5
$\begingroup$

Let's assume I have some quantum program I tested and tuned on a simulator and that now I want to see how it actually performs on real quantum hardware. I test it on IBM Superconducting devices and then I want to move to a different vendor offering a different technology, say trapped ions available on Microsoft Azure or Amazon Braket.

Does it make sense to make a comparison ? What differences should I pay attention to ? What should be the purpose of this experiment ?

$\endgroup$
0

1 Answer 1

6
$\begingroup$

At this stage it most certainly matters which technology you use, and in particular which circuit you wish to execute.

Although all of your mentioned technologies implement universal gates, the topology of superconducting devices is quite different and potentially more limiting from those of trapped ions.

For example it’s my understanding that trapped ions can have (almost) all-to-all connectivity, meaning any qubit within a line can act on (almost) any other qubit on the line. External laser pulses target the two qubits within the line.

But superconducting qubits have a particular topology where one qubit can only act on some nearest neighbor qubits through particularly fabricated microwave couplers. This topology is referenced in IBM's commentary about a "heavy hexagonal lattice".

Accordingly, superconducting devices may use a lot of SWAP gates to move qubits around; this can impact overall gate fidelity.

If your circuit has qubit [1] acting as a control for the negation of qubit [23], for example, then a trapped-ion system may be able to have qubit [1] act directly on qubit [23] with a CNOT gate. But, a superconducting system may require that qubit [23] SWAPS with qubit [22], which SWAPS with qubit [21]... before qubit [3] is SWAPPED with qubit [2], and then qubits [1] and [2] can finally do the CNOT gate.

In such a dramatic example all of these SWAP gates really impact the overall fidelity of the superconducting system. Of course you could always attempt to redesign a circuit so that qubits [1] and [23] don't have to be doing a CNOT gate therebetween; that is the art of proper transpiling. The Qiskit team has posted some videos about this layout and transpiling for their superconducting qubit architecture.

$\endgroup$
2
  • $\begingroup$ Your claim that "the native gate fidelity of superconducting qubits are better than those of trapped ions" is questionable at best; AFAIK the best two qubit gate fidelities have been demonstrated with trapped ions (e.g., journals.aps.org/prl/abstract/10.1103/PhysRevLett.117.060505) $\endgroup$
    – Rococo
    Dec 5, 2022 at 1:21
  • 1
    $\begingroup$ Thanks; edited. $\endgroup$ Dec 5, 2022 at 1:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.