[hwangyw]

• Adds a PWGJE task for charged b-jet fragmentation and groomed substructure analysis using the GNN b-jet tagging information from o2-analysis-je-jet-taggerhf
• Produces fragmentation (z, jT, r) distributions and Soft Drop groomed substructure (zg, Rg, nSD) distributions in MC

[github-actions]

O2 linter results: ❌ 0 errors, ⚠️ 6 warnings, 🔕 0 disabled

[nzardosh]

Hi, out of curiosity are you making any attempts to reconstruct the b-hadron itself? We have shown numerous times that not reconstructing the b hadron leads to a significant smearing of the substructure and one should not do substructure with b-tagged jets without a reconstructed hadron. Also without reconstructing the b-hadron the fragmentation observable doesnt really make sense since you are missing the most significant contributor which is the B-hadron

[hwangyw]

Hi, out of curiosity are you making any attempts to reconstruct the b-hadron itself? We have shown numerous times that not reconstructing the b hadron leads to a significant smearing of the substructure and one should not do substructure with b-tagged jets without a reconstructed hadron. Also without reconstructing the b-hadron the fragmentation observable doesnt really make sense since you are missing the most significant contributor which is the B-hadron

Hello, thank you for the comment. I understand that B hadron decays can affect these observables. For the fragmentation observables, the current implementation considers all charged-particle constituents of jets, similar to the LHCb charged-hadron measurement (https://inspirehep.net/literature/3082668). This is a baseline implementation, not a finalized analysis definition. The impact of B hadron decays on these observables will first be evaluated.
The same applies to the substructure task, and no final physics interpretation will be drawn from it at this stage. In the next stage, I am planning to address the treatment of B hadron decay products and include B hadron reconstruction. The exact reconstruction strategy has not yet been finalized.

[nzardosh]

Hi, out of curiosity are you making any attempts to reconstruct the b-hadron itself? We have shown numerous times that not reconstructing the b hadron leads to a significant smearing of the substructure and one should not do substructure with b-tagged jets without a reconstructed hadron. Also without reconstructing the b-hadron the fragmentation observable doesnt really make sense since you are missing the most significant contributor which is the B-hadron

Hello, thank you for the comment. I understand that B hadron decays can affect these observables. For the fragmentation observables, the current implementation considers all charged-particle constituents of jets, similar to the LHCb charged-hadron measurement (https://inspirehep.net/literature/3082668). This is a baseline implementation, not a finalized analysis definition. The impact of B hadron decays on these observables will first be evaluated. The same applies to the substructure task, and no final physics interpretation will be drawn from it at this stage. In the next stage, I am planning to address the treatment of B hadron decay products and include B hadron reconstruction. The exact reconstruction strategy has not yet been finalized.

Hi, thanks for your reply :) the impact on the substructure has already been demonstrated numerous times and we have shown that the decay daughters fully smear substructure properties across a variety of observables. I think for such analyses the question is quite settled that not reconstructing the beauty hadron leads to incorrect results (you can look here at section 4.2 of this paper https://arxiv.org/pdf/2511.10666 as an example). CMS has started using a ML based technique to try to reconstruct beauty-hadrons from partial decay information but its not clear to me if this can be done without good muon identification. This is an idea that might be worth looking in to at ALICE though.

In terms of the fragmentation of other charged constituents one could indeed do this but the trouble is that you include the decay daughters of the beauty hadron which have significant momentum but will give you inaccurate values of z. So here you also get incorrect results (not sure why LHCb doesn't account for this but I don't think their method is correct)

It is good to hear that you want to look into B-hadron reconstruction and the code for this already exists in the framework (you can look at the B0 or B+ jet finders that we have). I am not sure if we have enough statistics for such measurements but you could reimplement your code in this direction. However this would mean using a different paradigm than the current b-tagging you have implemented, so would require a substantial rewriting of your task.

To me personally the methodology behind this analysis idea will not work in its current implementation and I would recommend it not being carried out on data using the computing resources. Perhaps before committing this code it could be useful to check the impact on your observables of interest in simulations. However the decision with this should lie with @jaimenorman and Wenqing as convenors so I tagged them here to advise on how to proceed.

Thanks and sorry for any inconvenience this causes

[hwangyw]

Hi, out of curiosity are you making any attempts to reconstruct the b-hadron itself? We have shown numerous times that not reconstructing the b hadron leads to a significant smearing of the substructure and one should not do substructure with b-tagged jets without a reconstructed hadron. Also without reconstructing the b-hadron the fragmentation observable doesnt really make sense since you are missing the most significant contributor which is the B-hadron

Hello, thank you for the comment. I understand that B hadron decays can affect these observables. For the fragmentation observables, the current implementation considers all charged-particle constituents of jets, similar to the LHCb charged-hadron measurement (https://inspirehep.net/literature/3082668). This is a baseline implementation, not a finalized analysis definition. The impact of B hadron decays on these observables will first be evaluated. The same applies to the substructure task, and no final physics interpretation will be drawn from it at this stage. In the next stage, I am planning to address the treatment of B hadron decay products and include B hadron reconstruction. The exact reconstruction strategy has not yet been finalized.

Hi, thanks for your reply :) the impact on the substructure has already been demonstrated numerous times and we have shown that the decay daughters fully smear substructure properties across a variety of observables. I think for such analyses the question is quite settled that not reconstructing the beauty hadron leads to incorrect results (you can look here at section 4.2 of this paper https://arxiv.org/pdf/2511.10666 as an example). CMS has started using a ML based technique to try to reconstruct beauty-hadrons from partial decay information but its not clear to me if this can be done without good muon identification. This is an idea that might be worth looking in to at ALICE though.

In terms of the fragmentation of other charged constituents one could indeed do this but the trouble is that you include the decay daughters of the beauty hadron which have significant momentum but will give you inaccurate values of z. So here you also get incorrect results (not sure why LHCb doesn't account for this but I don't think their method is correct)

It is good to hear that you want to look into B-hadron reconstruction and the code for this already exists in the framework (you can look at the B0 or B+ jet finders that we have). I am not sure if we have enough statistics for such measurements but you could reimplement your code in this direction. However this would mean using a different paradigm than the current b-tagging you have implemented, so would require a substantial rewriting of your task.

To me personally the methodology behind this analysis idea will not work in its current implementation and I would recommend it not being carried out on data using the computing resources. Perhaps before committing this code it could be useful to check the impact on your observables of interest in simulations. However the decision with this should lie with @jaimenorman and Wenqing as convenors so I tagged them here to advise on how to proceed.

Thanks and sorry for any inconvenience this causes

Thanks for the detailed explanation and the reference. I will first quantify the impact of B hadron decays on these observables in simulation and investigate possible B hadron reconstruction strategies.