Short answer: Transmon qubits lose their quantum state far sooner than the bulk properties of their materials would allow, and researchers trace much of that loss to surfaces, interfaces and contaminants. Replacing niobium with tantalum in the device was reported in 2021 to give lifetimes and dynamical-decoupling coherence times above 0.3 milliseconds. Later measurements show that two-level-system defects, on surfaces and in the bulk, still dominate the loss. Materials work is now a main route to better superconducting qubits.
Why materials matter
A 2021 Nature Communications paper says building large, useful systems from transmon qubits will need significant improvements in relaxation and coherence times, which are orders of magnitude shorter than the limits set by bulk properties of the constituent materials. The authors infer that relaxation likely originates from uncontrolled surfaces, interfaces and contaminants. Earlier efforts to improve lifetimes focused mainly on designs that minimise the contribution of surfaces, and significant improvements in two-dimensional transmon lifetimes had remained elusive for several years.
The tantalum result
In that paper the authors fabricate two-dimensional transmon qubits in which tantalum replaces niobium, and report lifetimes and coherence times with dynamical decoupling exceeding 0.3 milliseconds. They observe increased lifetimes in seventeen devices, which they say indicates the improvement is robust and paves the way for higher gate fidelities in multi-qubit processors. A separate 2021 paper in npj Quantum Information compares transmons made from different superconducting materials, and reports that tantalum stands out systematically. It identifies material interfaces, namely metal-air, metal-substrate and metal-metal, as important loss channels, and says tantalum films with a dry etching process look promising for fabricating long-lived multi-qubit circuits. Dry etching is described as suitable for large-scale quantum circuits. That paper also states that the major source of decoherence comes from two-level systems.
What limits tantalum devices
A 2023 Physical Review X paper measures superconducting tantalum resonators to separate the origins of loss in state-of-the-art tantalum devices. It describes tantalum-based qubits as having record lifetimes exceeding 0.3 ms and says overall gate fidelities are limited by coherence times set by microwave dielectric loss. By studying loss against temperature, microwave photon number and device geometry, the authors find the losses are dominated by several types of saturable two-level systems, with evidence that both surface and bulk two-level systems contribute. They show that surface two-level systems can be altered with chemical processing, and extract the linear absorption from different surface sources under four surface conditions. At single-photon powers, the relevant regime for qubits, they measure resonators with internal quality factors from 5 to 15 million, comparable to the best reported qubits. In these devices the surface and bulk contributions are comparable, so they conclude that systematic improvements in materials on both fronts are needed to improve qubit coherence further.
What this does not tell you
These papers report results from particular fabrication processes and devices, with lifetimes of a few tenths of a millisecond at best in the qubit results cited. They do not show that a tantalum processor at scale will keep these figures. Resonator quality factors are a proxy for qubit coherence and not the same thing.
FAQ
Why do transmon qubits lose coherence?
Much of the loss is traced to surfaces, interfaces, contaminants and two-level-system defects.
What did tantalum change?
Replacing niobium with tantalum gave lifetimes and coherence times above 0.3 ms in the 2021 Nature Communications paper.
Is the problem solved?
No. The 2023 paper finds surface and bulk defects still contribute comparably.
Can chemistry help?
The 2023 paper shows surface two-level systems can be changed with chemical processing.
Does this show a scaled processor?
No. The papers cover individual devices and resonators.
All field notes