If you open up almost any superconducting quantum computer built in the last decade, from IBM's processors to Google's, you will find the same basic ingredient: a qubit called the transmon. It won the field by being simple, forgiving of stray electric charge, and easy enough to fabricate in large grids. But the transmon has a persistent weakness, and a growing number of researchers think a stranger cousin called the fluxonium is the way past it.
Why the transmon leaves something on the table
A transmon stores quantum information in the two lowest energy states of an electrical circuit. The trouble is that those two states sit uncomfortably close to a third, higher state. The gap between the first pair and the next rung of the ladder, a quantity engineers call anharmonicity, is small. That small gap forces control pulses to be gentle and slow so they do not accidentally kick the qubit out of its computational space, a failure mode known as leakage. It also puts a floor on how tightly qubits can be packed on the frequency dial before they start interfering.
The fluxonium takes a different approach to the same circuit problem. Invented around 2009 in Vladimir Manucharyan's lab, it adds a giant inductance built from a long chain of Josephson junctions wired in series, sometimes a hundred or more. That so-called superinductance reshapes the energy landscape dramatically. Instead of a shallow ladder, the fluxonium's lowest two states can be pushed far apart from the rest, giving it enormous anharmonicity. Leakage becomes far less of a worry, and control pulses can be more aggressive without spilling the qubit into unwanted states.
Longer memory, cleaner gates
The payoff shows up in two numbers researchers care about most. The first is coherence time, how long a qubit remembers its state before noise scrambles it. Fluxonium devices have posted relaxation and coherence times reaching hundreds of microseconds and, in some experiments, well past a millisecond, comfortably longer than typical transmons. Part of the reason is that the fluxonium's operating transition sits at a much lower frequency, often below one gigahertz, which changes how the circuit couples to the noisy environment around it.
The second number is gate fidelity, the chance that an operation does what it was supposed to. Academic groups have demonstrated single-qubit gates with error rates in the range of one in ten thousand and two-qubit gates crossing the 99.9 percent mark. Those are the kinds of figures error-correction schemes are hungry for, because every fraction of a percent of extra fidelity translates into fewer physical qubits needed to protect one logical qubit.
The catch nobody ignores
Fluxonium is not a free lunch. That long chain of junctions is harder to fabricate reliably than a transmon's single junction, and any defect along the chain can spoil the device. The low operating frequency, so helpful for coherence, creates its own headache: at the millikelvin temperatures inside a dilution refrigerator, thermal energy can still populate a low-lying state, meaning the qubit does not always start cleanly in its ground state. Resetting and reading out a low-frequency qubit calls for control electronics and readout tricks that differ from the transmon playbook the industry has spent years perfecting.
Control complexity is the broader concern. A whole ecosystem of microwave hardware, calibration routines, and packaging has grown up around transmons. Switching to fluxonium means rebuilding parts of that stack, and hardware teams do not abandon a working recipe lightly.
Who is betting on it
The interest is no longer confined to physics departments. Atlantic Quantum, a startup spun out of MIT, has made fluxonium the centerpiece of its architecture, pairing it with cryogenic control electronics meant to sit inside the fridge alongside the chip. Manucharyan's group, now at EPFL, continues to push the coherence and gate records. Larger players have run fluxonium experiments too, treating it as a serious candidate rather than a curiosity. The appeal is straightforward: if error correction is going to demand millions of physical qubits, every improvement at the single-qubit level compounds across the whole machine.
None of this guarantees the fluxonium unseats the transmon. The transmon's head start in manufacturing maturity is enormous, and quantum computing has a habit of rewarding whichever approach scales cleanest rather than whichever looks best on paper. But the fluxonium is a reminder that the qubit at the heart of superconducting quantum computing is not settled science. The circuit that dominates today was itself a refinement of earlier designs, and the search for something quieter and more forgiving is still very much underway.
