12
$\begingroup$

Transmon and Xmon qubits are two types of superconducting charge qubits that seem to be often used in superconducting quantum devices. However, I wasn't able to easily find direct comparisons between them. The Xmon architecture seems (1304.2322) to have been introduced by Martinis' group, as an alternative to the transmon qubit, so I would expect the former architecture to be better in at least some respects. On the other hand, it also seems (cond-mat/0703002 and 0712.3581 seem to be the relevant references) that the devices from IBM use transmon qubits.

What are the main differences between the two, from a practical point of view (in other words, when and why does one prefer one over the other)?

$\endgroup$
10
$\begingroup$

The transmon is a Josephson junction and capacitor in parallel. Originally, transmons were differential circuits, i.e. two transmons on the same chip were not galvanically connected in any way. In other words, transmons didn't share a ground reference. Furthermore, in the early days, transmons were almost always embedded into the middle of a harmonic resonator. The resonator, often referred to as a "bus resonator", was used to couple multiple qubits together, i.e. qubits embedded in the same resonator could couple to each other.

The important differences with the xmon were that

  1. The xmon was grounded. Each xmon on a chip connects to a common ground plane with a nominally fixed voltage.

  2. The xmon was not embedded into a resonator. Instead of coupling through a resonator, each xmon couples through a direct capacitance to each of its neighbors.

Nowadays, several research groups build qubits without the bus resonator and call them "transmons".


Much more could be written. If someone leaves a comment asking for more details on any particular aspect of the difference between transmon and xmon, I will write more.

History of the name

Rob Schoelkopf told me the story of where the name "transmon" came from while we were at the Les Houches summer school on "Quantum Machines". The charge qubit suffered from low frequency noisey charge fluctuations that lead to dephasing. To get around the problem, Professor Schoelkopf thought to shunt the junction with a bit of transmission line. The line would be a short circuit at dc, allowing low frequency charge to equalize, but it would be a high impedance at the qubit's resonance frequency allowing the resonance to remain. The combination of a transmission line with the junction plasmon mode lead to the name "transmon".

In the end, it turned out that a capacitor was simpler than a transmission line and served a purpose equivalent to the transmission line, so the qubit wound up being a capacitor in parallel with the junction. However, the name "transmon" had already stuck (or maybe "capmon" just didn't sound as good).

$\endgroup$
  • $\begingroup$ could you say a few words about the advantages of one with respect to the other? Is it just a matter that each is more suitable for some application? $\endgroup$ – glS Jul 27 '18 at 11:16
  • $\begingroup$ @glS As you can see in the edited answer, there may not really not a well-defined distinction between "xmon" and "transmon" any more, so it's hard to answer. $\endgroup$ – DanielSank Sep 8 '18 at 6:23
  • $\begingroup$ @DanielSank Would you be able to comment on the difference between a differential transmon and one that is not? The advantages that one might have over the other? That is, ignoring the xmon aspect of not coupling to other qubits via a bus, just the difference for the qubit itself in using an island (junction) coupled to a reservoir vs two coupled islands? $\endgroup$ – user129412 Jan 3 at 15:30
  • $\begingroup$ Because I seem to recall that the differential Cooper pair box aimed to be immune to correlated charge noise (but the transmon doesn't really care all that much about charge noise) and to get rid of the large reservoir in which non-equilibrium QPs are generated more often than on the island (but that seemed to be in vain as the islands are still poisoned, plus transmon islands are much bigger than CPB islands in general so that also does not seem all that relevant) $\endgroup$ – user129412 Jan 3 at 15:32
5
$\begingroup$

In one sense, the Xmon qubit is a transmon qubit, in that they both operate in the $E_J>>E_c$ regime of the CPB Hamiltonian and take advantage of the exponentially suppressed charge noise vs. polynomial decrease in anharmonicity effect discussed in (Koch, 2007). You could work out the dynamics of a superconducting qubit-resonator system without ever knowing whether the equations were describing an Xmon or a transmon, so functionally its hard to differentiate the Xmon.

On the other hand, there are a lot of important design differences introduced in the Xmon: The qubit is grounded (mentioned above), the qubit is no longer embedded in the resonator, its conveniently tunable, the lifetime is enhanced (although the ibmqx3 chip that IBM uses for its quantum experience has qubits with $T_1\approx40 \space \mu s$ which matches the original Xmon lifetime). Also, the Xmon's shape is a great match for a surface code architecture that requires a tight-packed grid of qubits.

Practically, there are a lot of other transmon designs that offer some of the same benefits of the Xmon. So "transmon vs. Xmon" isn't the general question to ask; just go with the design that's got the best lifetimes and maybe tunability.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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