Minimum LU–LC counterexample for graph states
- Field
- Topic
Problem
What is the least number of qubits for which two graph states can be locally unitary equivalent without being locally Clifford equivalent? For a simple graph \(G=(V,E)\) with \(V=\{1,\ldots,n\}\), define its graph state by
For graphs \(G\) and \(H\) on \(n\) vertices, use the state convention in Eq. (1) and write
where \(\mathcal C_1\) is the single-qubit Clifford group. Since \(\mathcal C_1\subset U(2)\), LC equivalence implies LU equivalence. Define the minimum counterexample size by
Determine the integer in Eq. (3).
Source
Krüger and Werner record the original conjecture that LU equivalence of graph states always implies LC equivalence [KW05]. After Ji, Chen, Wei, and Ying constructed a 27-qubit counterexample, Claudet isolated and resolved the minimum-size question in Eq. (3) [JCWY10], [Cla26].
Progress
Krüger and Werner posed the universal implication \(G\sim_{\mathrm{LU}}H\Rightarrow G\sim_{\mathrm{LC}}H\). This formulation did not include a minimum counterexample size [KW05].
Ji, Chen, Wei, and Ying disproved the universal implication by constructing LU-equivalent but non-LC-equivalent graph states on 27 qubits. Their construction proves \(n_{\min}\leq27\) but does not by itself exclude smaller counterexamples [JCWY10].
Claudet proved that LU and LC equivalence coincide for every pair of graph states on at most 26 qubits. Combining this lower bound with the 27-qubit construction gives
\begin{equation} n_{\min}=27. \tag{4} \end{equation}
Comment
Equation (4) concerns the minimum size at which LU and LC equivalence can differ; it does not say that 27 qubits are required for LU–LC equivalence. The original universal conjecture was already resolved negatively by the 27-qubit construction. As of September 2026, the matching lower bound through 26 qubits is contained in a recent arXiv v1 preprint, so the archived solved status records its theorem rather than its peer-review history.