How to use Sherpa 1.0.8 together with Athena (12.0.2 or higher).

Sherpa is a MC generator that at the moment has to be run outside the Athena framework. It can write out events in a native ASCII format (so-called Blobs), that can be read back into Athena using the ReadSherpa _i package, which is present in Athena since 12.0.2. It is recommended to use 12.0.4 or higher since there was a problem with ReadSherpa_i crashing (of the order of once per 10k events) in earlier versions.

Installing and modifying Sherpa 1.0.8

In order to install Sherpa 1.0.8 go to and download version 1.0.8. Make sure to also download and apply the patches. There is one final thing that has to be changed:

In SHERPA-1.0.8/Tools replace the file Input_Output_Handler.C with this one: Input_Output_Handler.C: This is necessary to set a higher precision for the filestream that Sherpa uses when writing out files. The default precision can lead to negative mass squares due to rounding errors. In future Sherpa versions it will be possible to set the precision at run-time.

Now go ahead and compile Sherpa:

./TOOLS/makeinstall -t -r

If the compilation fails, probably some requirements (gcc-version, autoconf etc) are not satisfied. Please refer to the Sherpa users guide or to the Sherpa-team for assistance.

Configuring Sherpa

Sherpa is steered by a number of .dat files. All of them are explained in detail in the Sherpa manual. Here are some hints on possible modifications:


By default, Sherpa uses Lund for the fragmentation. It also uses Lund for decays of unstable hadrons, but it also has an own decay package. See below for details.


The parameters here are the Sherpa Lund-tune. It is not advised by the Sherpa authors to change anything, since the parton shower has been tuned together with these parameters. However, when using Lund also for hadron decays, it is necessary to add the line
MSTJ(22) = 2
This sets all hadrons with ctau<10mm to be stable. This is the same setting as the Atlas default for Pythia.


Per default no underlying event is turned on. To change this, set:
MI_HANDLER      = Amisic                  ! Amisic / None
In addition, studies (Link ?) have shown that the underlying event description can be optimized by setting:
SCALE_MIN           = 2.57              ! minimum scale


In this file, four columns might be of interest:
Mass and Width
Here you can set the mass and width of any particle that is used in Sherpa. Useful e.g. for generating Higgs signal events with different masses and widths.
This switch determines if the particle is considered when Sherpa calculates the matrix-elements for the selected processes. Useful e.g. for turning off Z and gamma when generating Higgs signal events.
set it to 0 for the tau to get unstable taus (details in Fragmentation.dat).
This flag actually determines which quarks are included in Sherpas jet-container (93). Per default, only u,d,s,c,g are included. If you want to have the b's in there as well (from the b-pdf), set m_on to 0 for the b.


Here specify the processes to be simulated. For the matrix-element-parton-shower-matching please refer to the Sherpa manual. It is also possible to specify decay-chains.


Here you can set the number of events to be generated. Other flags that might be useful are:
specifies a directory where the results of the integration steps are stored. Has to be created first.
If this flag is given, Sherpa will write out events to a file sherpa.evts. After a certain number of events, a new file sherpa.1.evts is created and so on. After how many events a new file is opened can be steered by
FILE_SIZE = 5000
, here 5000. In addiction one can give random seeds:
RANDOM_SEED = 1234 6789
Refer to the Sherpa manual for the range of allowed seeds.

VERY IMPORTANT NOTE when giving the random seeds at the command line you have (a) to leave out the spaces adjacent to the '=' and (b) it was found out that only the first seed is used by Sherpa (and correspondingly also the older random-generator). In summary if you provided the seeds in Run.dat everything is fine, if you provided them on the command line and did not change the first number you regenerated the same events again !

UPDATE when giving two seeds at the command line, you have to give them in quotation marks, e.g.:

RANDOM_SEED="1234 6789"

All these parameters can also be set on the command-line, to facilitate working in a batch-environment.


Take care to specify the correct kinematic cuts. E.g.
JetFinder   sqr(20/14000) 1.
specifies the matching scale to be 20 GeV.
Mass 11 -11 66 116
cuts on the mass of the electron-positron pair from the hard process to be between 66 and 116 GeV. Setting the lower mass cut i.e. to 0 GeV would be very bad for the integration because of the divergent contribution of the virtual photon.

Running Sherpa.

In order to run Sherpa, change to SHERPA-1.0.8/Run/. There is a subdirectory LHC which contains a set of steering files that can be used for LHC processes (after the above modifications are done).

