[Rivet-svn] rivet: Add a better jet smearing resolution parametrisation, bas...

[Rivet Mercurial]

details:   https://rivet.hepforge.org/hg/rivet/rev/6917cfb1876e
branches:  
changeset: 5989:6917cfb1876e
user:      Andy Buckley <andy at insectnation.org>
date:      Mon Sep 04 15:43:49 2017 +0100
description:
Add a better jet smearing resolution parametrisation, based on GAMBIT code from Matthias Danninger.

diffs (33 lines):

--- a/ChangeLog	Fri Sep 01 12:57:34 2017 +0100
+++ b/ChangeLog	Mon Sep 04 15:43:49 2017 +0100
@@ -1,3 +1,7 @@
+2017-09-04  Andy Buckley  <andy.buckley at cern.ch>
+
+	* Add a better jet smearing resolution parametrisation, based on GAMBIT code from Matthias Danninger.
+
 2017-08-16  Andy Buckley  <andy.buckley at cern.ch>
 
 	* Protect make-plots against NaNs in error band values (patch from Dmitry Kalinkin).
--- a/include/Rivet/Tools/SmearingFunctions.hh	Fri Sep 01 12:57:34 2017 +0100
+++ b/include/Rivet/Tools/SmearingFunctions.hh	Mon Sep 04 15:43:49 2017 +0100
@@ -520,10 +520,17 @@
 
 
   /// ATLAS Run 1 jet smearing
-  /// @todo This is a cluster-level flat 3% resolution, I think, and smearing is suboptimal: improve!
   inline Jet JET_SMEAR_ATLAS_RUN1(const Jet& j) {
-    // Const fractional resolution for now
-    static const double resolution = 0.03;
+    // Jet energy resolution lookup
+    //   Implemented by Matthias Danninger for GAMBIT, based roughly on
+    //   https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/CONFNOTES/ATLAS-CONF-2015-017/
+    //   Parameterisation can be still improved, but eta dependence is minimal
+    /// @todo Also need a JES uncertainty component?
+    static const vector<double> binedges_pt = {0., 50., 70., 100., 150., 200., 1000., 10000.};
+    static const vector<double> jer = {0.145, 0.115, 0.095, 0.075, 0.07, 0.05, 0.04, 0.04}; //< note overflow value
+    const int ipt = binIndex(j.pT()/GeV, binedges_pt, true);
+    if (ipt < 0) return j;
+    const double resolution = jer->at(ipt);
 
     // Smear by a Gaussian centered on 1 with width given by the (fractional) resolution
     /// @todo Is this the best way to smear? Should we preserve the energy, or pT, or direction?