Physicists have implemented a quantum computing device on superconductors using a modular principle, placing four separate nodes on a flat substrate, the configuration of which can be changed arbitrarily. The scientists achieved an average fidelity of 96 percent for the results of two-qubit gates and a quality of 98.74 percent for the preparation of an entangled two-qubit state. The new method for creating quantum devices will be useful for the efficient scaling of systems of unlimited size. The results of the study were published in Physical Review X.
For quantum computing to be of practical use, scientists need devices containing a large number of qubits, each of which in turn must be connected to all the others. However, the implementation of a full-fledged device of such an architecture has not yet been presented: for example, in the case of superconductors, it is physically impossible to connect all the qubits to each other on one plane - the loss of coherence and the dephasing effect come into play.
At the same time, researchers do not abandon attempts to bypass physical limitations and use an approach similar to the organization of an Ethernet network - superconducting qubits are connected using a central routing element. To do this, physicists use a common resonator bus or connect computing modules with a multimode ring resonator. Such approaches suffer from other drawbacks - for example, a long switching time between nodes, which in order of magnitude approaches the coherence time of superconducting qubits.
American physicists led by Andrew Cleland from the University of Chicago have proposed a new modular architecture for a quantum processor, connecting each qubit of the system to a superconducting quantum interference device.
To do this, the scientists created a circuit of four superconducting qubits and four independently controlled nodes, which were galvanically connected to a central capacitor. This provided all-to-all qubit communication by selectively connecting pairs of qubits via appropriate switches controlled by dynamically adjusting the magnetic flux. The individual qubit modules, fabricated on sapphire substrates, were equipped with individual output contacts and placed on a common motherboard, which was cooled to 10 millikelvin.
Physicists demonstrated the functionality and protocol of the switches using a quantum CZ gate as an example. The test results showed that the average execution time of the gate was approximately 46 nanoseconds with a fidelity of 96 percent with an error of about 0.08 percent. The scientists then conducted numerical simulations and found that the theoretical limit of the accuracy of two-qubit gates on their device is close to 99 percent. After that, the researchers generated entangled Greenberger-Horn-Zeilinger states involving first two, then three and four qubits. The final degree of state coincidence for the first case was 98.74 percent, for the second - 88.15 percent, and for the third already 75.18 percent. The physicists associated this decrease in the result with the decoherence of individual qubits during their idle periods, while the algorithm worked with other elements of the circuit.
The authors of the work noted that the connection method they proposed could be improved by increasing the coherence time of the qubits and could subsequently be used to connect more than four modules simultaneously into star-like structures.
When it comes to connecting ion devices into a single quantum computer, scientists use slightly different approaches, mostly related to the physical movement of ions. For example, we have already written about how physicists have learned to transfer ions between chips at record speed.