Optimizing two-qubit gates for ultracold fermions in optical lattices

This paper presents an optimized approach for high-fidelity two-qubit collision gates in ultracold fermionic Lithium atoms within optical lattices, utilizing a momentum-dependent interaction model that surpasses standard Fermi-Hubbard simulations to enable tailored applications in quantum chemistry and simulation.

The Gist

Imagine you are trying to teach two tiny, invisible dancers (atoms) to perform a perfectly synchronized routine on a stage made of light. This is the core challenge of quantum computing using ultracold atoms.

This paper is essentially a "choreography manual" for these atomic dancers. The authors, a team of physicists from Germany and Italy, have developed a new, smarter way to program the lights and lasers that guide these atoms, making their dance moves (quantum gates) faster, more precise, and less likely to make mistakes.

Here is the breakdown of their work using everyday analogies:

1. The Stage: The Optical Lattice

Think of the optical lattice as a giant, invisible egg carton made of laser beams.

• The Atoms: The "dancers" are ultracold Lithium atoms. They are so cold they barely move on their own.
• The Goal: We want to move two specific atoms so they bump into each other in a very controlled way. This "bump" creates a quantum entanglement, a spooky connection where the two atoms become a single unit. This is the "two-qubit gate," the fundamental building block of a quantum computer.

2. The Old Way vs. The New Way

The Old Way (Fermi-Hubbard Simulation):
Previously, scientists tried to predict how these atoms would move using a simplified map. Imagine trying to navigate a city using a 2D paper map. It's fast, but it misses the hills, valleys, and traffic jams. In physics terms, this old map ignored the "momentum" (how fast and in what direction the atoms are moving) and treated the interaction between atoms as a simple, static rule.

The New Way (1D Confinement Simulation):
The authors built a 3D GPS for the atoms.

• They realized that when atoms are squeezed into a narrow lane (a "double-well" potential), their speed and direction matter immensely.
• The Discovery: They found that the "force" of the collision depends on how the atoms are moving.
  • Analogy: Imagine two cars merging onto a highway. If they are coming from the same lane and merging side-by-side, it's one kind of interaction. If they are coming from opposite directions and have to cross paths, the interaction is totally different.
  • The old map couldn't see this difference. The new map sees it clearly. This allows the scientists to tune the lasers differently depending on whether the atoms start in the same "lane" or opposite "lanes."

3. The Choreography: Optimization

The paper describes a process of optimization, which is like a coach refining a dance routine.

• The Controls: The coaches control the "laser intensity" (how bright the stage lights are) and the "scattering length" (how sticky the atoms are when they touch).
• The Trial and Error: Instead of guessing, they used a powerful computer algorithm (like a super-smart AI coach) to simulate millions of dance routines in seconds. It tweaked the laser pulses until it found the perfect sequence that makes the atoms dance exactly as needed with almost zero mistakes.
• The Result: They achieved a "high-fidelity" gate, meaning the dance is 99%+ perfect. They also found that by treating the "same-lane" and "opposite-lane" cases separately, they could get even better results for specific applications (like simulating chemical reactions vs. building a computer).

4. Dealing with Real-World Messiness

In a perfect world, the stage lights are steady, and the atoms are perfect. In the real world, lasers flicker, and the stage might be slightly crooked.

• Robustness Check: The authors tested their new dance routine against "messy" conditions.
  • The Asymmetric Stage: What if the left side of the stage is slightly higher than the right?
  • The Wrong Scattering: What if the atoms are slightly stickier or less sticky than expected?
  • The Extra Dancer: What if a third atom accidentally joins the dance?
• The Verdict: Their optimized routine is surprisingly tough. Even with these imperfections, the atoms still manage to perform the dance correctly. This is crucial because it means their method isn't just a theoretical dream; it can actually work in a real laboratory.

Why Does This Matter?

• For Quantum Computers: We need millions of these perfect "dance moves" to build a useful quantum computer. This paper shows a way to make those moves faster and more reliable.
• For Chemistry: By simulating how atoms interact with high precision, we can use these quantum computers to design new medicines or materials, essentially acting as a "virtual lab" for chemistry.

In Summary:
The authors took a complex problem—teaching atoms to dance on a stage of light—and solved it by building a better map (simulation) and a smarter coach (optimization). They realized that the atoms' speed matters, and by accounting for that, they created a dance routine that is robust, fast, and ready for the real world.

Technical Summary

1. Problem Statement

The realization of fast, high-fidelity two-qubit gates is a critical bottleneck for scaling neutral-atom quantum computers and simulators. While collisional gates (using s-wave scattering) offer inherent noise robustness, optimizing them for fermionic systems (specifically $^6$Li) presents challenges:

• Simulation Complexity: Accurate simulation of interacting atoms in optical lattices typically requires solving the Schrödinger equation in high-dimensional spaces, which grows exponentially with the number of particles and dimensions.
• Model Limitations: Previous approaches often relied on the Fermi-Hubbard model, which assumes a tight-binding approximation and neglects momentum-dependent interaction effects.
• Experimental Constraints: Real-world implementations face limitations such as finite laser switching speeds, transfer function distortions (bandwidth limits), and uncertainties in interaction strengths (scattering lengths).
• Initial State Dependence: The interaction dynamics differ significantly depending on whether atoms start in the same subwell or separate subwells, yet previous optimizations often treated these cases uniformly.

2. Methodology

The authors propose a comprehensive framework combining advanced numerical simulation, gradient-based optimization, and realistic experimental modeling.

