4K = -269°C, 50mK = -273,1°C, 20mK = -273,13°C
Minus 273 degrees Celsius – colder than anywhere in the universe. Classical electronic systems are not designed for this extreme cold. However, this is supposed to change. Jülich researchers are working on microchips that function without interference in these frosty conditions. They are to control the computers of the future: quantum computers with millions of qubits that only work at such temperatures.
A shiny, red-gold structure hangs in a frame about 2.5 metres high, linking metal spirals and many curved metallic cables over four levels. The elaborate web in the laboratory of the 2nd Institute of Physics A at RWTH Aachen University is part of an experiment that could lead the way for the universal quantum computers of the future. Their computing power promises to outperform today’s supercomputers like a racing car outperforms a tractor.
Picture above: 4K = -269°C, 50mK = -273,1°C, 20mK = -273,13°C
The shimmering construction is a cooling device in a class of its own: a cryostat. By mixing two different types of the noble gas helium, it can cool to a temperature close to absolute zero, that is, to almost minus 273.15 degrees Celsius, which is colder than ever measured in the universe. This is how extreme most qubits – the sensitive computing units of quantum computers – need it to be. Only then do they run stably and almost error-free. The many electrical cables are bent to prevent them from tearing from the extreme cold. Together with the cooling coils and the frame, they give the cryostat its typical appearance of a quirky chandelier. “Photographers love the motif – it has almost become the symbol of a quantum computer,” says physicist René Otten from the JARA Institute for Quantum Information. This makes him smile, as “the quantum computer itself, the quantum processor, is only very small, sitting quite inconspicuously at the very bottom of the cryostat.”
Too many cables
This quantum processor is controlled via the numerous wires. This, however, is precisely where a problem lies that complicates the path to future quantum computers involving millions of qubits: to control and read out the qubits, researchers use conventional electronics, which is not designed for temperatures below minus 40 degrees Celsius. For this reason, it is outside the cryostat. This means that the electronic devices generate signals at room temperature, which then run through the cryostat via the wires and finally arrive at the cryogenically cooled quantum processor.
“This works with 50 qubits, which is what quantum computers from IBM and Google use today, for example, or even with 100 qubits, but not for quantum computers in ten years’ time or more, which are expected to have over a million qubits,” says Dr. Carsten Degenhardt from the Jülich Central Institute for Engineering, Electronics and Analytics (ZEA-2), convinced. “In a few years, we could be in the same position with quantum computers as were the operators of conventional computing systems in the late 1950s.” Back then, engineers were designing ever more complex circuits to increase the computing power of computers: solder joints and the wiring of the individual components to one another were increasing to such an extent that both the space required and the probability of connection errors rose enormously. The situation could turn out similarly with qubits: the more there are to be controlled, the more wires will be needed. This is not only a problem of space. The wires are prone to interference; the more wires, the more interference. Furthermore, the wires give off heat, which seeps into the cryostat – poison for the qubits, which need the cold.
In the 1950s, the tangle of cables was eliminated by the invention of the integrated circuit, the microchip: a semiconductor wafer on which the electronic components – transistors, resistors, capacitors – were mounted, along with their connections. Now, a microchip is again expected to fix this and make the electronic wires almost superfluous. The challenge this time: the chip has to work at temperatures of around minus 273 degrees Celsius and convert digital inputs into signals for the quantum computer just as reliably as at room temperature. However, such unexpectedly tolerant components are unpurchasable.
Frosty new territory
Five years ago, therefore, the experts at ZEA-2 began working on their own chips. First, it was necessary to clarify the special requirements that a quantum processor presents to such a chip. For this purpose, the ZEA-2 team contacted the physicists led by Prof. Hendrik Bluhm at the JARA Institute for Quantum Information. Bluhm is working on a “Made in Germany” semiconductor quantum processor in the QUASAR project. Close cooperation developed between the JARA researchers, especially René Otten, and ZEA-2. “Together we determined, for example, the level and increments of the voltages that the chip is supposed to generate, plus the maximum electrical power acceptable for it to consume to achieve this without heating up the qubits too much. These are the basics for a circuit diagram,” says Dr.-Ing. Patrick Vliex, who is responsible for designing the prototype at ZEA-2.
Up to this point, Vliex had largely moved along the familiar paths of room temperature electronics. With the next step, however, he entered unknown territory. In order to create the circuit diagram and convert it into concrete construction instructions for the microchip – that is, for the chip design – he needed precise information about the behaviour of the individual microelectronic components such as resistors and transistors. This information is available, albeit only down to minus 40 degrees Celsius. “It was completely unclear whether transistors, for example, would work at all near absolute zero, that is, at more than 230 degrees Celsius less. That’s why we first carried out tests to that effect,” says Carsten Degenhardt. Fortunately, the construction elements proved to be unexpectedly tolerant: they behaved differently at very low temperatures compared to room temperature, but still fulfilled their function. The scientists thus decided to continue working on the design of the chip and carry out simulations.
Simulations can be used, for example, to check the voltage signals that the designed chip produces on the basis of digital inputs. If something is amiss, the chip layout can be changed without incurring large costs before production starts. The ZEA-2 tested the new chip as well with the help of simulations. “We were aware, however, that these simulations reflect our cryochip to no more than a limited extent, as they only include the models and data available to a maximum of minus 40 degrees Celsius,” says Vliex. Still, the results of these simulations were similar enough to those of the cryogenic test measurements for them to proceed. Besides, in designing the chip, Patrick Vliex was able to take into account some of the possible deviations in the behaviour of the components at around minus 273 degrees, for example by providing additional transistors that can be switched on if necessary.
“It was completely unclear whether transistors, for example, would work at all near absolute zero.”
Dr. Carsten Degenhardt
Qubit meets chip
After about a year of development, the Jülich electrical engineers finally commissioned a chip factory to produce a real chip from the finished design. Then came the waiting: “About half a year later, the chip factory delivered several copies of the prototype, and with that, everything was ready for our first joint test run with the Aachen researchers,” Vliex says. In the meantime, the colleagues in Aachen had continued researching their semiconducting qubit. The qubit is generated in a semiconducting wafer consisting of several layers (see “an open race“). The wafer is mounted on a circuit board. René Otten also placed ZEA-2’s low-temperature chip on this circuit board.
In March 2022, the time had come: René Otten closed the heat-insulating cover of the cryostat and started the experiment in the Aachen physics building. The following measurements, which lasted several days, showed that the signals from the chip in the cryostat were driving the qubit chip as desired. “That worked perfectly. The next step will be to check whether the chip also reads out signals as desired. In the long term, we want to establish a process with which we can design cryoelectronic chips in the same way as we already do room temperature microchips today. This would also include adapting the existing models for simulations,” says Degenhardt. Together with partners from research and industry, this is already to be implemented in another quantum computer project: in the QSolid project launched in 2022.
The shiny, red-gold wire web for the electronic control system could therefore soon be largely dispensed with. The cryostat would then probably be less photogenic. On the other hand, the entry into a realm of technology that is expected to enable a million-qubit quantum computer in ten or twenty years will then be made.
Photos: Forschungszentrum Jülich/Ralf-Uwe Limbach, René Otten; Illustration: Andrzej Koston