Even the best quantum processors are useless without the right control system. This is precisely what Quantum Machines specializes in, aiming to make a significant push towards scaling quantum processors to hundreds of thousands of qubits.
At the heart of quantum computers are quantum processors (QPUs), where qubits are generated. In the case of superconducting qubits, these must be cooled to a few thousandths of a degree above absolute zero. Indeed, many of us are familiar with images of such cooling systems, which resemble a gilded chandelier.
Qubits must be controlled and manipulated in order to program and operate quantum systems. This is achieved through quantum gates, which can only be set up and run calculations using control signals. The gates are usually generated via microwaves which interact with the qubits. Moving the control signals through the cooling units to the QPU requires many wires. While this can be achieved to some level using conventional electronics, the control system must be tailored for quantum to make the control and readout process as effective as possible.
Anyone with expertise in measurement technology, especially HF measurement technology, has a considerable advantage here. This is why test and measurement technology companies already have a niche in this space. However, applying the same HF measurement methods as used in traditional applications is not necessarily beneficial.
To effectively control and read QPUs, there's a need for a new approach. That's why startups have emerged in the field of quantum computing that specialize in controlling and implementing qubits. One such company is Quantum Machines, a startup founded in Israel in 2018 that is developing the hardware and software infrastructure required to run quantum computers. This includes everything from the control electronics at the gate level up to the qubit in the quantum processor.
"It all starts with the arbitrary waveform generator, but it has to be significantly modified in order to enable quantum computers to actually compute," says Jörn Höpfner, Scientific Business Development Manager at Quantum Machines.
"The actual quantum computers run in the QPUs. But to turn the quanta into qubits, they must be controlled, manipulated, and read. We do that with our control platform," explains Jörn Höpfner. "But that's difficult because the quantum states into which the qubits must be placed are extremely sensitive." So sensitive that for a long time, it was unclear whether for every "readout process," an outside interference would destroy the entanglement, making the practical use of quantum computers unattainable.
Now, we at least know that it works. However, there are many different ways to implement qubits. The most common today are probably superconducting qubits, ion traps, and quantum dots in semiconductors. Still, there are many other applications like optical qubits, NV centers, neutral atoms, carbon nanotubes, and topological qubits. They all have advantages and disadvantages. It's unclear who will get there first or whether there will be qubits based on different technologies for different applications. "The nice thing is that it doesn't matter for us - we can work with all quantum technologies," says Jörn Höpfner.
But back to the arbitrary waveform generator (AWG). Measurement device manufacturers typically build devices for operating conditions where one or two channels are usually sufficient. However, many channels are required in quantum computers. "That's why we build control systems with up to ten channels and integrate the electronic evaluation unit. That reduces the per-channel cost," says Höpfner.
There are also additional differences. With conventional AWGs, the signals are stored digitally to memory. However, with the Quantum Machines solution, pulses are generated via an FPGA. Generated pulses ensure that the logical qubits ultimately execute the desired algorithms. "Generating pulses is our core expertise. They map the logic gates in the QPU. Our devices generate the signals in real time and can be influenced by the values interpreted by the readout unit. That's not possible with classical AWGs." Of course, all of this has to happen very quickly: The response time is just 200ns which is within the coherence time of the qubit."
"The FPGAs are pre-programmed, "which also differentiates us from other manufacturers," says Höpfner. Quantum Machines relieves users of the time-consuming programming work: "Something that would previously take scientists several months to accomplish. Now, they can start running experiments within a few days." The company uses FPGAs with 18 cores, allowing them to run 18 measurements in parallel.
“With our Quantum Orchestration Platform, not only can users leverage the full potential of their QPUs, but they can significantly accelerate their development processes. The path to scaling to the thousands of qubits is now wide open.”Jörn Höpfner, Quantum Machines
Quantum Machines has integrated the hardware and software into its Quantum Orchestration Platform (QOP). That includes the QUA programming language and the OPX+. According to Höpfner, with QUA, Quantum Machines is the first company to combine quantum processing at the pulse level with classical processing. Quantum Machines developed the OPX+ to adapt programs to the hardware on which a specific quantum computer is based via the Pulse Processing Unit's assembly language.
This allows complex quantum algorithms to be executed, including error correction and multi-qubit calibration, among other things. Users can also seamlessly program using QUA, a python-based language; the FPGAs are already pre-programmed by Quantum Machines. "This not only allows users to leverage the full potential of their QPUs, but they can significantly speed up their development processes and open the path to scaling to thousands of qubits," says Höpfner. He's convinced this approach clearly differentiates Quantum Machines from the competition.
Quantum Machines have now even gone one step further on the path from gate-level to qubit. In order to access the qubits in the quantum processor, appropriate electronics are essential, including low-noise voltage generators and electromechanics like cables and connectors through the cryostat.
That's why in March of this year, Quantum Machines bought Danish company QDevil, which specializes in those exact components and now forms the quantum electronics division of Quantum Machines. It's the perfect complement to what Quantum Machines has developed so far and significantly strengthens its Quantum Orchestration Platform (QOP).
But despite all the excitement around quantum computing and the advantages of his own system, he believes there's still a lot of improvement to be made: "A lot of people still think about the classical AWGs first; the fact that better options exist isn't common knowledge yet." And, of course, there's a need for onsite training and support: "Our customer success team consists of physicists, most of whom have doctorate degrees. They attend the installations to make sure everything runs perfectly, demonstrating how the platform integration works in the customer's own laboratory, including programming and running functional experiments."
The feedback is encouraging. The Quantum Orchestration Platform is already being used by a large number of users. Accordingly, the company has set no small goal for itself: "We want to dominate the quantum world!"
NOTE: This article was originally published in the German magazine Markt & Technik. The original article can be found here on page 17.
Thanks for subscribing!