#195 closed defect (fixed)
nonworking cuts for identical particles in inital and final state?
| Reported by: | sschmidt | Owned by: | kilian |
|---|---|---|---|
| Priority: | P1 | Milestone: | |
| Component: | core | Version: | 2.0.0rc2 |
| Severity: | blocker | Keywords: | cuts |
| Cc: |
Description
When simulating processes with identical particles in the initial and final state (like e+e- -> e+ e- mu+ mu- or u U -> u, U, e1, E1) there might be a problem concerning the cuts. When appliying the cuts I found reasonable
cuts = all Pt > 10 GeV [finals]
and all Dist > 1 [finals, finals]
no stable Integration was achieved. Therefor I added all kinds of cuts and ended up with (for u U -> u, U, e1, E1)
cuts = all Pt > 10 GeV [finals]
and all Dist > 1 [finals,finals]
and all E > 1 GeV [finals]
and all 5 degree < Theta < 175 degree [finals]
and all 5 degree < Theta < 175 degree [finals,finals]
and all M2 > (10 GeV)^2 [combine [finals, finals]]
and all M2 < -(10 GeV)^2 [combine [incoming u, u]]
and all M2 < -(10 GeV)^2 [combine [incoming U, U]]
and all Pt > 10 GeV [combine[u,U]]
and all Pt > 10 GeV [combine[e1,E1]]
and all E > 10 GeV [combine[e1,E1]]
and all E > 10 GeV [combine[u,U]]
but still the integration was unstable.
Any ideas which singularity I'm missing?
Attachments (1)
Change History (9)
comment:1 Changed 14 years ago by
| Priority: | P3 → P1 |
|---|
comment:2 Changed 14 years ago by
comment:3 Changed 14 years ago by
Forgot to mention: without identical particles in initial and final state, e.g. e+e- -> mu+ mu+ mu- mu-, the integration works fine.
comment:4 Changed 14 years ago by
| Status: | new → assigned |
|---|
Doing some further analysis it looks like the phase space setup is not the culprit, inserting the W1 phase space file doesn't help. Maybe the t-channel mappings.
comment:5 Changed 14 years ago by
No, with same phase space results are similar, fortunately (wrong number of calls in previous test).
W2 drops phs channels with a Coulomb singularity. Looking into the code ...
comment:6 Changed 14 years ago by
Status after r1746:
W1 for e- e+ -> e- e+ mu- mu+ with cuts, zero masses:
! WHIZARD run for process eemm:
!=============================================================================
! It Calls Integral[fb] Error[fb] Err[%] Acc Eff[%] Chi2 N[It]
!-----------------------------------------------------------------------------
! Reading cut configuration data from file whizard.cut1
! Replacing default cuts by user-defined cuts.
cut M of 3 within 1.00000E+01 1.00000E+04
cut M of 5 within 1.00000E+01 1.00000E+04
cut M of 6 within 1.00000E+01 1.00000E+04
cut M of 9 within 1.00000E+01 1.00000E+04
cut M of 10 within 1.00000E+01 1.00000E+04
cut M of 12 within 1.00000E+01 1.00000E+04
cut pT of 1 within 1.00000E+01 1.00000E+04
cut pT of 2 within 1.00000E+01 1.00000E+04
cut pT of 4 within 1.00000E+01 1.00000E+04
cut pT of 8 within 1.00000E+01 1.00000E+04
! Preparing (fixed weights): 1 sample of 20000 calls ...
1 20000 4.8507895E+01 1.08E+01 22.27 31.49* 0.29 0.00 1
!-----------------------------------------------------------------------------
! Adapting (variable wgts.): 9 samples of 20000 calls ...
2 20000 3.2897175E+01 6.34E+00 19.28 27.26* 0.28
3 20000 5.7625126E+01 4.54E+00 7.88 11.15* 0.40
4 20000 5.8788191E+01 2.93E+00 4.98 7.04* 0.51
5 20000 6.4934654E+01 2.43E+00 3.75 5.30* 0.78
6 20000 6.3402611E+01 1.64E+00 2.58 3.65* 1.52
7 20000 6.1065987E+01 1.54E+00 2.53 3.58* 1.46
8 20000 6.3492507E+01 1.91E+00 3.01 4.25 1.02
9 20000 6.4694908E+01 2.44E+00 3.77 5.33 0.78
10 20000 6.0399532E+01 1.39E+00 2.29 3.24* 1.92
!-----------------------------------------------------------------------------
! Integrating (fixed wgts.): 3 samples of 100000 calls ...
