Hadron tensor framework & SuSAv2 model implementation #55

sjgardiner · 2019-08-28T22:44:10Z

This feature branch refactors the hadron tensor table code used by the Nieves MEC model into a more flexible and general framework. Key new features are:

Tensor table files are now self-describing, with locations configurable via XML (as opposed to being hard-coded)
New QE+MEC models based on the SuSAv2 approach serve as a proof of concept

See materials on GENIE docDB for more information. After building the susav2 feature branch, SuSAv2 QE+MEC events can be generated using the GTEST19_10b_00_000 tune identifier.

…or for the Valencia MEC model

hadronic tensor using a pre-interpolated TableEntry object. This will be helpful for the differential cross section implementation.

…adronX

type

…he new HadronTensorPool and HadronTensorI classes

ValenciaHadronTensorI inherit from it. The TabulatedHadronTensor class is now a subclass of ValenciaHadronTensorI called TabulatedValenciaHadronTensor.

done for the Nieves MEC model

… the HadronTensors.xml configuration file accordingly.

tensor objects (which may involve reading in a large file) as necessary (lazy initialization).

…dron tensor machinery (which will eventually be used by interaction modes other than MEC) into it. Adjust the build scripts accordingly.

Now this is done via the new HybridXSecAlgorithm class, which uses SuSAv2QELPXSec and LwlynSmithQELCCPXSec as sub-algorithms.

store a null pointer in the owned map to avoid redundant failed lookups in the future.

…imum differential cross section in MECGenerator::SelectSuSALeptonKinematics()

MECUtils source file. This will allow it to be re-used in other cases where we need to scan for the maximum differential xsec in the same phase space (e.g., for SUSAv2 1p1h)

CCQE+CCMEC and NCEL+NCMEC lists)

used for EMQE+EMMEC

…ntial cross section. This avoids problems with missing the minimum Q^2 value. Rejection sampling is still quite inefficient for EMMEC, but it's usable.

SuSAv2

Q^2 cut in the SuSAv2 QEL and MEC cross sections

This reverts commit f4d18d0.

QE response was a factor of two too large.

…SAv2 QE + MEC cross sections.

than scaling by A, we should scale by the number of nucleons matching the struck nucleon. This gets the ratio of initial-state p vs. n right when integrating splines.

QELEventGeneratorSuSA. Occasional problems were found when the cache was used, but there is a negligble impact on performance when the maximum diff xsec is directly calculated for each event. Reduce the safety factor for this class to 1.2 in the configuration XML file. Modify the choice of the layer boundaries in genie::utils::mec::GetMaxXSecTlctl so that the low Q^2 cut and the maximum Tl value are both always included. This will ensure that the maximum diff xsec is always found even when there are problems with round-off at the kinematic boundaries.

…RM in the GTEST19_10b_00_000 tune. This was done previously in the main EventGenerator.xml file, but the custom one for this tune was not updated.

genie::KNOTunedQPMDISPXSec class

genie::utils::mec::GetMaxXSecTlctl. This adjustment addresses a problem which shows up only at high energies.

differential cross section when generating SuSAv2 events.

mroda88

It works for me. Tested a number of nuclei and I have plots for the comparisons.

sjgardiner changed the title ~~Hadron tensor framework & SuSAv2 model implementation~~ Not merge - Hadron tensor framework & SuSAv2 model implementation Aug 28, 2019

sjgardiner force-pushed the susav2 branch 5 times, most recently from 0153637 to a096f65 Compare November 4, 2019 20:36

sjgardiner force-pushed the susav2 branch 2 times, most recently from 0e45869 to 6f54d91 Compare July 2, 2020 14:37

candreop added this to the 3.2.0 release milestone Jul 6, 2020

sjgardiner added 21 commits July 7, 2020 15:16

Fix incorrect variable names for compiler flags in Makefile.include

f536743

Add .gitignore

87ed770

Begin refactorization of the MEC hadron tensor calculation

cd28407

Finish first draft of the classes needed to compute the hadronic tens…

48fdae3

…or for the Valencia MEC model

Add protected helper functions that will compute W1, W2, etc. for the

d6fcc31

hadronic tensor using a pre-interpolated TableEntry object. This will be helpful for the differential cross section implementation.

Fix include guards, add code to compute the differential cross section

9ec1262

Switch to using the native GENIE functions to manipulate PDG codes

3c26b47

Add the HadronTensorPool class

5a31e6a

Fix a few minor bugs and rename the classes from ValenciaHadronX -> H…

8cf3375

…adronX

Add electron cross section calculation

d68b151

Add a value for electron hadronic tensors to the HadronTensorType_t enum

a4ad1fd

type

Refactor the NievesSimoVacasMECPXSec2016 cross section model to use t…

d0fac4b

…he new HadronTensorPool and HadronTensorI classes

Add the QE_Electron hadron tensor type

68ee566

Create a new base class called HadronTensorI, and make

6a2ef96

ValenciaHadronTensorI inherit from it. The TabulatedHadronTensor class is now a subclass of ValenciaHadronTensorI called TabulatedValenciaHadronTensor.

