(Primary) Vertex mismatch - wrong smearing?

[saracreates]

Hi!

While trying to debug the discrepancy in the impact parameter d0 between full and fast sim, see CLDconfig issue, I have found an other bug.

I'm looking at Huu data (/eos/experiment/fcc/prod/fcc/ee/test_spring2024/240gev/Huu/CLD_o2_v05/rec) and different production campaigns:

1. 00016553: done by @BrieucF in June 2024
2. 00016774: done by @kjvbrt in Jan 2025 using the newest key4hep version, trying to keep everything the same
3. 00016808: done by @kjvbrt in Feb 2025 with the intent to solve the mentioned issue in the CLDconfig

First, I have investigated 1) with the conclusion, that the MC primary vertex (PV) in fast and full simulation match, but the reconstructed PV is off, again see the mentioned CLDconfig issue.

There are two ways how I define "MC PV" in full sim (assuming H->uu events):

1. I look at the Higgs vertex
2. I look at the vertex of the Higgs daughters (u ubar; they are the same for both quarks)
   If everything goes right, the two definitions should be the same which is also observed in data 1).

The issue(s):
a) Data 2) and 3) have a problem: the two vertex definitions do not match!! Def. 1) gives the PV distribution as expected and seen in fast sim, while def. 2) gives results completely off (see distribution). It is interesting to note, that $y$ and $z$ seem to be smeared twice. One gets the wider distribution if the expected values (as seen in fast sim) are multiplied by a factor $\sqrt(2)$.

If I print out the vertex positions of the Higgs and its daughters in data 2) and 3), they are not the same!! If I print them in 1) they are - as expected - the same.

Any ideas what is going wrong here?

I suggest looking into the vertex smearing that is done in dd4hep... it might be that this is applied twice for $y$ and $z$. I have no idea what is happening with $x$.

Also, why do the Higgs and the daughters have different vertices? Something must be wrong here.

What are the differences between data 1) and 2)/3) that this bug occurs?

b) Even when fixing the beamspot constrains, discussed here, the reconstructed PV does not get fixed (to be clear: no improvement going from data 2) to data 3) ). Maybe this should be further discusses in the other issue but I still wanted to mention it here too.

Thanks for your help!
Sara

[BrieucF]

Hi, sorry I could not closely follow this up lately.

It looks like the problem is indeed that the vertex smearing is applied twice. Once with the generator (https://github.com/HEP-FCC/FCCDIRAC/pull/9/files#diff-0efff741aad577826c9540d1cc544677bc6ea0d5f98ae4203168ee99a31c915cR82), and once with ddsim (https://github.com/HEP-FCC/FCCDIRAC/pull/9/files#diff-b9a052b09f8d39eb1ba558bf262fce50a421e030336b828691446688c38ed1f5R6).

To fix the 'double smearing' @kjvbrt, apply this change from the whizhard/pythia config: HEP-FCC/FCCDIRAC@95b3776#diff-b01bbc5f55d44eac1440ab10c37511d4ba454490a8333d4cd693022c942b5f03L49

[kjvbrt]

Thx @BrieucF, will regenerate new files from scratch :)

[andresailer]

Re-use the SIM files used for 00016553 ?

[kjvbrt]

Re-use the SIM files used for 00016553 ?

These samples use different version of the stack, hence incompatible versions of PODIO/EDM4hep.

[andresailer]

Re-use the SIM files used for 00016553 ?

These samples use different version of the stack, hence incompatible versions of PODIO/EDM4hep.

Ahh, ok.