11 300000 6.2716486E+01 3.84E-01 0.61 3.35 0.64 3.28 3
!-----------------------------------------------------------------------------
The analogous W2 run:
| Integrating process 'eemm' | Generating phase space configuration ... | ... found 65 phase space channels. | Wrote phase-space configuration file 'eemm.phs'. | Applying user-defined cuts. | Using partonic energy as event scale. | iterations = 10:20000, 3:100000 | Creating grids | 20000 calls, 65 channels, 8 dimensions, 20 bins, stratified = T |=============================================================================| | It Calls Integral[fb] Error[fb] Err[%] Acc Eff[%] Chi2 N[It] | |=============================================================================| 1 20000 5.0357945E+01 1.52E+01 30.23 42.75* 0.55 2 20000 5.4118571E+01 7.17E+00 13.25 18.73* 0.41 3 20000 5.7732888E+01 4.41E+00 7.64 10.80* 0.95 4 20000 7.1026126E+01 4.70E+00 6.61 9.35* 0.89 5 20000 6.6568446E+01 4.70E+00 7.06 9.98 0.77 6 20000 6.2322346E+01 2.63E+00 4.21 5.96* 1.33 7 20000 6.0898616E+01 2.25E+00 3.69 5.21* 1.64 8 20000 6.3485224E+01 2.66E+00 4.18 5.92 1.78 9 20000 6.2857848E+01 2.17E+00 3.46 4.89* 1.63 10 20000 6.6377969E+01 3.88E+00 5.84 8.26 1.51 |-----------------------------------------------------------------------------| 10 200000 6.2758661E+01 1.04E+00 1.65 7.39 1.51 0.98 10 |-----------------------------------------------------------------------------| 11 100000 6.3585410E+01 8.73E-01 1.37 4.34* 1.14 12 100000 6.3960119E+01 8.76E-01 1.37 4.33* 0.98 13 100000 6.3568811E+01 8.47E-01 1.33 4.21* 1.09 |=============================================================================| 13 300000 6.3701364E+01 5.00E-01 0.78 4.29 1.09 0.06 3 |=============================================================================|
This is ok. (Note the chi2 value of the W1 run!)
comment:7 Changed 14 years ago by
| Resolution: | → fixed |
|---|---|
| Status: | assigned → closed |
With r1747, the no-cut case (with nonzero masses) also works well, if the user sets the mapping scale parameters accordingly. (Actually, the adaptation fails later due to numerical instabilities if the ratio sqrts/me is too high, but this is unavoidable.) W1 had some heuristics to set these scale parameters, but this is black magic.
(The user who removes cuts all should know what he is doing, some sentences in the manual would be appropriate.)
Closing the ticket for now.
I cross-checked with W1.93 for the process
There even without applying any user-defined cuts a stable integration is achieved
1 20000 7.3230860E+02 3.57E+02 48.80 69.02* 0.15 0.00 1 !----------------------------------------------------------------------------- ! Adapting (variable wgts.): 10 samples of 20000 calls ... 2 20000 1.2000781E+02 2.09E+01 17.41 24.63* 0.32 3 20000 5.0038262E+02 8.04E+01 16.06 22.71* 0.28 4 20000 5.5816724E+02 5.95E+01 10.67 15.08* 0.24 5 20000 9.3638237E+02 1.48E+02 15.77 22.30 0.18 6 20000 9.8169766E+02 6.42E+01 6.54 9.24* 0.33 7 20000 9.9291687E+02 3.57E+01 3.59 5.08* 0.65 8 20000 9.8156231E+02 2.14E+01 2.18 3.08* 1.93 9 20000 9.5766213E+02 1.73E+01 1.81 2.56* 2.44 10 20000 9.5253297E+02 1.68E+01 1.76 2.49* 1.87 11 20000 9.5701570E+02 1.59E+01 1.66 2.35* 2.54 !----------------------------------------------------------------------------- ! Integrating (fixed wgts.): 3 samples of 20000 calls ... 12 60000 9.6946841E+02 8.40E+00 0.87 2.12* 1.54 0.02 3The automatically defined cuts are
I implemented these cuts in W2 (at least I hope so):
cuts = all M2 > (10 GeV)^2 [combine[e1,E1]] and all M2 < -(10 GeV)^2 [combine[incoming e1, e1]] and all M2 < -(10 GeV)^2 [combine[incoming E1, E1]]but the integration remains unstable
Am I doing something wrong?