Sandra questions 170715

Dear Sandra, Many thanks for your questions and comments. Here are our answers.

1)Formula 1 . I see that you use the PDG M mass instead of the reconstructed one; Iwould then assume that you correct the p_T accordingly but I do not see anydiscussion about it.

Answer: using the B PDG mass is common as you can verify in AN-14-045, AN-12-044, and refs. [8-10] of our AN. The reconstructed mass is never used in the definition of the proper decay time (or length). On the other hand we use the corrected pt values obtained from the mass constraints of the intermediate resonances.


Since this work will be the CMS Run1 legacy paper for B-hadron lifetime analyses, I think that this is ''the proper time" to discuss and set a standard approach. Or, differently, can you point me to some already available CMS material and studies about that?

A couple of comments:

1.) the two Analysis Notes you referred to do not include any discussion about the reason, the benefit and the implications of using the M_PDG.

2) LHCb uses B-hadron reconstructed mass, see for instance the recent Phys.Lett.B736 (2014) 154-162: measurement od the $\Xi_b$ and $\Omega_b$ baryon lifetimes.

3. I understand tthe rationale of using constaints on the intermediate resonances (J/Psi, f0 etc refitting momenta..); I am wondering about the need and reason of using the B-hadron PDG mass ( which would dimply additional momenetum corrections)


Reply: We checked the difference between the lifetimes obtained using both approaches in the B+ -> Jpsi K decay channel (the highest statistics channel) and found ctau(using reco mass) - ctau(using PDG mass) = 0.00 microns, and hence the approaches are equivalent in this case. We can mention in the next version of the note that we performed this check and that the difference is negligible.

2) Sec 2.2 MC simulations: No list of MC datasets is provided.Does it mean that you use private MC samples (at least for some decays) ? If so,have they been centrally validated? Anyhow, I suppose you have checked the data/MC agreement on the main variables of your analysis; some distributions should be added in the AN.

Answer: Official Bd to J/psi K* MC is used for the efficiency studies. We will update the note accordingly. However, due to the flatness of the efficiency as reported in our note and also in AN-14-045 (Bs to J/psi phi) the analysis results do not depend on the MC simulation.


Well, this is not exactly what I read in the AN -14-045 (7.1.2 pag 26 (and formula 11) "...that we have an almost flat efficiency distribution. Nevertheless, we fit a not straightforward function in order to get the structures still present in the efficiency shape"

In addition,on how can you assume it flat, a priori, for all your hadrons and final states?


Reply: In the section 21.2 (pag 55) of AN -14-045, they mention: "In the final fit no proper decay time corrections are involved. It means that we assume the proper decay time efficiency to be flat. [...] Then we want to estimate what is the systematic error due to that assumption.". It just means that they do not use the efficiency function for their nominal fit. This is reasonable, since the change in the lifetime when they use the function or not is less than 1 micron, as we have also verified for all our decay channels. On the other hand, attempting to use the efficiency curve (for nominal fits, each crosscheck or systematic estimations) would take about 10 times more time. This is almost prohibited for samples such as B+->Jpsi K. So we assume flatness and pay the consequences for assuming this, having to add a corresponding systematic uncertainty.

With respect to your original question, we certainly checked the agreement between data and MC for this official MC sample: see backup of

We can include these plots in the note.

3) Line 104: Can you confirm that you will remove the tracks coming from thereconstructed b-hadrons out the PV in your final selection?

Answer: No, we did not, however for the estimation of the systematic uncertainty due to the PV selection we removed the tracks and found that the difference is negligible. This will be included in the next version of the note.

4) Line 114: Which is your muon selection?

Answer: Global Muon. This will be included in the note.


Why don't you use the standard Muon POG recommended selection ,ie. for 2012 data ,"softMuon"?


Reply: Because the softMuon selection biases the lifetime several microns when we studied it in MC (and the effect is reproduced in data). The bias was tracked down to the IP requirements in the softMuon selection. Global muon and PF do not have those problems. Similar observations were reported at a given moment by Giacomo and Theri.

5) Line 130: Which is your track selection?

Answer: We used the tracks from the container thePATTrackHandle. In addition we required a pt>0.5 for each track and excluded those tracks associated to muons.


Can you briefly recall what this container contains (just the main selection criteria)?


Reply: We are looking for what are the criteria to form the container, as soon as we know it we will update this point.

6) Line 139: Do you implement a Kinematic Vertex Fit with a double constraint for theB0 and Lambda_b decays?

