wiki:MadGraphSamples
Last modified 6 years ago Last modified on 05/04/12 18:31:24

MadGraph Samples for LHC physics

Aim of the project

Produce a database of validated LHC parton-level (and possibly hadron-level) MC events with MadGraph and the corresponding codes for large scale production. Main SM backgrounds and signals should be included, as well as representative BSM samples.

Motivation

MC's samples will play a fundamental role in the path to making discoveries at the LHC. Latest generation of the MC's have improved on several important aspects: from a more accurate simulation of events with multi-jet final states (with ME/PS matching) to full flexibility on new physics models implementation.

The increasein the MC powers is paralleled by the larger complexity of the codes both at the computational and user level, leading to the need of a continuous and reliable support from the MC developers. Normally, many issues arise in the dey-to-day use of such codes, some of which become essential in the final experimental analyses.

We propose to provide samples and codes for key SM and BSM processes at the LHC.

For each process we plan to publish:

  1. Small-size (<1M) parton-level and hadron level (Pythia) sample.
  2. Associated (frozen) code for large scale production over the grid.

The samples and the associated codes would be therefore validated by the MC authors and directly used by experimentalists or as a reference (comparison) with dedicated analysis generations.

A public set of samples to be used as a common reference by all the LHC community (exp and th) could be very useful for other reasons. For instance, they could form a base for future comparisons between experiments or make the interactions with theorists (both model builders and MC developers) easier.

People

  1. MadGraph Team (including Steve Mrenna)
  2. Roberto Chierici & Paolo Bartalini (CMS)
  3. Interested: Jon Butterworth & Osamu Jinnouch (ATLAS)

Tools

Modus Operandi

We have identified three phases:

  1. Agreement on the definition of the most important samples (all)
    • List of processes
    • Trigger or acceptance cuts
    • SM Parameters
    • Benchmark points and models for BSM
  2. MC codes and samples generation and validation (mg team + anybody interested)
  3. Large scale samples generation with (full) detector simulation (CMS+ATLAS)

Matching validation will be performed with MatchChecker.

Definition of the most important samples

A sample is defined by specifying:

  • Model
  • Initial state and final state particles
  • Couplings (EW, and QCD)
  • Phase space region

Information for MG/ME generation is collected in the banner through which a the sample can be fully reproduced.

