difference in tau pT between madspin and madgraph decay syntax

Hi,

Thanks for the report, I'm trying to reproduce your plot now.
Can I ask you which version of MG5aMC are you using?

Cheers,

Olivier

Other question, do you have the full log for the first call to madspin (the problematic one)

I'm in particular interested in the line

INFO: Total number of events written: 10000/10000
INFO: Average number of trial points per production event: 14.8521
INFO: Branching ratio to allowed decays: 1
INFO: Number of events with weights larger than max_weight: 0
INFO: Number of subprocesses 22
INFO: Number of failures when restoring the Monte Carlo masses: 0

to see if you have the same as me or if you have some warning at that stage.

Cheers,

Olivier

Hi Oliver,

Thanks a lot for the reply and sorry my response is very slow.

I'm using version 2.9.3.

For the few lines you pointed out, I think what I got was identical to yours:

INFO: Total number of events written: 5500/5500
INFO: Average number of trial points per production event: 19.5289090909
INFO: Branching ratio to allowed decays: 1
INFO: Number of events with weights larger than max_weight: 0
INFO: Number of subprocesses 22
INFO: Number of failures when restoring the Monte Carlo masses: 0

The full set of madspin messages are attached below:

INFO: Extracting the banner ...
INFO: process: p p > t t~ t t~ QED=2
INFO: options: QCD=4
INFO: detected model: sm. Loading...
set lhapdf to /srv/workDir/lhapdf-config
^[[1;34m/cvmfs/atlas.cern.ch/repo/sw/software/21.6/sw/lcg/releases/LCG_88b/MCGenerators//gosam_contrib/2.0-779ba/x86_64-centos7-gcc62-opt/lib does not seem to correspond to a valid ninja lib . Please ent\
er the full PATH/TO/ninja/lib .
You will NOT be able to run ninja otherwise.
^[[0m
^[[1;34m/cvmfs/atlas.cern.ch/repo/sw/software/21.6/sw/lcg/releases/LCG_88b/MCGenerators//collier/1.1-92381/x86_64-centos7-gcc62-opt does not seem to correspond to a valid collier lib . Please enter the f\
ull PATH/TO/collier/lib .
You will NOT be able to run collier otherwise.
^[[0m
set fastjet to /cvmfs/atlas.cern.ch/repo/sw/software/21.6/sw/lcg/releases/LCG_88b/fastjet/3.2.0/x86_64-centos7-gcc62-opt/bin/fastjet-config
set lhapdf to /srv/workDir/lhapdf-config
set max_weight_ps_point 400 # number of PS to estimate the maximum for each event
set Nevents_for_max_weight 75
set BW_cut 15
set seed 123457
decay t > w+ b, w+ > all all
decay t~ > w- b~, w- > all all
launch
INFO: We need to recalculate the branching fractions for t~,w-,w+,t
INFO: using the FeynRules formula present in the model (arXiv:1402.1178)
INFO: load particles
INFO: load vertices
INFO: Restrict model /cvmfs/atlas.cern.ch/repo/sw/software/21.6/sw/lcg/releases/MCGenerators/madgraph5amc/2.9.3.atlas-a5843/x86_64-centos7-gcc62-opt/models/sm with file ../../../cvmfs/atlas.cern.ch/repo/\
