⚛️ quantum physics

Entanglement dynamics and performance of two-qubit gates for superconducting qubits under non-Markovian effects

This paper employs numerically exact simulations to investigate how non-Markovian reservoir correlations and qubit-reservoir interactions influence entanglement dynamics, the validity of the rotating wave approximation, and the performance of two-qubit gates in superconducting quantum architectures.

Original authors: Kiyoto Nakamura, Joachim Ankerhold

Published 2026-04-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Kiyoto Nakamura, Joachim Ankerhold

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build a super-advanced library where books (quantum information) can be in two places at once and talk to each other instantly. This is the dream of a quantum computer. The "books" in this library are tiny particles called qubits.

However, there's a problem. These qubits are incredibly fragile. They are like delicate glass sculptures sitting in a room full of invisible, jittery ghosts (noise). These ghosts bump into the qubits, causing them to drop their special "quantum magic" (entanglement) and turn into ordinary, boring objects.

This paper is like a high-tech detective report written by scientists at Ulm University. They used a super-powerful computer simulation to watch exactly how these ghosts mess with the qubits, specifically looking at two main things:

How the qubits lose their magic (disentanglement).
How well we can perform tricks (gates) to make them talk to each other.

Here is the breakdown of their findings, explained simply:

1. The "Lazy" Approximation vs. The Real Deal

For a long time, scientists used a "lazy" shortcut to predict how these ghosts affect qubits. They assumed the ghosts only push the qubits in one direction (like a gentle breeze). They ignored the fact that sometimes the ghosts push back hard or in weird, counter-intuitive ways.

The Analogy: Imagine trying to predict how a boat moves in the ocean. The "lazy" method assumes the waves only push the boat forward. The "real" method accounts for the fact that waves also crash against the side, pull the boat back, and create complex swirls.
The Discovery: The scientists found that this "lazy" shortcut is actually quite dangerous. When the connection between the qubit and the noise is strong, ignoring those "backwards pushes" leads to big mistakes. It's like thinking your boat is safe when it's actually about to capsize. They proved that to build a reliable quantum computer, we must account for these complex, "counter-rotating" pushes.

2. The "Sudden Death" of Magic

One of the most famous phenomena in quantum physics is "Entanglement Sudden Death" (ESD). This is when two qubits, which are perfectly linked, suddenly snap that link and become totally separate, even if they haven't been touched.

The Analogy: Imagine two dancers holding hands perfectly in sync. Suddenly, without warning, their hands let go, and they stop dancing.
The Discovery: The paper shows that this "sudden death" happens much faster and more often than the "lazy" methods predicted. The "ghosts" (noise) don't just slowly fade the magic away; they can cut the cord instantly. This is crucial because if you don't know when the magic will die, you can't build a computer that lasts long enough to do math.

3. The "Memory" of the Noise

Usually, scientists assume noise is "forgetful" (Markovian). It hits the qubit and immediately forgets it happened. But in reality, the noise has a memory (Non-Markovian).

The Analogy: Imagine you shout in a canyon. A "forgetful" noise would just be the shout. A "remembering" noise is an echo. The echo comes back and hits you again, changing how you feel.
The Discovery: The noise in these superconducting qubits echoes. When the qubit interacts with the noise, the noise remembers that interaction and pushes back later. This "echo" can actually help or hurt the qubit depending on the timing. The scientists found that if you try to ignore these echoes (by resetting the system), you get the wrong answer. You have to listen to the whole conversation, not just the first sentence.

4. Testing the Tricks (Gates)

To make a quantum computer work, you have to perform "gates" (tricks) like flipping a switch or swapping information between qubits. The scientists tested two different ways to perform a common sequence of tricks (Hadamard + CNOT).

