Parameter scan

[fiedl]

We do not know the actual ice properties of the hole ice, yet.

How do the angular acceptance plots look like for different hole ice parameters?

• Import script.
• Document script.
• Parallelise such that it can run as array job on the cluster.
• Run rough scan on the cluster.
• Run detailed scan on the best region of the parameter space. -> Run a with better resolution and more photons.
• Run detailed scan on the region of Martin's best parameters.
• Run a with better resolution and more photons.
• Plot results.
• Document results in README.

Results

[2018-07-21 16:38:53] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_reference_curve_scan_contours.py ~/hole-ice-study/results/parameter_scan
best values:
  LLH: -633.315809421
  esca = 1.3m
  r = 1.0 r_dom

[2018-07-21 16:44:04] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_angular_acceptance.py ../../results/parameter_scan/esca1.0_r1.0rdom_abs100

[fiedl]

The angular acceptance plots do not look too well. In fact, they do not show any of the expected hole ice effects.

This is a plot with extreme hole ice parameters, i.e. almost instant scattering and absorption within the hole ice:

A detailed scan of the right hand side region where the photons come from below the detector, i.e. where the hole ice effect should be max. Also increased distance to 10m to increase the effect. But this does not look too convincing as well:

[fiedl]

There have been issues when running the simulation on the gpu cluster that needed to be resolved first: #14, #16

Now, the plot looks a lot more like expected. This is for abs=0.1, sca=0.1, dst=1.0m:

[fiedl]

• There should be no positive corrections for interaction factor 0.1. #17 needs to be fixed before running an accurate scan.
• Also Implement missing test cases for apply_hole_ice_correction() #20 should be done before running a detailed scan.
• Optionally consider What if scattering or absorption factors are greater than one? #2.

[fiedl]

Results of rough parameter scan: https://github.com/fiedl/diplomarbeit/commit/891bcbdc5d9f2c1aca0c75d73e18b988e6f4e5d5

[fiedl]

Parameter scan over scattering length and column radius

• The definition of hole-ice parameters has been changed from relative to absolute values (Different hole-ice correction mechanism: Generic propagation through different media #45)
• New parameter scan over scattering length and column radius
• Absorption length is not to be scanned and needs to be set to a high value in order to have absorption not interfering

[fiedl]

Run on the cluster

[2018-03-12 15:29:08] fiedl@wgs16 /afs/ifh.de/group/amanda/scratch/fiedl/hole-ice-study master ⚡
▶ git pull

[2018-03-12 15:29:21] fiedl@wgs16 /afs/ifh.de/group/amanda/scratch/fiedl/clsim sf/hole-ice-2018
▶ git pull

[2018-03-12 15:29:50] fiedl@wgs16 /afs/ifh.de/group/amanda/scratch/fiedl/software/icecube-simulation-V05-00-07/debug_build
▶ ./env-shell.sh

[2018-03-12 15:30:46] fiedl@wgs16 /afs/ifh.de/group/amanda/scratch/fiedl/software/icecube-simulation-V05-00-07/debug_build/clsim
▶ make

[2018-03-12 15:36:21] fiedl@wgs16 /afs/ifh.de/group/amanda/scratch/fiedl/hole-ice-study/scripts/ParameterScan master ⚡
▶ ./run.rb --submit-to-cluster

[fiedl]

best fit to reference plot:

{:cluster=>true,
 :hole_ice_scattering_length=>0.09000000000000008,
 :hole_ice_absorption_length=>100.0,
 :hole_ice_radius=>0.1651,
 :distance=>1.0,
 :plane_wave=>true,
 :number_of_photons=>100000.0,
 :number_of_runs=>2.0,
 :number_of_parallel_runs=>2.0,
 :angles=>[0, 10, 20, 30, 40, 50, 60, 70, 90, 120, 140, 150, 160, 170, 180],
 :run_id=>"Run-2018-hiHoh7uf",
 :started_at=>"2018-03-13 23:34:34 +0100",
 :gcd_file=>
  "$I3_TESTDATA/sim/GeoCalibDetectorStatus_IC86.55380_corrected.i3.gz",
 :ice_model_file=>"$I3_SRC/clsim/resources/ice/spice_mie",
 :seed=>123456,
 :hole_ice_cylinder_positions=>[[-256.02301025390625, -521.281982421875, 0]],
 :hole_ice_cylinder_radii=>[0.1651],
 :cylinder_scattering_lengths=>[0.09000000000000008],
 :cylinder_absorption_lengths=>[100.0],
 :hole_ice_radius_in_dom_radii=>1.0,
 :hole_ice_effective_scattering_length=>1.5,
 :dom_index=>[1, 1],
 :dom_position=>[-256.02301025390625, -521.281982421875, 500],
 :hole_ice=>:simulation,
 # ...
}

[fiedl]

Contour plot

Contour plot (#37) of parameter scan agreements with reference plot (#10, #6)

fiedl@fiedl-mbp ~/hole-ice-study/scripts/AngularAcceptance/lib master ⚡
▶ ./plot.rb ~/hole-ice-study/results/parameter_scan
▶ python create_contour_plot.py parameter_scan.txt

[fiedl]

Simulation performance

For each parameter set, 30e5 photons are propagated (1e5 photons per run, 2 runs per angle, 15 angles).