SM parameters
  • Masses -> latest best values. LO parameters for EW.
  • PDF set and alphaS : cteq?
  • Scales: for matched samples these are automatically set by the code.
Standard Model Samples
QCD Jets
Process Stars Couplings Binning Phase space region Matching Event files Remarks
jets (2) 3. QCD only HT bins (100-250 !GeV, 250-500 !GeV, 500-1000 !GeV, 1000-inf !GeV) abs(eta(j))<5, Ptj>20 !GeV 0,1,2,3,4+ light jets are u,d,c,s,b,g; Need to veto the first gluon splitting into bb in the PS
bb~ + jets 3. QCD only HT bins (100-250 !GeV, 250-500 !GeV, 500-1000 !GeV, 1000-inf !GeV) abs(eta(j))<5, Ptj>20 !GeV 0,1,2,3+ massive b; Need to veto the first gluon splitting into bb in the PS
bb~ bb~+ jets 1. QCD only none 0,1+ massive b
Vector Boson
Process Stars Couplings Binning Phase space region Matching Event files Remarks
W (-> l v)+ jets 3 EW=2 + QCD none all 0,1,2,3,4+ W=W+,W- ; l=(e,mu,tau)
Z /a (-> l+l-)+ jets 3 EW=2 + QCD none m(l+l-)>50 !GeV 0,1,2,3,4+ photon is included ; l=(e,mu,tau)
Z /a (-> l+l-)+ jets 3 EW=2 + QCD none 50>m(l+l-)>10 !GeV, HT>30 !GeV 0,1,2,3,4+ photon is included ; l=(e,mu,tau)
Z (-> vv)+ jets 3 EW=2 + QCD none pt(Z)>50 !GeV 0,1,2,3,4+
V (-> l l')+ QQ~ 1 EW=2+QCD none all no V=W+,W-,Z ; l=(e,mu,tau,v), (Z->vv included) Q=b, c
a + jets 3 EW=1 + QCD HT bins (40-100 !GeV, 100-200 !GeV, 200-inf !GeV) pt(a)>20 !GeV, abs(eta(a))<2.5, !DeltaR(a,jet)>0.3 0,1,2,3,4+
a + QQ~ + jets 1 EW=1 + QCD none pt(a)>20 !GeV, abs(eta(a))<2.5, !DeltaR(a,jet)>0.3 0,1,2+ photon; Q=b
Vector Bosons
Process Stars Couplings Binning Phase space region Matching Event files Remarks
VV(-> 4l)+ jets 3 EW=2+QCD none m(l+l-)>10 !GeV; m(lv)>30 !GeV 0,1+ V=W+,W-,Z l=(e,mu,tau,v)
VV (-> 4l) + QQ~ 1 EW=1 + QCD none all no V=W+,W-,Z l=(e,mu,tau,v), Q=b
aV(-> 2l)+ jets 1 EW=2+QCD none pt(a)>20 !GeV, !DeltaR(a,jet)>0.3 0,1+ V=W+,W-,Z l=(e,mu,tau,v)
a a + jets 1 EW=2+QCD none pt(a)>20 !GeV, abs(eta(a))<2.5, !DeltaR(a,jet)>0.3 0,1,2+ photon
a a + QQ~ + jets 1 EW=1 + QCD none pt(a)>20 !GeV, abs(eta(a))<2.5, !DeltaR(a,jet)>0.3 no photon; Q=b
V V V 3 EW=3 none all no V=W+,W-,Z
a a a 3 EW=3 none pt(a)>20 !GeV, abs(eta(a))<2.5, !DeltaR(a,jet)>0.3 no
Top
Process Stars Couplings Binning Phase space region Matching Event files Remarks
tt + jets ' 3 QCD only none all 0,1,2,3+ top decays into everything. Done with DECAY
tt + bb~ 3 QCD only none all no top decays into everything. Done with DECAY
tjb 3 EW=4, QCD=1 none all no t-channel, b massive, leptonic top decay
tj 3 EW=4, QCD=0 none all no t-channel, leptonic top decay
tb 3 EW=4, QCD=0 none all no s-channel, b massive, leptonic top decay
tW 3 EW=5, QCD=1 none all no tW-channel, inclusive top and W decays
tWb 3 EW=5, QCD=2 none all no tW-channel, b-massive, doub-res diagram subtraction, inclusive top and W decays
Higgs
Process Stars Couplings Binning Phase space region Matching Event files Remarks
Higgs + jets 3 QCD only none all 0,1,2,3+ HEFT, mh=115,125,135,145,160,180,200 !GeV
Higgs + 2 jets 3 EW only none all no matching mh=115,125,135,145,160,180,200 !GeV includes V (->jj) + Higgs
tt~ + Higgs 3 QCD=2,EW=1 none all 0,1+ mh=115,125,135,145,160,180,200 !GeV
V (-> l l') + Higgs + jets 3 EW=3 + QCD none all 0,1,2 mh=115,125,135,145,160,180,200 !GeV
New Physics Samples
  • MSSM : we use the Snowmass benchmark points
  • UserModel : Z' into leptons
  • TopBSM : New resonances in the ttbar invariant mass spectrum
  • TwoHiggsDoublet : Alternative EWSB scenarios

Some LHE files and cards used for validation

Please note that these files need to be passed through the Pythia package found on the MG/ME download page. For all jet multiplicities below the highest multiplicity wanted, IEXCFILE=1 must be specified in the pythia_card.dat. For the highest multiplicity wanted, IEXCFILE=0 must be specified.

W+ > l+ vl

Sample events xqcut LHE file Banner
W+ 0jet 100k 10 lhe_file banner
W+ 1jet 100k 10 lhe_file banner
W+ 2jet 100k 10 lhe_file banner
W+ 3jet 100k 10 lhe_file banner
W+ 4jet 100k 10 lhe_file banner

---

W- > l- vl~

Sample events xqcut LHE file Banner
W- 0jet 100k 10 lhe_file banner
* W- 1jet* 100k 10 lhe_file banner
W- 2jet 100k 10 lhe_file banner
W- 3jet 100k 10 lhe_file banner
W- 4jet 100k 10 lhe_file banner

---

Z/a > l+ l-

Sample events xqcut M(l+l-) LHE file Banner
Z/a 0jet 100k 10 50 lhe_file banner
Z/a 1jet 100k 10 50 lhe_file banner
*Z/a 2jets 100k 10 50 lhe_file banner
Z/a 3jets 100k 10 50 lhe_file banner
Z/a 4jets 5x 20k 10 50 lhe_file1, lhe_file2, lhe_file3, lhe_file4, lhe_file5 banner

Z > v v~

Sample events xqcut LHE file Banner
Z + 1jet 100k 10 lhe_file banner
Z + 2jets 100k 10 lhe_file banner
Z + 3jets 100k 10 lhe_file banner
Z + 4jets 100k 10 lhe_file banner

ttbar+ jets:

Please specify QCUT=30 (or as indicated below) in the pythia_card.dat.

Sample events xqcut QCUT LHE file Banner
ttbar+ 0 jet 100k 20 30 lhe_file banner
ttbar+ 1 jet 100k 20 30 lhe_file banner
ttbar+ 2jets 100k 20 30 lhe_file banner
ttbar+ 3jets 100k 20 30 lhe_file banner

MC samples generation and validation

People involved in the generation/validation should report progress in the DevelopmentArea (you have to be registered in the team to access this page).

Sample codes for the grid

LHC @ 10 !TeV - Gridpacks used for production

Please note that all cross-sections in the following do include the correct BR already. Typically, the results is given for all the lepton families included.

Non matched processes
GridPack xsec
V+QQ, i.e. Z,W+ or W- + bb~ or cc~ (massive) gridpack 289 pb
W+ and W- + c quark (massive) gridpack1487 pb
single top s-channelIncluding leptonic top decay, $\mu_R=\mu_F=m_{top}$gridpack1.66 pb
single top t-channel pp>tj (p=u,d,s,c,b,g)Including leptonic top decay, $\mu_R^2=\mu_F^2=m_{top}^2+Q^2$, initial state b-quark gridpack 43.7 pb
single top t-channel pp>tbj (p=u,d,s,c,g)Including leptonic top decay, $\mu_R^2=\mu_F^2=m_{top}^2+Q^2$, massive b-quark gridpack 27.6 pb
single top W associate pp>tW (p=u,d,s,c,b,g)Inclusive top and W decays, $\mu_R=\mu_F=m_{top}$, initial state b-quark gridpack 27.5 pb
single top W associate pp>tbW (p=u,d,s,c,g)Inclusive top and W decays, $\mu_R=\mu_F=m_{top}$, double resonant diagrams (ttbar) subtracted, massive b-quark ar.gz gridpack 17.1 pb
Matched processes
GridPack Efficiency (xqcut,qcut) xsec after matching
ttbar +0,1,2,3
Including integration grids and with (inclusive) top decay using DECAY. Generation time for 1000 events: about 1h30 on a 3.00 GHz P4 desktop machine.
mini-soup (4 flavors) with decay ~33% (20,30)317 pb
mini-soup (5 flavors) with decay ~33% (20,30)317 pb
W into leptons + 0,1,2,3,4j (5 flavors)
running time: ~30-40 min for 1000 events
mini-soup~45%(10,15)~40nb
Z/a into leptons + 0,1,2,3,4j (5 flavors) with mll>50 GeV
running time: ~90 min for 1000 events
mini-soup ~40%(10,15)~3.7 nb
Z/a into leptons + 0,1,2,3,4j (5 flavors) with 10<mll<50 GeV and Sum over jet pt's>30 GeV
running time: ~90 min for 1000 events
mini-soup to be tested (xqcut=5,qcut=10?)to check
Z into nunu + 0,1,2,3,4j (5 flavors)
running time: ?
mini-soup to be tested~11.1 nb
VV into leptons + 0,1j (5 flavors)
running time: ?
mini-soup to be tested11.8 pb
QCD: 2,3,4j (5 flavors)
pay attention to the change of xqcut (and then Qcut has to be changed) between Slice 1,2 and 3,4: qcut=30 for slice 1,2 and 60 for 3,4. This is done in order to increase the efficiency
(100,250) GeV ~34%(20,30) ~15 microbarn
(250,500) GeV ~20% (20,30)~400 nanobarn
(500,1000) GeV ~30%(40,60)~14 nanobarn
(1000,...) ~20%(40,60)~370 picobarn
b-enriched :bb+0,1,2,3;b+1,2,3,b~+1,2,3
pay attention to the change of xqcut (and then Qcut has to be changed) between Slice 1,2 and 3,4: qcut=30 for slice 1,2,3 and 60 for 4. This is done in order to increase the efficiency
(100,250) GeV ~28% (20,30)~450 nanobarn
(250,500) GeV ~15% (20,30) ~15 nanobarn
(500,1000) GeV ~10% (20,30)~700 picobarn
(1000,inf) GeV ~15% (40,60)~13 picobarn
a+1,2,3,4 j (5 flavor)
cross section are given before matching
(40,100) GeV~11% (5,10)~ 368.7 nb
(100,200)~?% (5,10)~103.6 nb
(200,inf)~?% (10,15)~10.49 nb

LHC @ 14 !TeV - Gridpacks used for production

Please note that all cross-sections in the following do include the correct BR already. Typically, the results is given for all the lepton families included.

Non matched processes
GridPack xsec
single top s-channelIncluding leptonic top decay, $\mu_R=\mu_F=m_{top}$ gridpack 2.65 pb
single top t-channel pp>tj (p=u,d,s,c,b,g)Including leptonic top decay, $\mu_R^2=\mu_F^2=m_{top}^2+Q^2$, initial state b-quark gridpack 82.8 pb
single top t-channel pp>tbj (p=u,d,s,c,g)Including leptonic top decay, $\mu_R^2=\mu_F^2=m_{top}^2+Q^2$, massive b-quark gridpack 55.6 pb
single top W associate pp>tW (p=u,d,s,c,b,g)Inclusive top and W decays, $\mu_R=\mu_F=m_{top}$, initial state b-quark gridpack 62.4 pb
single top W associate pp>tbW (p=u,d,s,c,g)Inclusive top and W decays, $\mu_R=\mu_F=m_{top}$, double resonant diagrams (ttbar) subtracted, massive b-quark gridpack 41.5 pb
Matched processes
GridPack Efficiency (xqcut,qcut) xsec after matching
ttbar +0,1,2,3
Including integration grids and with (inclusive) top decay using DECAY. Generation time for 1000 events: about 1h30 on a 3.00 GHz P4 desktop machine.=
4-flavor ~27% (20,30)~750 pb
5-flavor ~27% (20,30)~750 pb
W into leptons + 0,1,2,3,4j (5 flavors)
running time: ~30-40 min for 1000 events
gridpack.tar.gz~40% (10,15)60 nb
Z/a into leptons + 0,1,2,3,4j (5 flavors) with mll>50 GeV
running time: ~90 min for 1000 events
gridpack.tar.gz~45% (10,15)~7 nb
Z/a into leptons + 0,1,2,3,4j (5 flavors) with 10<mll<50 GeV and Sum over jet pt's>30 GeV
running time: ~90 min for 1000 events
to come to be tested (xqcut=5,qcut=10?)to check
Z into nunu + 0,1,2,3,4j (5 flavors)
running time: ?
mini-soup to be tested~19.5 nb
VV into leptons + 0,1j (5 flavors)
running time: ?
mini-soup to be tested19.3 pb
QCD: 2,3,4j (5 flavors)
(100,250) GeV~30% (20,30)~24 mb
(250,500) GeV~18% (20,30)~770 nb
(500,1000) GeV~12% (20,30)~36 nb
(1000,inf) GeV~9% (20,30)~1 nb
b-enriched :bb+0,1,2,3;b+1,2,3,b~+1,2,3
(100,250) GeV~50% (20,30)~900 nb
(250,500) GeV~30% (20,30)~50 nb
(500,1000) GeV~20% (20,30)~4 nb
(1000,inf) GeV~20% (20,30)~0.15 nb
a+1,2,3,4 j (5 flavor)
cross section are given before matching
(40,100)~11% (5,10)~574 nb
(100,200)~18% (5,10)~174 nb
(200,inf)~27% (10,15)~19 nb

