Physics
Generation for Atlas Data Challenges
Continuously updated/always incomplete
Last
update November 19 2004
Production has
finished.
In releases prior to
8.0.4 there is a confict between tauola and atlfast which prevents them
being run in the same job. This is fixed in 8.0.4 and later. There are
no other code changes between 8.0.1 and 8.0.x that affect event
generation.
Note that the
samples made with MC@NLO contain weighted events. You must use the
event weight (which is found in EventInfo) when making plots (hf1(your
thing, wt)). Some versions of atlfast do not use the weights in the
ntuples. These ntuples are therefore unphysical
This page deals only with generation issues.
For simulation and reconstruction go here
Ian Hinchliffe
Generator
Software
status in releases
Contact persons
HLT Streams
DC2
General comments
- Important
note regarding Pythia.
Note
that the events with Pythia (6.221) have tuned underlying event.
Do not
change the structure functions or any of the following parameters.
There is an interplay between these parameters and the structure
functions which means that they cannot be varied in a simple way.
MSTP(82)=4,
PARP(82)=1.8 PARP(84)=0.5.
We are using mstp(128)=0 in DC2. mstp(128)=2 was used in DC1 and gives
a simpler structure. This change was requested by physics users.
- Note that all output is to AthenaPool. AthenaRoot should not be
used . Here is a JobOptions
that read the Pool File, runs atlfast and
puts out an Ntuple.
- Herwig. We have a version
of Herwig including Jimmy running in atlas release 8.0.1. It is under
active testing/tuning. Note that some samples must use herwig. Note that Jimmy will
not be used for DC2. Some of the samples below will be rerun later
using Jimmy.
- Tauola/Photos. These should probably be on for processes that
care about the details of tau decay, such as certain Higgs samples.
There is a problem with the Herwig/Taoula interface that affects the
SUSY events (see below)
- Some events will be filtered before being passed to G4. In most
cases, the generated events will be written and the filter run in the
simulation job. The event numbers below refer to the simulated
events. The generated event samples will be larger.
- Approved JobOptions for Event
Generation appear in CVS under
Physics/GeneratorOptionsDC2/share. A tag of this will be placed in the
release that the files run with. To be safe you are recomended to check
out the latest tag. No compilation is needed as this package only
contains runtime files. You
can see the head of CVS here
- Production is done with 8.0.4 or
later (due to the atlfast/tauola conflict, mentioned above)
This 10M event sample will satisfy
the requirement of many trigger streams and a fairly rich sample
without
having too much bias in the physics. There are two classes of
events. Events in class B have a much larger cross-section. The
numbers below are the number of events and the bias that must be put back to restore the
true cross-sections.: sigma =Nevents/bias.
We have too many top and too few W_jets events for example. Its normalized to the
Z rate. Note that within each class the biases
are very similar, so that the unbiased stream is almost realistic within each class. There is
no way to avoid the large difference in biases between the two classes. If
the
class
B events are filtered out, then a
reasonable physics sample is left.
Events will be generated and saved as Pool files.
Samples that require
filtering will have the filter applied after the Pool file is read and
before simulation is run. The samples that are to be filtered in this
way are B4, B5 and H1-H9. Larger samples must be generated in these
cases.
Some of the samples need external event files at generation time. These
are
A0, A4, A5 A7. In these cases, sample files are in the cvs area.
For
intructions where to the the rest, please read this
file
Note that
this is the final layout (not the initial layout) and that single
particle samples are not listed here.
Production of these samples is expected
to begin in tomorrrow.
After the events are generated, the
atlfast ntuples are looked at by someone in the appropriate physics
group. As part of this I am asking people to fill in the HLT table below .
After the "sign off" events go to simulation.
Class
A:
A0:
1M top (bias=2.92 )
MC@NLO (+herwig) used . Tauola is on. THIS SET IS DONE.
