Good LTC Does Not Imply Balanced Cut Rank¶
Claim/Theorem¶
Even asymptotically good locally testable codes do not, by themselves, force linear balanced-cut rank-connectivity.
Let \(A_N\) and \(B_N\) be two families of linear codes of equal blocklength \(N\) such that each family is:
- asymptotically good, and
- locally testable with constant query complexity and constant soundness.
Form the direct-sum family
\[
C_n \;=\; A_N \oplus B_N
\]
on disjoint coordinate sets, with total length \(n=2N\).
Then:
- \(C_n\) remains asymptotically good, because
\[
k(C_n)=k(A_N)+k(B_N)=\Theta(n),
\qquad
d(C_n)=\min(d(A_N),d(B_N))=\Theta(n).
\]
- \(C_n\) remains locally testable with constant query complexity and constant soundness: choose one component uniformly at random and run its local tester on that component. The rejection probability is the average of the two component rejection probabilities, hence still lower-bounds distance to \(C_n\) up to a constant factor.
- For the balanced partition \(L\) consisting of the coordinates of the first summand, [[cross-cut-stabilizer-rank-rank-formula.md]] gives
\[
\lambda_{C_n}(L)=0.
\]
Equivalently, for the associated stabilizer space,
\[
\chi_L(\mathcal S_n)=0.
\]
Therefore local testability does not fix the balanced-cut connectivity problem left open by [[good-code-parameters-do-not-imply-cut-rank.md]]. Any successful route from local testability to Conjecture 3 must use more than LTC soundness alone, for example some irreducibility, connectedness, or expansion property that rules out direct-sum style decompositions.
Dependencies¶
- [[cross-cut-stabilizer-rank-rank-formula.md]]
- [[good-code-parameters-do-not-imply-cut-rank.md]]
- [[quantum-tanner-constituent-ltc.md]]
Conflicts/Gaps¶
- This is again a disconnected counterexample. It does not rule out the possibility that a connected expander-style LTC family must have large balanced-cut rank-connectivity.
- The node shows that local testability alone is insufficient, but it does not identify the extra hypothesis that would make the implication true.
- The example concerns classical codes. Translating any eventual positive theorem back to the quantum stabilizer-space frontier will still require additional work.
Sources¶
10.48550/arXiv.2202.13641