sw/software/21.6/sw/lcg/releases/MCGenerators/madgraph5amc/2.9.3.atlas-a5843/x86_64-centos7-gcc62-opt/models/sm/restrict_default.dat .
INFO: Run "set stdout_level DEBUG" before import for more information.
^[[1;34mWARNING: The LO estimate for the width of particle 6 ^[[0m
^[[1;34mWARNING: differs from the one in the banner by 10 percent ^[[0m
INFO:
INFO: decay channels for t~ : ( width = 1.476561 GeV )
INFO: BR d1 d2
INFO: 1.000000e+00 b~ w-
INFO:
INFO:
INFO: decay channels for w- : ( width = 2.046347 GeV )
INFO: BR d1 d2
INFO: 3.333605e-01 s c~
INFO: 3.333605e-01 d u~
INFO: 1.111202e-01 mu- vm~
INFO: 1.111202e-01 e- ve~
INFO: 1.110387e-01 ta- vt~
INFO:
INFO:
INFO: decay channels for w+ : ( width = 2.046347 GeV )
INFO: BR d1 d2
INFO: 3.333605e-01 s~ c
INFO: 3.333605e-01 d~ u
INFO: 1.111202e-01 mu+ vm
INFO: 1.111202e-01 e+ ve
INFO: 1.110387e-01 ta+ vt
INFO:
INFO:
INFO: decay channels for t : ( width = 1.476561 GeV )
INFO: BR d1 d2
INFO: 1.000000e+00 b w+
INFO:
INFO: generating the production square matrix element
INFO: generate p p > t t~ t t~ QED=2 QCD=4;
INFO: Done 11.19
INFO: generating the full matrix element squared (with decay)
INFO: generate p p > t t~ t t~ QED=2 QCD=4, (t~ > b~ w- , w- > all all QCD=99), (t > b w+ , w+ > all all QCD=99) --no_warning=duplicate;
INFO: Done 2841
INFO: generate matrix element for decay only (1 - > N).
INFO: output standalone_msF /srv/workDir/PROC_sm_0/decay_me
INFO: Done 4.192
INFO: Compiling code
INFO: detect independant decays
INFO: Done in 1.58508992195s
INFO:
INFO: Estimating the maximum weight
INFO: *****************************
INFO: Probing the first 75 events
INFO: with 400 phase space points
INFO:
INFO: Event 1/75 : 7m 2s
INFO: Event 6/75 : 43m 56s
INFO: Event 11/75 : 1h 21m
INFO: Event 16/75 : 1h 52m
INFO: Event 21/75 : 2h 25m
INFO: Event 26/75 : 3h 2m
INFO: Event 31/75 : 3h 37m
INFO: Event 36/75 : 4h 15m
INFO: Event 41/75 : 4h 48m
INFO: Event 46/75 : 5h 18m
INFO: Event 51/75 : 5h 54m
INFO: Event 56/75 : 6h 25m
INFO: Event 61/75 : 6h 55m
INFO: Event 66/75 : 7h 30m
INFO: Event 71/75 : 8h 2m
INFO:
INFO: Decaying the events...
INFO: Event nb 1000 8m 26s
INFO: Event nb 2000 10m 58s
INFO: Event nb 3000 12m 19s
INFO: Event nb 4000 14m 1s
INFO: Event nb 5000 15m 31s
INFO: Total number of events written: 5500/5500
INFO: Average number of trial points per production event: 19.5289090909
INFO: Branching ratio to allowed decays: 1
INFO: Number of events with weights larger than max_weight: 0
INFO: Number of subprocesses 22
INFO: Number of failures when restoring the Monte Carlo masses: 0
INFO: The decayed event file has been moved to the following location:
INFO: /srv/workDir/PROC_sm_0/Events/run_01_decayed_1/unweighted_events.lhe.gz
INFO: MadSpin Done
INFO: storing files of previous run
INFO: Done

You do only have 5500 events? (so only around 1000 tau lepton)
Are you sure that this is not jsut a statistical fluctuation of the result?
Did you try at higher statistics?

