⚛️ lattice

Physics-informed quantum algorithms for glueball-like excitations in a $\mathbb{Z}_2$ lattice gauge theory

This paper presents a physics-informed quantum computing framework for a (2+1)-dimensional $\mathbb{Z}_2$ lattice gauge theory that variationally prepares the gauge vacuum and employs Wilson-loop quantum subspace expansion, eigenvector continuation, and quench dynamics to systematically construct and characterize localized glueball-like excitations, offering a transferable approach for future non-Abelian studies.

Original authors: Dan-Bo Zhang

Published 2026-08-27

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Dan-Bo Zhang

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

In the deepest layers of the universe, forces that hold matter together behave in ways that are difficult to picture. In the theory describing the strong force, which binds the smallest particles inside atoms, there are particles made entirely of force itself, without any ordinary matter inside them. These are called glueballs. They are like knots of pure energy, formed because the force carriers can grab onto each other and loop back, creating a closed, self-contained structure. For decades, scientists have tried to find these elusive objects in particle accelerators, but they are hard to spot because they often mix with other types of particles. To understand them better, researchers use powerful computers to simulate the rules of the strong force on a grid, a method known as lattice gauge theory. However, these simulations are incredibly difficult, requiring massive computing power to track the complex dance of fields and to distinguish a true glueball from the background noise of the vacuum.

A researcher has now developed a new way to use quantum computers to study these force knots, specifically within a simplified model that mimics the behavior of the strong force. Instead of trying to simulate the full complexity of the real world immediately, they focused on a two-dimensional grid where the rules are slightly simpler but still capture the essential feature of confinement: the tendency of force lines to snap back into closed loops rather than spreading out. In this model, they created a "glueball-like" particle, which is a localized, closed loop of force that sits on top of the empty space, or vacuum. The researcher did not just guess where these particles might be; they built a step-by-step quantum algorithm to first prepare the perfect empty space, and then to carefully construct and measure the properties of the force loops that appear when energy is added to that space.

The process began with the vacuum itself. In this quantum world, the empty space is not truly empty; it is a seething sea of potential loops. To prepare this state on a quantum computer, the researcher used a two-stage approach. First, they created a basic structure that contained the right kind of loops, similar to laying down a foundation of bricks. Then, they applied a refinement process that added the missing connections and correlations between these loops, ensuring the vacuum looked exactly as the laws of physics demanded. This was crucial because if the starting point is wrong, any particle built on top of it will be distorted. Once they had a high-quality vacuum, they needed to find the particles. Instead of searching through every possible way the system could vibrate, which would take too long, they used a method that builds the particle out of specific, physically meaningful pieces. They took the vacuum and applied operators that create closed loops of different sizes and shapes, then used a mathematical technique to sort through these possibilities to find the lowest energy states. This allowed them to identify the lightest, most stable force knot and measure its energy with high precision.

One of the most significant challenges in this field is that as the force gets stronger, these loops tend to grow larger and more complex, making them harder to simulate. The researcher found that if they tried to describe the particle using only a small set of fixed loop shapes, their results would become inaccurate as the particle grew. To solve this, they introduced a clever compression technique. They realized that the way the particle changes as the force strength varies is smooth and predictable. By looking at how the vacuum behaves at nearby force strengths, they could infer how the particle should look at the target strength without needing to explicitly calculate every single large loop. This allowed them to capture the "dressing" of the particle—the way it is surrounded by a cloud of fluctuating fields—using far fewer resources than before.

With these tools in place, the researcher was able to measure the actual size of the force knot. They defined a radius based on how far the force extends from the center of the loop. In the strongest part of the confined region, the particle was very compact, roughly the size of a single basic loop unit. As they moved toward a point where the confinement weakens, the particle grew larger, expanding to about twice that size before hitting the limits of the simulated grid. This growth was not just a change in energy; it was a physical spreading of the force field. The researcher also looked at how these loops are created when the system is suddenly disturbed, a process known as a quench. They found that when the force is weak, the system resists forming large loops, but when the force is strong, large, complex loops appear naturally and abundantly.

The study confirms that this quantum approach can successfully prepare a vacuum, construct specific excitations, and measure their internal structure without needing to simulate the entire universe at once. The researcher is careful to note that while their model is a simplified version of the real strong force, the methods they developed are designed to be transferable. The way they built the vacuum and compressed the information about the particle's size and shape provides a blueprint that can be applied to more complex, realistic models of the strong force. By proving that these steps work in a controlled, simpler environment, they have laid the groundwork for future quantum simulations that could one day help physicists identify the true nature of glueballs in the real world, turning a theoretical concept into a measurable reality.

Technical Summary: Physics-Informed Quantum Algorithms for Glueball-like Excitations in a Z2 Lattice Gauge Theory

Problem Statement
The paper addresses the challenge of performing glueball spectroscopy on quantum computers. While numerical lattice gauge theory has successfully established the pure-gauge glueball spectrum in non-Abelian theories (like QCD), quantum computing approaches require specific strategies to prepare a correlated gauge vacuum and systematically construct low-lying pure-gauge excitations. The author notes that while recent trapped-ion simulations have observed closed-loop excitations in a $(2+1)$ -dimensional $Z_2$ lattice gauge theory, a general algorithmic framework is needed to prepare the vacuum, construct controlled excitations, and reveal their internal structure without relying on generic many-body searches. The study focuses on a pure $(2+1)$ D $Z_2$ lattice gauge theory, using the term "glueball-like" to describe localized, closed-flux excitations on the confining side of this Abelian model, distinguishing them from the non-Abelian glueballs of QCD.

