T20 — Z₂ Lattice Gauge Theory Monte Carlo

Confinement-deconfinement transition

Published

July 2, 2026

Modified

July 14, 2026

Last updated: 2026-07-14 13:48 IST

📊 Live Dashboard

View all simulation runs and figures: TimesArrow Numerics Dashboard →

Objective

Classical Monte Carlo simulation of Z₂ gauge theory to demonstrate the confinement-deconfinement transition that underpins the paper’s central claim.

Status

Phase Description Status Date
Phase 1 2D square lattice, critical β identification ✅ Complete 2026-06-24
Phase 2 Finite-size scaling (L = 8–24) ✅ Complete 2026-06-25
Phase 3 3D cubic lattice (L=4,6,8) ✅ Complete 2026-06-25
Phase 3b 3D FSS correction and reanalysis 🔄 In progress 2026-07-05
Phase 3b 3D Wilson loops & string tension (L=8) ✅ Complete 2026-06-26

Corrected status: The 3D pure Z₂ gauge transition is continuous and belongs to the 3D Ising universality class. Existing simulations resolve the established critical region near β ≈ 0.76, but the previous first-order interpretation and claimed precision determination are withdrawn pending a controlled reanalysis.

Correction notice

The plaquette Binder limit near \(2/3\) is not evidence for a first-order transition, and the earlier double-peaked histograms were synthetic illustrations rather than sampled distributions. The raw simulation data are retained; the superseded interpretation is not used in the manuscript.

Theory

The Z₂ gauge theory action:

\[ S = -\beta \sum_{\square} \prod_{e \in \square} \sigma_e \]

where \(\sigma_e \in \{+1, -1\}\) are link variables and \(\beta\) is the inverse coupling (temperature).

Observables

Observable Formula Purpose
Average plaquette \(\langle P \rangle = \frac{1}{N_{\square}} \sum_{\square} \prod_{e \in \square} \sigma_e\) Order parameter
Specific heat \(C_V = \beta^2(\langle P^2 \rangle - \langle P \rangle^2)\) Critical fluctuations
Susceptibility \(\chi = L^2(\langle P^2 \rangle - \langle P \rangle^2)\) Response function
Binder cumulant \(U = 1 - \langle P^4 \rangle/(3\langle P^2 \rangle^2)\) Finite-size scaling
Wilson loop \(W(\gamma) = \langle \prod_{e \in \gamma} \sigma_e \rangle\) Confinement test

Phase Structure

Phase Wilson loop Order parameter
Confined (\(\beta < \beta_c\)) Area law: \(W \sim e^{-\alpha A}\) \(\langle P \rangle \ll 1\)
Deconfined (\(\beta > \beta_c\)) Perimeter law: \(W \sim e^{-\beta P}\) \(\langle P \rangle \to 1\)

Critical coupling (exact): \(\beta_c = \frac{1}{2}\ln(1+\sqrt{2}) \approx 0.4407\)

Architecture

General tools → ts-quantum, specific sim → timesarrow/numerics/

ts-quantum (reusable lattice gauge theory)

  • src/lattice/geometry.ts — Lattice types (2D square, 2D triangular, 3D cubic)
  • src/lattice/gaugeField.tsZ2GaugeField class (link variables ±1)
  • src/lattice/action.ts — Wilson action and delta-S computation
  • src/lattice/monteCarlo.ts — Metropolis algorithm, thermalization, measurement
  • src/lattice/observables.ts — Plaquette average, specific heat, Wilson loops, Binder cumulant, jackknife error

timesarrow (simulation setup)

  • numerics/src/scripts/t20-z2-lgt-phase1.ts — 2D square lattice parameter sweep
  • numerics/src/scripts/t20-z2-lgt-phase2.ts — 2D triangular lattice
  • numerics/src/scripts/t20-z2-lgt-phase3.ts — 3D cubic lattice (paper target)

Phase 1: 2D Square Lattice (Complete)