Cheers,

Olivier

> On 16 Jun 2022, at 22:55, Yang Qin <email address hidden> wrote:
>
> Question #702049 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/702049
>
> Status: Needs information => Open
>
> Yang Qin gave more information on the question:
> Hi Oliver,
>
> Thanks a lot for the reply and sorry my response is very slow.
>
> I'm using version 2.9.3.
>
> For the few lines you pointed out, I think what I got was identical to
> yours:
>
> INFO: Total number of events written: 5500/5500
> INFO: Average number of trial points per production event: 19.5289090909
> INFO: Branching ratio to allowed decays: 1
> INFO: Number of events with weights larger than max_weight: 0
> INFO: Number of subprocesses 22
> INFO: Number of failures when restoring the Monte Carlo masses: 0
>
>
> The full set of madspin messages are attached below:
>
> INFO: Extracting the banner ...
> INFO: process: p p > t t~ t t~ QED=2
> INFO: options: QCD=4
> INFO: detected model: sm. Loading...
> set lhapdf to /srv/workDir/lhapdf-config
> ^[[1;34m/cvmfs/atlas.cern.ch/repo/sw/software/21.6/sw/lcg/releases/LCG_88b/MCGenerators//gosam_contrib/2.0-779ba/x86_64-centos7-gcc62-opt/lib does not seem to correspond to a valid ninja lib . Please ent\
> er the full PATH/TO/ninja/lib .
> You will NOT be able to run ninja otherwise.
> ^[[0m
> ^[[1;34m/cvmfs/atlas.cern.ch/repo/sw/software/21.6/sw/lcg/releases/LCG_88b/MCGenerators//collier/1.1-92381/x86_64-centos7-gcc62-opt does not seem to correspond to a valid collier lib . Please enter the f\
> ull PATH/TO/collier/lib .
> You will NOT be able to run collier otherwise.
> ^[[0m
> set fastjet to /cvmfs/atlas.cern.ch/repo/sw/software/21.6/sw/lcg/releases/LCG_88b/fastjet/3.2.0/x86_64-centos7-gcc62-opt/bin/fastjet-config
> set lhapdf to /srv/workDir/lhapdf-config
> set max_weight_ps_point 400 # number of PS to estimate the maximum for each event
> set Nevents_for_max_weight 75
> set BW_cut 15
> set seed 123457
> decay t > w+ b, w+ > all all
> decay t~ > w- b~, w- > all all
> launch
> INFO: We need to recalculate the branching fractions for t~,w-,w+,t
> INFO: using the FeynRules formula present in the model (arXiv:1402.1178)
> INFO: load particles
> INFO: load vertices
> INFO: Restrict model /cvmfs/atlas.cern.ch/repo/sw/software/21.6/sw/lcg/releases/MCGenerators/madgraph5amc/2.9.3.atlas-a5843/x86_64-centos7-gcc62-opt/models/sm with file ../../../cvmfs/atlas.cern.ch/repo/\
> sw/software/21.6/sw/lcg/releases/MCGenerators/madgraph5amc/2.9.3.atlas-a5843/x86_64-centos7-gcc62-opt/models/sm/restrict_default.dat .
> INFO: Run "set stdout_level DEBUG" before import for more information.
> ^[[1;34mWARNING: The LO estimate for the width of particle 6 ^[[0m
> ^[[1;34mWARNING: differs from the one in the banner by 10 percent ^[[0m
> INFO:
> INFO: decay channels for t~ : ( width = 1.476561 GeV )
> INFO: BR d1 d2
> INFO: 1.000000e+00 b~ w-
> INFO:
> INFO:
> INFO: decay channels for w- : ( width = 2.046347 GeV )
> INFO: BR d1 d2
> INFO: 3.333605e-01 s c~
> INFO: 3.333605e-01 d u~
> INFO: 1.111202e-01 mu- vm~
> INFO: 1.111202e-01 e- ve~
> INFO: 1.110387e-01 ta- vt~
> INFO:
> INFO:
> INFO: decay channels for w+ : ( width = 2.046347 GeV )
> INFO: BR d1 d2
> INFO: 3.333605e-01 s~ c
> INFO: 3.333605e-01 d~ u
> INFO: 1.111202e-01 mu+ vm
> INFO: 1.111202e-01 e+ ve
> INFO: 1.110387e-01 ta+ vt
> INFO:
> INFO:
> INFO: decay channels for t : ( width = 1.476561 GeV )
> INFO: BR d1 d2
> INFO: 1.000000e+00 b w+
> INFO:
> INFO: generating the production square matrix element
> INFO: generate p p > t t~ t t~ QED=2 QCD=4;
> INFO: Done 11.19
> INFO: generating the full matrix element squared (with decay)
> INFO: generate p p > t t~ t t~ QED=2 QCD=4, (t~ > b~ w- , w- > all all QCD=99), (t > b w+ , w+ > all all QCD=99) --no_warning=duplicate;
> INFO: Done 2841
> INFO: generate matrix element for decay only (1 - > N).
> INFO: output standalone_msF /srv/workDir/PROC_sm_0/decay_me
> INFO: Done 4.192
> INFO: Compiling code
> INFO: detect independant decays
> INFO: Done in 1.58508992195s
> INFO:
> INFO: Estimating the maximum weight
> INFO: *****************************
> INFO: Probing the first 75 events
> INFO: with 400 phase space points
> INFO:
> INFO: Event 1/75 : 7m 2s
> INFO: Event 6/75 : 43m 56s
> INFO: Event 11/75 : 1h 21m
> INFO: Event 16/75 : 1h 52m
> INFO: Event 21/75 : 2h 25m
> INFO: Event 26/75 : 3h 2m
> INFO: Event 31/75 : 3h 37m
> INFO: Event 36/75 : 4h 15m
> INFO: Event 41/75 : 4h 48m
> INFO: Event 46/75 : 5h 18m
> INFO: Event 51/75 : 5h 54m
> INFO: Event 56/75 : 6h 25m
> INFO: Event 61/75 : 6h 55m
> INFO: Event 66/75 : 7h 30m
> INFO: Event 71/75 : 8h 2m
> INFO:
> INFO: Decaying the events...
> INFO: Event nb 1000 8m 26s
> INFO: Event nb 2000 10m 58s
> INFO: Event nb 3000 12m 19s
> INFO: Event nb 4000 14m 1s
> INFO: Event nb 5000 15m 31s
> INFO: Total number of events written: 5500/5500
> INFO: Average number of trial points per production event: 19.5289090909
> INFO: Branching ratio to allowed decays: 1
> INFO: Number of events with weights larger than max_weight: 0
> INFO: Number of subprocesses 22
> INFO: Number of failures when restoring the Monte Carlo masses: 0
> INFO: The decayed event file has been moved to the following location:
> INFO: /srv/workDir/PROC_sm_0/Events/run_01_decayed_1/unweighted_events.lhe.gz
> INFO: MadSpin Done
> INFO: storing files of previous run
> INFO: Done
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

