A Hierarchy of Spectral Gap Certificates for Frustration-Free Spin Systems

This paper introduces a general hierarchy of semidefinite programming optimization problems that provides rigorous, increasingly tight lower bounds on the spectral gaps of frustration-free quantum Hamiltonians in the thermodynamic limit, significantly outperforming existing finite-size methods like Knabe's bound in both accuracy and parameter range.

The Gist

Imagine you are trying to figure out how "stiff" a giant, invisible spring is. In the world of quantum physics, this "spring" is a system of particles (like atoms in a chain) interacting with each other. Physicists want to know if there is a spectral gap.

Think of the spectral gap as the minimum amount of energy required to "jiggle" the system out of its calm, resting state (the ground state) into a slightly excited state.

• Gapped: The system is stiff. You need a specific, non-zero amount of energy to make it move. It's like a heavy door that needs a firm push to open.
• Gapless: The system is floppy. You can jiggle it with almost zero energy. It's like a door on a broken hinge that swings with a breeze.

Knowing if a system is gapped or gapless is crucial. It tells us if the material is a conductor or an insulator, how fast a quantum computer can run, and how information travels through the system.

The Problem: The "Infinite" Puzzle

The problem is that these quantum systems are theoretically infinite. You can't build an infinite chain of atoms in a lab, and you can't simulate an infinite one on a computer (it would take forever).

For decades, scientists have used "finite-size criteria" to guess the answer.

• The Old Way (The "Knabe" Method): Imagine you want to know if a massive, infinite wall is made of solid brick or just paper. The old method says: "Let's build a small 10-foot section of the wall. If that section is solid, we assume the whole infinite wall is solid."
• The Flaw: Sometimes, a small section looks solid, but the infinite wall has a hidden weak spot further down. The old methods often gave very loose estimates (e.g., "It's definitely stiff, maybe 0.2 units of stiffness") or failed to detect the gap entirely in tricky situations. They were like trying to guess the weight of a whale by weighing a single fish.

The New Solution: A "Smart Ladder"

The authors of this paper, led by Kshiti Sneh Rai and Flavio Baccari, have built a hierarchy of spectral gap certificates. Think of this as a smart ladder where each rung gives you a better, more accurate answer.

Here is how their method works, using a simple analogy:

1. The "Positive Sum" Rule

The core idea relies on a mathematical trick. To prove a system is stiff (gapped), you have to show that a specific mathematical equation (involving the energy of the system) is always "positive" (like saying a balance scale never tips the wrong way).

• The Challenge: Checking this for an infinite system is impossible.
• The Trick: Instead of checking the whole infinite system, they break the problem down into small, manageable chunks (like looking at 3 atoms, then 4, then 5).

2. The Optimization Ladder

The authors created a computer program that climbs a ladder of complexity:

• Rung 1 (Level n=3): The computer looks at groups of 3 atoms. It tries to find a proof that the system is stiff based only on how those 3 atoms interact. If it finds a proof, it gives a lower bound (a "minimum guarantee") for the stiffness.
• Rung 2 (Level n=4): The computer looks at groups of 4 atoms. It has more information, so it can find a tighter, better proof.
• Rung 3 (Level n=5): It looks at 5 atoms, and so on.

As you climb higher up the ladder (increasing the number of atoms in the local group), the "guarantee" of stiffness gets closer and closer to the true answer.

3. Why It's Better

The old methods were like trying to solve a puzzle by only looking at specific, pre-chosen pieces. The authors' method is like having a robot that tries every possible way to assemble those pieces to find the strongest proof.

• It includes the old methods: The old methods are actually just "special cases" of this new method. If the old method found a gap, this new method will find it too, but usually with a much stronger number.
• It's flexible: The old methods often required the atoms to interact in very specific, rigid ways (mathematically called "projectors"). This new method works even when the interactions are messy or complex.

Real-World Results: The "AKLT" Chain

To test their ladder, they looked at a famous quantum model called the AKLT chain (a specific type of magnetic spin chain).

• The Old Record: For years, the best proof said the gap was at least 0.248.
• The New Result: Using their ladder (specifically looking at groups of 6 atoms), they proved the gap is at least 0.349.
• The Reality: The actual gap is known to be around 0.350.
• The Takeaway: Their method got incredibly close to the true answer, far surpassing the old methods.

They also tested it on other models where the system is on the verge of becoming "floppy" (critical). The old methods said "We can't tell if it's stiff or floppy," while their method successfully proved it was still stiff, even in regions where other methods failed.

The Bottom Line

This paper introduces a powerful new tool for quantum physicists. Instead of guessing the properties of infinite quantum systems based on small, imperfect samples, they now have a systematic, automated way to get better and better proofs the more computing power they throw at it.

It's like upgrading from a blurry, low-resolution map to a high-definition satellite image. You can finally see the terrain clearly, proving that the "door" to the quantum world is indeed stiff, and telling you exactly how hard you need to push to open it.

Technical Summary

1. Problem Statement

The central challenge addressed is the rigorous estimation of the spectral gap (the energy difference between the ground state and the first excited state) of quantum many-body Hamiltonians in the thermodynamic limit ($N \to \infty$).

• Difficulty: Determining if a system is gapped or gapless is computationally hard and, in general dimensions, undecidable.
• Context: The authors focus on frustration-free Hamiltonians, where the global ground state minimizes every local interaction term individually.
• Limitations of Existing Methods:
  • Martingale Methods: Powerful analytically but often fail to provide tight numerical lower bounds.
  • Finite-Size Criteria (e.g., Knabe, Gosset–Mozgunov): These relate the infinite-system gap to the gap of a finite open-boundary system. While computationally feasible, they often yield loose lower bounds and require the interaction terms to be projectors (or involve approximations that lose spectral information).

