Problem with b decays: B hadron lifetime much too short in 2/3 of all cases?

[Martin Flechl]

I am producing ttbar->bqqbtaunu events, where the first tau in the event is forced to decay hadronically. Sherpa event files are read in with the ATLAS Sherpa interface and then analyzed.

After detector simulation, I observed a b-tagging efficiency which is a factor 3 lower than expected. A quick check by b-tagging experts indicates that the (truth) B hadron decay length is very small in about 2/3 of all cases.

The Run.dat is attached. The native hadron decay packages has been used but the same effect is observed when running with the Lund one. The whole Decaydata directory has been copied to the Run directory and only the tau decay info has been modified.

Martin

[Frank Siegert]

Hi Martin,

as a first test I checked all B0, B+ and B_s lifetimes in events produced with your Run.dat, and they look fine to me. Now before I go on to check the issue of decay lengths (taking the boosts into account), I would like to have some more information on what exactly you are seeing vs. what you are expecting when you say "B hadron decay length is very small in about 2/3 of all cases".

Thanks, Frank

[devivie@lal.in2p3.fr]

Hello,

I had a quick look at Martin's events. For the time being, I only considered charged lowest lying B mesons (|pdgId| = 521). The decay length seems very small for 70% of their decay. For the remaining 30%, a quick-and-dirty fit seems to give c*tau ~ 500 um, which is as expected.

I have the impression that the difference between the two populations comes from the origin of the B mesons: for the "low lifetime" one, they originates from the decay of excited state (mainly B*(+-), |pdgId| = 523); for the "normal lifetime" one, they originate directly from the fragmentation of the b-quark.

I attached three plots :

o L = decay_length / (beta*gamma). With a cut L > 0.01, I can observe c*tau = 500 um. But this concerns only ~ 30% of the B mesons.

o the |pdgId| of the "first mother" of the B meson. Dots are for L < 0.01 and the histogram is for L > 0.01. One can see peaks at 523 and < 100 respectively.

o the approximate transverse impact parameter of tracks from the B decay. Dots are for tracks from B whose "first mother" |pdgId| is 523, and the histogram is for tracks from B whose "first mother" |pdgId| is below 100. (Distributions are normalized.)

I am not very confident with my code. Maybe I cannot consider the two populations in the same way and all the above observations are wrong. Nevertheless is it possible that B from direct fragmentation and B from excited state decay are treated differently, with a problem in the later case ?

Best,

Jean-Baptiste

[Frank Siegert]

Hi Jean-Baptiste,

thanks for the detailed description, now I know what to look for. I will keep you updated about my findings.

Frank

[Frank Siegert]

Hi Martin and Jean-Baptiste,

finally a resolution for this bug report, sorry that it took another week. The problem was, that the vertex position was calculated in the wrong stage of the hadron decay, namely when the blob was still in its CMS. This explains why you only saw it for B's from decays, and not for B's from fragmentation.

I am attaching a patch that solves the issue. It will also be provided as an official patch on our download page.

Thanks again for your detailed bug report.

Frank

[Frank Siegert]

Having looked at the decay length code, I realised, that there might be another source of B's which don't have the correct decay length set. Some B mesons decay partonically, with subsequent showering and hadronisation of the decay products. During this process, the decay vertex position seems to get lost. I have to check further into this, and might provide an additional patch to fix this (separate) issue.

[anonymous]

Hello Franck,

Just to confirm, indeed, looking again at Martin's events, and only at charged B originating directly from b fragmentation, the decay length seems fine in all cases, but the subsequent quasi stable charged particles (pions, kaons, protons) have a small impact parameter when the charged B decay through partons. I attached a plot of this transverse IP, for all B+- from b fragmentation (upper), for those decaying via partons (middle plot) and those decaying directly through hadrons (lower plot).

Sorry, I should have seen this earlier, but I did not go further after my first observations of the impact of the B origin.

Best,

Jean-Baptiste

[Frank Siegert]

Thanks for the confirmation. I have prepared a patch, which should solve the problem with the partonic decays I described. Can you try it out, and let me know, whether you still see sources of strange B decay lengths?

patch -p0 < partonic-decay-vertex-position.patch

This has to be applied in addition to the other patch!

Thanks, Frank

PS: I plan to upload this patch as official before tomorrow morning, such that GENSER can build a Sherpa version 1.1.3p including these two patches tomorrow (I have notified them already). If you can provide any findings on whether the patch works or doesn't by then, please write them here.

[Frank Siegert]

I have added the patch to our official download page, and asked GENSER to install a Sherpa 1.1.3p containing the two patches.

PS: The patch attached to this bug report applies to version 1.1.2, while the one on the download page applies to 1.1.3.

[Martin Flechl]

Hi Frank,

Just to confirm that I tested the patch and that everything looks fine now (as a simple check, by comparing a SHERPA and a PYTHIA ttbar sample run through ATLAS detector simulation), the problem seems solved. Sorry that I could not provide feedback more quickly and thanks for your efforts!

Cheers, Martin

[Stefan Hoeche]

Milestone old deleted