Error-corrected phase estimation averaged over variable grids on a trapped-ion quantum computer: hyperacuity spectra of a CO molecule adsorbed onto $\chi$-Fe$_5$C$_2$

This paper introduces and experimentally validates on a trapped-ion quantum computer a novel "QPE averaged over variable grids" (QAVG) method that combines low-resolution quantum phase estimation with origin shifts and continuous parametrization to accurately reconstruct the excitation spectra of a CO molecule on a $\chi$-Fe$_5$C$_2$ surface, effectively overcoming hardware noise and spectral leakage to enable robust early-fault-tolerant quantum simulations.

The Gist

The Big Picture: Tuning a Radio in a Noisy Room

Imagine you are trying to tune an old-fashioned radio to find a specific music station. You want to hear the song clearly, but two things are making it hard:

1. The dial is coarse: The numbers on the dial jump in big steps (like 100, 105, 110), so you can't land exactly on 103.5.
2. The room is noisy: There is static and background chatter that makes the signal fuzzy.

This is exactly the problem scientists face when using Quantum Computers to study how molecules work. They want to know the exact "energy notes" (spectra) a molecule plays, but current quantum computers are like that coarse, noisy radio. They can't get a perfect reading, and the "static" (errors) often tricks the computer into thinking it found the right note when it hasn't.

The Solution: The "Vernier" Trick (QAVG)

The authors of this paper propose a clever new method called QAVG (Quantum Phase Estimation Averaged over Variable Grids).

Think of it like a Vernier caliper (a tool used by mechanics to measure tiny distances more precisely than a standard ruler).

• The Old Way: You take one measurement with the ruler. If the object is slightly off the line, you guess.
• The QAVG Way: You take the same measurement, but you shift the ruler slightly to the left, then slightly to the right, then slightly up, and so on. You do this many times.

By combining all these slightly shifted measurements, the computer can "triangulate" the true position of the energy level. Even if the ruler is coarse and the room is noisy, the pattern of the shifts reveals the exact answer with much higher precision than a single measurement could ever provide.

The Experiment: A Molecule on a Metal Surface

To test this, the researchers didn't just use a simple math problem; they simulated a real-world chemical scenario:

• The Scene: A Carbon Monoxide (CO) molecule sticking to a specific type of iron-carbide surface (used in making fuels).
• The Goal: Figure out exactly how the electrons in that molecule behave when they get excited. This is crucial for understanding how industrial catalysts work.

They built a simplified model of this interaction (a "dimer" model) and ran it on a Quantinuum H2-2, which is a real, physical quantum computer that uses trapped ions (electrically charged atoms held in place by magnetic fields).

Two Types of "Listening"

The team tested their method in two different ways:

1. Physical Circuits (The Direct Approach): They ran the experiment directly on the raw hardware. It's like listening to the radio with no special equipment.
2. Logical Circuits (The Error-Corrected Approach): This is the more impressive part. They used a "Steane code," which is a way of grouping seven physical qubits (the basic units of the computer) together to act as one single, protected "logical" qubit.
   • Analogy: Imagine you have a fragile message written on a piece of paper. Instead of sending just one copy, you send seven copies. If one gets torn or smudged, the computer looks at the other six to figure out what the original message said and fixes the error.
   • They even used a "flag" system to catch errors as they happened and threw out the bad data (shots) before it corrupted the result.

The Results: Seeing the Invisible

The results were surprising and successful:

• Beating the Noise: Even though the "logical" circuits were noisier and more complex than the direct ones, the QAVG method managed to reconstruct the molecule's energy spectrum with incredible accuracy.
• Smoothing the Bumps: When the computer tries to find the best answer, it often gets stuck in "local minima"—think of it as a hiker getting stuck in a small valley and thinking it's the bottom of the mountain. The QAVG method, by averaging over all those shifted grids, smoothed out the landscape. It turned a bumpy, confusing terrain into a smooth slope, allowing the computer to easily find the true bottom (the correct answer).
• Hyperacuity: The paper calls this "hyperacuity." Just as human eyes can detect a tiny gap between two lines that is smaller than the width of a single cell in our retina (by using multiple cells together), this method detects energy levels more precisely than the computer's hardware resolution should theoretically allow.

The Bottom Line

This paper proves that you don't need a perfect, futuristic quantum computer to get useful scientific results today. By using a smart mathematical trick (shifting the grid and averaging) and combining it with error correction, researchers can extract high-precision data about complex molecules from current, imperfect hardware.

It's a roadmap for the "early-fault-tolerant" era: a time where we can do serious science even before we have perfect, error-free quantum computers.

Technical Summary

Technical Summary: Error-corrected phase estimation averaged over variable grids on a trapped-ion quantum computer

Problem Statement
Quantum Phase Estimation (QPE) is a fundamental algorithm for extracting excitation spectra of many-electron systems, offering advantages over classical methods for calculating quantities like one-particle Green's functions (GFs) without requiring diagonalization of huge Hilbert subspaces. However, practical implementation on current hardware is hindered by two primary factors: low grid resolution (limited by the number of ancillary qubits) and environmental noise. Furthermore, standard QPE cost landscapes often suffer from "spectral leakage," creating rugged landscapes with multiple local minima that trap optimization algorithms, particularly when using a single energy origin grid.

Methodology
The authors propose and implement QPE averaged over variable grids (QAVG), a vernier-type approach designed to reconstruct spectra accurately despite low-resolution grids and noise. The methodology involves an end-to-end workflow applied to a model system of a CO molecule adsorbed onto a $\chi$-Fe$_5$C$_2$ surface (a catalyst relevant to Fischer–Tropsch synthesis).

1. Classical Pre-processing:

   • Density Functional Theory (DFT) calculations were performed on the CO/$\chi$-Fe$_5$C$_2$ system to obtain electronic structures.
   • Maximally Localized Wannier Orbitals (MLWOs) were constructed, identifying five relevant orbitals (two $2\pi^*$-derived from CO and three $d$-derived from Fe).
   • To fit current quantum hardware constraints, a reduced Hubbard dimer model was derived, mapping the five MLWOs to a two-orbital system ($p$ and $d$ types). This model was further reduced to a single-qubit representation by focusing on specific electron- and hole-excited subspaces ($n_e=3$ and $n_e=1$).

2. The QAVG Algorithm:

   • Instead of using a single energy origin, QAVG employs multiple energy origin shifts ($E_{orig}^{(s)}$) smaller than the grid resolution ($1/t_0$).
   • The approach utilizes physically motivated continuous parametrization. Unknown transition amplitudes are parametrized using angles ($\vartheta$) to satisfy orthonormality conditions derived from natural orbitals.
   • A cost function is defined as the $L_1$ distance between observed histograms (from multiple origin shifts) and theoretical probability distributions. Minimizing this averaged cost function allows for the reconstruction of the Green's function and the Density of States (DOS).

3. Experimental Implementation:

   • Experiments were conducted on the Quantinuum H2-2 trapped-ion quantum computer.
   • Physical Circuits: Standard QPE circuits were executed using both 3-ancilla and single-ancilla (mid-circuit measurement) configurations.
   • Logical Circuits: Logical QPE circuits were implemented using the Steane [[7, 1, 3]] error-correcting code.
     • Logical states were prepared using a fault-tolerant (FT) encoding circuit with flag qubits.
     • Logical $z$-rotations were implemented non-FT (as transversal implementation is impossible for non-Clifford gates in this code), but offline bit-flip correction (BFC) was applied to the measured classical bit strings to correct single-qubit errors.

Key Results

• Spectral Reconstruction: The QAVG approach successfully reconstructed the many-electron DOS for the dimer model. The results from both physical and logical circuits showed deviations significantly smaller than the nominal QPE grid resolution.
• Noise Robustness: Even when using noisy histograms from logical circuits (which involved more physical qubits and non-FT operations), the reconstructed spectra agreed well with noiseless simulations and Full Configuration Interaction (FCI) benchmarks.
• Optimization Stability: The cost landscapes averaged over the shifted grids were found to be significantly smoother than those from single-reference grids. The averaging process substantially suppressed local minima caused by spectral leakage, facilitating the convergence of optimization algorithms (Nelder-Mead) to the global minimum.
• Logical vs. Physical: While logical circuits introduced more noise due to the lack of online error correction for non-Clifford gates, the QAVG method remained robust. The logical QPE results were comparable to physical QPE results, demonstrating the viability of error-corrected sampling even with imperfect logical gates.

Significance and Claims
The paper claims that QAVG provides a robust route to quantum simulations of correlated spectra suitable for the era of early-fault-tolerant quantum computers (early-FTQC).

• Overcoming Resolution Limits: By combining low-resolution QPE with multiple origin shifts and continuous parametrization, the method achieves "hyperacuity" (a term borrowed from vision science), reconstructing fine spectral features that would otherwise be unresolvable with the given grid spacing.
• Mitigating Optimization Barriers: The approach addresses the "curse of dimensionality" in cost landscapes by smoothing out local minima, making parameter optimization more reliable for larger systems.
• Practicality for Early-FTQC: The demonstration on logical qubits using the Steane code, despite the use of non-FT logical rotations and offline correction only, suggests that useful spectral information can be extracted before the hardware achieves full, genuine fault tolerance. The authors posit that this approach will become even more powerful as hardware scales to support larger, more realistic models (e.g., the full 17-orbital model mentioned as a future possibility).