Switch to using a single "file" XML attribute for each hadron tensor

ab98b9e

Swap the interpolation order in BLI2DNonUnifObjectGrid to match what is

6380d5a

done for the Nieves MEC model

Use "_EM" instead of "_Electron" in HadronTensorType_t enumerated types

0395170

Add new versions of all of the Nieves MEC hadron tensor files. Update…

689808e

… the HadronTensors.xml configuration file accordingly.

Tweak the HadronTensorPool class so that it will only construct hadron

9cddc2e

tensor objects (which may involve reading in a large file) as necessary (lazy initialization).

Create a new src/Physics/HadronTensors/ folder and move all of the ha…

618ec58

…dron tensor machinery (which will eventually be used by interaction modes other than MEC) into it. Adjust the build scripts accordingly.

Remove the deprecated MECHadronTensor singleton class

6177a7b

sjgardiner added 22 commits July 7, 2020 15:16

Rename QELKinematicsGeneratorSuSA to QELEventGeneratorSuSA

dac59ef

Rework delegation of free nucleon cross section calculations for SuSAv2.

a58a411

Now this is done via the new HybridXSecAlgorithm class, which uses SuSAv2QELPXSec and LwlynSmithQELCCPXSec as sub-algorithms.

In cases where a TabulatedHadronTensorModelI fails to load a table file,

c1117e7

store a null pointer in the owned map to avoid redundant failed lookups in the future.

Use a brute-force scan instead of a hard-coded expression for the max…

cca30fb

…imum differential cross section in MECGenerator::SelectSuSALeptonKinematics()

Move the GetMaxXSecTlctl() out of the MECGenerator class and into the

efc1134

MECUtils source file. This will allow it to be re-used in other cases where we need to scan for the maximum differential xsec in the same phase space (e.g., for SUSAv2 1p1h)

Add the "EMQE+EMMEC" event generator list (similar to the existing

bf6a9d0

CCQE+CCMEC and NCEL+NCMEC lists)

Tweak test tune S19_10b_00_000 configuration so that SuSAv2 can also be

13a44dd

used for EMQE+EMMEC

Switch to working with Tl and Q^2 to estimate the maximum MEC differe…

60070ac

…ntial cross section. This avoids problems with missing the minimum Q^2 value. Rejection sampling is still quite inefficient for EMMEC, but it's usable.

Use approach similar to SuSAv2 MEC to get maximum 1p1h cross section for

fffc8ec

SuSAv2

Switch to using min Q^2 = 0.4

1d3824b

Enable RES-EM and DIS-EM for the S19_10b_00_00 tune. Enforce the global

3375b34

Q^2 cut in the SuSAv2 QEL and MEC cross sections

Revert "Switch to using min Q^2 = 0.4"

9642362

This reverts commit f4d18d0.

Fix library ordering problem in genie-config

9956ad4

Fix a problem in the hadron tensor table for SuSAv2 EMQE. The transverse

e368bf0

QE response was a factor of two too large.

Make minor adjustments to the ValidProcess member function for the Su…

2f9a661

…SAv2 QE + MEC cross sections.

Remove unneeded headers from HybridXSecAlgorithm.cxx

ade4816

Rename the temporary SuSAv2 testing tune to "GTEST19_10b"

6314c85

Fix issue with scaling of nuclear EMQE cross section for SuSAv2. Rather

6e19d72

than scaling by A, we should scale by the number of nucleons matching the struck nucleon. This gets the ratio of initial-state p vs. n right when integrating splines.

Add an extra Q-value correction for the SuSAv2 EM MEC model

eb23294

Remove PauliBlocker from the event generation workflow for QEL-CC-CHA…

0bd978f

…RM in the GTEST19_10b_00_000 tune. This was done previously in the main EventGenerator.xml file, but the custom one for this tune was not updated.

Update the DIS model used by the GTEST19_10b_00_000 tune to use the

14a6592

genie::KNOTunedQPMDISPXSec class

sjgardiner force-pushed the susav2 branch from 6f54d91 to 14a6592 Compare July 7, 2020 20:19

sjgardiner added 2 commits July 9, 2020 14:25

Choose the phase space boundaries more carefully for the scan in

a10fc08

genie::utils::mec::GetMaxXSecTlctl. This adjustment addresses a problem which shows up only at high energies.

Switch to using XML-configurable tolerances for checking the maximum

02ab7ab

differential cross section when generating SuSAv2 events.

sjgardiner force-pushed the susav2 branch from 17efd22 to 02ab7ab Compare July 9, 2020 23:31

sjgardiner changed the title ~~Not merge - Hadron tensor framework & SuSAv2 model implementation~~ Hadron tensor framework & SuSAv2 model implementation Jul 10, 2020

sjgardiner self-assigned this Jul 10, 2020

mroda88 approved these changes Jul 15, 2020

View reviewed changes

mroda88 merged commit e9e9239 into GENIE-MC:master Jul 15, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Hadron tensor framework & SuSAv2 model implementation #55

Hadron tensor framework & SuSAv2 model implementation #55

sjgardiner commented Aug 28, 2019 •

edited

Loading

mroda88 left a comment

Hadron tensor framework & SuSAv2 model implementation #55

Hadron tensor framework & SuSAv2 model implementation #55

Conversation

sjgardiner commented Aug 28, 2019 • edited Loading

mroda88 left a comment

Choose a reason for hiding this comment

sjgardiner commented Aug 28, 2019 •

edited

Loading