Quantum computers are becoming more and more powerful, and the best systems can already keep up with the fastest supercomputers for special calculations. Some of these machines, which use the bizarre laws of quantum physics for their calculations, have now found their way into data centers. However, despite these great advances, we are still a long way from developing a universally applicable and fault-tolerant quantum computer that can solve all kinds of tasks reliably and with sufficiently low error rates.
The quantum physical information units, the quantum bits, which are mainly realized through superconducting microwave resonators and stored charged and neutral atoms, are still quite fragile and susceptible to failure. Thermal radiation and unwanted stray electric fields can destroy the states of the quantum bits, which inevitably leads to calculation errors. Scalability also poses difficulties: the performance of quantum processors cannot be easily increased by simply increasing the number of quantum bits. The more qubits there are, the more difficult it becomes to shield and control each one from disruptive influences.
Many research groups have therefore long seen quantum computers made of silicon as an alternative, especially to the superconducting systems favored by IBM and Google. This also includes developers from well-known chip manufacturers. Extremely compact and stable quantum bits can be created in the semiconductor material, which can be manipulated and programmed with electric fields. Another plus point: Silicon-based quantum bits can be produced using common semiconductor technologies that are also used to produce classic processors, and they can be housed on a silicon chip. This makes it possible to scale quantum systems in a relatively simple way.
An American research group from HRL Laboratories in Malibu, California, has now been able to overcome another hurdle that previously stood in the way of the practical suitability of semiconductor quantum systems. She has developed a silicon quantum processor made of 18 quantum bits that calculates as reliably as the superconducting quantum bits of IBM or Google systems already do. While three years ago the typical error rate of a two-qubit system made of silicon was around four percent, the researchers led by Jacob Blumoff were able to reduce the error rate by more than a factor of ten.
The American scientists use the spins of individual electrons, which are isolated in tiny silicon structures, so-called quantum dots, like a kind of cage, as quantum bits. Spins are ideal quantum bits: they can “point” up and down at the same time or exist as a superposition of both states. The spins can be controlled using electrical pulses alone.
Blumoff's researchers have managed to eliminate a main source that disrupts quantum calculations and causes major errors: the heat that is normally transferred to a deep-cooled quantum chip via copper cables from the external control electronics. As a result, the quantum bits are heated despite thermal shielding. Electron spins suddenly flip and coherence times, which are crucial for computing to be as error-free as possible, shorten.
Blumoff's researchers have now gotten the problem under control with a new computer architecture. Instead of installing the control electronics and wiring outside the cryostat, as is usual, the control processor and the quantum chip are located together inside the cooling device, which is kept at freezing temperatures with liquid helium. Both components - their temperatures differ by a few Kelvin - are connected to each other via a superconducting ribbon cable made of niobium wires. This has the advantage that all control signals are transmitted to the quantum bits without interference, but no heat can flow from the control processor to the colder quantum dots.
The quantum computer works largely autonomously. Only a keyboard and a monitor are required to address the control electronics, which in turn programs the quantum processor. Test runs with special quantum algorithms showed an error rate of only 0.2 percent, write Blumoff's researchers in the journal Nature. Long underestimated as a platform, the silicon-based quantum processor can keep up with the large quantum systems from Google and Co., at least when it comes to the reliability of the calculations.