Productions with 1 multiplicity / sample

tt~ +jets:
Process EXCL. Xsec (Qcut=30) INCL. xsec (Qcut=30) Efficiency of the matching
tt~309 pb/52% (excl)
tt~+ 1 jet222 pb/28% (excl)
tt~+ 2 jets100 pb/14% (excl)
tt~+ 3 jets /63 pb14% (incl)

Differences of cross sections between 4 and 5 flavor are too small to be taken into account here. Notes

  • xqcut used: 20 GeV -> Qcut=30 GeV to be used in pythia_card.dat
  • for the 0,1,2 jets multiplicities, use IEXCFILE=1
W+ + light jets only (udscg)

Notes

  • xqcut used: 10 GeV -> Qcut=15 GeV to be used in pythia_card.dat
  • for all multiplicities EXCEPT the highest, use IEXCFILE=1
Process GridPack EXCL. Xsec (Qcut=15) INCL. xsec (Qcut=15)
(w+>leptons) +0 jet gridpack.tar.gz 22.5 nb
(w+>leptons) +1 jets gridpack.tar.gz 8.1 nb
(w+>leptons) +2 jets gridpack.tar.gz 2.8 nb
(w+>leptons) +3 jets gridpack.tar.gz 1 nb 1.6 nb
(w+>leptons) +4 jets to come
W- + light jets only (udscg)
Process GridPack EXCL. Xsec (Qcut=15) INCL. xsec (Qcut=15)
(w->leptons) +0 jet gridpack.tar.gz 16.5 nb
(w->leptons) +1 jets gridpack.tar.gz 5.7 nb
(w->leptons) +2 jets gridpack.tar.gz nb
(w->leptons) +3 jets gridpack.tar.gz nb
(w->leptons) +4 jets to come

Notes

  • xqcut used: 10 GeV -> Qcut=15 GeV to be used in pythia_card.dat
  • for all multiplicities EXCEPT the highest, use IEXCFILE=1

Other notes

For the pythia step treatement the following parameter could be usefull:

  • MSTP(81)=0 !use virtuality-ordered showers
  • MSTP(5)=109 ! tune to use the Underlying Events for PDF CTEQ6L1

Typical run time

Reference machine is a recent macbook running Leopard, intel CPU 2.16 GHz Dual Core 2 (but only one CPU is used of course), 4Mb of L2 cache and 1Go of RAM. The process is (W+>e+ve)+jets at the LHC, with all parameters and cuts set to their default values. The warming up time is for a rather slow 24*1GHz cluster (FYNU cluster at UCL). The compiled gridpack sizes are for static linking only, dynamic linking should decrease them a lot. Also, the run time is not expected to scale linearly with the number of essay diagrams, in particular for complicated processes. Generating 10K W+2j events could take only 2h, not 12h!

Process total # diag. # subproc Warming up time Gridpack size Compilation Time Compiled gridpack size Time to generate 1000 events
(W+>e+ve)+0j 2 2 2min 0.9Mb 10s 1.1 Mb 5s
(W+>e+ve)+1j 12 6 5min 1.0Mb 30s 1.7 Mb 1min40
(W+>e+ve)+2j 188 50 20min 2.8 Mb 2min10 8.5 Mb 69min
(W+>e+ve)+3j 1848 ? 112min 14 Mb 6min 22 Mb ?