Answer: Yes, However this is done in two steps. First the Lambda^0 or K^0_s vertex mass constrained fit is performed, and then this is used as an input for the vertex mass constrained fit of the mumu+Lambda^0(K^0_s) with the mumu mass constrained to the J/psi.

7) Line 156: Are the values of the transverse significance and impact parameterstudied/optimized on MC samples?

Answer: These cuts come from the V0Producer. From which we obtained the Lambda^0's and K^0_s's

8) I have not seen mass distributions for f0, K* etc from b-hadron decays. It wouldbe interesting to look at them to judge the quality of the selected resonances.

Answer: We will include these plots in the next version of the note.

10) Line 182: What do you exactly mean with “weight the decay length shape”?

Answer: We mean to allow enough events from each sideband to model the bkg under the signal peak. We will clarify this in the next version of the note.

11) Line 205:. I would suggest to quantify the “much less than 1” in term of thesignificance of the parameter returned by the fit : zero within 1 (2) sigma? Or noimprovement in chi2 adding the parameter ( if it is the case).

Answer: We will rewrite this sentence in the next version of the note. The slope is fixed when its fitted value is consistent with 0 within its uncertainty (1 sigma) and the chi2 does not improve.

12) Line 212: JpsiK contamination . Can you explain how exactly you study thiscontamination?

Answer: From the f^0 to pipi we selected the leading pt pion and we assigned to it the mass of the kaon, this is combined with the J/psi to construct the B^+ to J/psi K^+. A fit to this mass distribution is performed to obtain the B^+ signal and its width. We selected the B^+ candidates 3 sigmas around the signal peak and observed the distribution of these events in the J/psi+f^0 mass distribution, from which we obtain the template for the B^+ contamination.

13) Is the contribution in the data consistent with the MC prediction?

Answer: We never estimated this in MC. However this number is fixed from the fit to the B^+ signal in data, as you can see in figure 25.

14) I would have thought that the f0 mass cut on the two track combination (and the4-track vertex CL) made it negligible . Can you comment more?

Answer: As we can see from figure 25 it is not (these pictures include all cuts for the B_s to J/psi f^0)

15) At any rate, more in general I think you should explain in more detail how yourreflection peaks are obtained ( appendix E). For instance:

Fig 23 caption says: “when different mass assumption on the final state tracks(originally forming the f0) are assumed”. How are tthe green histograms obtained?

How is B+->JpsiK+ (Jpsi+ 1 track, no “two track forming resonances”)) treated ?

I might misinterpret what you do here; more information might be useful, such aswhether

you start from selecting B0, Bs , B+ (in which final states ?) and then change massassignment or you select first Jpsi pipi, change mass assignment etc..

(same for Fig.24)

Answer: All these studies for the different b hadrons are done in a similar way to what we explained in answer to question 12. We began with our J/psi+ pipi candidates and we changed mass assignment for the pipi tracks (for example KK,Kpi, leading pt pion as a kaon), this to look for B^+ to J/psiK^+, B^0 to J/psiKpi, B_s to J/psiKK. We will clarify this in the next version of the note.

Fig 22. The caption says that the contributions are studied from an inclusive Bs->JpiX . Do the B, B0 Lamba_b come from the “other” b hadrons and are automaticallynormalized to the same luminosity? I think it would be better to add some comments.

Answer: As the caption says this is inclusive B to J/psi X MC at generation level without any specific luminosity normalization. The normalization for each b hadron depends on the branching ratios specified in EvtGen.


What is the integrated luminosity of the MC sample? In other words, how can I compare the contamination yields in Fig. 22 with what we do have in the 2011+2012 data?


Reply: We want to insist that this MC sample is at generation level, and remark that we only use it for illustration purposes. So we believe you cannot compare directly this MC and the data, as you propose, using this figure (need reco efficiencies for each component), even if we manage to calculate the integrated lumi (this plot was obtained in AN-12-363). This study was useful for us to understand the possible contaminations that we expect in data. The yields of these contaminations are obtained directly from the fit to data (some of them are considered all together).

16) Table 2 ( and others) : you report a large background component with the samelifetime as the signal. Can you elaborate a bit more on it?(clarifications about “reflections” might help understanding it as well).

Answer: From the fit to the sidebands in the different b hadron decay channels we have studied, we have observed a long-lived bkg component, which can not be explain only with the "reflections" assumption, there are some J/psi's which come from misreconstructed b hadrons but there are also some others which come from decay chains of other particles, both of them can produce these long-lived bkg, but as we said it is difficult to quantify each of them and we decided just to model them by using the sidebands events.