The Finding:
- Speed is key: The faster you do the tricks, the better. Every second you wait (called "idling") gives the ghosts more time to mess things up.
- The "Goldilocks" Noise: They tested different types of noise (some fast and chaotic, some slow and deep). Surprisingly, the "medium" noise (not too fast, not too slow) actually allowed the qubits to hold their magic the longest. It's like finding the perfect temperature for baking a cake—not too hot, not too cold.

5. The Takeaway for the Future

The main message of this paper is: Stop using the simple, lazy math.

If we want to build a quantum computer that actually works, we need to use these super-precise simulations that account for:

The complex, "backwards" pushes of the noise.
The "echoes" (memory) of the environment.
The fact that noise can kill entanglement instantly.

In a nutshell: Building a quantum computer is like trying to keep a soap bubble alive in a hurricane. The old maps (approximations) told us the hurricane was a gentle breeze. This paper says, "No, it's a tornado with memory, and if you don't respect the wind, your bubble will pop before you even start." By using their new, precise maps, engineers can finally start building a computer that survives the storm.

1. Problem Statement

The performance of superconducting quantum computers has improved significantly, yet residual noise from environmental reservoirs remains a critical barrier to high-fidelity operations. Current standard methods for modeling open quantum systems, such as the Bloch-Redfield and Lindblad equations, rely on the Born-Markov approximation and the Rotating Wave Approximation (RWA). These approximations assume weak coupling, memoryless (Markovian) environments, and neglect counter-rotating terms in the system-reservoir interaction.

As gate fidelities approach the thresholds required for fault tolerance, these approximations become insufficient. Specifically:

The validity of the RWA in the presence of strong coupling or specific noise spectra is unclear.
The role of non-Markovian memory effects (retarded reservoir feedback) on entanglement generation and destruction during actual gate sequences is not fully understood.
There is a lack of rigorous, non-perturbative simulations for multi-qubit gate sequences that account for realistic noise spectra (e.g., $1/f$ noise, quasiparticle noise) and finite temperatures.

2. Methodology

The authors employ a numerically exact simulation technique known as the Free-Pole Hierarchical Equations of Motion (FP-HEOM), combined with Tensor-Train (TT) representations to manage computational complexity.

Model: A two-qubit architecture where each qubit couples independently to its own reservoir (noise source). The system Hamiltonian includes time-dependent single-qubit driving fields ( $\Omega_j(t)$ ) and a direct qubit-qubit coupling ( $J(t)$ ) for two-qubit gates.
Noise Spectra: Two types of reservoirs are investigated:
1. Lorentzian noise: Models a damped Rabi model (qubit in an imperfect cavity).
2. Broadband noise: Models macroscopic reservoirs with spectral densities $J(\omega) \propto |\omega|^s$ , covering Ohmic ( $s=1$ ), sub-Ohmic ( $0<s<1$ ), and deep sub-Ohmic regimes.
Approximations Tested: The study explicitly compares results obtained with the RWA (neglecting counter-rotating terms $\hat{\sigma}^+ \hat{B}^\dagger + \hat{\sigma}^- \hat{B}$ ) versus without the RWA (full Caldeira-Leggett model).
Metrics: Performance is quantified using Concurrence (entanglement measure) and Fidelity (closeness to ideal unitary evolution).

3. Key Contributions and Results

I. Validity of the Rotating Wave Approximation (RWA)

The authors analyzed the disentanglement dynamics of maximally entangled Bell states under various noise conditions.

Counter-Rotating Terms are Crucial: The RWA introduces significant errors, particularly for the Bell state $|\Phi\rangle = (|10\rangle + |01\rangle)/\sqrt{2}$ . The RWA incorrectly predicts zero population for the doubly excited state $|11\rangle$ , whereas the exact method shows that counter-rotating terms induce two-photon excitation processes ( $|01\rangle \to |11\rangle$ ).
Entanglement Sudden Death (ESD): In broadband noise scenarios, the exact method (without RWA) predicts Entanglement Sudden Death (concurrence reaching exactly zero in finite time), a phenomenon absent in RWA calculations where concurrence decays asymptotically.
Detuning Effects: Under the RWA, the dynamics are symmetric with respect to the sign of frequency detuning. The exact method reveals an asymmetric dependence on detuning, indicating that the system is more sensitive to low-frequency noise modes in the strong-coupling regime.
Error Magnitude: For gate fidelities required in state-of-the-art quantum computing ( $>0.999$ ), the errors introduced by the RWA (ranging from $10^{-3}$ to $>10^{-1}$ ) are unacceptable for high-precision predictions.

