Entangling gate performance and fidelity limits with neutral atom Förster resonances

This paper establishes a new fidelity bound for neutral-atom entangling gates near Förster resonances by developing a two-eigenstate model that accounts for coupled interaction channels, demonstrating that properly managing exchange dynamics can improve predicted gate fidelities by up to two orders of magnitude and saturate a theoretical limit approximately 40% higher than previous estimates.

The Gist

Imagine you are trying to teach two strangers (atoms) to dance a perfect, synchronized tango. In the world of quantum computing, this "dance" is called an entangling gate, and it's the fundamental move needed to build powerful quantum computers.

For a long time, scientists have been trying to get these atoms to dance using a special trick called Rydberg interactions. Think of this as turning the atoms into giant, fluffy balloons (Rydberg states) that can feel each other's presence from far away.

The Old Way: The "One-Step" Dance

Previously, researchers analyzed this dance by assuming the atoms only had one way to interact. They treated the interaction like a simple, single-lane highway. If the atoms got too close, they would bump into each other (a "blockade"), and that bump was the only thing that mattered.

The problem? Real atoms are more complex. Sometimes, instead of just one lane, there are two lanes that are perfectly balanced. This happens at a special point called a Förster resonance. It's like a dance floor where two different dance moves happen at the exact same time, perfectly in sync.

The New Discovery: The "Two-Step" Dance

This paper says: "Stop pretending there's only one lane! If you ignore the second lane, you're missing a huge part of the dance."

The authors found that when you acknowledge both lanes (the two eigenstates), a magical thing happens:

1. The "Dark" Partner: One of the dance moves is "bright" (easy to see and control), and the other is "dark" (invisible to the laser).
2. The Cancellation Trick: Because the atoms can swap energy between these two lanes, the errors that usually ruin the dance cancel each other out. It's like two people pushing a swing in opposite directions at the exact right moment; the swing stays perfectly still, or in this case, the "mistakes" disappear.

The Results: A Much Better Dance

By using this new understanding, the authors did two main things:

1. They found a new speed limit for perfection.
They calculated the absolute best possible score (fidelity) you can get for this dance.

• The Old Limit: Based on the single-lane model, the best you could hope for was a certain level of perfection.
• The New Limit: By using the two-lane model, they proved you can actually get about 40% better than the old limit. It's like realizing you can run a marathon 40% faster because you found a shortcut everyone else missed.

2. They designed a new dance routine.
They created a specific sequence of laser pulses (a "rank-two" gate) that takes full advantage of this two-lane system.

• The Routine: It involves exciting the atoms to two different states at once, letting them swap energy in the middle, and then bringing them back.
• The Outcome: This routine hits that new, higher speed limit. It's the most efficient way to get these atoms to entangle.

What About the Old Routines?

The paper also looked at the old, standard dance routines (like the "π-2π-π" gate) that people are currently using.

• The Surprise: When they re-evaluated these old routines using the new "two-lane" math, the predicted performance jumped up dramatically—sometimes by 100 times (two orders of magnitude).
• The Lesson: Even if you don't change your hardware, just understanding that the "two-lane" physics exists means your current computers are likely performing much better than we thought. However, if you design new computers, you must use the new math, or you will be optimizing for a world that doesn't exist.

The Catch (The "Hardware" Cost)

To get the full 40% boost from the new "rank-two" routine, you need a slightly more complex setup. Instead of using one laser to control the atoms, you need two lasers to control two different states simultaneously.

• Analogy: It's like upgrading from a bicycle with one gear to a bicycle with two gears. It's a bit more complex to build, but it lets you go much faster and smoother on the same terrain.

Summary

In short, this paper says: Stop simplifying the physics. When atoms interact via Förster resonances, they have a hidden "dark" partner that helps cancel out errors. By acknowledging this, we can design gates that are significantly more accurate, and we realize that our current estimates of how well these quantum computers work have been too pessimistic.

Technical Summary

Technical Summary: Entangling Gate Performance and Fidelity Limits with Neutral Atom Förster Resonances

Problem Statement
Neutral-atom quantum processors utilizing Rydberg interactions have achieved two-qubit gate fidelities exceeding fault-tolerance thresholds. However, further fidelity improvements are hindered by two primary error channels: decay from Rydberg states and coherent leakage into off-target states within the Rydberg manifold due to finite blockade. Standard analyses typically model these systems using a "one-eigenstate" approximation, where all Rydberg interactions are condensed into a single effective state $|rr\rangle$. This paper argues that this simplification fails to capture the physics of Förster resonances, where resonant dipole-dipole interactions create symmetric, bifurcated interaction channels. The single-channel model does not preserve the state dynamics crucial for accurate gate fidelity calculations, potentially leading to significant underestimations of achievable performance.

