Readout-Induced Leakage in Superconducting Circuits with Nonlinear Couplings

This paper demonstrates that while native nonlinear qubit-resonator couplings offer theoretical advantages for superconducting circuits, they do not inherently eliminate drive-induced leakage and can actually exacerbate it without careful device engineering, such as optimizing spectral placement and eliminating parasitic modes.

The Gist

Imagine you are trying to listen to a very quiet whisper (the quantum bit, or "qubit") in a noisy room. To hear it clearly, you need to shout a bit louder (increase the "readout power"). However, if you shout too loud, you accidentally startle the whisperer, causing them to jump up and run away into a different room entirely. In the world of quantum computing, this "running away" is called leakage. Once the qubit leaves its designated "computing room," it creates errors that are very hard to fix.

This paper investigates a new way to build these quantum listening devices. The researchers wanted to see if a specific, fancy design could stop the qubit from running away, even when you shout loudly.

Here is the breakdown of their findings using simple analogies:

1. The Old Way vs. The New Idea

• The Old Way (Linear Coupling): Think of the qubit and the listening device as two people holding hands. If you shake one hand (send a signal), the other person feels it immediately. This is simple, but if you shake too hard, you might pull the person off their feet (leakage).
• The New Idea (Nonlinear Coupling): The researchers tried a "smart" connection. Imagine the two people are connected by a complex system of springs and pulleys designed so that shaking one person only makes them wiggle, not jump. Theoretically, this should act like a safety net, preventing the qubit from ever leaving its seat, no matter how hard you push.

2. The Surprise: The Safety Net Has Holes

The researchers built a device using this "smart" connection (specifically called a mediated cosine-cosine coupling). They expected it to be perfect. Instead, they found something tricky:

• The Hidden Room: To make this smart connection work, they had to add a third person to the room (an "auxiliary mode" or mediator).
• The New Problem: While the smart connection stopped some types of jumping, the presence of this third person created new pathways for the qubit to escape. It's like building a fancy door to keep a cat in, but in doing so, you accidentally installed a secret tunnel that the cat can now use to get out.
• The Result: The "smart" design didn't automatically fix the problem. In fact, if the room wasn't designed perfectly, it made the leakage worse than the old, simple way.

3. The "Goldilocks" Frequency

The most striking discovery was about timing and tuning.

Imagine you are trying to push a child on a swing. If you push at exactly the right rhythm, the swing goes high. If you push at the wrong rhythm, nothing happens.

• The researchers found that the "leakage" depends entirely on the exact pitch (frequency) of the signal they use to listen to the qubit.
• They tested two setups that were almost identical—only the "pitch" of the signal differed by less than 7% (a tiny difference, like the difference between a C note and a C-sharp on a piano).
• The Shock: In the first setup, the qubit leaked out of the room 20 times more often than in the second setup, even though the hardware was the same.
• The Lesson: You cannot just say, "We have a good design." You have to tune the design to the exact frequency you plan to use. A design that works perfectly at one frequency might be a disaster at a slightly different one.

4. The Takeaway

The paper concludes that while these fancy "nonlinear" designs are promising, they are not magic. They don't automatically solve the leakage problem.

• It's like High-End Audio Engineering: Just because you have a high-quality speaker doesn't mean it will sound good in every room. You have to account for every echo, every wall, and every piece of furniture (every "mode" in the circuit).
• The Warning: If you build a quantum computer using these new methods, you can't just rely on the theory. You have to map out every single "room" and "tunnel" in your device and ensure your signal frequency doesn't accidentally hit a "leakage trap."

In short: The new "smart" connections are a great idea, but they are incredibly sensitive. If you don't tune them perfectly to the exact frequency you are using, they can actually make the quantum computer less reliable than the old, simpler methods. The key to success isn't just the design; it's the precise engineering of every single frequency involved.

Technical Summary

Technical Summary: Readout-Induced Leakage in Superconducting Circuits with Nonlinear Couplings

Problem Statement
In superconducting quantum processors, achieving fast, high-fidelity quantum non-demolition (QND) readout is a critical bottleneck for fault-tolerant quantum error correction. While increasing readout drive power improves the signal-to-noise ratio and reduces readout duration, it inevitably triggers drive-induced unwanted transitions (DUST). Specifically, multiphoton resonances can excite the qubit from computational states into non-computational "leakage" states. These leakage events generate correlated errors that are particularly detrimental to error correction protocols.

Conventional linear hybridization between transmon qubits and readout resonators is known to suffer from these issues. To mitigate this, researchers have proposed native nonlinear coupling schemes (e.g., balanced cross-Kerr, native cosine-cosine, and mediated cosine-cosine couplings). These schemes theoretically offer intrinsic Purcell protection and stricter selection rules that should suppress multiphoton transitions. However, it remains unclear whether these benefits persist in realistic devices where additional electromagnetic modes (auxiliary or parasitic) are inevitably present.