In order to run, Sherpa has to be run twice:

./Sherpa PATH=LHC/
cd LHC
cd ..
./Sherpa PATH=LHC/
The first pass is necessary to write out the necessary libraries, which are then compiled. On the second pass the cross-sections are integrated, written into the RESULT_DIRECTORY (if it is specified and exists) and the requested number of events is generated. The integration step has to be done only once.

Using Sherpa to decay tau-leptons

In the default Sherpa settings (Lund for hadron decays), taus are stable. However one can decay taus (and only taus) by a few modifications:

In SHERPA-1.0.8/Run there is a directory

There is a file HadronDecays.dat. Change it to the following (it might be good to back it up first):
!-- Files for hadron decay tables ----------------------

! particle -> directory/ name.dat do not forget the / and space
15    -> TauDecays/         TauDecays.dat;

In addition, change in Fragmentation.dat:

DECAYMODEL    = Hadrons       ! Lund or Hadrons
and in Particle.dat set the tau to be unstable:
kf       Mass   Width   3*e  Y    SU(3)  2*Spin maj  on stbl m_on  Name
1        .01     .0    -1    -1      1      1    0    1   1   0     d
2        .005    .0     2     1      1      1    0    1   1   0     u
3        .2      .0    -1    -1      1      1    0    1   1   0     s
4       1.42     .0     2     1      1      1    0    1   1   0     c
5       4.7      .0    -1    -1      1      1    0    1   1   0     b
6     174.3      1.5    2     1      1      1    0    1   1   1     t
11      .000511  .0    -3    -1      0      1    0    1   1   0     e-
12      .0       .0     0     1      0      1    0    1   1   0     nu_e
13      .105     .0    -3    -1      0      1    0    1   1   0     mu-
14      .0       .0     0     1      0      1    0    1   1   0     nu_mu
15     1.777    2.36E-12 -3  -1      0      1    0    1   0   1     tau-
Now all decays except the tau-lepton are handled by Lund and the taus are decayed by Sherpa itself.

One can also create a complete new directory for the hadron decays and then set in Fragmentation.dat:

DECAYPATH     = Decaydata_onlytau/ ! directory of decay data files (note: / at the end)
IF the directory is a subdirectory of SHERPA-1.0.8/Run or one can give the complete path to it.

Steering Sherpas tau-decays:

In the DECAYPATH directory (as above), just edit
and comment out or delete all decays that are not wanted, i.e. if you want only leptonic decays, it should read like:
# outgoing part.         | BR    | DC-file

# leptonic
{16,11,-12}      | 0.1784        | Tau_Electron.dat;
{16,13,-14}      | 0.1736        | Tau_Muon.dat;

Changes for using Sherpa 1.0.9 with the ReadSherpa\_i interface

Sherpa 1.0.9 was released on December 21st 2006. It contained an important bug-fix for the radiation off heavy quarks in the initial state which mainly affected the b-associated higgs production, see below. Some things have changed in 1.0.9. First of all, the old .dat-files are not all compatible to the new version, so it is recommended to start again from the example .dat files provided in the Sherpa-release. Here a list of the most important changes:

0. Sherpa generates, by default, 1000 events per file. When 1000 events are written to the file, a new file is created and chained to the previous one. Sherpa 1.0.8 did not save the extension ".evts" in the chained file. Therefore, ReadSherpa\_i places the extension and opens the next file as appropriate. Sherpa 1.0.9 "corrected" this situation and places the trailing ".evts" to the chained filename. As a result, ReadSherpa\_i will attempt to open a file with the extension ".evts.evts" when inputting files generated with Sherpa V1.0.9

There is no elegant workaround, except to generate 1000 events per file, and invoke ReadSherpa\_i for each file (specifying 1000 events). This is fixed in Release 13. One can change the number of events per file via the FILESIZE=XXXX directive in Run.dat. The number of events processed by ReadSherpa_i must be modified accordingly. It is important to specify the 1000 ( or whatever number ) in the job options file. Otherwise, ReadSherpa\_i will attempt to chain to the next non-existent file with ".evts.evts" extension.

1. Lund.dat: Do not set MSTJ(22)=2 anymore. Instead there is a switch

that should be set in Fragmentation.dat. This sets the maximum ctau in mm for a particle to be decayed. It works for the HADRONS and the Lund model as well.

2. Output precision. The output precision can now be set at runtime. Just add the switch

to get 11 digits.

