Confinement-Higgs transition in a disordered gauge theory and the accuracy threshold for quantum memory

I. INTRODUCTION

• Quantum-memory accuracy thresholds map to confinement-Higgs transitions in disordered Z_2 gauge models, with errors below p_c protectable in the large-block limit.The three-dimensional model incorporates qubit errors and noisy syndrome measurements; perfect measurements reduce it to the two-dimensional RBIM.Source: §I · pp. 1–2
• The critical concentration pc0 = .0293 ± .0002 on the T = 0 axis provides a lower bound on the accuracy threshold pc in the three-dimensional Z2 gauge model.An earlier analytic argument established that pc0 ≥ .0114.Source: §I · p. 2
• The 2D RBIM T=0 calculation finds p_c0=.1031 ± .0001, agreeing with Kawashima and Rieger but disagreeing with the conjecture p_c=p_c0.Earlier work reported p_c=.1093±.0002 on the Nishimori line.Source: §I · p. 2

A. Random-bond Ising model

• The RBIM order parameter detects whether excited bonds can separate far from quenched antiferromagnetic bonds, producing arbitrarily large closed domain-wall loops.The cycle D=E+E′ is formed from the two chains and controls long-distance spin correlations.Source: A. Random-bond Ising model · p. 3
• Along the Nishimori line K=K_p, antiferromagnetic and excited-bond chains share a probability distribution, and thermal and quenched disorder have equal strength.The spin-glass order parameter coincides with the ferromagnetic order parameter on this line.Source: A. Random-bond Ising model · p. 4
• The conjectured vertical phase boundary below the Nishimori point lacks rigorous justification, while a reentrant boundary remains possible.The Nishimori-line entropy singularity motivates verticality but does not establish that the phase boundary must coincide with it.Source: A. Random-bond Ising model · pp. 4–5, Figure 2 · p. 4

B. Random-plaquette gauge model

• In the RPGM, wrong-sign plaquettes form quenched Dirac strings bounded by magnetic monopoles, while excited plaquettes form flux-tube chains with the same boundaries.Their sum is a closed cycle whose large loops represent magnetic disorder.Source: B. Random-plaquette gauge model · pp. 5–6
• The Wilson loop obeys perimeter-law decay in the Higgs phase and area-law decay in the confinement phase.The distinction reflects whether thermally excited flux tubes remain near or wander far from the quenched wrong-sign plaquettes.Source: B. Random-plaquette gauge model · pp. 5–6
• On the Nishimori line K=K_p, the RPGM has enhanced symmetry, and its free energy equals the Shannon entropy of magnetic-monopole distributions up to a nonsingular term.The entropy is singular at p=p_c.Source: B. Random-plaquette gauge model · p. 6
• The RPGM has no gauge-glass phase on the Nishimori line because averaged Wilson-loop moments cross from perimeter to area law at the same point.The equality of the relevant disorder averages prevents distinct typical-sample and sample-to-sample transitions there.Source: B. Random-plaquette gauge model · p. 6
• If a Nishimori-line point is confined, the corresponding point at any temperature is also confined, so the sub-Nishimori phase boundary is either vertical or reentrant.This conclusion follows from the stated inequality and the absence of gauge-glass behavior on the Nishimori line.Source: B. Random-plaquette gauge model · p. 6

C. Further generalizations

• In d dimensions, RBIM order probes codimension-1 excited chains deviating from quenched chains, whereas RPGM order probes codimension-2 flux-tube chains with shared codimension-3 boundaries.The corresponding closed objects are domain-wall cycles for RBIM and flux-tube cycles for RPGM.Source: C. Further generalizations · pp. 6–7
• For rank-r antisymmetric tensor fields, disorder and excitations occupy codimension-(r+1) chains bounded by common codimension-(r+2) frustration chains.A Wilson-loop analogue detects whether thermally fluctuating chains depart far from quenched chains.Source: C. Further generalizations · p. 7
• These generalized models retain enhanced symmetry on the Nishimori line K=K_p, allowing analogous properties to be derived.Source: C. Further generalizations · p. 7

III. ACCURACY THRESHOLD FOR QUANTUM MEMORY

• RBIM and RPGM phase behavior provides a framework for relating disordered spin and gauge systems to the performance of topological quantum memory.The paper reprises the main ideas of this connection from prior work.Source: §III · p. 7