A. Simulation Technique: "Leapfrog" Method

To overcome the computational cost of standard split-step Fourier methods (which scale as $O(N \log N)$ and suffer in high dimensions), the authors developed an alternative "leapfrog" method:

• Approach: Based on a Numerov-type ansatz applied to the time dimension rather than spatial dimensions. It approximates the time derivative of the wavefunction using a discrete third-order approximation in real space with periodic boundary conditions.
• Advantage: Computation time scales approximately linearly with the number of grid points ($N$), making it significantly faster for time-dependent problems and multi-particle simulations compared to the Wannier-state projection methods used in prior work.
• 1D Confinement: The system is modeled as a 1D confinement where transverse dynamics ($y, z$) are frozen. The interaction is described by an effective 1D pseudo-potential $U_{1D}\delta(x_1 - x_2)$, derived from confinement-induced resonances.

B. Optimization Strategy

The authors employ a gradient-based optimization (BFGS algorithm) to minimize gate infidelity. The process is structured in a multi-stage flowchart (Fig. 4):

1. State-to-State Pre-optimization: Optimize laser intensities ($V_s(t), V_l(t)$) for single-atom transitions.
2. Two-Atom Pre-optimization: Optimize the effective scattering length ($a_{1D}$) for two atoms using the pre-optimized pulses.
3. Joint Gate Optimization: Simultaneously optimize $V_s(t)$, $V_l(t)$, and $a_{1D}$ for the full two-qubit gate, using previous results as initial guesses to reduce computational cost.
4. Case Separation: Crucially, the authors perform separate optimizations for two distinct initial conditions:
   • Atoms starting in different subwells ($w_{LR}$ or $w_{RL}$).
   • Atoms starting in the same subwell ($w_{LL}$ or $w_{RR}$).

C. Experimental Realism

The optimization incorporates realistic experimental constraints:

• Transfer Functions: The control signals are filtered through a complex Butterworth transfer function to model the bandwidth limitations of the acousto-optic modulators (AOMs) and the delay between electrical and optical signals.
• Control Parameters: The optimization controls the amplitude of short ($V_s$) and long ($V_l$) lattice potentials while keeping the relative phase constant (except for robustness checks).

3. Key Contributions

1. Novel Simulation Algorithm: Introduction of the "leapfrog" numerical method, which offers a speedup for time-dependent quantum simulations while maintaining high precision, validated against experimental Rabi oscillation data.
2. Discovery of Momentum Dependence: The 1D confinement simulation reveals a momentum dependence in the interaction energy that is invisible in the standard Fermi-Hubbard model.
   • Atoms starting in separate subwells have opposite momenta, leading to stronger effective interactions.
   • Atoms starting in the same subwell have similar momenta, leading to weaker interactions.
3. Tailored Optimization: Demonstration that optimizing gates separately for "same-subwell" and "different-subwell" initial states yields significantly higher fidelities than a combined optimization.
4. Robustness Analysis: Comprehensive analysis of gate performance under:
   • Asymmetric lattices (relative phase errors).
   • Uncertainties in scattering length (interaction energy).
   • Laser intensity fluctuations.
   • State preparation errors (e.g., three-particle collisions).

4. Results

• Gate Fidelity:
  • The optimized gates achieve high fidelities with infidelities as low as $10^{-3}$ to $10^{-4}$ for gate times $\tau \geq 300$ $\mu$s.
  • Separate Optimization: For atoms starting in different subwells, separate optimization improves precision by two orders of magnitude compared to combined optimization for longer gate times.
  • Speed Limit: The empirical quantum speed limit is found to be $\tau/\alpha \approx 300/\pi$ $\mu$s, with errors around 1% at this limit.
• Momentum Dependence Impact:
  • The study confirms that the interaction strength is higher for atoms starting in separate subwells ($w_{LR}$) compared to the same subwell ($w_{LL}$).
  • This necessitates different optimal scattering lengths ($a_{1D}$) for the two cases (e.g., $a_{1D} \approx -11925 a_0$ was found optimal for specific parameters, but the paper notes this value is too high for different subwells and too low for same subwells if not optimized separately).
• Robustness:
  • The gates are robust against small relative phase errors ($\phi$) and scattering length deviations ($\Delta a_{1D}$).
  • Interestingly, gates with lower initial infidelity (higher precision) are slightly more sensitive to control noise than those with higher initial infidelity.
  • The gates remain functional even in the presence of three-body collisions (state preparation errors), though the dynamics differ from the intended two-body evolution.

5. Significance and Applications

• Quantum Simulation & Computing: The ability to generate high-fidelity entangling gates for fermionic $^6$Li atoms in optical lattices is a major step toward scalable quantum processors. The robustness against experimental imperfections makes these protocols viable for near-term experiments.
• Quantum Chemistry: The paper highlights that "same-subwell" initialization is critical for simulating molecular interactions (quantum chemistry), while "different-subwell" initialization is standard for lattice-based quantum simulation. The proposed method allows for tailored gate sequences for these specific applications, rather than a one-size-fits-all approach.
• Methodological Advancement: The "leapfrog" simulation technique provides a scalable tool for future research in quantum control, enabling the simulation of larger systems and more complex time-dependent potentials without the prohibitive cost of full 2D/3D simulations or the inaccuracies of tight-binding approximations.

In conclusion, this work bridges the gap between theoretical optimal control and experimental reality by developing faster simulation tools, uncovering momentum-dependent physics in collisional gates, and providing robust, tailored gate sequences for fermionic quantum technologies.