10. Results and Analysis
10.1. Result-directory naming
output_folder is a base name. PhonoMC scans sibling directories and creates
the next numeric suffix, for example Cross_100nm_0 followed by
Cross_100nm_1. Record the exact printed result path in batch workflows.
10.2. convergence.txt
The header begins with # and defines all columns. The current columns are:
timestepCompleted solver step.
time_psSimulated elapsed time in ps.
T_1 ... T_nGrid temperatures in K, in the same index order as
grid_centers.csv.heatfluxParticle-weighted average heat flux along the inferred transport axis in W/m².
kappa_intConductivity from a least-squares grid-temperature gradient.
kappa_effConductivity from reservoir endpoint temperatures and full domain length.
absorbed,injected,recovered,netPer-step reservoir and recovery bookkeeping.
netis reservoir injection minus absorption.hs_injected/hs_injected_energy_evLocal-heat-source particle count and injected energy for the step.
10.3. Temperature convergence
Look for a statistically stationary profile rather than a perfectly flat trace. Increase simulation time if the profile is still drifting. Increase particle count if the steady trace is dominated by noise.
For a nominally one-dimensional case, plot several representative T_i
traces and inspect the final spatial profile using grid_centers.csv.
10.4. Particle balance
Reservoir-driven steady cases should not exhibit a persistent unexplained
particle drift. Compare cumulative absorption and injection in summary.txt
and inspect recovered events. Frequent recovery indicates a geometry or
collision-tolerance problem, not a normal convergence mechanism.
10.5. Conductivity interpretation
Compare kappa_int and kappa_eff after the temperature profile has
settled. Differences can arise from boundary jumps, nonlinearity, insufficient
grid resolution, or sampling noise.
Repeat a conductivity calculation with:
more particles
a smaller time step
longer simulated time
at least one finer grid
Report the convergence study alongside the selected conductivity value.
10.6. summary.txt diagnostics
The summary includes normalized input, geometry, grid and boundary counts, runtime, OpenMP settings, rough-scattering selection statistics, escaped particle recovery, reservoir totals, and heat-source injection totals.
Large rough-boundary fallback counts deserve investigation. They can indicate that no frequency-compatible reflected mode was available for many events.
10.7. One-dimensional plots
python3 tools/plot_convergence.py RESULT_DIRECTORY \
--max-temp-lines 40 --show-legend
Outputs:
plots_1d/temperature_vs_time.pngplots_1d/heatflux_vs_time.pngplots_1d/kappa_vs_time.png
10.8. Three-dimensional plots
python3 tools/plot_temperature_3d.py \
--input INPUT.toml \
--results RESULT_DIRECTORY \
--x-slice-rel 0.5 \
--y-slice-rel 0.5
Use slice images to check whether the heat-source region, periodic directions, and thermal contacts produce the intended spatial pattern.