A. Toric codes

• The toric code uses local site and plaquette check operators to define a four-dimensional code space encoding two qubits.There are 2·(L^2−1) independent checks constraining 2L^2 physical qubits.Source: A. Toric codes · pp. 7–8
• Z chains create electric-charge defects at their endpoints, while X chains on the dual lattice create magnetic-flux defects at their endpoints.These defects correspond to check operators with eigenvalue −1.Source: A. Toric codes · p. 8, Figure 4 · p. 8
• Homologically trivial cycles are products of check operators and act trivially, whereas nontrivial torus cycles implement encoded operations on the protected qubits.The two fundamental lattice cycles define encoded Z operations, and the dual cycles define encoded X operations.Source: A. Toric codes · p. 8, Figure 5 · p. 8
• Bit-flip and phase-flip errors can be corrected independently, so analyzing Z errors suffices because Y errors combine one of each.The dual-lattice construction handles X errors analogously.Source: A. Toric codes · p. 8

B. Perfect measurements and the random-bond Ising model

• Recovery succeeds when the inferred chain differs from the actual error chain by a homologically trivial cycle; a nontrivial cycle corrupts encoded information.The syndrome determines only the boundary of the error chain, leaving its homology class to infer.Source: B. Perfect measurements and the… · p. 9
• The accuracy threshold is the error rate below which failure probability approaches zero for large code size, while above it failure approaches a nonzero constant.This threshold quantifies the hardware reliability required for robust quantum memory.Source: B. Perfect measurements and the… · p. 9
• The toric-code recovery problem maps to the two-dimensional RBIM along the Nishimori line, with error chains represented by antiferromagnetic bonds and candidate recoveries by thermal configurations.The cycle formed by the actual and candidate chains is ordered below the transition and develops large loops at the Nishimori point.Source: B. Perfect measurements and the… · p. 9
• Choosing the most likely recovery chain minimizes energy and succeeds for p < pc0, with pc0 ≤ pc because energy minimization need not be optimal.The energy-minimization procedure is efficiently implementable using Edmonds perfect matching and requires no prior knowledge of p.Source: B. Perfect measurements and the… · p. 9

C. Imperfect measurement and the random-plaquette gauge model

• Measurement errors become timelike bonds while qubit errors become spacelike bonds, producing a three-dimensional error chain whose boundary matches the syndrome chain.Horizontal bonds occur with probability p, vertical bonds with probability q, and both error types are represented in spacetime.Source: C. Imperfect measurement and the… · pp. 9–10, Figure 6 · p. 10
• Recovery succeeds when the inferred error chain differs from the actual chain by a homologically trivial cycle on T^2 × R.This extends the perfect-measurement homology criterion to faulty measurement histories.Source: C. Imperfect measurement and the… · p. 10
• The RPGM is anisotropic when qubit and measurement error rates differ, with spacelike and timelike couplings set by their respective Nishimori relations.The spacelike coupling obeys Kspace = Kq, while the timelike coupling obeys Ktime = Kp.Source: C. Imperfect measurement and the… · p. 10
• For small p and q, flux loops remain suppressed and recovery succeeds with high probability; at a critical curve, flux tubes condense and disorder the gauge system.In the isotropic case p = q, the Nishimori-point concentration equals the toric-code accuracy threshold for X, Z, and measurement errors.Source: C. Imperfect measurement and the… · p. 10
• The T = 0 critical concentration pc0 provides a lower bound on the full threshold pc for the three-dimensional RPGM.The paper cites the rigorous bound pc0 ≥ .0114 and estimates pc0 numerically.Source: C. Imperfect measurement and the… · p. 10

A. Method

• The RBIM ground-state minimization is polynomial-time via Edmonds perfect matching, enabling direct T = 0 simulations and estimates of pc0 and ν0.The same strategy applies to the three-dimensional RPGM, where the disorder defines a dual one-chain bounded by magnetic monopoles.Source: A. Method · p. 11
• The failure probability is expected to approach zero below pc0 and 3/4 above pc0 for the RBIM, motivating finite-size scaling in L/ξ.The crossing of failure-probability curves estimates pc0, while slope scaling extracts ν0.Source: A. Method · p. 11
• Finite-size corrections and detectable scaling violations require a modified ansatz, with even-L data judged more reliable than odd-L data.Odd sizes show enhanced finite-size effects because their failure probability is smaller.Source: A. Method · p. 11
• For each disorder sample, the ground state is obtained by finding a minimal chain with the same boundary as the wrong-sign disorder, then testing whether the resulting cycle is homologically trivial.For the RPGM, eight homology classes occur on the three-torus.Source: A. Method · p. 11
• In the three-dimensional RBIM and four-dimensional RPGM, ground-state construction requires minimal surfaces with fixed boundaries, an NP-hard problem that appears computationally intractable.This limits the direct numerical approach to the lower-dimensional cases studied here.Source: A. Method · p. 11