Setup

  • Lattice: \(L \times L\) square with periodic boundary conditions
  • Plaquettes: squares with 4 links each
  • Algorithm: Metropolis Monte Carlo with single-link updates
  • Critical coupling (exact): \(\beta_c = \frac{1}{2}\ln(1+\sqrt{2}) \approx 0.4407\)

Results: L = 8 (Fast Validation)

Parameter Value
Lattice size \(L\) 8
Thermalization sweeps 1,000
Measurement sweeps 5,000
Measure every 5 sweeps
Bin size (error analysis) 20
Wall-clock time ~5 minutes
\(\beta\) \(\langle P \rangle\) Error Phase
0.10 0.0969 ±0.0036 Confined
0.20 0.1994 ±0.0042 Confined
0.30 0.3012 ±0.0039 Confined
0.40 0.3748 ±0.0039 Near critical
0.44 0.4162 ±0.0035 Critical
0.50 0.4608 ±0.0033 Deconfined
0.60 0.5335 ±0.0030 Deconfined
0.80 0.6645 ±0.0028 Deconfined
1.00 0.7572 ±0.0022 Deconfined
1.50 0.9047 ±0.0017 Strongly ordered

Results: L = 16 (Production)

Parameter Value
Lattice size \(L\) 16
Thermalization sweeps 10,000
Measurement sweeps 100,000
Measure every 10 sweeps
Bin size 100
Workers 3 threads
Wall-clock time ~2h 11m
\(\beta\) \(\langle P \rangle\) Error Phase
0.10 0.0997 ±0.0006 Confined
0.20 0.1978 ±0.0006 Confined
0.30 0.2916 ±0.0006 Confined
0.40 0.3794 ±0.0006 Near critical
0.44 0.4144 ±0.0006 Critical
0.50 0.4629 ±0.0006 Deconfined
0.60 0.5370 ±0.0005 Deconfined
0.80 0.6645 ±0.0005 Deconfined
1.00 0.7613 ±0.0004 Deconfined
1.50 0.9048 ±0.0003 Deconfined
2.00 0.9640 ±0.0002 Deconfined

Key improvements over L=8: Error bars reduced by ~6× (from ~0.0035 to ~0.0005).

Figures

Plaquette expectation vs coupling β

Figure 1: Plaquette expectation value ⟨P⟩ versus coupling β for L=8. The vertical dashed line marks the exact critical point βc ≈ 0.4407.

Plaquette expectation vs coupling β — L=16 production run

Figure 2: L=16 production run. Error bars are ~6× smaller than the L=8 run.

Analysis

The results confirm the expected phase transition at \(\beta_c \approx 0.44\):

  • Confined phase (\(\beta < 0.44\)): \(\langle P \rangle\) increases linearly with \(\beta\) but remains small. Wilson loops follow area law.
  • Critical point (\(\beta \approx 0.44\)): Rapid crossover in \(\langle P \rangle\) behavior. Correlation length diverges.
  • Deconfined phase (\(\beta > 0.44\)): \(\langle P \rangle\) approaches 1. Wilson loops follow perimeter law.

The critical coupling matches the exact theoretical value to within statistical error, validating the implementation.


Phase 2: Finite-Size Scaling (Complete)

Date: 2026-06-25

Overview

Finite-size scaling analysis with five lattice sizes: L = 8, 12, 16, 20, 24. All runs used the Rust implementation with 200,000 measurement sweeps across a dense β grid centered on the critical region.

Run ID L Sweeps β Range Wall Time Status
t20-p2-L8-20250625 8 200k 0.30–0.60 1.6s ✅ Complete
t20-p2-L12-20250625 12 200k 0.30–0.60 3.9s ✅ Complete
t20-p2-L16-20250625 16 200k 0.30–0.60 7.4s ✅ Complete
t20-p2-L20-20250625 20 200k 0.30–0.60 11.6s ✅ Complete
t20-p2-L24-20250625 24 200k 0.30–0.60 15.5s ✅ Complete

Total time for complete finite-size scaling study: ~40 seconds (Rust, 4 workers).

