we are not counting on it. We are not inventing new materials. We do process modification on existing process, we are just tuning them. The process is simpler than CMOS, it’s not semiconductors, it’s multilayer conductors deposited on silicon with Josephson junctions, capacitors and resistor resistors.”
“We do need specific materials but we are a fabless company and we are working with a partnership with a tier one fab company with their material on a process we are finalising now. We did a lot of work to be compatible with their material,” he said.
The company has developed a full 3D model of all the chips, the flip chip technology with the digital controllers connecting to the qubits (see above), and the funding from Viola Ventures will allow the architecture to be proven at a small scale before scaling up.
“We have a long way to go, but we think we can present a full prototype in less than two years in the lab. Then we can scale.” The company is initially focussing on transmon qubits as these are the most robust, says Cohen
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