Numerical analysis

• The failure probability was sampled over p and multiple linear sizes, with even and odd sizes analyzed separately because even-L values lie systematically higher.Each plotted data point averages 10^6 samples.Source: B. Random-bond Ising model · p. 11, Figure 7 · p. 12, Figure 8 · p. 12

Scaling analysis

• After subtracting the nonuniversal correction, data for L from 2 to 42 collapse onto a single scaling line with only small residual violations.Agreement between independently fitted even and odd datasets supports the large-L extrapolation and the reported statistical errors.Source: Figure 9 · p. 12, B. Random-bond Ising model · pp. 12–13

Numerical analysis

• pc0 = .10309 ± .00003 from the even-size fit, while independent even and odd analyses reveal a nonnegligible finite-size effect.The modified scaling ansatz includes an even/odd nonuniversal correction to Pfail.Source: B. Random-bond Ising model · p. 12
• ν0 = 1.463 ± .006, with Dodd = −.053 ± .003 and µodd = 2.1 ± .3 from odd-size data.Source: B. Random-bond Ising model · p. 12

Interpretation

• The measured pc0 disagrees with the Nishimori-point pc, contrary to the Nishimori–Kitatani conjecture that the low-temperature phase boundary is vertical.The paper also notes evidence for a reentrant phase diagram and consistency between ν0 and the Nishimori-point exponent ν.Source: B. Random-bond Ising model · p. 13

• Failure-probability data were generated across L = 9–16 and p values near .029, then analyzed separately for even and odd L because even-L values were systematically higher.Source: C. Random-plaquette gauge model · p. 13
• The failure probability Pfail was measured for linear sizes L = 10, 12, 14, 16 using 10^6 samples per data point.Source: Figure 10 · p. 13
• pc0 = .02937 ± .00002 and ν0 = 0.974 ± .026 for even L, while odd L gives pc0 = .02900 ± .00001 and ν0 = 1.025 ± .016.Source: C. Random-plaquette gauge model · p. 13
• The even- and odd-L fit results are incompatible, indicating a nonnegligible finite-size effect.Source: C. Random-plaquette gauge model · p. 13
• The even-L data collapse onto a single curve when plotted against x = (p − pc0)L1/ν0, indicating small scaling violations.Source: Figure 11 · p. 13, C. Random-plaquette gauge model · p. 14

• At zero temperature, the RBIM failure test constructs a minimum-energy chain with the same boundary, and failure occurs when the resulting cycle is homologically nontrivial.Source: D. The failure probability at finite… · p. 14
• At nonzero temperature, success is determined by whether the trivial homology class has the lowest free energy; otherwise the sample is a failure.Source: D. The failure probability at finite… · p. 14
• The partition-function ratio for each nontrivial homology class is a thermal expectation value of a domain-wall-inserting observable around a representative cycle.Source: D. The failure probability at finite… · p. 14
• Direct Monte Carlo can fail because rare configurations with excited bonds may dominate the expectation value despite Boltzmann suppression.Source: D. The failure probability at finite… · p. 14
• To make Monte Carlo estimates tractable, Z(E,h)/Z(E,e) is factorized along chains that add one bond at a time, so each ratio becomes a single-bond expectation value.Source: D. The failure probability at finite… · p. 14
• The method can estimate Pfail(p,T) for the RBIM and, in principle, along the Nishimori relation, but the analogous RPGM calculation was not attempted.Source: D. The failure probability at finite… · p. 14

• The RBIM phase boundary is reentrant rather than vertical below the Nishimori line, with long-range order restored at low nonzero temperature for pc0 < p < pc before disappearing again.Source: §V · pp. 14–15
• For pc0 < p < pc, the ground state can remain outside the most likely error homology class with probability bounded away from zero as L →∞.Source: §V · p. 15
• The failure probability is a homological observable with a simple finite-size-scaling form, enabling accurate estimates of pc0 and ν0 using relatively modest computational resources.Source: §V · p. 15
• The measured zero-temperature thresholds are pc0 ≈ .1031 for the RBIM and pc0 ≈ .0293 for the RPGM, above their rigorous lower bounds.Source: §V · p. 15
• The RPGM result implies arbitrarily high-fidelity storage when independently distributed qubit and syndrome-measurement error probabilities are each below 2.9%.Source: §V · p. 15
• The critical exponent is ν0 ≈ 1.46 in the RBIM and ν0 ≈ 1.0 in the RPGM, and ν0 determines code-block size for a target fidelity near pc0.Source: §V · p. 15
• The paper presents quantum-memory threshold analysis as a statistical-physics approach that also offers a perspective on local spin and gauge systems with strong quenched disorder.Source: §V · p. 15