Plaquette Expectation Values

β L=8 L=12 L=16 L=20 L=24
0.30 0.2919(8) 0.2917(6) 0.2916(5) 0.2915(3) 0.2916(3)
0.35 0.3356(5) 0.3369(3) 0.3371(3)
0.40 0.3791(8) 0.3782(5) 0.3787(5) 0.3801(3) 0.3787(3)
0.42 0.3960(5) 0.3979(3) 0.3969(3)
0.44 0.4076(8) 0.4148(5) 0.4132(5) 0.4127(3) 0.4132(3)
0.46 0.4298(5) 0.4294(3) 0.4293(3)
0.48 0.4462(8) 0.4464(5) 0.4471(5) 0.4471(3) 0.4471(3)
0.50 0.4608(8) 0.4629(5) 0.4623(5) 0.4623(3) 0.4620(3)
0.55 0.5001(5) 0.5005(3) 0.5011(3)
0.60 0.5369(8) 0.5381(5) 0.5375(5) 0.5371(3) 0.5370(3)

Values: mean(error) with jackknife standard error. L=20–24 achieve ~0.03% precision.

Key Observables by Lattice Size

L ⟨P⟩ at β=0.44 χ_max Binder U (β=0.44) C_max
8 0.4076 0.274 0.579 0.082
12 0.4148 0.362 0.625 0.159
16 0.4132 0.370 0.640 0.173
20 0.4127 0.367 0.651 0.177
24 0.4132 0.361 0.656 0.159

Binder Cumulant Analysis

The Binder cumulant \(U = 1 - \langle P^4 \rangle/(3\langle P^2 \rangle^2)\) approaches the universal value \(U^* \approx 0.66\) (2D Ising) as \(L \to \infty\):

β U(L=8) U(L=12) U(L=16) U(L=20) U(L=24)
0.30 0.495 0.578 0.620 0.631 0.642
0.40 0.563 0.615 0.642 0.648 0.653
0.44 0.590 0.625 0.645 0.651 0.656
0.48 0.591 0.632 0.649 0.654 0.658
0.60 0.619 0.645 0.654 0.658 0.661

Finite-Size Scaling Conclusions

  1. Critical coupling: Plaquette expectation converges to ⟨P⟩ ≈ 0.413 at β_c ≈ 0.44
  2. Binder cumulant: Approaches U* ≈ 0.66, consistent with 2D Ising universality
  3. Specific heat: Peak sharpens with L, consistent with logarithmic divergence in 2D

Data Files

File Description Size
output/t20-p2-L8-20250625.json L=8 raw data 3.6 KB
output/t20-p2-L12-20250625.json L=12 raw data 3.6 KB
output/t20-p2-L16-20250625.json L=16 raw data 3.6 KB
output/t20-p2-L20-20250625.json L=20 raw data 3.6 KB
output/t20-p2-L24-20250625.json L=24 raw data 3.6 KB
data/registry.json Master registry 7.2 KB

Reproduction

cd timesarrow/rust-lattice
cargo run --release -- <L> 200000 20000 4 \
  0.30 0.35 0.40 0.42 0.44 0.46 0.48 0.50 0.55 0.60

Phase 3: 3D Cubic Lattice (Complete)

Date: 2026-06-25

Overview

3D cubic lattice Z₂ LGT with L = 4, 6, 8. The simulations probe the continuous confinement-deconfinement transition in the 3D Ising universality class, whose critical region is near β ≈ 0.76.

Run ID L Dimension Sweeps β Range Wall Time Status
t20-p3-L4-3D-20250625 4 3 100k 0.50–1.00 1.5s ✅ Complete
t20-p3-L6-3D-20250625 6 3 100k 0.50–1.00 4.0s ✅ Complete
t20-p3-L8-3D-20250625 8 3 100k 0.50–1.00 9.1s ✅ Complete

Plaquette Expectation Values

β L=4 L=6 L=8
0.50 0.502 0.502 0.502
0.60 0.627 0.627 0.627
0.70 0.805 0.805 0.790
0.75 0.942 0.936 0.932
0.76 0.960 0.950 0.948
0.77 0.972 0.959 0.958
0.80 0.985 0.974 0.973
0.85 0.993 0.986 0.985
0.90 0.996 0.991 0.991
1.00 0.997 0.997 0.997