II. Dynamics During and After Two-Qubit Gates ( $\sqrt{iSWAP^\dagger}$ )

The study simulated the generation of entanglement via a $\sqrt{iSWAP^\dagger}$ gate followed by an idle period.

Spectral Exponent Dependence: The spectral exponent $s$ $s$ dictates performance.
- Ohmic ( $s=1$ ): Fast reservoir dynamics prevent entanglement generation in the short-time domain due to rapid population of the $|11\rangle$ state.
- Deep Sub-Ohmic ( $s=1/8$ ): Sluggish reservoir dynamics lead to faster disentanglement and intense oscillations (memory effects) after the gate.
- Intermediate ( $s=1/2$ ): This regime yielded the best performance, balancing low-frequency noise suppression and manageable memory effects.
Gate Time vs. Coupling: When the system-reservoir coupling dominates the qubit-qubit coupling (long gate times), entanglement begins to decay during the gate operation.
Memory Effects: The state of the system after the gate is strongly correlated with the dynamics during the gate. Resetting the reservoir (factorizing the state) after the gate leads to a significant underestimation of concurrence, proving that non-Markovian memory effects are essential for accurate prediction.
Heterogeneous Environments: When qubits couple to reservoirs with different spectral exponents, the real part of the off-diagonal density matrix element ( $\rho_{23}$ ) grows, and the system exhibits damped oscillations. This behavior can be detected via simple readout of specific matrix elements.

III. Performance of Gate Sequences (Hadamard + CNOT)

The authors analyzed a realistic sequence of a Hadamard gate and a CNOT gate using two different decomposition schemes.

Sequence Length: Shorter sequences consistently outperform longer ones. Inserting "idling" (buffer) times to separate pulses transiently recovers fidelity but ultimately degrades the final performance due to the longer total exposure to noise.
Initial State Sensitivity: Sequences starting from $|10\rangle$ or $|00\rangle$ performed better than those starting from $|11\rangle$ or $|01\rangle$ . This is attributed to the difficulty of maintaining the population of the highly excited $|11\rangle$ state against relaxation processes.
Non-Markovianity vs. Monotonicity: The study challenges the assumption that monotonic decay of infidelity implies Markovian dynamics. Even when total infidelity increases monotonically, partial fidelities of individual qubits can oscillate, indicating underlying non-Markovian dynamics. Standard Lindblad simulations failed to capture these oscillations and overestimated fidelity.

4. Significance and Conclusion

This paper provides a rigorous benchmark for the simulation of superconducting qubits, demonstrating that approximate methods (RWA, Markovian, Lindblad) are insufficient for high-precision quantum device characterization.

Practical Implication: For next-generation quantum devices aiming for fault tolerance, simulations must include counter-rotating terms and non-Markovian memory effects.
Design Guidance: The results suggest that optimizing the spectral density of the environment (targeting intermediate spectral exponents like $s=1/2$ ) and minimizing gate times are critical strategies.
Methodological Advance: The successful application of FP-HEOM with Tensor-Train decomposition to multi-qubit gate sequences establishes a new standard for simulating open quantum dynamics without relying on perturbative or Markovian assumptions.

The authors conclude that while approximate methods may suffice for qualitative analysis, the "minute effects" of residual degrees of freedom are now the dominant factor limiting the performance of high-fidelity quantum gates, necessitating the exact numerical approaches presented in this work.