Short answer: Flip-chip assembly puts a qubit chip face to face with a control or wiring chip, which adds routes for wiring as processors grow. A 2022 Quantum Science and Technology paper integrated transmon qubits into bump-bonded flip-chip modules, reporting coherence times exceeding 90 microseconds, single-qubit fidelities above 99.9% and two-qubit fidelities above 98.6%. A 2021 Applied Physics Letters paper shows qubits on separate dies coupled without a galvanic connection. These are laboratory demonstrations.
Why flip-chip
The 2021 abstract says flip-chip architectures have recently enabled significant scaling-up of multi-qubit circuits and have been used to assemble hybrid systems that combine different substrates, for example in quantum acoustics experiments. The standard method uses superconducting galvanic connections between two substrates, typically with indium wafer-bonding systems. The authors say these give highly reliable, temperature-cyclable assemblies but are expensive, somewhat inflexible in design, and require robust substrates that can withstand the large compressive forces needed to cold-weld the indium bonds.
2022: bump-bonded modules
The 2022 paper integrates single and coupled superconducting transmon qubits into flip-chip modules. Each module has two chips, one quantum chip and one control chip, bump-bonded together. The authors report time-averaged coherence times exceeding 90 microseconds, single-qubit gate fidelities above 99.9% and two-qubit gate fidelities above 98.6%. They also present design methods and discuss how sensitive device parameters are to variation in interchip spacing. They state that the extra flip-chip fabrication steps do not degrade qubit performance compared with their baseline state-of-the-art single-chip planar circuits. They say the technique could be extended to processors with hundreds of qubits in one module, since it offers wiring access to all qubits and couplers. That extension is the authors' projection and was not shown.
2021: no galvanic connection
The Applied Physics Letters paper takes the opposite approach to bonding. Assembling dies with very low-force contacts and air-dried adhesives is much simpler, though it gives no galvanic contact between the dies. The authors demonstrate that this can reliably couple superconducting qubit circuits with the qubits on separate dies. They demonstrate full vector control of each qubit on each of the two dies, with high-fidelity single-shot readout, entanglement-generating excitation swaps, and a benchmarked controlled-Z gate between the two qubits. They say it exemplifies a simple, inexpensive assembly method for two-plus-one-dimensional circuit integration that supports delicate or unusually shaped substrates.
A 2025 thesis
A 2025 Chalmers University of Technology doctoral thesis titled "Flip-chip Integrated Superconducting Quantum Processors" was listed as a source. Its abstract begins by describing the challenge of scaling up qubit counts without degrading their performance, but the abstract was cut off, so the thesis is not described further.
Limits
The two experiments involve a few qubits and use opposite bonding strategies, so they are not variations of the same technique. The 2021 paper reports no fidelity figures in the readable abstract. Statements about hundreds of qubits per module are projections.
FAQ
What is a flip-chip module?
Two chips bonded face to face, such as a quantum chip and a control chip.
Did flip-chip bonding hurt the qubits?
The 2022 authors report no degradation compared with their single-chip baseline.
How good were the gates?
Single-qubit fidelity above 99.9% and two-qubit fidelity above 98.6%, per the 2022 abstract.
Is a galvanic connection required?
The 2021 paper shows qubits on separate dies coupled without one.
Does this give a processor with hundreds of qubits?
The authors say it could be extended that way. It was not demonstrated.
All field notes