As this looks more like fluctuations, the leading factor appears to be the specific cluster worker the job has run on.

[fiedl]

[2018-07-20 22:38:09] fiedl@kepler00 /afs/ifh.de/group/amanda/scratch/fiedl/hole-ice-study/scripts/ParameterScan
▶ git pull
▶ ./run.rb --submit-to-cluster

[2018-07-21 16:28:46] fiedl@wgs00 /afs/ifh.de/group/amanda/scratch/fiedl/hole-ice-study/scripts/ParameterScan
▶ rm -r $SCRATCH/hole-ice-study/results/parameter_scan/esca*
▶ mv cluster-results/esca* $SCRATCH/hole-ice-study/results/parameter_scan/

[fiedl]

[2018-07-21 16:36:20] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./convert_options_txt_to_json.rb ~/hole-ice-study/results/parameter_scan
▶ ./calculate_reference_plot_llhs.py ~/hole-ice-study/results/parameter_scan

[fiedl]

[2018-07-21 16:38:53] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_reference_curve_scan_contours.py ~/hole-ice-study/results/parameter_scan
best values:
  LLH: -633.315809421
  esca = 1.3m
  r = 1.0 r_dom

[fiedl]

Best agreement:

[2018-07-21 16:40:21] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_angular_acceptance.py ../../results/parameter_scan/esca1.3_r1.0rdom_abs100

[2018-07-21 16:44:04] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_angular_acceptance.py ../../results/parameter_scan/esca1.0_r1.0rdom_abs100

[fiedl]

bottom right corner: strong hole ice

[2018-07-21 16:45:31] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_angular_acceptance.py ../../results/parameter_scan/esca0.2_r1.8rdom_abs100

[fiedl]

top left corner: weak hole ice

[2018-07-21 17:16:31] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_angular_acceptance.py ../../results/parameter_scan/esca3.0_r0.1rdom_abs100/

[fiedl]

upper right: when increasing the radius while increasing the scattering length, one effect strengthens while the other weakens the hole-ice effect. Thus, this is still a good agreement:

[2018-07-21 17:17:30] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_angular_acceptance.py ../../results/parameter_scan/esca2.8_r1.5rdom_abs100

[fiedl]

Another scan with pencil beam

As performance is now good, I'm retrying the pencil-beam scan.

• Prepare run.rb for pencil beams
• Rescan
• Plot contours
• Plot best angular-acceptance plot
• Make sure to reset run.rb later to --plane-wave mode

[fiedl]

[2018-07-21 17:37:46] fiedl@kepler00 /afs/ifh.de/group/amanda/scratch/fiedl/hole-ice-study/scripts/ParameterScan
▶ git pull
▶ ./run.rb --submit-to-cluster

[2018-07-21 18:43:24] fiedl@kepler00 /afs/ifh.de/group/amanda/scratch/fiedl/hole-ice-study/scripts/ParameterScan
▶ rm -r ../../results/parameter_scan_pencil_beam/esca*
▶ mv cluster-results/esca* ../../results/parameter_scan_pencil_beam/

[fiedl]

[2018-07-21 18:56:19] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./convert_options_txt_to_json.rb ../../results/parameter_scan_pencil_beam
▶ ./calculate_reference_plot_llhs.py ../../results/parameter_scan_pencil_beam

[fiedl]

[2018-07-21 19:00:51] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_reference_curve_scan_contours.py ~/hole-ice-study/results/parameter_scan_pencil_beam
best values:
  LLH: -5575.61508961
  esca = 3.5m
  r = 2.5 r_dom

Best values:

[2018-07-21 19:06:55] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_angular_acceptance.py ../../results/parameter_scan_pencil_beam/esca3.5_r2.5rdom_abs100

[2018-07-21 19:07:43] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./plot_angular_acceptance.py ../../results/parameter_scan_pencil_beam/esca2.5_r1.5rdom_abs100

[fiedl]

Starting photons from farther away

[2018-07-24 15:41:35] fiedl@trex01 /afs/ifh.de/group/amanda/scratch/fiedl/hole-ice-study/scripts/AngularAcceptance
▶ ./run.rb --cluster --hole-ice-scattering-length=0.06 --hole-ice-radius=0.1651 --distance=2.0 --number-of-photons=1e6 --number-of-runs=1 --number-of-parallel-runs=1 --angles=0,10,20,30,40,50,60,70,90,120,140,150,160,170,180 --plane-wave

[fiedl]

[2018-07-24 16:06:33] fiedl@fiedl-mbp ~/hole-ice-study/scripts/lib master ⚡
▶ ./convert_options_txt_to_json.rb ../AngularAcceptance/results/current
▶ ./plot_angular_acceptance.py ../AngularAcceptance/results/current

One would need another gauging simulation. But this plot already shows that on the left-hand side there are still too many hits.