Methodology
The authors employ a combination of numerical simulations and experimental validation to investigate the efficacy of nonlinear coupling schemes against readout-induced leakage.

1. Numerical Simulation:

   • The authors utilize Floquet steady-state simulations to analyze three specific coupling schemes: (i) ideal $\cos \hat{\phi}_q \cos \hat{\phi}_r$ coupling, (ii) balanced cross-Kerr interaction, and (iii) effective $\cos \hat{\phi}_q \cos \hat{\phi}_r$ coupling mediated by an auxiliary mode.
   • They benchmark these against a standard linearly coupled transmon-resonator system.
   • The simulations sweep the drive frequency ($\omega_d$) and amplitude (parameterized by induced ac-Stark shift) to map the "multiphoton resonance landscape." They quantify leakage using a hybridization parameter $\Theta$, which measures the deviation of Floquet modes from ideal displaced states.
   • The study explicitly includes auxiliary modes (mediator modes) and parasitic capacitances to simulate realistic device conditions.

2. Experimental Validation:

   • The team experiments with a "dimon" device, a multimodal circuit implementing a mediated effective cosine-cosine coupling. This device features a quadrupolar qubit mode linearly uncoupled from a dipolar mediator mode, which is linearly coupled to a readout resonator.
   • Spectroscopy: They perform pump-probe spectroscopy to map drive-induced unwanted state transitions (DUST) across a 4 GHz frequency window. They distinguish between "forbidden" and "allowed" transitions by utilizing both symmetric and asymmetric drive ports to test selection rules.
   • Leakage Benchmarking: They conduct repeated readout experiments interleaved with stochastic qubit operations (identity and bit-flip). By analyzing the decay of correlation between the input sequence and readout outcomes, they extract leakage ($L_\uparrow$) and seepage ($L_\downarrow$) rates.
   • Frequency Sensitivity Test: Two independent experiments are performed on the same device with readout resonator frequencies differing by less than 7% (7.513 GHz vs. 7.025 GHz) to isolate the impact of readout frequency choice on leakage.

Key Contributions and Results

• Limitations of Nonlinear Coupling: The study demonstrates that while ideal nonlinear couplings (specifically the native cosine-cosine interaction) exhibit a sparser frequency landscape of multiphoton resonances compared to linear coupling, this protection is fragile. In realistic implementations involving auxiliary modes, the benefit of nonlinear coupling is often negated. The presence of mediator modes introduces a dense spectrum of new multiphoton transitions, effectively increasing the number of leakage channels.
• Role of Auxiliary Modes: The mediated cosine-cosine coupling, intended to provide protection, actually introduces additional joint-excitation pathways (e.g., transitions labeled $[x, y:n]$ involving both qubit and mediator modes). The authors show that these transitions are highly sensitive to the frequency and anharmonicity of the auxiliary mode.
• Frequency Sensitivity: A critical finding is that readout-induced leakage is extremely sensitive to the choice of readout frequency. The authors demonstrate that varying the readout frequency by less than 7% can alter the leakage rate by more than an order of magnitude. This sensitivity persists regardless of whether the coupling is linear or nonlinear.
• Experimental Quantification: In the benchmarking experiments, the "Exp 1" configuration (readout at 7.513 GHz) exhibited a leakage rate of 2.1% to 6.3% depending on power, whereas the "Exp 2" configuration (readout at 7.025 GHz) reduced leakage to 0.1% while maintaining similar readout fidelity. This confirms that leakage is not a fixed property of the coupling scheme but is highly dependent on spectral placement.
• Parasitic Processes: The study identifies that inelastic scattering involving parasitic package modes and two-level system (TLS) defects further complicates the resonance landscape, creating additional leakage channels that are not suppressed by the symmetry of the qubit design.

Significance and Claims
The paper establishes that the theoretical advantages of native nonlinear couplings for qubit readout are not automatically realized in physical devices. The authors argue that:

1. Design Necessity: Merely engineering a nonlinear coupling term is insufficient. Successful implementation requires the explicit modeling and spectral placement of all relevant auxiliary and parasitic modes.
2. Analogy to Filter Synthesis: Engineering nonlinear couplings for readout is analogous to high-order filter synthesis; while additional modes offer greater control, they make system performance increasingly sensitive to mode frequencies, nonlinearities, and inter-mode couplings.
3. Comprehensive Characterization: A single-frequency characterization is inadequate for evaluating nonlinear coupling schemes. A comprehensive assessment requires systematic frequency sweeps in both numerical simulations and spectroscopy experiments to identify and avoid "accidental" multiphoton resonances.

The work concludes that while nonlinear couplings offer a promising path toward intrinsic Purcell protection, their practical utility is contingent upon rigorous device design that accounts for the complex, multi-mode nature of realistic superconducting circuits.