Changes between Version 64 and Version 65 of Reweight

Timestamp:

Mar 25, 2022, 11:45:09 AM (2 years ago)

Author:

Olivier Mattelaer

Comment:

--

Reweight

@@ -169 +169 @@
 
    1. The re-weighting assumes that the matrix-element is symmetric for identical particles. This is a limitation for process like
-      generate p p > Z j j, z > j j. In general such issues are automatically detected but in some particular case the re-weighting can fail to detect it and produce result which might be un-expected.
+      generate p p > Z j j, z > j j. Starting at 3.4.0, the re-weighting is done as an average of the various combination to alleviate the ambiguity. 
+      This should give a good approximation as long as the Narrow-Width-Approximation is valid. 
+      You can change the parameter "identical_particle_in_prod_and_decay" to "max" to only consider the highest weight evaluated, or set it to "crash" to make the code to crash.
+
+      The option for version older than 3.4.0 is still valid: you can set the parameter "keep_ordering" to  True (via "change keep_ordering True"). This assumes that particle written in the leshouches file are in the exact same order as the one used by madgraph (looks at the Feynman diagram for the the ordering) --use with great care-- (new in 2.7.0)
+
    2. The ordering of the particles in the event file is not guaranteed and information about internal propagator not save to use. Consequently process like
       "p p > e+ e- X, X > mu+ mu-" and "p p > mu+ mu- Y, Y > e+ e-" can not be distinguish. The re-weighting will systematically crash when this happens.
 
-   Work around for those limitations can be done by setting some special flag in the reweight_card and carefully generating the input event:
-   a. '''use_eventid True''': This allow to use the event id flag of the leshouche event file to distinguish different process. Within MG5aMC, you should use the "@" syntax in order to put such eventid in the leshouches file: for example "generate p p > e+ e- X, X > mu+ mu- @1; add process p p > mu+ mu- Y, Y > e+ e- @2" (new in 2.7.4)
-   b. '''change keep_ordering True''': This assumes that particle written in the leshouches file are in the exact same order as the one used by madgraph (looks at the Feynman diagram for the the ordering) --use with great care-- (new in 2.7.0)
-
+   Work around for this limitation: '''change use_eventid True''': This allow to use the event id flag of the leshouche event file to distinguish different process. Within MG5aMC, you should use the "@" syntax in order to put such eventid in the leshouches file: for example "generate p p > e+ e- X, X > mu+ mu- @1; add process p p > mu+ mu- Y, Y > e+ e- @2" (new in 2.7.4)
+  
 
  == Content of the reweight_card
@@ -194 +197 @@
    9. '''change virtual_path PATH''': Allow to use an external library of standalone matrix-element (see details [#Generateexternallibrary here]) for the new theory for the virtual contribution. '''new in 2.5.6'''
    10. '''change systematics OPTS''': '''only for ouptut mode '2.0' ''', allows to run the systematics computation on the new flight with options "OPTS". Note that this assume that the orignal and final theory have the SAME power of alpha_s at born level. '''new in 2.5.6'''
-   11. '''change use_eventid''': see above
-   12. '''change keep_ordering'': see above --dangerous flag-- 
+   11. '''change use_eventid''': see above (new in 2.7.4)
+   12. '''change keep_ordering'': see above --dangerous flag-- (new in 2.7.0)
+   13. '''change identical_particle_in_prod_and_decay [average|max|crash]''': handling the case where particle identity is not secure (see above) (new in 3.4.0) 
 
  2. '''Benchmark definition''':[[BR]]