Integrate continuum interactions into cmontecarlo #670

chvogl · 2016-11-24T15:57:41Z

This PR imports the core routines of the continuum interaction procedure developed by @chvogl during his master's project. This PR constitutes only the first step, thus the immediate goal is not yet to have a fully fledged Tardis-continuum version but to import the necessary changes step by step in a series of PRs. The goal of the current one can be summarized as follows:

include core C routines of continuum interaction scheme
add necessary C unit tests for new additions and update and complete existing C unit tests
ensure that normal Tardis operation is not affected by changes

For example for cooling by free-bound emission a continuum for emission is selected (i.e. a continuum_id).

- remove deprecated npz files with free-free emissivities - refactor test_sample_nu_free_free and related fixtures to adjust to new data format

To allow easier testing the fixture get_rkstate is added. This fixture allows to prepare the mt_state of the random number generator such that requested random numbers are generated.

In addition the flag compute_chi_bf is added to prevent the recomputation of the continuum opacities in cases where no physical interaction occurs (e.g. passing a line).

The previous implementation of the temporary bf-opacity storage (chi_bf_tmp_partial) as a member of the storage structure is not thread safe. This can be solved by assigning an individual temporary bf-opacity storage to each packet.

- add data structure for saving bound-free cross sections - function bf_cross_section returns bound-free cross sections using linear interpolation on tabulated data - add test for the calculation of bound-free opacities

Includes estimators for: -photoionization and stimulated recombination rate coefficients -bound-free heating rate coefficients -free-free heating rate coefficients -stimulated recombination cooling coefficients

A k_packet represents a packet of thermal energy. Assuming radiative equilibrium k_packets are converted in situ back to r_packets. This involves selecting the type of the cooling process (ff- or bf-emission, collisional excitation or ionization), as well as selecting the individual cooling process (e.g collisional excitation to level i).

The fixture model_3lvlatom is added, which provides macro atom data (references, transition probabilities etc.) for a simple 3 level atom (including internal jumps and branching).

-move testing for close lines into a function of its own -move handling of line emission events into a a function of its own

K-packets can be treated as a macro atom level. Having two frameworks for sampling from discrete probabilities is redundant.

Tests the correct selection of a bound-free continuum for absorption.

wkerzendorf · 2016-11-24T16:55:38Z

tardis/montecarlo/struct.py

@@ -114,6 +135,18 @@ class StorageModel(Structure):
 CONTINUUM_OFF = 0
 CONTINUUM_ON = 1

+# Variables corresponding to `emission_type` enum.
+BB_EMISSION = -1


There is a Python Enum package that does that. Maybe we want to use it here: https://pypi.python.org/pypi/enum34 - it will become a built-in with python 3.4

wkerzendorf · 2016-11-25T09:59:23Z

tardis/montecarlo/src/cmontecarlo.c

@@ -94,6 +94,51 @@ line_search (const double *nu, double nu_insert, int64_t number_of_lines,
  return ret_val;
 }

+tardis_error_t


see if this is needed.

unoebauer · 2016-12-05T10:48:11Z

tardis/montecarlo/montecarlo.pyx

@@ -74,7 +79,8 @@ cdef extern from "src/cmontecarlo.h":
        double inverse_sigma_thomson
        double inner_boundary_albedo
        int_type_t reflective_inner_boundary
-        double *chi_bf_tmp_partial
+        photo_xsect_1level ** photo_xsect


Hmm - double pointers. Do we really want these?

I'm not familiar with the exact usecase but I think double pointers can easily and should be avoided here.

Using double pointers can be easily avoided by using an array of structures instead of an array of pointers to structures. Do you think this is preferable? Since the structures are small (two pointers and an integer) this should be ok. Alternatively it is possible to store the cross section data (frequencies and cross sections) consecutively in two arrays. In this case a reference array is needed to know where the data for a certain level starts and ends.

If possible, you should store the data as a structure of arrays.

As an example this would change double *nu = storage->photo_xsect[continuum_id]->nu; to double *nu = storage->photo_xsect->nu[continuum_id];

Let's leave this for now! Minor issue, which we can address later on if we want.

unoebauer · 2016-12-05T10:54:58Z

tardis/montecarlo/src/cmontecarlo.c

+        / (storage->photo_xsect[continuum_id]->nu[result] - storage->photo_xsect[continuum_id]->nu[result-1])
+        * (storage->photo_xsect[continuum_id]->x_sect[result] - storage->photo_xsect[continuum_id]->x_sect[result-1]);
+      return bf_xsect;
+    }


Hydrogenic approximation should be included

unoebauer · 2016-12-05T11:05:53Z

Would be nice to have a brief description and list of milestone, so people see where this PR is heading to.

unoebauer · 2016-12-20T10:51:30Z

I'd like this PR to get merged before the Christmas break! Can you identify the missing parts for doing so, @chvogl?

chvogl added 26 commits October 18, 2016 13:47

Add function for sampling frequency of free-free emission

006ba69

Add test for sampling the frequency of free-free emission

0ca9e48

Add function for sampling frequency of bound-free emission

43b1b00

Add function for sampling of an individual cooling process

6b9b45f

For example for cooling by free-bound emission a continuum for emission is selected (i.e. a continuum_id).