Don't you find worrisome that your background has the same lifetime as the signal?

Mis-reconstructed b-hadrons or Jpsi from other decay chain are combined wih "combinatorial tracks" and would tend to have shortt lifetime..


Reply: Your reasoning is not verified in sidebands (not all the bkg is combinatorial, as we mentioned there could be decay chains of unrelated particles which can produce long-lived components). We do not find that worrisome since all the purpose of performing a multidimensional fit (mass, ct, etc.) is that it can easily separate signal and background components using (mainly the) mass information that, otherwise, would be undistinguishable if using only ct. This technique is verified with ToyMC.

17) Line 279: what does simultaneous lifetime measurements mean?

Answer: We will remove "simultaneous". We meant lifetime measurements.

18) Line 282: I am not sure I understand what” a fast trigger simulation thatanticipate a negligible trigger effect.” means.

Answer: We will rephrase this sentence. We were trying to explain the toy MC studies we performed and described in appendix D.

19) Line 305: As discussed in last Tuesday meeting you should calibrate your s scale on a not -displaced trigger sample.

Answer: Yes, we are waiting for Sara to send the unbias sample.

20) Fig 10 & 11. It would be better to have a common X scale for homogeneous plots ( atleast for visual inspection).

Answer: Ok we will look for time to do this.

21) Line 400:. I would then suggest to rephrase as “ our study shows that the PV choice.”

Answer: Ok.

22) Line 402 : Did you perform a closure test only on the B0 signal and on 2011 MC?

Answer: Yes, unfortunately for the B_d to J/psiK* we only have 2011 MC.

23) Fig.14 caption : can you comment more on the different lifetime values you obtainwith a trigger unbiased and displaced trigger selection and, in particular, on the455+/-1.4 micron ( input 460 ) from a “trigger unbiased selection”.

Answer: Plot at the center confirms that after all selection requirements (triggers, 200 microns cut, etc) and assumed flat efficiency, the lifetime fit result is unbiased (and Appendix D shows that the efficiency is flat). The right plot shows that fixing s=1 does not change significantly the lifetime fit result. The left plot, in which we did not choose trigger (neither apply triggers cuts) includes the prompt region (<200 microns), where the efficency might not be flat (due to various effects). And this maybe the reason of the difference with the input value, we could study this effect, however, this a region we do not use in our analysis.

24) Line 407 : You apply cut also on the sigma_lambda. It is not clear to me if the cut is considered in your closure test.

Aswer: Yes, we do.

25) Fig.15 : I am not sure I understand the normalization: the text and the figurecaption seem to say something different. In the denominator do you have events notfiring the displaced triggers ( complementary events wrt the numerator) or eventsfiring non-displaced triggers? I think that the proper normalization should be “allthe events”.

Answer: In the denominator we have the events that did not fire the displaced triggers. We will clarify this in the next version of the note. Also, we are working on an additional test using the higher statistics B^+ MC.


Can you then elaborate more on the meaning of this distribution?. It is not a proper "efficiency" ...


Reply: This has been clarified in the last version of the note.

26) Fig 15. What are the units on the X axis, i.e, how many lifetimes does your plotcover?

Answer: The units are cm and the plot covers approx. 10 lifetimes. We will re-do these plots to include the units.

27) I would not conclude from this plot that your efficiency is flat., or, saying inother words, you cannot assume it flat for a precise lifetime measurement.

his assumption should be verified on very high statistics MC and, indeed, you have a too big uncertainty on the beta parameter to reach such a strong conclusion.Line 430: How do you propagate the big error of eq 28 into your analysis?

Answer: In fact due to the poor MC statistics used for these plots we decided to perform the toy MC studies described in appendix D.1, in a similar way as it was done for the B_s to J/psi phi CP-violation analysis. From this we concluded that the efficiency is flat. The closure test (center plot, figure 14,) in MC confirms this assumption. In addition the lifetime measurement of the B^+ to J/psi K^+ was originally considered as a cross-check to confirm in data this assumption, due to the high statistics of this channel, and the well determined this lifetime is. As we can see from our lifetime results, this assumption of the flatness of the efficiency is confirm by the B^+ lifetime measurement.

28) Eq 29 and discussion afterward. the question about fixed PDG MC and not p_Tre-correction comes about again ; consequently the following discussion is not thatclear to me.

Answer: See answer to question 1.

-- JhovannyMejia - 2015-07-22

Edit | Attach | Watch | Print version | History: r5 < r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r5 - 2015-08-01 - JhovannyMejia
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright &© 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback