TL;DR
Researchers have demonstrated a way to perform ultra-fast, high-fidelity entangling gates between fluxonium qubits by ditching standard capacitive coupling in favor of inductive coupling. By combining a "native" ZZ-interaction with a sinusoidal noise-filtering technique, they achieved a 20 ns Controlled-Z (CZ) gate with 99.53% fidelity, overcoming a long-standing trade-off between noise protection and gate speed in fluxonium circuits.
What Problem Does This Paper Address?
Fluxonium qubits are a rising star in quantum computing because they are exceptionally "quiet"—their design protects them from charge noise, leading to very long coherence times. However, this protection makes them difficult to "talk" to.
Most researchers use capacitive coupling (connecting qubits via electric fields), but for fluxonium, this restricts interactions to "XX-type" (state swapping). To perform a Controlled-Z (CZ) gate, which is essential for universal quantum computing, researchers usually have to temporarily "push" the qubit into higher-energy states that are short-lived, or use slow microwave-driven tricks. Both methods limit the speed and fidelity of the gate. The challenge is finding a way to get a fast, strong entangling interaction without sacrificing the fluxonium's signature noise protection.
Methods
The authors propose and demonstrate two key innovations:
- Inductive Coupling: Instead of capacitors, the team used a galvanic connection (shared Josephson junctions) to couple the qubits via magnetic fields. This creates a "native" ZZ-interaction that can be tuned simply by adjusting the external magnetic flux applied to the loops.
- Continuous Dynamical Decoupling: To activate the ZZ-interaction, the qubits must be moved away from their "sweet spot" (the flux-insensitive position). This normally makes them highly vulnerable to magnetic flux noise. To solve this, the authors applied a sinusoidal flux modulation at 50 MHz. This modulation effectively "filters" out low-frequency noise, protecting the qubit's coherence even while the gate is being performed.
Datasets
As an experimental physics paper, this work does not use traditional "datasets." Instead, the approach was evaluated using:
- Spectroscopy: To map out the energy levels and interaction strengths.
- Ramsey Interferometry: To measure coherence times ($T_2$) under different modulation schemes.
- Randomized Benchmarking (RB): The gold standard for measuring gate fidelity, performed over a 30-hour window to demonstrate stability.
Key Insights and Ideas
- ZZ as a Feature, Not a Bug: In many qubit types (like transmons), ZZ-interactions are considered parasitic errors. Here, the authors leverage a "native" ZZ-interaction as the primary mechanism for entanglement, allowing for gates that are first-order in coupling strength (i.e., very fast).
- Active Noise Filtering: The "aha moment" is the realization that you don't have to stay at the flux-insensitive "sweet spot" to have a high-fidelity gate. By modulating the flux at high frequencies, you can create a "synthetic" protection that allows you to operate in high-interaction regimes that were previously considered too noisy.
- Adiabaticity is Key: By using a sinusoidal drive rather than square pulses, the system stays in its "instantaneous energy eigenstate," preventing errors caused by jumping to unwanted higher-energy levels.
Key Results
- Gate Speed: The CZ gate was performed in just 20 ns, which is remarkably fast for fluxonium-based architectures.
- Fidelity: They achieved a mean CZ gate fidelity of 99.53%, placing it among the top-performing two-qubit gates for this platform.
- Coherence Boost: The dynamical decoupling scheme improved the dephasing time ($T_2$) by more than an order of magnitude compared to standard unmodulated pulses when operating away from the sweet spot.
- Stability: The gate performance remained stable over 30 hours of continuous monitoring, despite the presence of fluctuating environmental defects (TLS).
Limitations and Future Work
- Pulse Distortions: The researchers noted significant flux-pulse distortions that required adding a 20 ns "idle" time after each gate to let the system settle. This effectively doubled the total gate cycle time. Future work will involve better pulse-shaping and hardware calibration to eliminate this overhead.
- Two-Level Systems (TLS): Like all superconducting qubits, performance is still limited by microscopic defects in the materials (TLS poisoning) that cause unpredictable dips in coherence.
- Scalability: While this work demonstrates a single pair, scaling to many qubits will likely require tunable couplers (like inductively coupled SQUIDs) to turn the "always-on" interactions off when the qubits are idling.
Summary updated