3. MI.dat: Amisic is still truned off by default - the underlying event model is unchanged in 1.0.9 wrt. 1.0.8, so the same tune

SCALE_MIN           = 2.57
can be changed as a starting point.

4. ME.dat: The switches for the scle schemes have changed extensively, just refer to the sherpa manual to what they mean now.

Setups for already requested CSC-samples

Samples 5356 and 5357 (bbA, A to mumu, m_A=110 GeV, tan(beta)=45 and m_A=200 GeV, tan(beta)=30)

WARNING ! There was a bug in Sherpa 1.0.8 that affected the b-associated higgs production. The problem was that b's in the initial state didn't radiate properly, which led to unphysical kinematical distributions (unphysical peak at 0 transverse momentum of the higgs). Samples 5356 and 5357 were redone with Sherpa 1.0.9 where the problem was fixed properly. Documentation of the changed setup will be made available shortly. If needed contact me at

These samples were done with Sherpa 1.0.8. A standard model higgs boson was used where the masses (110 GeV for 5356 and 200 GeV for 5357) were set in Particle.dat. The difference between a standard model higgs (scalar) and the pseudoscalar A of the MSSM are small (effects of spincorrelations scale as m_b/m_higgs). The only difference is the width which was manually set in Particle.dat. The width was taken from Feynhiggs 2.4.1 for a A boson with tan(beta)=45 (for 5356) and tan(beta)=30 (for 5357).

Changes to the standard input files:

1. Lund.dat as above.

2. MI.dat as above.

3. Fragmentation.dat, modified as above to use Sherpa to decay all taus (all tau decays), all other decays are then handled by Lund.

4. Particle.dat: Set taus unstable, set m_on=0 to have b's included in the jet-Container.

5. Model.dat: Set Yukawa-coupling to c-quark to 0 to ensure that the Higgs is radiated off b-quarks

6. ME.dat: Set Yukawa masses to running (should mostly only affect the cross-section).

To run the attached setup (setup5356.tar and setup5357.tar), just untar the tarball and say

/PATH/TO/SHERPA/Sherpa PATH=. RESULT_DIRECTORY=./Results/ SHERPA_OUTPUT=sherpa0108.nicename._0000X EVENTS=whatever

WARNING ! Setups were removed and will be updated with the ones for Sherpa 1.0.9 shortly.

The libraries are not included in the tarball, just tell me if you need them.

Feynmangraphs included in Sherpa.

Sherpa can write the Feynmangraphs included in the selected processes. Just fill in
in the process declaration. It will write out .tex-files that have to be processed by the Feynmf-package and Metafont. The graphs for bbh are attached.

Sample 5358 (bbA, A to tautau, m_A=130 GeV, tan(beta)=20)

WARNING ! This sample was affected in the same way as 5356 and 5357 and was changed accordingly. Don't use Sherpa 1.0.8 for b-associated higgs production.

Setup is mostly identical to the one for 5356 and 5357, main difference is that the taus are forced to decay leptonically.

Sample 8185 (Z->mumu plus 0 to 3 jets).

Setup is mostly identical to 5356 and 5357, expect:

- no decay chain is given, instead the Higgs is turned off in Particles.dat, and Z and Photon are switched on (giving correct gamma-Z-interference).

- Selector.dat contains a mass cut of 60 GeV for the invariant dimuon-mass.

- b-Quarks are again included in the jet-container.

Feynmangraphs are again attached.

-- MarkusWarsinsky - 25 Sep 2006

Topic attachments
I Attachment History Action Size Date Who Comment
C source code filec Input_Output_Handler.C r1 manage 14.9 K 2006-09-25 - 16:59 MarkusWarsinsky Input_Output_Handler.C
Unknown file formatgz graphs_Zto3jets.tar.gz r1 manage 2715.3 K 2006-11-10 - 16:00 MarkusWarsinsky Feynmangraphs included by Sherpa for 8185.
Unknown file formatgz graphs_bbh.tar.gz r1 manage 88.9 K 2006-11-10 - 15:59 MarkusWarsinsky Feynmangraphs included by Sherpa for 5356 and 5357.
Unknown file formatgz setup8185.tar.gz r1 manage 421.4 K 2006-11-10 - 15:50 MarkusWarsinsky Setup for CSC sample 8185
Edit | Attach | Watch | Print version | History: r8 < r7 < r6 < r5 < r4 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r8 - 2007-10-02 - AndreasLudwig
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback