Superconducting qubits are microwave circuits containing Josephson junctions that behave as controllable artificial atoms, and they are the most engineering-mature solid-state route to a quantum processor, but the 2026 literature is still dominated by the discovery of new decoherence and readout-failure mechanisms rather than by evidence of a solved path to scale.
Summary
A superconducting qubit is an electrical circuit, cooled to millikelvin temperatures, in which one or more Josephson junctions supply the nonlinearity that turns a harmonic microwave resonator into an anharmonic system with addressable, well-separated energy levels. Because the level spacing is unequal, a microwave pulse can drive one transition without exciting the rest, so the lowest two levels act as a qubit. Control and measurement are done with microwave cavities under the framework of circuit quantum electrodynamics, which emerged in the early 2000s and made these circuits leading candidates for scalable quantum devices; the two dominant designs today are the transmon and the fluxonium. Readout is normally dispersive: a probe tone on a coupled resonator picks up a qubit-state-dependent frequency shift, ideally a projective, quantum non-demolition measurement.
The parameters that decide the technology are coherence time, single- and two-qubit gate fidelity, readout fidelity and its non-destructiveness, qubit count with useful connectivity, and the wiring and power overhead per qubit. Each of these is currently set by a different physical obstacle. Coherence is attacked by two-level-system (TLS) defects in the amorphous materials and tunnel barriers of the device, by 1/f flux and charge noise, by residual microwave dissipation in the superconductor itself, and by ionising radiation that generates phonons and quasiparticles. Readout is attacked by measurement-induced state transitions, where a strong probe drive pushes the qubit out of the computational subspace. Scaling is attacked by input/output: conventional architectures need independent microwave cables per qubit, and dense cable arrays impose prohibitive footprint, thermal load and cost constraints at hundreds of qubits.
Most of the current research effort splits into two directions. The first is materials and error physics: an empirical scaling relation between microwave dissipation and superfluid density across amorphous films, resonators, 3D cavities and transmons points to an intrinsic bulk loss channel from nonequilibrium quasiparticles, independent of surface dielectric loss, and sets a material-determined coherence limit. The second is circuit design: intrinsically protected qubits such as the cos(2φ) circuit, which allows only paired Cooper-pair tunnelling and so suppresses charge-induced errors, plus more speculative proposals using Andreev-spin hybrids, altermagnetic Josephson junctions and fractional vortices in 0-π junctions qubit states based on fractional vortices in 0 pi josephson.
Alongside computation, the same circuits are used as quantum simulators and as hybrid platforms, coupling qubits to mechanical resonators, acting as single-photon switching elements, and hosting engineered many-body physics such as quantum solitons in capacitively coupled transmon arrays.
Evidence base
Open questions
- Does the superfluid-density dissipation bound represent a genuine floor for transmon coherence, or can material or geometry choices circumvent the trapped-quasiparticle channel it attributes the loss to?
- How common are TLS defects that couple to two or more qubits at once, and do the resulting correlated, non-Markovian error patterns break the independence assumptions of standard error-correction thresholds?
- Can radon plate-out during fabrication and packaging be controlled cheaply enough that its area-scaling alpha background does not set an upper limit on usable die size?
- Does the 0.19% gate-fidelity penalty of optical I/O hold as the number of multiplexed channels grows from two towards hundreds, and what is the net cryogenic thermal budget compared with coaxial wiring?
- Do protected designs such as cos(2φ) ever escape the residual first-harmonic 1/f flux-noise limit measured at the flux symmetry point, or does each new protection mechanism simply expose a different dominant channel?
Assessment drafted 2026-08-31 from up to 18 KB sources using the technology-scorecard framework; scores are a draft read pending review.