I generated in total 100k events, but they were done in batches. The log file is only from one of the subjobs. The discrepancy I saw was much larger than the stat uncertainties.

Cheers, Quake

I have generated a sample of 100k events decayed with MadSpin.
and the plot for the electron and tau shows the same PT shape.

I'm now running the full decay with madgraph to see if I see a difference in that case, but that generation is quite slow so it will still take time before having result.

Cheers,

Olivier

> On 25 Jun 2022, at 15:10, Yang Qin <email address hidden> wrote:
>
> Question #702049 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/702049
>
> Status: Answered => Open
>
> Yang Qin is still having a problem:
> Hi Olivier,
>
> I generated in total 100k events, but they were done in batches. The log
> file is only from one of the subjobs. The discrepancy I saw was much
> larger than the stat uncertainties.
>
> Cheers, Quake
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

Indeed the full decay with MG takes ages... But your results using madspin already seems to show some difference from mine: I got different tau pT compared to e/mu pT distributions in this case. Whereas they are very similar in the case of full decay in madgraph.

Any tips what could be causing this?

As I mentioned in the original post, the difference would go away when I generate top to tau decays exclusively. i.e. explicitly require top to all decay to tau. In another word, when I use madspin to decay the tops, the tau pT distribution would be different if we generate all top decay inclusively comparing to only generating top to tau decay.

