Changelog

• Exchanged mpiFFT4py with mpi4py-fft (cleaner, simpler interface by same group)
• Support for adiabatic EOS (with gamma = 5/3 hardcoded at the moment)
• Added script for higher order turbulent flow analysis, see below
• Data in column major format stored in HDF5 files is only read by proc 0 and then distributed. Data stored in row major format is read in parallel by all processes.

Requirements

• mpi4py-fft package (incl dependencies) homepage/download
• h5py package
• yt package (optional - can be used to read initial data)
• palettable package (optional - used for colors in the sample plotting notebook)

General usage

The run_analysis.py script is the main file that needs to be run.

In general the following parameters are available

• --res RES set linear resolution of the cubic box
• --type {transfer,flow,unit-test} set analysis type
  • transfer for energy transfer analysis
  • flow for analysis of turbulence statistics
  • unit-test for some preliminary unit tests
• --data_type {Enzo,AthenaPP,AthenaHDFC,Athena} set data cube type
  • Enzo reads Enzo data using yt frontend
  • AthenaPP reads Athena++/K-Athena data using yt frontend
  • AthenaHDFC reads Athena data that has been converted to hdf5 data
  • Athena reads Athena data using yt frontend
  • to add more options see section "Adding new simulation output"
• --data_path DATA_PATH set data location
• --outfile OUTFILE set file to store results (should be .pkl for transfer and .hdf5 for flow)
• --extrema_file EXTREMA_FILE Path to pickled python dictionary containing minimum and maximum values for quantities (used for creating histograms with a fixed [global] bounds)
  • Style of dictionary is for example. {'rho' : [ 0., 10]}
  • All quantities are always binned to the min and max values of the individual snapshot.
  • If no dictionary is found under the given path (e.g., by setting it to a nonexisting file/path) no histograms with global bounds will be created
• -b enable magnetic fields
• -forced output is actively forced
• --eos {isothermal,adiabatic} set equation of state
• --gamma GAMMA set adiabatic gamma index
• -approx-isothermal assume c_s^2 / gamma = p/rho = 1
• --terms {All,Int,UU,BUT,BUP,UBT,UBPb,BB,BUPbb,UBPbb,SS,SU,US,PU,FU} set energy transfer terms to analyze
• --binning {log,lin,test} set binning used in energy transfer analysis
  • lin leads to linearly equally spaced bins with boundaries at $k = 0.5,1.5,2.5,...,Res/2$
  • log leads to logarithmically equally spaced bins with boundaries at $k = 0, 4 * 2^{(i - 1)/4},Res/2$
  • test leads to bins used for regression testing, i.e. $k = 0.5,1.5,2.5,16.0,26.5,28.5,32.0$
• --kernels choose one or more real space convolution kernels to be used in filtering
  • Box for a box car/top hat filter (implementation needs update)
  • Sharp for a sharp spectral filter
  • Gauss for a smooth Gaussian filter

Energy transfer analysis

Sample usage

Use the run_analysis.py script with the --flow transfer option. For example (to run the transfer analysis on the regression data set),

srun -n 8 python ./run_analysis.py --terms All FU PU BUPbb UBPbb --res 128 --data_path DD0024/data0024 --data_type Enzo --binning test --type transfer --outfile test-out.pkl --eos adiabatic --gamma 1.0001  -forced -b

Turbulent flow analysis

Features

• Uses MPI with slab decomposition
• For a given scalar field
  • higher order statistical moments for arb. fields incl. mean, rms, variance, standard deviation, skewness, kurtosis, minimum, maximum, absolute minimum, absolute maximum
  • 1d histograms with automatic and given bounds
• For two scalar field
  • correlation coefficient
  • 2d histograms with automatic and given bounds
• Decomposion of vector fields in harmonic, solenoidal and compressive modes
• Power spectra
  • for total, solenoidal and compressive components
  • with different normalization: no weighting, surface average, shell average
  • with different definitions of kinetic energy density
  • $E(k) = \sqrt(\rho u)^2$ (Grete, et al., 2017)
  • $E(k) = |\overline{\rho u_l}|^2 / 2\rho_l$ (Sadek & Aluie, 2018)
    • must specify convolution kernel type with --kernels
• Dispersion measures, rotation measures and line of sight magnetic field along all axes

Usage

Use the run_analysis.py script with the --flow flow option. For example (to analyze a driven turbulence hydro simulation with an isothermal equation of state), For example,

srun -n 8 python ./run_analysis.py --res 256 --data_path /PATH/TO/SIM/DUMP  --data_type Athena --type flow --eos isothermal --outfile /PATH/TO/OUTFILE.hdf5 -forced --kernels Gauss

Sample analysis

The notebook sample_flow_analysis.ipynb contains examples on how to analyze/visualize/plot the data from the flow analysis.

Current limitations

• data is assumed to be in a periodic box with side length 1 and equal grid spacing
• pressure is calculated with $c_s = 1$ in the isothermal EOS case
• wavenumber are implicitly normalized (k = 1, ...)
• only slab decomposition (of equal size) among MPI processes
• units are hard coded/implicitly assumes
  • velocity has units of speed of sound (with $c_s = 1$)
  • magnetic field includes $1/\sqrt{4 \pi}$

Adding new simulation output

• Edit run_analysis.py and add another option to the --data_type argument
• Edit IOhelperFuncs.py and make sure that the field variables are the strings that are availble for that particular dump within it, i.e. they should be present in the ds.field_list array.