A0a: Note that 0.5M of these
top events
will be also be simulated with misalignment
A1:
1M Z (to ee),
no pt
cut. (bias=1) Pythia . THIS SET IS BEING
GENERATED: TEST FILE
HAS BEEN VALIDATED
A2:
1M Z (to mumu ),
no pt
cut. (bias=1), Pythia . THIS SET IS DONE. TEST FILE HAS BEEN VALIDATED
A3: 1M Z (to tau tau),
no pt
cut. (bias=1), Pythia. Note that Tauola is on THIS SET IS DONE. TEST FILE HAS BEEN VALIDATED
A4: 1M W (to leptons)
MC@NLO (+herwig) used as
this will give good representation of W+2 jets. For this sample the W
decays have to be generated separately. Therefore this sample consists
of six subsamples for W+ decaying to e, mu and tau and W- decaying to
e, mu and tau. There should be 167K of each. THIS SET IS DONE
using 8.0.5 or 8.0.4 with GeneratorOptionsDC2-00-00-11
A5:
0.5M Z+jet sample. Z
decays to leptons only (including neutrinos). Made with Madcup and
hadronized with Pythia. THIS SET IS
READY TO START GENERATION.TEST FILE
HAS BEEN VALIDATED
A6: 0.25M dijets (pt>600)
(bias=1.36) THIS SET IS READY TO
START GENERATION
A7: 0.25M W(to leptons)
+4 jets with Alpgen and Pythia Approximate cuts
are p_T>15 GeV for jets and leptons THIS SET IS DONE
A8: 0.5M QCD events filtered to
provide a sample with at least 1 b-jet with
pt>15 and eta<2.5 Note that no
additional filtering is required after generation. THIS
SET IS DONE. TEST
FILE HAS BEEN VALIDATED
A9: 200K
susy coannihilation point. THIS SET IS DONE.
(note that tauola is OFF due to a possible
problem with the tauola interface in susy events). The Physics/GeneratorOptionsDC2/share files
changed on June 11 (tag GeneratorOptionsDC2-00-00-15 or later)
A10:
100K Higgs to tau tau THIS SET IS DONE. NOTE THAT TAG GeneratorOptionsDC2-00-00-20 or later MUST
BE USED)
A11:
50K DC1 susy. This is data
set 2315 from DC1. Used primarily for G4 validation. Old
generated events need conversion to force K and Lambda to be
stable. THIS SET IS DONE
Class
B:
B1
1 M jets (pt>180) (bias=
0.001) THIS SET IS DONE
B2:
0.2 M gamma+jet
(p_t>20)
(bias=0.00125) THIS SET IS DONE. TEST FILE
HAS BEEN VALIDATED
B3 : 0.25M
bb->B->(mu6mu6) THIS
SET IS DONE
B4: 1M jets with
(p_T>17), prefiltering to provide
sample of
e/gamma fakes. bias depends on Filter. THIS SET IS DONE, The filter
passes 5%:
therefore 20M events need to be generated. The
generator threshold is lowered to 15 GeV so the filter is less
efficient than DC1. TEST FILE
HAS BEEN VALIDATED
B5: 50K gamma_jet prefiltering
to provide sample of gamma- gamma fakes.
bias depends on Filter.THIS SET IS DONE, The filter passes 6%: therefore almost 1M events need to be
generated. The generator
threshold is lowered to 15 GeV so the filter is less efficient than
DC1. TEST FILE
HAS BEEN VALIDATED
Miscellaneous Higgs samples (these
have smaller cross sections than class A and therefore have a much
larger bias in the mixed sample.
H1:
40K Higgs to 4 leptons. Higgs
mass 130 GeV. Pythia. Events will be filtered to give all leptons with
p_T>5 GeV and |eta| < 2.7. The filter is 61% efficient. 70K
events will be generated. THIS SET IS DONE
H2 40K
Higgs to 4 leptons. Higgs mass 180 GeV. Pythia. Events will be
filtered to give all leptons with p_T>5 GeV and |eta| < 2.7. The
filter is 72% efficient. 60K events will be generated. THIS SET IS DONE
H3 15K
Higgs to gamma+gamma. Higgs mass 120 GeV. Pythia. Events will be
filtered to give photons with p_T>20 GeV and |eta| < 2.7. The
filter is 74% efficient. 25K events will be generated. THIS SET IS DONE
H4 15K
Higgs to WW. Both W decays to e or mu. Vector boson fusion process.
Higgs mass 170 GeV. Pythia. The filter passes 72% of the events. 25K
will be generated.THIS SET IS DONE
H5:
15K
Higgs to WW. One W decays to e or mu, the other to hadrons Vector
boson fusion process. Higgs mass 170 GeV. Pythia. Events will be
filtered so that lepton passes trigger. The
filter passes 44%. 50K will be generated. THIS SET IS DONE
H6: 15K
Higgs to tau tau. Both taus decay to e or mu. Vector
boson fusion process. Higgs mass 115 GeV. Pythia. Events will be
filtered so that leptons pass trigger.
Events will be
filtered so that leptons pass trigger. The
filter passes 9%. 200K will be generated.THIS SET IS DONE
H7: 15K
Higgs to tau tau. One tau decays to e or mu, the other to hadrons .