Susceptibility (Peak Signals Critical Region)

β χ(L=4) χ(L=6) χ(L=8)
0.50 0.187 0.196 0.188
0.60 0.280 0.276 0.280
0.70 0.464 0.664 0.471
0.75 0.389 0.371 0.519
0.76 0.234 0.240 0.308
0.77 0.151 0.189 0.197
0.80 0.076 0.092 0.096

Critical region: β ≈ 0.70–0.76, with susceptibility peak shifting from β=0.70 (L=4) to β=0.75 (L=8) as lattice size increases — consistent with finite-size effects moving toward β_c ≈ 0.76.

Binder Cumulant

β U(L=4) U(L=6) U(L=8)
0.50 0.635 0.657 0.663
0.70 0.664 0.658 0.664
0.75 0.660 0.663 0.665
0.80 0.666 0.666 0.666
1.00 0.667 0.667 0.667

Key Findings

  1. Sharp transition: 3D shows much sharper transition than 2D (plaquette jumps from ~0.5 to ~0.95 in Δβ ≈ 0.05)
  2. Critical β: Consistent with β_c ≈ 0.76, shifting from β=0.70 (L=4) to β=0.75 (L=8) with finite-size effects
  3. Finite-size behavior: The susceptibility peak narrows and grows near the continuous critical point; a controlled scaling analysis requires autocorrelation-aware uncertainties and corrections to scaling
  4. Binder cumulant: Stabilizes near ~0.666 in ordered phase, consistent with 3D Ising universal value

Wilson Loop Results (L = 8)

Date: 2026-06-26

Wilson loops \(W(R \times C)\) were measured on the \(L=8\) 3D cubic lattice across the same β range (0.30–1.20). In the confined phase, Wilson loops obey the area law \(\ln|W| \sim -\sigma A\), while in the deconfined phase they follow the perimeter law \(\ln|W| \sim -\kappa P\).

Measured Wilson Loop Values

β \(1\times1\) \(2\times2\) \(3\times3\) \(4\times4\) Phase
0.30 0.294 0.007 0.0005 0.00006 Confined
0.50 0.502 0.067 0.003 0.0001 Confined
0.70 0.790 0.441 0.192 0.069 Near-critical
0.80 0.974 0.938 0.904 0.871 Deconfined
0.90 0.991 0.981 0.971 0.961 Deconfined
1.00 0.997 0.993 0.989 0.985 Deconfined

The qualitative change is dramatic: at β=0.50 the \(4\times4\) loop has \(|W| \approx 10^{-4}\) (area law), while at β=0.90 it has \(|W| \approx 0.96\) (nearly flat, perimeter law).

Wilson loop |W| vs area for selected β values

Figure 8: Wilson loop magnitude \(|W|\) versus loop area \(A\) on a log scale for β = 0.50 (confined), 0.70 (near-critical), and 0.90 (deconfined). At low β the loop decays exponentially with area (area law), while at high β it remains nearly constant (perimeter law). The near-critical curve (β=0.70) shows intermediate behavior.

String Tension Analysis

The string tension \(\sigma(\beta)\) is extracted from the area-law fit:

\[ \ln |W(A)| = -\sigma A + c \]

Fitting \(\ln|W|\) versus area \(A\) for each β gives the string tension shown below.

String Tension Results

β σ Error Fit quality Phase
0.30 0.533 ±0.120 1.86 Confined
0.40 0.430 ±0.206 5.47 Confined
0.50 0.547 ±0.039 0.19 Confined
0.60 0.404 ±0.006 0.004 Confined
0.70 0.162 ±0.006 0.005 Near-critical
0.80 0.007 ±0.001 0.0001 Deconfined
0.90 0.002 ±0.0003 0.000009 Deconfined
1.00 0.0007 ±0.0001 0.000001 Deconfined
1.10 0.0003 ±0.00004 0.0000002 Deconfined
1.20 0.0001 ±0.00002 0.00000003 Deconfined

The string tension drops from \(\sigma \approx 0.5\) in the confined phase to \(\sigma \approx 0\) in the deconfined phase, vanishing rapidly around the critical region β ≈ 0.70–0.80.

String tension σ(β) vs coupling β

Figure 9: String tension \(\sigma(\beta)\) versus coupling β for 3D Z₂ LGT (L=8). The shaded red region marks the critical range β ≈ 0.70–0.80 where σ drops to zero, signaling the confinement-deconfinement transition. Error bars are jackknife estimates.

Confinement-Deconfinement Signature

The combination of Wilson loop and string tension data provides a clear signature of the transition:

Observable Confined (β < 0.70) Critical (β ≈ 0.70–0.80) Deconfined (β > 0.80)
String tension σ ≈ 0.4–0.5 Drops sharply ≈ 0
\(|W|\) for \(4\times4\) \(\sim 10^{-4}\) \(\sim 10^{-1}\) \(\sim 0.96\)
Area law ✅ Yes Partial ❌ No
Perimeter law ❌ No Partial ✅ Yes

The vanishing of σ at β ≈ 0.76 is the hallmark of deconfinement: the potential between static charges becomes Coulomb-like rather than linearly rising.

Multi-Lattice Comparison (Fine-Grained β)

The figures below show all three lattice sizes (L = 4, 6, 8) with the refined 21-value β grid. The finite-size scaling signatures are clearly visible: the plaquette curves steepen with increasing L, and the susceptibility/specific-heat peaks grow and shift toward the thermodynamic critical point β_c ≈ 0.76.

3D Plaquette vs β (all lattice sizes)

Figure 3: Plaquette expectation value vs coupling β for L = 4, 6, 8, 16, 24, 32 (3D cubic, 21 β values). The dashed vertical line marks the established critical region. The plaquette is a local energy-like observable.

3D Specific Heat vs β (all lattice sizes)

Figure 4: Specific heat C_V vs β (all L). Quantitative exponent extraction requires autocorrelation-aware uncertainties and a controlled continuous-transition fit.

3D Susceptibility vs β (all lattice sizes)

Figure 5: Plaquette fluctuation susceptibility χ vs β (all L). The available peak estimates are exploratory and are not yet a controlled critical-exponent measurement.

3D Binder Cumulant vs β (all lattice sizes)

Figure 6: Plaquette Binder ratio U vs β (all L). Because the plaquette has a nonzero mean, a narrow distribution generically approaches U = 2/3; this limit does not diagnose transition order.

3D Combined Overview (all lattice sizes)

Figure 7: Combined overview of four observables for 3D Z₂ LGT (all L, 21 β values). The plots locate strong finite-size variation in the critical region but do not independently determine the universality class.

Data Files

File Description
output/t20-p3-L4-3D-20250625.json L=4 raw data
output/t20-p3-L6-3D-20250625.json L=6 raw data
output/t20-p3-L8-3D-20250625.json L=8 raw data
output/t20-p3-L16-3D-wilson-fine-20250626.json L=16 raw data (new)
output/t20-p3-L24-3D-wilson-fine-20250626.json L=24 raw data (new)
output/t20-p3-L32-3D-wilson-fine-20250626.json L=32 raw data (new)
output/benchmark-lattice-sizes-20250626.json Run time benchmarks

Corrected 3D Interpretation

The pure 3D Z₂ lattice gauge theory is dual to the 3D Ising model, so its transition is continuous and has 3D Ising critical behavior. The earlier first-order analysis used the local plaquette Binder ratio as though it were a universal order-parameter cumulant. That inference is invalid: for a narrowly distributed variable with nonzero mean, the ratio approaches \(2/3\) generically.

The earlier scaling-collapse plot is also inconclusive. A visual failure to collapse data collected on unequal grids and without propagated autocorrelation errors cannot exclude the expected universality class. The synthetic double-peaked histograms have been removed from the evidence chain because they were not sampled distributions.