So I tried in total 4 setups: (madspin or not) x (top decay inclusive or exclusively to tau).
It appears the only setup that gives a different prediction is when I use madspin and generate top decay inclusively.

Cheers, Quake

Well since I do not have your plot, neither those from the decay syntax of MadGraph (to see if the difference between the plot are much bigger in one case compare to the other) I can not comment too much yet.

A big difference in my madspin setup compare to yours is that I did one run of 100k events rather than ~20 of ~5k events.
A single run means a single log to check for potential issue in the various max_weight to compute. Also the precision (and in general absolute value) of the max_weight is automatically increase for larger number of events. Those fine tuning may explain your difference --if some max-weight are under-estimate in your case-- (but you likely should have seen it in the log if that was the case).

Cheers,

Olivier

So now I do have the plot for the standard madevent (even if it fails to reach 100k and only generated 50k events)

Multiple conclusion here:
1) the normalization of madevent and madspin are quite different:
madevent: 0.008172 ± 7e-06
madspin: 0.009547 ± 8.2e-06
The reason here should be associated to the running of alpha_s and the choice of dynamical scale which is not the same in the two setup.

2) scale variation are indeed large (but consistent between the two mode of computation):
madevent:
scale variation: +72.2% -38.8%
madspin:
# scale variation: +72.2% -38.8%

For example, using shat as central scale the cross-section would be much smaller:
madevent: 0.00337009 +54.2% -33.1%
madspin: 0.00287056 +54.3% -33.1%

3) The pt shape of electron/tau are all quite close to each other. Now, given the high dependency in the scale for the total cross-section and the fact that the scale choice is not (and can not be ) the same in the two cases the comparison should take into account the scale uncertainty bin per bin (and the information is in the file to do the associated plot). Given the CKKW algorithm used for the computation of the scale the fact that the tau is massive can impact the computation in madevent run (but not in madspin).

One could redo the comparison in fixed scale such that the two run could be compared more precisely but given the time for the madevent run, I will not do such validation run.

Cheers,

Olivier

Hi Olivier,

Thanks for the follow-up. It might be some issue with my setup. I have attached some plots showing the issue here:
https://drive.google.com/file/d/1p5qQ4DIK23JFgiRiRFWSvuyAKSQi9o2e/view?usp=sharing

The issue appeared to be just in the decay part (as all other top related plots all look consistent), so naively I would expect something is wrong with the madspin part. Could you advise where I could look into?

Did you use the same generation lines as I specified above? In particular I wonder if the line for madspin 'w+ > all all' could be problematic? because the distributions become consistent if I specify w decaying explicitly to tau.

Cheers, Quake

PS: I don't think this is related to the fact that I'm running in batches. The setup with madspin run pretty fast. So I did generate events with madspin all within one single job, but the result was the same. In fact the issue was identified when I was generating all events at once.

The command I did used are the following one for madspin:

generate p p > t t~ t t~ QED=2
output
launch
madspin=ON
decay t > w+ b, w+ > all all
decay t~ > w- b~, w- > all all
set nevents 100k

and the following one for the decay chain syntax:

define allW = j l+ l- vl vl~ ta+ ta- vt vt~
generate p p > t t~ t t~ QED=2 QCD=4, (t > b allW allW QCD=0 QED=2), (t~ > b~ allW allW QCD=0 QED=2)
output
launch
set cut_decays F
set nevents 100k

I'm using 2.9.11.

Now looking at the list of bug fixed between 2.9.3 and 2.9.11, a couple of them are related to MadSpin.
In particular one solved by 2.9.7 is related to wrong BR within MadSpin:
https://bugs.launchpad.net/mg5amcnlo/+bug/1951120
Since here you have decay with 0/1/2/3 tau a bias in the branching ration pick might bias the ratio between 0/1/2/3 tau and therefore the pt plot... I would not say that the argument that you faced that exact bug is very convincing but given that I do not reproduce your issue, I would advise you to update to the latest Long Term Stable version.

Cheers,

Olivier

Hi Olivier,

Thanks very much for the suggestion. I will give the new version a go and will come back to this.

Cheers, Quake

Hi Oliver,

Sorry for the long turn-around time...

I have tried to switch to the latest versions of MG (I tried both 2.9.9 and 2.9.11). But I got the same result as before. I think I followed the exact generation setup as you did. And I was using the default scales etc. In particular, in the case of MadSpin, I still get different PT distributions from tau, compared to e/mu. (which I didn't see in case of decaying top using madgraph.) I don't think the difference is any fluctuation, as this has been very systematic since the first set of tests/samples I produced.

I have included below the plot comparing the e/mu/tau pT distributions that I produced using mg2.9.11, together with the setup I used:
https://drive.google.com/drive/folders/16jR-pkb1Cm1phZA2_Di1HNnNXlT3S-4v?usp=sharing

Did you see similar level of difference in the lepton pT distributions? If not, what could be possibly wrong with my setup?

Cheers, Quake

Can you include scale uncertainty in your plot?

As said above, the scale computation is not (can not be) the same in madevent and madspin.
So All differences should be taken within scale uncertainty. (I reported them to be 70% for the total cross-section).
Now for shape, the dependence is typically smaller (since you remove the uncertainty on the shape).

This being said, this looks much better than the previous plot and this I can believe that both are correct and consistent with each other.

Cheers,

Olivier

Hi Olivier,

Thanks for the follow-up.

I can add scale uncertainties in my plot. But maybe let me ask a few (maybe naive) questions first...

I suppose a significant part of the scale variations will affect the top kinematics, which will then be propagated to the leptons. I have seen that the top kinematics are very consistent in all setups. The only difference I saw was in the tau pT. So I suppose I could compare the lepton pT with consistently varied scales (e.g. e/mu/tau pT with the same scale variation), and see if that changes the different between tau/e/mu? Just to understand what's sensible to compare. As I think it's unfair to make a direct comparison with the distribution with varied scales, because that also contains variation in the top quark kinematics.

If I understand correctly, MadGraph and MadSpin are using different scales to decay the top and W? Is there a way to manually set the scales in MadSpin so that it's consistent with what's in the MadGraph? I'm using the default scale options in MadGraph.

Not sure this is related to alpha_S, as the difference is in the top decay, which is purely EW right? I would assume that top decay is well understood and there's a "best" setup for the scale choices but maybe not? So what exactly is the difference between the dynamical scale choices between MadGraph and MadSpin?

So CKKW is implemented by default in MadEvent? How does this affect the tau pT? I thought this affects the QCD radiations, which would have shown up in the top kinematics already. Plus I'm doing everything at parton level before any shower. But maybe I misunderstood something here?

I have actually tested to set tau mass to 0 in case of decaying top using MadGraph. but this resulted in almost 0 difference.

But in any case, when I'm decaying the top quarks with MadGraph, I always get very consistent pT spectrum between the different flavours of leptons:
https://drive.google.com/file/d/176BD0qvyykKcQMoDiMB8TG9Hkp-dU8uZ/view?usp=sharing
I believe this makes much more sense?

Sorry the questions are a bit scattered around. I appreciate your patience.

Cheers, Quake

Hi,

> If I understand correctly, MadGraph and MadSpin are using different
> scales to decay the top and W? Is there a way to manually set the scales
> in MadSpin so that it's consistent with what's in the MadGraph? I'm
> using the default scale options in MadGraph.

With the default scale (using the CKKW prescription to determine the scale)
this is not possible.

With any of the pre-defined setup of scale this is not possible (but fixed scale obviously)

Now if you use HT/2 for madevent, then you can perform a re-weighitng of the madspin sample to match that
HT/2 scale scale and then you can compare your madspin weighted sample against your madevent sample with the exact same scale choice.

> I would assume that top decay is well
> understood and there's a "best" setup for the scale choices but maybe
> not?

I do not think that one can really speak of a "best" setup. The choice of the formula for the scale is equivalent to choose a prior. Then of course people tend to select scale choice that seems to better reproduce data.
But on the theory side, I do not know any real "best" setup.

> So CKKW is implemented by default in MadEvent? How does this affect the
> tau pT? I thought this affects the QCD radiations, which would have
> shown up in the top kinematics already. Plus I'm doing everything at
> parton level before any shower. But maybe I misunderstood something
> here?

as said above the CKKW algorithm to determine the scale (also use in MLM)
is used to determine the renormalization scale in computation.
This is not the full CKKW algorithm, (i.e no merging is applied, no sudakov is computed, ..)
and in particular all radiation of jet/gluon are using the renormalization scale.

Now some have argued that event without merging, one can use multiple scale (for alphas/PDF) via the CKKW algorithm
to have a better QCD description and this is available in MadGraph if you set ckkw parameter on "1"

> I suppose a significant part of the scale variations will affect the top
> kinematics, which will then be propagated to the leptons. I have seen
> that the top kinematics are very consistent in all setups.

I do not know all the detail here, but this statement might be missleading
- it is very difficult to fully check a 6 dimensional histograms.
- you do have important spin-correlation into the decay here, so the picture of factorizing the part in two might not be 100% valid
- You do have distribution that does not 100% match (look at the M(t t~) for example)

This being said, I can be wrong and this might not be scale related.
But this is my best bet for the moment.

Cheers,

Olivier

> On 3 Aug 2022, at 18:15, Yang Qin <email address hidden> wrote:
>
> Question #702049 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/702049
>
> Status: Answered => Open
>
> Yang Qin is still having a problem:
> Hi Olivier,
>
> Thanks for the follow-up.
>
> I can add scale uncertainties in my plot. But maybe let me ask a few
> (maybe naive) questions first...
>
> I suppose a significant part of the scale variations will affect the top
> kinematics, which will then be propagated to the leptons. I have seen
> that the top kinematics are very consistent in all setups. The only
> difference I saw was in the tau pT. So I suppose I could compare the
> lepton pT with consistently varied scales (e.g. e/mu/tau pT with the
> same scale variation), and see if that changes the different between
> tau/e/mu? Just to understand what's sensible to compare. As I think it's
> unfair to make a direct comparison with the distribution with varied
> scales, because that also contains variation in the top quark
> kinematics.
>
> If I understand correctly, MadGraph and MadSpin are using different
> scales to decay the top and W? Is there a way to manually set the scales
> in MadSpin so that it's consistent with what's in the MadGraph? I'm
> using the default scale options in MadGraph.
>
> Not sure this is related to alpha_S, as the difference is in the top
> decay, which is purely EW right? I would assume that top decay is well
> understood and there's a "best" setup for the scale choices but maybe
> not? So what exactly is the difference between the dynamical scale
> choices between MadGraph and MadSpin?
>
> So CKKW is implemented by default in MadEvent? How does this affect the
> tau pT? I thought this affects the QCD radiations, which would have
> shown up in the top kinematics already. Plus I'm doing everything at
> parton level before any shower. But maybe I misunderstood something
> here?
>
> I have actually tested to set tau mass to 0 in case of decaying top
> using MadGraph. but this resulted in almost 0 difference.
>
> But in any case, when I'm decaying the top quarks with MadGraph, I always get very consistent pT spectrum between the different flavours of leptons:
> https://drive.google.com/file/d/176BD0qvyykKcQMoDiMB8TG9Hkp-dU8uZ/view?usp=sharing
> I believe this makes much more sense?
>
> Sorry the questions are a bit scattered around. I appreciate your
> patience.
>
> Cheers, Quake
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.