Simulation Data

All runs: ring topology, N=64, stochastic Euler-Maruyama integration, dt=0.001, 5×105 steps (discard first 105). Natural frequencies drawn from N(0, 0.05). Reported r is time-averaged over the last 104 steps.

Run set A: fixed noise, varying coupling

σ = 0.05, K stepped from 0.5 to 3.0 in increments of 0.5.

K⟨r⟩σ(r)Slip countMax phase diff
0.50.1870.0628472.91π
1.00.2930.0745232.44π
1.50.4580.0582411.83π
2.00.7210.031580.92π
2.50.9040.01470.37π
3.00.9710.00600.14π

Run set B: fixed coupling, varying noise

K = 2.0, σ stepped from 0.01 to 0.15.

σ⟨r⟩σ(r)Slip count
0.010.9630.0040
0.030.8920.0159
0.050.7210.03158
0.080.5110.052203
0.120.3040.069478
0.150.1910.081713

Run set C: parity check

K = 1.75, σ = 0.04. Comparing even and odd N. Each run 3×.

N⟨r⟩ (run 1)⟨r⟩ (run 2)⟨r⟩ (run 3)Slips (mean)
320.5310.5440.538119
330.4870.4760.491164
640.5830.5900.57992
650.5040.5110.498148

Parity effect is real — odd-N configurations consistently underperform even-N by roughly 12-15% in ⟨r⟩ and 50% more slips.

Run parameters

# coupled-osc run --topology ring --n 64 --K 2.0 --sigma 0.05 \
    --dt 0.001 --steps 500000 --burn-in 100000 --seed 20260714 \
    --output run_005k_64ring_K2.0_s0.05.h5

Raw HDF5 files available on request. Python analysis scripts in the repo.


Data generated 2026-07-14. Reproducible with the commit tagged v0.3.1.