We previously reported on a 72 qubit superconducting chip named Bristlecone that Google first described at the APS conference in March 2018. But in recent technical conferences, Google has been talking about a new 54 qubit chip codenamed Sycamore that may actually have more quantum computing performance than Bristlecone.
Sycamore appears to contain several improvements over the previous design. Google mentions that the 2-qubit gates are superfast (10-20ns.). They also described a new parameterized 2-qubit gate called the fSim gate at APS 2019 that may offer higher overall performance than the standard CNOT or CZ gates for certain applications such as error correction, quantum supremacy, and pre-error correction applications including quantum chemistry and quantum approximate optimization algorithms. Another improvement is the addition of tunable couplers. The previous design had a limitation that simultaneous gate operations on neighboring qubits might interfere with each other. To avoid this, it required staggering the gate operations so that neighboring interactions didn’t occur at the same time. But this slowed down performance and the addition of the tunable couplers eliminated this. As can be seen in the diagram below, this can double the amount of gate operations that can happen at one time. (A video from TQC 2019 with additional explanation about this can be found here.)
Google has taken an interesting engineering approach to develop their technology. Rather than just selecting some static parameters like qubit count, T1 and T2 times, etc. and optimizing the design around those, they developed an application benchmark called Quantum Supremacy which they first described in 2016. They have been working every since to create a technology which demonstrates that their quantum computer can beat the best classical computer on this particular benchmark.
We do expect Google to officially announce successful completion of this Quantum Supremacy experiment soon. But this reiterates the lesson that many have said before. A raw qubit count should not be the only measure of how powerful a quantum computer is. Other factors including qubit quality, noise characteristics, qubit connectivity, native gates supported and other will play a big factor in the application level performance of the machine.
September 20, 2019