⚛️ quantum physics

Scalable Ground-State Certification of Quantum Spin Systems via Structured Noncommutative Polynomial Optimization

This paper demonstrates that leveraging the inherent structures of quantum spin systems within a noncommutative polynomial optimization framework significantly mitigates the scalability limitations of semidefinite programming relaxations, enabling the computation of rigorous ground-state bounds for systems as large as $16\times16$ square lattices.

Original authors: Jie Wang, David Jansen, Irénée Frerot, Marc-Olivier Renou, Victor Magron, Antonio Acín

Published 2026-04-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jie Wang, David Jansen, Irénée Frerot, Marc-Olivier Renou, Victor Magron, Antonio Acín

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

Imagine you are trying to find the absolute lowest point in a vast, foggy, and mountainous landscape. In the world of quantum physics, this "lowest point" is called the ground state, and it represents the most stable, lowest-energy configuration of a system made of many interacting particles (like tiny magnets called spins).

Knowing exactly where this lowest point is helps scientists understand everything from how superconductors work to why magnets stick to your fridge. However, finding this point is incredibly hard because the landscape is so complex that even the most powerful supercomputers struggle to map it out.

This paper presents a new, super-smart way to navigate this foggy landscape. Here is the breakdown of what they did, using simple analogies.

1. The Old Way: Guessing and Checking

Traditionally, scientists used a method called "variational calculation." Imagine you are trying to find the bottom of a valley, but you can only see a few feet in front of you. You guess a spot, check the height, and if it's lower, you move there. You keep doing this until you think you can't go any lower.

The Problem: You might get stuck in a small dip (a local minimum) and think you've found the bottom, when actually, there's a much deeper valley just over the next hill. You can never be 100% sure you found the true bottom.

2. The New Way: Building a Safety Net

The authors in this paper use a mathematical technique called Noncommutative Polynomial Optimization. Think of this not as guessing, but as building a safety net underneath the landscape.

Instead of trying to find the exact bottom, they calculate a "floor" that the ground state cannot go below.

If they can prove the ground state energy is higher than -0.5, and another method proves it's lower than -0.4, they know the answer is somewhere between -0.5 and -0.4.
This gives them certainty. They aren't just guessing; they are mathematically proving the limits.

3. The Big Problem: The Net is Too Heavy

The problem with this "safety net" method is that it is incredibly heavy. As the system gets bigger (more particles), the math required to build the net explodes in size. It's like trying to build a safety net for a single person; that's easy. But trying to build one for a stadium full of people? The net becomes so massive that no computer can lift it.

Previously, this method could only handle small systems (like a 10x10 grid of particles). The authors wanted to handle much larger systems (like a 16x16 grid), but the math was too heavy.

4. The Solution: Folding the Net (Exploiting Structure)

The genius of this paper is realizing that the landscape isn't random; it has patterns and symmetries.

Imagine you are folding a giant, messy piece of paper to fit it into a small envelope.

Symmetry: If the landscape looks the same if you rotate it 90 degrees, you don't need to calculate the whole thing. You can calculate one quarter and just copy the result.
Sparsity: Most particles only talk to their immediate neighbors, not the ones across the room. The authors realized they could ignore the "long-distance" chatter in their math, drastically shrinking the size of the problem.
Sign Symmetries: They found that flipping the "colors" (positive/negative) of the particles in specific ways didn't change the physics. This allowed them to fold the math problem in half, then in half again.

By recognizing these patterns (which they call algebraic structures), they were able to fold that massive, heavy safety net down into something small enough for a standard computer to carry.

5. The Results: A Bigger, Clearer Picture

Because they folded the net so efficiently, they achieved two major things:

Bigger Systems: They successfully calculated bounds for a 16x16 grid of spins. Previous methods could only manage a 10x10 grid. That's a huge jump in scale.
Tighter Bounds: Their "safety net" is much closer to the actual ground than before. The gap between their "floor" and the "ceiling" (the best guesses from other methods) is tiny.

Why Does This Matter?

Think of this as upgrading from a blurry, low-resolution map of a city to a high-definition, 3D satellite view.

For Physicists: It gives them a way to verify if their other computer simulations are correct. If a simulation says the energy is lower than this paper's "floor," the simulation is wrong.
For the Future: It proves that we can use rigorous math to certify the behavior of complex quantum systems without needing a quantum computer to do the heavy lifting. It's a powerful tool for understanding the fundamental rules of our universe, from magnets to potential future quantum computers.

In short: The authors took a math problem that was too heavy to lift, realized the weight was distributed in a pattern, folded the pattern up, and suddenly, they could lift it and see much further than anyone else before.

1. Problem Statement

Determining the ground-state properties (energy and observable expectation values) of quantum many-body systems is a fundamental challenge in condensed matter physics.

Limitations of Existing Methods:
- Variational Methods (DMRG, PEPS, VMC, NNQS): These provide upper bounds on the ground-state energy but lack rigorous lower bounds. Their accuracy is limited by the expressiveness of the chosen wavefunction ansatz.
- Exact Diagonalization: Provides exact results but is computationally restricted to very small systems (typically $N < 50$ spins).
- Quantum Monte Carlo (QMC): Often suffers from the "sign problem" in frustrated or fermionic systems.
- Standard SDP Relaxations (NPA Hierarchy): While the Navascués-Pironio-Acín (NPA) hierarchy, based on noncommutative polynomial optimization, provides rigorous lower bounds on energy and bounds on observables, it suffers from severe scalability issues. The size of the resulting Semidefinite Programming (SDP) problems grows exponentially, limiting previous applications to small-to-medium systems (e.g., $10 \times 10$ lattices).

2. Methodology

The authors propose a framework that leverages the inherent algebraic and geometric structures of quantum spin systems to drastically reduce the size of the SDP relaxations while maintaining or improving accuracy.

