squeezing-codes

code relating to the analysis of squeezing codes [1].

usage

setting things up:

run Pkg.activate(".") and Pkg.Instantiate() in the project folder.

simulation_driver.jl: Run a single simulation with custom parameters.

examples:

julia --project=. simulation_driver.jl --mode=dw_growth --Lprl=100 --rule=R --p=0.05
julia --project=. simulation_driver.jl --mode=dw_history --Lprl=50 --T_evolve=200
julia --project=. simulation_driver.jl --mode=dw_roughness --Lprl=20 --n_trials=10

Main Parameters:

• mode: Simulation mode
  • "dw_growth" computes growth of interface roughness in time
  • "dw_roughness" computes late-time height fluctuations of interface
  • "dw_history" visualizes the growth of an interface
  • "history" visualizes the normal 2d dynamics
• rule: Rule name - "R", "F", or "M" (default: "R")
• Lprl: System size parallel to domain wall
• Lperp: System size perpendicular to domain wall
• p: Noise strength (default: 0.0)
• rotate: If true, rotate the system by 90 degrees (default: false)
• r: If true, reverse spins in initial configuration (default: false)
• save: If true, save data to file in data/ folder with pre-defined name (default: true for dw_growth and dw_roughness, false else)

Other Parameters:

• T_evolve: Number of time steps to evolve
• n_trials: Number of trials to average over (for dw_growth and dw_roughness modes)
• border: Border thickness for active region padding (default: 4)
• use_max_height: If true, defines height field using maximum height of flipped spin rather than sum (default: false)
• final_state: If true, saves only the final state as an image
• pretty: If true, makes animations with minimal visual clutter
• out_adj: Optional string to append to filename before .jld2 extension
• reset_center: If true, resets domain wall center when it drifts too far (default: true)

All parameters can be set via command-line arguments using --param=value syntax. Defaults are defined in the script; to run the script in the REPL, modify these values and run include("simulation_driver.jl).

run_batch.jl: Run parameter sweeps, either sequentially or as SLURM job arrays.

# Run all jobs sequentially (for testing)
julia --project=. run_batch.jl

# Submit as SLURM job array (see run_batch_slurm.sh)
sbatch run_batch_slurm.sh

Edit the parameter vectors in run_batch.jl to define the parameter sweep.

plotter.py: Plot simulation results from JLD2 files.

# Plot automatically from data/ directory
python3 plotter.py

# Plot specific files
python3 plotter.py data/R_dw_growth_Lprl50.jld2

# options: 
-instexp  # Plot instantaneous exponent
-allhists # Plot individual trial histories

testing that things work:

visualization:

1. julia --project=. simulation_driver.jl --mode=dw_history --Lprl=150 --Lperp=100 --T_evolve=300 --rule=R --p=.1
2. julia --project=. simulation_driver.jl --mode=dw_history --Lprl=150 --Lperp=100 --T_evolve=300 --rule=M --rotate=true
3. julia --project=. simulation_driver.jl --mode=dw_history --Lprl=150 --Lperp=100 --T_evolve=300 --rule=F --rotate=true --r=true
4. julia --project=. simulation_driver.jl --mode=history --Lprl=150 --Lperp=150 --rule=M --r=true

domain growth:

1. julia --project=. simulation_driver.jl --mode=dw_growth --Lprl=75 --rule=R --n_trials=25
2. python3 plotter.py data/R_dw_growth_Lprl75_n_trials25.jld2

domain roughness:

1. julia --project=. simulation_driver.jl --mode=dw_roughness --Lprl=20 --rule=F --n_trials=25
2. python3 plotter.py data/F_dw_roughness_Lprl20_n_trials25.jld2 (just plots a single data point; designed to be run with multiple files at different system sizes)

references

1. The original paper on squeezing codes