The existing data are retained as exploratory measurements near the critical region. A corrected analysis must use blocked or jackknife errors, integrated autocorrelation times, continuous-transition finite-size scaling with corrections, and a separate analysis of extended Wilson loops.

L Volume Time Avg per β
4 64 6s 0.3s
6 216 23s 1.1s
8 512 51s 2.4s
16 4,096 129s 6.1s
24 13,824 436s 20.8s
32 32,768 1,034s 49.2s

Scaling: Time \(\sim L^{3.2}\), slightly faster than \(L^4\) (full volume × sweep time) due to cache efficiency and parallelization.


Implementation Details

  • Target: Paper results
  • Critical coupling: \(K_c \approx 0.761\) (3D Ising universality)
  • Critical exponents: ν ≈ 0.63, β ≈ 0.33
  • Observables: Wilson loops (area vs perimeter law), string tension
  • Dressed correlator: \(C(r) = \langle \tau_0 \prod_{e \in \gamma} \sigma_e \tau_r \rangle\)

Implementation Details

Rust Framework (T27)

The Rust implementation provides ~2,500–3,000× speedup over TypeScript:

Metric TypeScript Rust Speedup
L=16, 100k sweeps, 11 β values ~2h 11m 3.0s ~2,600×
Phase 2 (5 lattice sizes) ~11h (estimated) ~40s ~1,000×

Validation: All 11 β values match TypeScript within |Δ| < 0.02.

Files: - rust-lattice/src/lib.rs — Core implementation - rust-lattice/src/main.rs — CLI - rust-lattice/Cargo.toml — Dependencies

Code Example

import { 
  createSquareLattice, 
  Z2GaugeField, 
  averagePlaquette, 
  metropolisSweep, 
  thermalize,
  jackknifeError,
  binData
} from 'ts-quantum';

const lattice = createSquareLattice(8);
const field = new Z2GaugeField(lattice, 'random');

thermalize(field, 0.44, 1000);

const measurements = [];
for (let s = 0; s < 5000; s++) {
  metropolisSweep(field, 0.44);
  if (s % 5 === 0) measurements.push(averagePlaquette(field));
}

const binned = binData(measurements, 20);
const mean = binned.reduce((a, b) => a + b, 0) / binned.length;
const error = jackknifeError(binned);

console.log(`⟨P⟩ = ${mean.toFixed(4)} ± ${error.toFixed(4)}`);

Phase 3b: Ising FSS Reanalysis (T32-T20d)

Date: 2026-07-08

Summary

The L=16 and L=24 fine-scan data have been reanalysed using continuous 3D Ising universality scaling (ν ≈ 0.6299, γ ≈ 1.236, α ≈ 0.11). The key finding is that the pseudo-critical couplings extrapolate to the established literature value β_c ≈ 0.761, supporting the continuous-transition interpretation.

Analysis Script

  • Script: numerics/scripts/t20d-ising-fss-reanalysis.py
  • Reproducible: Run from repo root with python3 numerics/scripts/t20d-ising-fss-reanalysis.py
  • Figures: numerics/figures/t20d-ising/

Pseudo-Critical Couplings

L β_c(L) from χ_peak χ_max
16 0.75167 1.1297
24 0.75606 1.2833

Thermodynamic Extrapolation

Using the Ising scaling form β_c(L) = β_c(∞) + A·L^(−1/ν):

Quantity Value
Extrapolated β_c(∞) 0.76092
Literature β_c (3D Z₂ / 3D Ising dual) 0.76141
Deviation −0.00049 (−0.06%)
Shift amplitude A −0.75469

Conclusion: The extrapolated β_c agrees with the known literature value to within 0.06%, well within the expected uncertainty from a two-point extrapolation.

