Abstract
The LU-LC conjecture is an important open problem concerning the structure of entanglement of states described in the stabilizer formalism. It states that two local unitary equivalent stabilizer states are also local Clifford equivalent. If this conjecture were true, the local equivalence of stabilizer states would be extremely easy to characterize. Unfortunately, however, based on the recent progress made by Gross and Van den Nest, we find that the conjecture is false.
Introduction
Quantum information theory seeks to understand the fundamental nature of entanglement. A central question is determining when two quantum states are locally equivalent—that is, transformable into one another using only operations on individual qubits. For the class of stabilizer states, which are foundational to quantum error correction and one-way quantum computing, the LU-LC conjecture proposed a simplifying symmetry: that local unitary (LU) equivalence implies the more restrictive local Clifford (LC) equivalence. If true, this would mean the complex landscape of general unitaries could be reduced to the discrete, group-theoretic structure of the Clifford group for these states, making the characterization of entanglement computationally efficient.
Methodology and the QFP Reduction
The breakthrough by Zhengfeng Ji and colleagues relied on transforming the quantum problem into a classical algebraic one. They utilized a reduction proposed by Gross and Van den Nest, which maps the LU-LC conjecture onto the Quadratic Form Phase (QFP) problem. This problem concerns whether certain complex phases in a quadratic function over a finite field can be restricted to the set {±1, ±i}.
The authors represented the problem using a subspace $S \subseteq \mathbb{F}_2^n$ defined by basis vectors $\xi^1, \dots, \xi^d$.
By simplifying the QFP problem into a system of Linear Congruence Equations (LCE) modulo 4, they were able to apply linear algebraic tools. Their analysis showed that the conjecture holds for low-rank cases where the rank $d \le 5$. However, they hypothesized that a counterexample could exist at $d=6$. To find it, they developed a "backward approach" algorithm that randomly generated potential solutions to the LCE and then verified if they constituted a valid counterexample to the original quantum problem.
Key Results: The $n=27$ Counterexample
The search yielded a definitive counterexample involving $n=27$ qubits and a rank-6 subspace. The authors provided the exact basis vectors and quadratic forms for two stabilizer states that are LU equivalent but not LC equivalent. The phases required for the local unitary transformation are powers of $e^{\pi i / 4}$, which fall outside the allowed Clifford phases.
This discovery explains why previous numerical searches, which typically focused on smaller systems (up to 12 qubits), failed to find a discrepancy. The divergence between LU and LC equivalence only appears once the system reaches a sufficient complexity threshold at rank 6. Interestingly, the two graph states used in the counterexample differ by only a single edge in their graph representation, yet this minimal difference is sufficient to break LC equivalence while maintaining LU equivalence.
Conclusion and Future Directions
The formal disproof of the LU-LC conjecture reveals that the local entanglement properties of stabilizer states are more nuanced than previously thought. While the Clifford group captures much of their behavior, it does not fully encompass all local unitary symmetries.
Several questions remain. While $n=27$ is the first known counterexample, it is not yet proven to be the smallest; a counterexample might still exist in the range of 13 to 26 qubits. Furthermore, the fundamental intuition for why this divergence occurs at rank 6 remains elusive. Most importantly, now that the LU-LC conjecture is known to be false, the community faces a new challenge: developing an efficient algorithm to decide general LU equivalence for stabilizer states.
Summary updated