2. Methodology

The authors propose a hierarchy of Semidefinite Programs (SDPs) to systematically search for gap certificates. The method generalizes existing finite-size criteria by automating the search for optimal decompositions of a specific operator inequality.

Core Mathematical Formulation

A frustration-free Hamiltonian $H$ has a spectral gap $\Delta \ge \delta$ if and only if the operator $H^2 - \delta H$ is positive semidefinite (PSD):
$$H^2 - \delta H \succeq 0$$
The goal is to maximize $\delta$ subject to this constraint. Since solving this exactly for an infinite system is intractable, the authors introduce two relaxation steps:

1. Geometric Locality Truncation:
   The operator $H^2$ contains non-local terms (products of interactions far apart). The authors define a truncated operator $[H^2]_n$ that includes only terms acting on sites within a distance $n$. Since the discarded terms are PSD (products of disjoint local terms), proving $[H^2]_n - \delta H \succeq 0$ is a sufficient condition for the original inequality, providing a valid lower bound.

2. Local Decomposition (The SDP):
   Instead of checking the global positivity of the truncated operator, the method searches for a local generating term $q_n$ (supported on $n$ sites) such that:
   $$[H^2]_n - \delta H = \sum_i q_{n,i}$$
   where $q_{n,i} \succeq 0$.

   • Translation Invariance: The problem is formulated under the assumption of translation invariance.
   • Degrees of Freedom: The authors characterize the non-uniqueness of the local generating term. If $g_n$ is a local term summing to the global operator, any other valid local term $q_n$ can be written as $q_n = g_n + \mathbb{1} \otimes Y_{n-1} - Y_{n-1} \otimes \mathbb{1}$, where $Y_{n-1}$ is an arbitrary operator on $n-1$ sites. The optimization searches over all possible $Y_{n-1}$ to find the one that makes $q_n$ PSD.

The Hierarchy

The method forms a hierarchy labeled by $n$ (the size of the local support):

• Level $n$: An SDP optimizing $\delta$ over $n$-body local terms.
• Convergence: As $n$ increases, the lower bound $\delta_{LTI}(n)$ monotonically increases:
  $$\delta_{LTI}(n) \le \delta_{LTI}(n+1) \le \dots \le \Delta_{\infty}$$
• Dual Formulation: The dual of this SDP corresponds to minimizing the energy of the first excited state over reduced density matrices (marginals) that satisfy local translation invariance.

3. Key Contributions

1. Generalization of Finite-Size Criteria: The authors prove that existing bounds (Knabe, Gosset–Mozgunov, and Martingale bounds) are feasible but not necessarily optimal solutions within their SDP hierarchy. Consequently, the proposed method is guaranteed to match or strictly outperform these existing methods for the same computational scale.
2. Removal of Projector Assumption: Unlike many finite-size criteria that require interaction terms to be projectors (which can degrade the bound), the LTI-SDP works directly with the original Hamiltonian terms, preserving more spectral information.
3. Algorithmic Automation: The method automates the search for the "appropriate" finite-size criterion, removing the need for manual, case-specific decomposition strategies.
4. Numerical Certification: The authors address the issue of numerical precision in SDP solvers, providing a refined approach to convert numerical solutions into rigorous mathematical proofs of a gap.

4. Numerical Results

The method was benchmarked on three 1D frustration-free models:

• AKLT Model (Spin-1 Chain):

  • The method achieved a lower bound of $\Delta \ge 0.34976$ (for $n=6$).
  • This is significantly tighter than the Knabe bound ($\approx 0.248$) and the Gosset–Mozgunov bound.
  • The result is extremely close to the exact diagonalization value ($\approx 0.350$), demonstrating near-optimal performance.

• Deformed Clock Models (Z3 Potts):

  • The method detected a gap in a significantly larger parameter region $(r, s)$ compared to Knabe and Gosset–Mozgunov bounds.
  • Specifically, the LTI-SDP with $n=5$ detected gaps in regions where the existing methods required $n=10$ (a much higher computational cost) to detect anything, or failed to detect a gap entirely.

• 1D Glauber Model:

  • This model approaches criticality at $\gamma = 1$.
  • Existing finite-size criteria could only detect a gap for $\gamma \lesssim 0.96$.
  • The LTI-SDP detected a gap orders of magnitude closer to the critical point ($\gamma \to 1$), demonstrating superior sensitivity near phase transitions.

5. Significance and Outlook

• Impact on Many-Body Physics: The method provides a powerful, systematic tool for rigorously proving the existence of spectral gaps, which is crucial for understanding quantum phases, entanglement scaling, and adiabatic state preparation.
• Superiority over Existing Tools: It represents a strict improvement over the "gold standard" finite-size criteria, offering both higher accuracy and broader applicability (e.g., non-projector Hamiltonians).
• Scalability: While currently demonstrated on 1D systems, the authors argue the method generalizes straightforwardly to higher dimensions. The main bottleneck is the computational cost of solving large SDPs, but the authors suggest that the required system size $n$ for convergence in 2D may be smaller than the system sizes required by traditional finite-size criteria.
• Open Questions: The completeness of the hierarchy (whether it converges to the exact gap for all gapped frustration-free systems) remains an open question, particularly in 2D where the spectral gap problem is undecidable. However, for 1D frustration-free systems, the hierarchy is expected to be complete.

In summary, this paper presents a rigorous, optimization-based framework that supersedes previous analytical and numerical techniques for certifying spectral gaps, offering unprecedented accuracy and range in detecting gapped phases in quantum spin systems.