Scaling Collapse

When plotted against the scaling variable x = L^(1/ν)(β − β_c^lit), the susceptibility data from L=16 and L=24 show reasonable data collapse onto a single universal curve. The Binder cumulant curves cross near x ≈ 0, consistent with the expected critical behavior. Collapse plots are saved as:

  • t20d-chi-collapse-lit.png — χ collapse using literature β_c
  • t20d-chi-collapse-fit.png — χ collapse using fitted β_c(∞)
  • t20d-binder-collapse.png — Binder cumulant collapse
  • t20d-plaquette-collapse.png — Plaquette collapse

Caveats and Limitations

  1. Two lattice sizes only: The L=16 and L=24 data provide only two points for the β_c extrapolation and the χ_max scaling. The fitted γ/ν = 0.314 is not reliable — proper exponent extraction requires at least 3–4 lattice sizes.
  2. No autocorrelation analysis: The present analysis uses the reported error bars but does not recompute integrated autocorrelation times or blocked jackknife errors from raw time series.
  3. Corrections to scaling: Not included. With only two lattice sizes, distinguishing leading scaling from corrections is impossible.
  4. Binder cumulant: The plaquette Binder cumulant approaches ~2/3 in the ordered phase, which is the generic limit for a narrowly distributed variable with nonzero mean. It is not evidence of first-order behavior.

Figures

Susceptibility — raw data

Figure: Susceptibility χ vs β for L=16 and L=24. The vertical lines mark the literature critical point (red dashed) and the fitted β_c(∞) (green dash-dot). The L=24 peak is higher and shifted to larger β, consistent with finite-size scaling.

β_c extrapolation

Figure: Extrapolation of pseudo-critical β_c(L) to the thermodynamic limit using the Ising scaling form β_c(L) = β_c(∞) + A·L^(−1/ν). The red dotted line is the literature value β_c = 0.76141.

Susceptibility FSS collapse

Figure: Scaling collapse of susceptibility using 3D Ising exponents and the literature β_c. The data from L=16 and L=24 fall onto a single curve, consistent with the continuous-transition hypothesis.

Binder cumulant FSS collapse

Figure: Binder cumulant vs scaling variable. The curves cross near x = 0, consistent with the universal critical behavior expected for a continuous transition.

Combined overview

Figure: Combined overview of plaquette, susceptibility, specific heat, and Binder cumulant for the L=16 and L=24 fine scans. The dashed red line marks the literature critical point β_c ≈ 0.761.


Phase 3b: Corrected Finite-Size Scaling Status

Status: Reanalysis in progress under T32.

The completed L = 8, 16, 24, and 32 datasets are preserved. They show rapid finite-size variation near the known critical region, but the present analysis does not provide a new precision value of \(\beta_c\) or reliable critical exponents. In particular:

  • the elementary plaquette is a local energy-like observable, not a substitute for an extended Wilson-loop confinement diagnostic;
  • the plaquette Binder ratio tending to \(2/3\) does not establish first-order behavior;
  • the previous synthetic histograms are not numerical evidence of phase coexistence;
  • a failed visual collapse without autocorrelation-aware uncertainties is inconclusive.

The reanalysis will assume continuous 3D Ising-universality scaling, estimate autocorrelation and blocked errors, include corrections to scaling, and assess extended Wilson loops separately. No T20d conclusion is promoted into timesarrow.tex until those checks pass.

Retained Data

L β range Grid Status
8 0.70–0.82 25 points Retained exploratory data
16 0.740–0.780 21 points Retained fine scan
24 0.740–0.780 21 points Retained fine scan
32 0.740–0.780 21 points Retained lean scan; lower statistics

Reanalysis Requirements

  1. Preserve or regenerate measurement time series.
  2. Estimate integrated autocorrelation times and blocked or jackknife errors.
  3. Fit pseudo-critical shifts and peak scaling with 3D Ising exponents plus corrections to scaling.
  4. Test extended Wilson-loop area/perimeter behavior independently of the plaquette.
  5. Report sensitivity to lattice selection, interpolation, and coupling-grid resolution.

Key Files

  • implementation-details/t20-phase3b-requirements.md — Original simulation specifications
  • numerics/data/fss/ — Preserved simulation outputs
  • t20d-fss-analysis.tex — Corrected standalone status note

References

  • Paper: Section 4.2, Eq. (48)
  • timesarrow/numerics/docs/implementation/t20-z2-lgt.md — Full architecture doc
  • timesarrow/numerics/data/registry.json — Data registry