Where are the physical gates in the Google processor?

Google's article Quantum supremacy using a programmable superconducting processor states that the processor "53 qubits, 1,113 single-qubit gates, 430 two-qubit gates, and a measurement on each qubit, for which we predict a total fidelity of 0.2%."

Where are the gates physically located? In the diagram, the qubits are the crosses and the qubits are the couplers.

• So the gates exist over time, in the 4th dimension?
– vy32
Jan 21 '20 at 14:39
• @Mark, can you make this an answer so that it can be improved and accepted?
– vy32
Jan 21 '20 at 14:39
• Yes, the quantum gates are sort of work in time dimension; quantum circuits are very different from classical circuits here. Jan 21 '20 at 14:50

In relating quantum computing to classical computing there may be a small conceptual hurdle that needs to be overcome. Although a classical $$\mathsf{NAND}$$ gate may be implemented in hardware (say CMOS with a set of N- and P-type transistors), the idea of a quantum gate such as a $$\mathsf{CNOT}$$ or an $$\mathsf{H}$$ gate used in quantum computing most often is a little more subtle.

Conventionally a quantum circuit including a number of quantum gates acting on qubits usually is meant to read with time progressing from left to right. Most quantum gates are not so much physical as in classical computing on CMOS, but are rather well-timed laser, or in the case of Sycamore, microwave pulses on the qubits/couplers. These pulses may excite a qubit from $$\vert 0\rangle$$ to $$\vert 1\rangle$$ or act as one qubit controlling another qubit, etc. In this sense, quantum gates are much more like opcodes than like a set of transistors arranged in parallel/series in the classical sense.

Accordingly the "1,113 single-qubit gates, 430 two-qubit gates" quoted in the Sycamore paper corresponds to these microwave pulses acting on the transmon qubits/couplers. My understanding is that in the case of Sycamore, different shapes of the microwave pulses generated by an automatic waveform generator acts on the qubits in different ways, thus realizing different gates at different times. In the classical setting we might have that assembly language opcodes are transcribed to machine code; in the quantum world we might have that quantum gates are transcribed to the shape of these pulses.

I believe that quantum gates acting on transmon qubits correspond to microwave pulses, and quantum gates acting on trapped ions correspond to similarly well-timed laser pulses.

Note that photons move at the speed of light, and I believe photonic quantum gates as proposed for example in Boson Sampling experiments include physical mirrors/beam splitters/etc. Thus, I think that photonic gates may be a bit more similar to classical gates, and may actually be implemented in hardware.

• So when Google performs a computation with this system, do they actually perform the same computation many times over, and then do some kind of post-processing to find the actual "result?"
– vy32
Jan 21 '20 at 16:34