A. Noncommutative Polynomial Optimization Framework

The ground-state energy $E_{GS}$ is formulated as a noncommutative polynomial optimization problem. The Hamiltonian $H$ is treated as a polynomial in Pauli matrices $\sigma^a_i$ . The problem is relaxed into a hierarchy of SDPs (the NPA hierarchy) where the constraint that the Hamiltonian is non-negative ( $H - \lambda \succeq 0$ ) is replaced by a Sum of Hermitian Squares (SOHS) decomposition.

B. Structural Exploitation for Scalability

The core innovation lies in reducing the dimension of the moment matrix $M_d(\ell)$ by exploiting specific symmetries and sparsity:

Sparsity (Local Monomial Basis): Instead of using a full basis of all monomials up to degree $d$ , the authors construct a sparse monomial basis containing only monomials supported on contiguous sites (nearest neighbors and short-range correlations). This significantly reduces the number of variables.
Sign Symmetries:
- Model Symmetry: The Heisenberg model is invariant under global sign flips of Pauli matrices (e.g., $\sigma_x \to -\sigma_x, \sigma_y \to -\sigma_y$ ). This allows the moment matrix to be block-diagonalized into four independent blocks based on the signature of the monomials.
- Hamiltonian Symmetry: Additional sign symmetries specific to the Hamiltonian further partition these blocks based on the parity of the monomial degree.
Conjugate Symmetry: Since the Hamiltonian is real, the linear functional $\ell$ can be restricted to real numbers, converting complex SDPs into real SDPs, halving the computational cost.
Translation Symmetry (Periodic Boundary Conditions): By exploiting the invariance under lattice translations, the moment matrix blocks become circulant matrices. These can be block-diagonalized via Discrete Fourier Transform (DFT), reducing the problem size from $O(L)$ to $O(1)$ per block (where $L$ is the lattice size). In 2D, this is applied twice (horizontal and vertical).
Permutation Symmetry: Invariance under the permutation of $x, y, z$ axes allows the merging of identical blocks, further reducing the number of constraints.
Mirror/Dihedral Symmetry: Reflection symmetries of the lattice are used to further constrain the solution space.

C. Strengthening the Relaxations

To improve the tightness of the bounds, the authors incorporate additional constraints:

Reduced Density Matrix (RDM) Positivity: Enforcing that reduced density matrices (up to $k$ -body) are Positive Semidefinite (PSD). Due to $U(1)$ symmetry (conservation of total magnetization), these matrices are also block-diagonalized.
State Optimality Conditions: Incorporating conditions derived from the fact that the ground state minimizes the energy (e.g., $\ell([H, f]) = 0$ and specific PSD constraints on commutators).

3. Key Contributions

Algorithmic Breakthrough: Demonstrated that fully exploiting algebraic structures (sign, conjugate, permutation, translation, and dihedral symmetries) can reduce the SDP size by orders of magnitude compared to standard approaches.
Scalability: Achieved ground-state certification for $16 \times 16$ square lattices ( $N=256$ spins), a significant leap from the previous limit of $10 \times 10$ ( $N=100$ ) in prior work [31].
Accuracy: Obtained significantly tighter bounds than previous SDP methods and competitive with variational methods (DMRG, NNQS) for 1D and 2D systems.
Dual Bounds: Successfully provided both lower bounds on energy and rigorous upper/lower bounds on observable expectation values (e.g., spin correlations, structure factors).

4. Numerical Results

The method was tested on four Heisenberg models:

Heisenberg Chain (1D):
- For $N=100$ , the new SDP lower bounds are within $0.0019\%$ of DMRG upper bounds, a massive improvement over the $0.0820\%$ gap of previous SDP methods.
J1-J2 Heisenberg Chain (1D, Frustrated):
- Improved bounds on ground-state energy and nearest/next-nearest neighbor correlations ( $C(1), C(2)$ ) across various frustration parameters ( $J_2$ ).
- Tight bounds on the structure factor $S(\pi, \pi)$ .
Square Lattice Heisenberg Model (2D):
- Scaled up to $16 \times 16$ lattices.
- For $L=16$ , the new lower bound is $-0.674580$, compared to the QMC upper bound of $-0.669976$ (gap $\approx 0.69\%$ ), significantly tighter than previous SDP results.
- Bounds on long-range spin correlations were computed, though the gap widens for highly non-local observables due to the local nature of the monomial basis.
Square Lattice J1-J2 Model (2D, Frustrated):
- For $10 \times 10$ lattices, the new method reduced the relative gap to DMRG/NNQS bounds by roughly half compared to previous SDP results (e.g., at $J_2=1.0$ , gap reduced from $4.08\%$ to $3.06\%$ ).

5. Significance

Rigorous Certification: This work provides a scalable, rigorous method to certify ground-state properties, offering a "gold standard" against which approximate variational methods (like DMRG or Neural Networks) can be benchmarked.
Overcoming the Sign Problem: Unlike QMC, this approach does not suffer from the sign problem in frustrated systems, making it a powerful tool for studying quantum spin liquids and other complex phases.
Bridging the Gap: By pushing the limits of SDP relaxations to $16 \times 16$ lattices, the method bridges the gap between small-system exact methods and large-system approximate methods, providing high-accuracy data for the thermodynamic limit extrapolation.
Future Directions: The authors suggest combining this method with DMRG (using coarse-graining maps) and utilizing the full $SU(2)$ symmetry for further reductions. They also propose using reinforcement learning to optimize the choice of monomial bases.

In summary, the paper establishes that structured noncommutative polynomial optimization is a viable and powerful tool for the rigorous study of large-scale quantum many-body systems, overcoming the historical scalability bottlenecks of SDP relaxations.