Add function to calculate the free-free opacity

734493b

Move data for C continuum tests into single hdf file

f1f3874

- remove deprecated npz files with free-free emissivities - refactor test_sample_nu_free_free and related fixtures to adjust to new data format

Add test for sampling of individual cooling processes

588e755

To allow easier testing the fixture get_rkstate is added. This fixture allows to prepare the mt_state of the random number generator such that requested random numbers are generated.

Add test for calculation of free-free opacity

7904bbc

Add test for montecarlo_continuum_event_handler

ae8e9ed

Restructure compute_distance2continuum

6e21e33

In addition the flag compute_chi_bf is added to prevent the recomputation of the continuum opacities in cases where no physical interaction occurs (e.g. passing a line).

Make chi_bf_tmp_partial a packet property

8f58233

The previous implementation of the temporary bf-opacity storage (chi_bf_tmp_partial) as a member of the storage structure is not thread safe. This can be solved by assigning an individual temporary bf-opacity storage to each packet.

Add better treatment of bound-free cross-sections and opacities

1e038fb

- add data structure for saving bound-free cross sections - function bf_cross_section returns bound-free cross sections using linear interpolation on tabulated data - add test for the calculation of bound-free opacities

Add estimators for continuum processes

f8ece86

Includes estimators for: -photoionization and stimulated recombination rate coefficients -bound-free heating rate coefficients -free-free heating rate coefficients -stimulated recombination cooling coefficients

Add function to handle continuum emission events

d04addf

Add test for cmontecarlo function get_event_handler

31ae2f0

Add test for cmontecarlo function macro_atom

56804a8

The fixture model_3lvlatom is added, which provides macro atom data (references, transition probabilities etc.) for a simple 3 level atom (including internal jumps and branching).

Restructure montecarlo_line_scatter function

4b16fd6

-move testing for close lines into a function of its own -move handling of line emission events into a a function of its own

Remove k_packet and associated content

0af724e

K-packets can be treated as a macro atom level. Having two frameworks for sampling from discrete probabilities is redundant.

Add test for test_for_close_line

0d8ee37

Rename next_interaction2process enum and remove deprecated items

e960386

Adapt macro_atom to handle continuum processes

159d839

Add correct handling of free-free absorption events

801a025

Fix formatting

51ef965

Add correct handling of bound-free absorption events

ceece25

Add test for montecarlo_bound_free_scatter

70a29c5

Tests the correct selection of a bound-free continuum for absorption.

wkerzendorf reviewed Nov 24, 2016

View reviewed changes

wkerzendorf reviewed Nov 25, 2016

View reviewed changes

Register continuum_compare_data.hdf as package_data

ecfefbf

unoebauer reviewed Dec 5, 2016

View reviewed changes

chvogl added 4 commits December 5, 2016 15:35

Add adiabatic cooling

a202ff0

Add hydrogenic approximation for bound-free cross sections

515f49f

Remove deprecated attributes from model fixture

66db9c8

Remove unnecessary padding of chi_bf_tmp_partial

b46ff14

chvogl added 3 commits December 20, 2016 13:11

Extend packet logging to continuum interactions

0062c62

Remove not yet relevant tests section

ad60e1c

Restructure montecarlo_continuum_event_handler

0770724

unoebauer approved these changes Jan 9, 2017

View reviewed changes

wkerzendorf merged commit cd8f195 into tardis-sn:master Jan 9, 2017

chvogl mentioned this pull request Jan 27, 2017

Missing openmp compatability of chi_bf_tmp_partial array #348

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Integrate continuum interactions into cmontecarlo #670

Integrate continuum interactions into cmontecarlo #670

chvogl commented Nov 24, 2016 •

edited

Loading

wkerzendorf Nov 24, 2016

wkerzendorf Nov 25, 2016

unoebauer Dec 5, 2016

yeganer Dec 5, 2016

chvogl Dec 6, 2016 •

edited

Loading

yeganer Dec 7, 2016

yeganer Dec 7, 2016

unoebauer Dec 20, 2016

unoebauer Dec 5, 2016

unoebauer commented Dec 5, 2016

unoebauer commented Dec 20, 2016

Integrate continuum interactions into cmontecarlo #670

Integrate continuum interactions into cmontecarlo #670

Conversation

chvogl commented Nov 24, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chvogl Dec 6, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

unoebauer commented Dec 5, 2016

unoebauer commented Dec 20, 2016

chvogl commented Nov 24, 2016 •

edited

Loading

chvogl Dec 6, 2016 •

edited

Loading