High-fidelity iSWAP gate with Double Transmon Coupler

This paper demonstrates a high-fidelity (99.827%) parametric iSWAP gate between two transmon qubits using a double transmon coupler, which enables fast two-qubit operations with suppressed crosstalk and robust static interaction cancellation without requiring numerical optimization.

The Gist

Imagine you are trying to get two very shy, high-strung dancers (called qubits) to perform a perfect, synchronized dance move called an iSWAP. In the world of quantum computing, this dance is crucial because it allows the dancers to swap places and share information, creating a "link" (entanglement) that is the engine of quantum power.

However, there's a big problem: these dancers are extremely sensitive. If they get too close, they accidentally bump into each other's personal space, causing them to stumble (errors). If they are too far apart, they can't dance at all. Usually, to make them dance, you have to pull them into a tight embrace, but this often causes them to trip over each other's feet (a problem called "crosstalk" or unwanted interactions).

This paper presents a clever new solution: a Double Transmon Coupler (DTC). Think of this coupler as a super-smart, invisible dance instructor standing between the two dancers.

Here is how the paper's breakthrough works, broken down into simple concepts:

1. The "Off" Switch: The Perfect Pause

Normally, when you aren't dancing, you want the dancers to be completely independent so they don't accidentally mess up their solo routines. In older systems, getting them to be truly independent was like trying to balance a pencil on its tip; you had to tune it perfectly, and even a tiny vibration would ruin it.

The new Double Transmon Coupler acts like a magic floor. When the dancers are in "rest mode," this floor has a special "cancellation point." It's like a noise-canceling headphone for the dancers. Even if they are close, the instructor (the coupler) creates a field that perfectly cancels out any accidental bumps or whispers between them. The paper shows that at this specific setting, the dancers are effectively invisible to each other, allowing them to rest without disturbing one another.

2. The "On" Switch: The Parametric Pulse

When it's time to dance, the instructor doesn't just push them together. Instead, the instructor taps a rhythm on the floor (a parametric flux modulation).

Think of it like a metronome. If you tap the floor at just the right speed (matching the difference in the dancers' natural rhythms), the dancers suddenly feel a strong magnetic pull to swap places. This happens incredibly fast (in just 40 nanoseconds, which is faster than a blink of an eye). Because the instructor only taps the rhythm when needed, the dancers don't have to change their natural rhythm or get dangerously close to each other the whole time. This avoids the "bumping" problems that happen in older methods.

3. The Challenge: The "Non-Commuting" Error

Here is the tricky part the paper solved. In the past, if the dancers made a mistake, you could just repeat the dance move to see how big the mistake was and fix it. But with this specific dance (iSWAP), the mistakes are weird.

Imagine if the dancers' mistake was that they were slightly out of step (a phase error) and they were slightly off-center (an amplitude error). If you tried to repeat the dance to measure the error, the "out of step" mistake would actually hide the "off-center" mistake, making it harder to fix. It's like trying to measure a wobble in a spinning top while the top is also tilting; the movements interfere with each other.

4. The Solution: Robust Phase Estimation

To fix this, the authors developed a new calibration routine called Robust Phase Estimation (RPE).

Instead of just repeating the dance, they created a compound routine. They told the dancers to do the swap, then spin, then swap again, then spin the other way. By arranging these moves in a specific sequence, they were able to "amplify" the specific errors they wanted to measure while canceling out the confusing parts.

It's like using a magnifying glass that only focuses on the wobble, ignoring the tilt. This allowed them to measure the errors with extreme precision without needing to run thousands of random tests or use complex computer simulations to guess the fix.

The Result

By using this smart instructor (the DTC) and the new measurement technique (RPE), the team achieved a dance performance that was 99.827% perfect.

• Speed: The dance took only 40 nanoseconds.
• Accuracy: The error rate was so low that the only thing stopping it from being 100% perfect was the natural "tiredness" of the dancers (decoherence), not the dance moves themselves.
• No "Tuning" Needed: The system didn't require hours of computer optimization to find the right settings; the calibration routine did it efficiently.

Why This Matters (According to the Paper)

The paper claims this is a major step forward because:

1. It's Modular: The "cancellation point" is built into the design of the instructor, so it works even if the dancers are slightly different sizes (frequency variations). You don't have to redesign the whole stage for every new pair of dancers.
2. It's Scalable: Because it reduces the risk of dancers bumping into each other when they aren't dancing, you can pack more dancers onto the same floor without them tripping over each other.
3. It's Fast and Clean: It achieves high speed and high accuracy without the messy "parasitic" interactions that usually plague fast quantum gates.

In short, the paper demonstrates a way to make two quantum bits swap information quickly and perfectly, using a new type of "instructor" that keeps them apart when they need to rest and brings them together only when they need to dance, all while using a new method to ensure the dance steps are perfectly calibrated.