Vector
boson fusion process. Higgs mass 115 GeV. Pythia. Events will be
filtered so that lepton passes trigger. The
filter passes 48%. 35K will be generated.THIS SET IS DONE (THERE IS A TYPO IN
THE OPTIONS IN RELEASE 8.0.4 USE 8.0.5 OR or 8.0.4 with
GeneratorOptionsDC2-00-00-11 or later)
H8:
15K MSSM Higgs producing bbA final state
with A decaying to mu mu. A mass 300 GeV Pythia. Events will be
filtered so that leptons pass trigger. The
filter passes 89%. therefore no filter will be applied in
this case NOTE THAT TAG GeneratorOptionsDC2-00-00-20 or later MUST
BE USED)
H9: 15K
MSSM Higgs producing bbA final state with A decaying to b b, A mass 500
GeV Pythia. Events will be
filtered so that there are two b's with |eta|< 2.5 and at
least 4 jets with pt>20 GeV. There is no
generator level filter for b's. Therefore these events are not filtered
as the jet filter is very efficient. NOTE THAT TAG GeneratorOptionsDC2-00-00-20 or later MUST
BE USED)
Pile
up
Pythia will be used for the min-bias pile
up events. Herwig/Jimmy cannot be tuned properly.
M1 : 500K min bias events with the
atlas tunings THIS SET IS DONE
Jet and
single particle
Calibration Samples
Once events are made, we can compare with table 13-11 of the HLT TDR to
ensure that all samples are being produced. The table shows the number
of events from each sample that will enter the various streams. This
table is approximate. It is from the atlfast nutples above. This table
will be completed as the samples are made. Note that the numbers apply
to the events AFTER FILTERING (in the cases where a filter is applied).
Bold face numbers in this table
are from the productions. The rest are estimates from pre-production.
Sample
|
production bias
|
e25
|
2e15
|
gam60
|
2gam20
|
mu20
|
2mu10
|
j400
|
3j165
|
4j110
|
j70ET70
|
tau35E45
|
2mu6
|
B1
|
|
0
|
0
|
1700
|
20
|
70
|
0
|
18000
|
9900
|
6600
|
4600
|
0
|
5
|
A6
|
|
0
|
0
|
1530
|
10
|
10
|
0
|
245000
|
89500
|
42550
|
17580
|
0
|
0
|
A4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
A0
|
|
130000
|
185000
|
0
|
0
|
153000
|
20500
|
3400
|
3700
|
6500
|
183600
|
21600
|
23400
|
A1-A3
|
|
662920
|
399410
|
810
|
240
|
748980
|
399720
|
150
|
0
|
20
|
11730
|
6160
|
410890
|
B2
|
|
0
|
0
|
8770
|
360
|
30
|
0
|
10
|
0
|
0
|
0
|
0
|
0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
A5
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
A7
|
|
|
|
|
|
|
|
|
|
|
|
|
|
total
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The "production bias" is calculated as follows:- [(number of events
produced/1M)*(cross section for Z to ee)/(cross section for process)].
It is the luminosity needed to produce the sample divided by the
luminosity to produced the Z to ee sample.
Physics Samples for DC2 post production
These are samples requested by Physics groups
Others
"Jet production" . This is reproduced from DC1. Pythia
parameters can be found here
. These produce QCD jets, top, bottom, single photon Z
and W. The minimum P_T is set to 17 GeV. There is no rapidity
restriction. K_L and Lambdas are stable. The random number seeds
(AtRndmGenSvc.Seeds = {"PYTHIA 4789899 989240512", "PYTHIA_INIT 820021
2347532"};) needs to be changed for each run. Other
sets for the jet samples with differant p_T thresholds are obtained by
changing the line "pysubs ckin 3
17." and replacing 17. by the appropriate number. This may
be used for testing only.
Other possibilities
0.5M W+gamma, pt>8 (all
W
decays) (bias =3.36) JobOptions
file 50K processed via atlfast Ntuple
1M W+Jets (pt>100, W to enu,
tau nu, mu nu and ud only) (bias= 0.63). JobOptions
file 100K processed via atlfast Ntuple
0.5M gamma gamma
(p_t>12)
(bias =0.88) JobOptions
file 50K processed via atlfast Ntuple
0.5M Z+jets, (Z to neutrinos,
p_t>50) (bias =1.03) JobOptions
file 50K processed via atlfast Ntuple
Contact persons
Data Challenge Coordinator
Gilbert Poulard