Methodology
The author proposes a physics-informed quantum-computing framework organized around the physical structure of the glueball-like state rather than a generic excitation space. The methodology consists of four main algorithmic components:

Gauge-Invariant Vacuum Preparation:
- Loop-Gas Ansatz: The process begins with a variational circuit $V(\theta)$ that interpolates between the solvable electric limit ( $x=1$ ) and the magnetic limit ( $x=0$ ). This "loop-gas" state is prepared using a local stabilizer construction (similar to toric-code preparation) involving CNOT gates and single-qubit rotations. It provides a gauge-invariant manifold containing coherent closed loops but lacks full correlation in the intermediate regime.
- Hamiltonian Variational Ansatz (HVA): To restore missing correlations, the loop-gas state is refined using an HVA. Variational layers are generated by the electric ( $H_E$ ) and magnetic ( $H_B$ ) terms of the Hamiltonian. This ensures gauge invariance is preserved at every layer while systematically improving the energy and state fidelity.
Wilson-Loop Subspace Expansion (QSE):
- Once the vacuum $|\Psi_0\rangle$ is prepared, low-lying excitations are constructed by acting on the vacuum with a set of gauge-invariant Wilson-loop operators $\{O_a\}$ .
- The Hamiltonian is projected onto this non-orthogonal subspace, yielding a generalized eigenvalue problem $H_{sub} \mathbf{c} = E S \mathbf{c}$ .
- This approach allows for the extraction of energies, symmetry properties (via reflection operators), and local loop content (motif densities) without reconstructing the full wavefunction.
Eigenvector Continuation (EC) for Compression:
- To address the exponential growth of the Wilson-loop basis required for larger, more complex excitations (especially near the critical point), the author employs eigenvector continuation.
- This technique utilizes ground states from nearby Hamiltonian parameters ( $x'$ ) as a training set. Since the vacuum at a nearby coupling already contains a coherent mixture of loop sectors (via perturbation theory), these states provide a compact representation of the parameter-dependent dressing of the closed-flux configurations.
- The final basis combines explicit Wilson-loop operators with these nearby vacuum states, capturing both local excitation directions and non-local dressing.
Structural Characterization and Dynamics:
- Bethe-Salpeter (BS) Radius: To quantify the spatial extent of the excitations, the author defines a BS-inspired transition amplitude between the interacting vacuum and the lightest excitation using a family of closed-loop probes. The root-mean-square radius is calculated from these amplitudes, normalized by the elementary plaquette radius.
- Quench Dynamics: The framework is applied to nonequilibrium production by initializing the system in the electric vacuum and suddenly quenching the Hamiltonian to a different coupling. The time evolution of local closed-loop motif counts is monitored to probe excitation production.

Key Results

Vacuum Preparation: On a $6 \times 4$ periodic lattice, the combination of the loop-gas ansatz and HVA refinement successfully prepares the vacuum with high fidelity, particularly at $x=0.2$ (near the critical point $x_c \approx 0.247$ ). The HVA layers significantly reduce infidelity compared to the loop-gas state alone.
Excitation Spectrum: Using a Wilson-loop basis with a cutoff of two plaquettes, the QSE method identifies a low-lying spectrum at $x=0.6$ . The results show a non-degenerate first excited state and a degenerate pair of higher states. The analysis of motif densities reveals that the first state is spatially uniform, while the degenerate pair exhibits distinct spatial symmetry patterns (even/odd under reflection).
Subspace Compression: The study demonstrates that while increasing the Wilson-loop basis size improves accuracy, it becomes computationally prohibitive. Eigenvector continuation (EC) effectively compresses this requirement. For the first excitation at $x=0.4$ , using EC with a small number of training vacua ( $C=3$ ) recovers the correct energy and spatial dressing, reducing the error significantly even with a minimal explicit loop basis.
Spatial Structure: The BS radius analysis shows that in the strong-electric limit ( $x=0.8$ ), the excitation size is close to one elementary loop ( $R_{BS}/r_0 \approx 1.06$ ). As the coupling decreases toward the critical point, the radius grows (reaching $\approx 2.04$ at $x=0.4$ ). The results indicate that EC captures much of this spatial broadening without requiring a large explicit loop basis. However, near $x \approx 0.35$ , the radius plateaus due to finite-size saturation on the $6 \times 4$ lattice, marking the limit where the localized particle picture begins to break down.
Quench Dynamics: Simulations of quenches from $x=1$ to $x=0.1$ and $x=0.6$ show that deep quenches (to the magnetic-dominated regime) generate substantial weight in larger loop motifs, whereas shallow quenches (to $x=0.6$ ) remain dominated by electric-field configurations with suppressed extended loops.

Significance and Claims
The paper claims that its primary contribution is a transferable algorithmic structure rather than a claim that the $Z_2$ model contains QCD gluons. The framework is built from gauge-invariant vacuum and excitation structures, making it naturally applicable to non-Abelian lattice gauge theories.

The author emphasizes that the method successfully prepares and resolves a well-defined class of "glueball-like" pure-gauge states by:

Organizing computation around physical structures (closed flux) rather than generic search spaces.
Using EC to compress the parameter-dependent dressing of excitations, mitigating the basis-size explosion near criticality.
Providing a direct measure of spatial extent via BS-inspired transition amplitudes, rather than inferring size from basis coefficients.

The work identifies a "favorable regime" where the low-lying states have support only on loop operators of bounded size, allowing for a polynomial-dimensional generalized eigenproblem. The author notes that the method may fail near criticality or for highly excited states where the relevant loop size grows extensively with the system size, causing the excitation to lose its particle-like character. The $Z_2$ model serves as a minimal laboratory to isolate and verify the preparation, compression, and structural observation components before transferring the architecture to non-Abelian digitizations (e.g., $SU(2)$ , $SU(3)$ ).

Technical Summary: Physics-Informed Quantum Algorithms for Glueball-like Excitations in a Z2 Lattice Gauge Theory

More like this