Technical Summary

1. Problem Statement

Entangling operations are fundamental to quantum information processing, but scaling superconducting quantum computers faces significant challenges:

• Frequency Collisions & Crosstalk: As the number of qubits increases, managing frequency collisions and suppressing parasitic crosstalk (specifically static ZZ interactions) between non-resonant qubits becomes difficult.
• Parametric Gate Limitations: While parametric gates offer a versatile solution for coupling off-resonant qubits without tuning them into resonance, achieving high fidelity (comparable to resonant gates) has been difficult. This is largely due to complex drive optimization requirements and the non-commuting nature of exchange interactions (iSWAP) with single-qubit errors (Stark shifts) and cross-Kerr (ZZ) interactions.
• Calibration Complexity: Traditional calibration methods (like randomized benchmarking or process tomography) are often quadratically sensitive to errors and require lengthy numerical optimization or repeated randomized benchmarking, which is inefficient for non-commuting error terms.

2. Methodology

The authors propose a novel architecture and calibration routine to overcome these hurdles:

A. Device Architecture: Double Transmon Coupler (DTC)

• Design: The system utilizes a Double Transmon Coupler (DTC), consisting of two auxiliary transmon qubits (Q3, Q4) coupled via a Josephson junction (JJ5) in a three-junction loop. This coupler sits between two data qubits (Q1, Q2).
• Static Decoupling: The DTC is designed with specific capacitive ($g_C$) and inductive ($g_L$) interactions that have opposite signs. By tuning the flux bias ($\Phi_c$), the authors identify a "cancellation point" where the net effective coupling between data qubits is zero ($g_{eff} = 0$).
• Robustness: Unlike single-transmon couplers that require fine-tuned capacitive coupling to cancel inductive coupling, the DTC's cancellation point is first-order insensitive to data qubit frequencies. This makes the design modular and robust against fabrication variations.
• Parametric Activation: To perform the gate, a high-frequency flux modulation is applied to the coupler. This modulates the effective coupling at the difference frequency of the data qubits ($\omega_2 - \omega_1$), enabling a fast, resonant exchange interaction (iSWAP) without detuning the qubits from their optimal "sweet spots."

B. Calibration Technique: Robust Phase Estimation (RPE)

• Challenge: Standard phase estimation fails for iSWAP because error terms (like Stark shifts) do not commute with the exchange interaction, causing state vector rotations that suppress error amplification upon repetition.
• Solution: The authors developed a generalized RPE routine using compound gates.
  • They construct specific sequences (e.g., $Y_{Q1} \cdot \text{iSWAP} \cdot Y_{Q2} \cdot \text{iSWAP}$) that convert unknown error parameters (sum and difference of Stark shifts, pump amplitude errors) into eigenphases.
  • These eigenphases can then be amplified linearly by repeating the sequence, allowing for Heisenberg-limited precision in error estimation.
  • This method eliminates the need for numerical pulse optimization or full process tomography for calibration.

3. Key Contributions

1. High-Fidelity Parametric iSWAP: Demonstration of a 40 ns iSWAP gate with 99.827% fidelity, achieved without numerical pulse optimization.
2. DTC Architecture Validation: Experimental confirmation that the DTC provides a robust "off" state with minimal static ZZ interaction ($g_{zz} \approx -2\pi \times 220$ kHz), effectively decoupling data qubits during idle times.
3. Novel Calibration Protocol: Introduction of a compound-gate-based RPE technique capable of calibrating non-commuting error terms in two-qubit gates, a significant advancement over previous methods limited to commuting errors (like CZ gates).
4. Scalability: The architecture allows for modular circuit layouts where the coupler parameters define the cancellation point, removing the need for redesign when coupling qubits with widely varying frequencies.

4. Key Results

• Gate Fidelity:
  • Process Tomography: Measured a gate fidelity of 99.827(±0.004)%.
  • Interleaved Randomized Benchmarking (RB): Confirmed a fidelity of 99.70%, consistent with tomography results.
• Static Decoupling:
  • Simultaneous RB showed that the depolarizing parameters of the two-qubit system matched the product of single-qubit parameters, confirming effective decoupling.
  • The residual $g_{zz}$ was measured at 220 kHz, which theoretical analysis attributes to higher-order energy levels of the coupler. Simulations suggest this can be further reduced by increasing the coupler frequency.
• Error Analysis:
  • The measured gate error is primarily limited by qubit decoherence (estimated theoretical limit $\approx 99.72\% - 99.84\%$), not by the gate mechanism or residual ZZ interactions.
  • The residual $g_{zz}$ contributes only $\approx 0.002\%$ to the error, well below the decoherence limit.
• Speed: The gate operates in 40 ns, enabling fast two-qubit operations.

5. Significance and Impact

• Resource Efficiency: The high fidelity and speed of the iSWAP gate, combined with the DTC's ability to suppress crosstalk, are crucial for implementing surface codes and reducing circuit depth in quantum algorithms (e.g., quantum chemistry, analog optimization).
• Scalability: The "cancellation flux" being internally defined by the coupler rather than dependent on specific qubit frequencies simplifies the design of large-scale quantum processors, reducing the risk of frequency collisions.
• Calibration Advancement: The RPE technique for non-commuting errors provides a generalizable framework for calibrating complex entangling gates across different qubit modalities, potentially accelerating the development of fault-tolerant quantum computers.

In conclusion, this work demonstrates that parametric gates can match the fidelity of resonant gates when paired with a robust coupler architecture (DTC) and advanced calibration techniques (RPE), paving the way for more efficient and scalable quantum information processing.