Methodology
The authors develop a systematic "two-eigenstate" model to describe the pair physics of atoms near a Förster resonance. In this regime, two Rydberg states $|a\rangle, |b\rangle$ and two nearby states $|\alpha\rangle, |\beta\rangle$ satisfy energy conservation ($\delta_A + \delta_B = 0$), enabling coherent population exchange $|ab\rangle \leftrightarrow |\alpha\beta\rangle$.

1. Hamiltonian Formulation: The authors define the interaction Hamiltonian in the basis $\{|ab\rangle, |\alpha\beta\rangle, |a\beta\rangle, |\alpha b\rangle\}$. Under the rotating wave approximation, the interaction reduces to a coupling $V$ between $|ab\rangle$ and $|\alpha\beta\rangle$, creating a "bright" and a "dark" state structure.
2. Fidelity Bound Derivation: The paper derives a fundamental lower bound for gate infidelity ($1-F$) based on the integrated Rydberg population time ($T_R$) required to generate entanglement. This is quantified by a figure-of-merit $\eta = V \int G(s) ds$, where $G(s)$ minimizes the Rydberg population over states with a specific entropy rate.
   • For standard rank-one pulse sequences (addressing only $|a\rangle$ and $|b\rangle$), the bound is $\eta \geq 1 + \pi/2$.
   • For rank-two pulse sequences (addressing $|a\rangle, |b\rangle, |\alpha\rangle, |\beta\rangle$), the authors prove a tighter bound of $\eta \geq \pi/2$.
3. Gate Protocols: The authors construct specific pulse sequences to saturate these bounds.
   • A rank-two $\pi - \text{gap} - \pi$ sequence is proposed to saturate the $\eta = \pi/2$ bound. This involves crossed rank-two $\pi$ pulses on atoms A and B, followed by a free evolution "gap" period.
   • Common rank-one protocols (Adiabatic Rapid Passage, $\pi-2\pi-\pi$, and Time Optimal gates) are re-evaluated using both the one-eigenstate and two-eigenstate models to compare predicted fidelities.

Key Results

1. Improved Fidelity Bound: By allowing coupling to both pair states in the resonance (rank-two access), the fundamental fidelity limit improves by approximately 40% compared to the single-channel model. The new bound is $F \leq 1 - (\pi/2)/(V \tau_R)$, where $V$ is the interaction strength and $\tau_R$ is the Rydberg lifetime.
2. Saturation of the Bound: The proposed rank-two $\pi - \text{gap} - \pi$ protocol realizes a $\sqrt{i\text{SWAP}}^\dagger$ gate that saturates the $\eta = \pi/2$ bound in the limit of infinite Rabi frequency ($\Omega \to \infty$).
3. Discrepancy in Standard Protocols: When evaluating standard rank-one protocols (ARP, $\pi-2\pi-\pi$, TO) near Förster resonances, the two-eigenstate model predicts fidelities up to two orders of magnitude higher than the one-eigenstate model.
   • Mechanism: In the two-eigenstate model, the Förster coupling $V$ continuously mixes $|ab\rangle$ and $|\alpha\beta\rangle$. This creates a dark-state structure that cancels linear AC-Stark phase errors (which scale as $\Omega/V$ in the one-eigenstate model) down to quadratic errors (scaling as $(\Omega/V)^2$).
4. Optimization Sensitivity: Gates numerically optimized under the one-eigenstate model perform significantly worse when applied to the actual two-eigenstate system. The authors note that local minima in the phase landscape make it unlikely that experimental optimizations starting from single-channel models will converge to the true maximum fidelity.

Significance and Claims
The paper claims that the exchange interaction between manifolds within Förster resonances can strongly attenuate entangling gate errors, a mechanism not predicted by standard single-channel models. Consequently, accurate gate fidelity calculations require atomic models that include these interaction channels.

The authors emphasize that:

• The improved fidelity bound ($\eta \geq \pi/2$) is directly attributable to the increased rank of the pulse sequence (rank-two access), not merely the dimensionality of the atomic model.
• For rank-one implementations (which require no additional hardware beyond addressing resonant levels), the two-eigenstate model still predicts significantly higher fidelities due to the error-suppression mechanism of the dark state.
• These improvements are most pronounced in the low-$V$ regime, which is critical for realizing long-range gates necessary for error-correction codes in neutral atom architectures.
• Simplified one-eigenstate descriptions fail to accurately predict gate performance for Rydberg pair states with multiple interaction channels, potentially leading to suboptimal experimental designs.