Missing D^0 DDX decays

[afernez]

Phoebe reminded us that for run1, some $DDX$ decays were missing in the cocktails for the $D^0$ sample--in their ANA, they noted the decays were "of the general topology $B\rightarrow D^{*}[\rightarrow DzX]DK^{*}$" (without a $\tau$). Checking the dec files (run2--identical to run1--11894600 and 12893600 for $D^0$, 11894610 and 12895400 for $D^{*}$), the problem seems to be that for some decays where one $D$ ends up as $K\pi (\pi{slow}/\gamma)$ and the other $D$ yields the $\mu$, the interchanged decay where the former $D$ yields the $\mu$ and the latter $D$ the $K\pi (\pi{slow}/\gamma)$ isn't included in the cocktail, but if the two $D$ mesons aren't the same (eg. not just $D^{*+}D^{*-}$) and both could potentially give the $\mu$ or $K\pi (\pi_{slow}/\gamma)$, the interchanged decay isn't identical and should be added to the cocktail.

Note: in the dec files, the $D$ mesons that could give either a $\mu$ or $K\pi (\pi{slow}/\gamma)$ are $D=D^{*+},D^{*0},D^0$, and the naming convention is such that a particle whose name includes "Other" will give a $\mu$ while the non-"Other" will yield $K\pi (\pi{slow}/\gamma)$.

Phoebe suggested we could potentially follow the same path as run1 and account for these missing decays by re-using comparable events from the $D^{*}$ sample. To implement this, I started writing a script in this repo that would manually add on the $D^{*}$ sample events (reconstructed as $D^0$--it's very nice that we had already reconstructed every event in both the $D^0$ and $D^{*}$ modes!) to the postprocessed $D^0$ sample muonic $DDX$ ntuples--postprocessing the events following the same workflow as normal $DDX$, but adding on some truth-matching code to select out the specific missing decays and also fake-passing the $D^{*}$ veto for half of the events (I'm not sure I understand this still... the run1 ANA notes that this is done "to mimic the $D^{*0}$ component which would otherwise still be missing in this approach"). However, I quickly found that the added truth-matching was cutting everything out, and it doesn't seem to be an accident.

The specific decays in the $D^0$ $DDX$ dec files that seem to be missing are

• MyOtherD*- MyD0 K*+
• MyOtherD*- MyD*0 K*+
• MyOtherD*- MyD0 K+ (not a $K^{*}$, but I don't see why this decay shouldn't be included too?)

but... these decays are also missing in the $D^{*}$ $DDX$ dec files!

Unless I'm fundamentally misunderstanding what the issue is and what run1 did, I don't think we can get events for these decays without requesting additional MC.

@yipengsun (just because I had said I would check my plan to implement this with you)

[yipengsun]

Naively, I think by "reusing", we just need to manually set the d0_dst_veto_deltam to a large value (e.g. 100) so that d0_dst_veto_ok is true (this means that the event is unlikely to be a D* event). In addition, in the truth-matching for the corresponding D0 decays, we now accept D* event as well.

Is my understanding above correct?

However, the main problem seems to be that the D* decays (MyOtherD*- MyD*0 K*+) are missing as well...

[afernez]

@hadjichris leading up to preparing a MC request, I'll try to re-summarize the problem and what I understand about what Phoebe/Greg did for run1, and then if you want to discuss, whenever's convenient for you we could meet about any questions or suggestions.

The decays that I believe are missing from the $D^0$ $DDX$ cocktails are what I listed in the first post: MyOtherD*- MyD0 K*+, MyOtherD*- MyD*0 K*+, and MyOtherD*- MyD0 K+. These would all belong to 11894600; I believe 12893600 is fine as-is. Probably a good first thing is to

• [x] Check that these 3 decays are the only ones missing, and that it makes sense that they all ought to be included in the $D^0$ cocktail

If these decays are generated with comparable stats to their analogous decays (eg. MyOtherD*- MyD0 K*+ same stats as the included MyD*- MyOtherD0 K*+), it would amount to about 7% of the previous $B^0 \rightarrow D^0 X_c [\rightarrow \mu \nu X']X$ (11894600) request, so 547.4M*0.07=38.3M tracker-only simulated events (numbers from table). It seemed to me like there was a slight preference for it, but for a MC request we'll need to

• [x] Decide if we'll only request the missing decays

Now, for run1 this issue wasn't addressed with additional MC, but I also think the solution didn't fully cover the missing decays. I haven't seen the code used to do this, so my understanding is still rough (you can read the paragraph 11.2.1 "Additional Simulation for the $D^0 \mu$ Sample" in the run1 ANA for a similar overview), but:

• From a tuple Greg provided that contains the ID information for the re-used decays, the $D^{*}$ cocktail decays they used to cover the missing $D^0$ decays seem to be $B^0 \rightarrow D^{*+}D^{(*)[-,0]}K^{*[0,-]}$ and $B^- \rightarrow D^{*+}D^{(*)-}K^{-}$ (I didn't bother checking who the $\mu$ mother was for all events, though that would would complete the story of which dec file decays are effectively being added). Notably, this doesn't cover the missing `MyOtherD- MyD0 K+`.
• Fake-passing the $D^{*}$ veto (if you aren't familiar, this veto is used to make the $D^0$ and $D^{*}$ sample mostly disjoint; see the run2 ANA 3.6 "Veto of Overlapping Candidates", for example) seemingly was used to take some events with a $D^{*+} (\rightarrow D^0 \pi_{slow}^-)$ and pretend that it was a $D^{*0}$ instead; I guess this was to compensate for the fact that the $D^{*}$ dec file $D^{*0}$ always decays with a $\mu$.

I think we've mostly agreed that the best way forward is to request more MC rather than following what run1 did, but any thoughts you have are useful too, so also we should

• [x] Agree that requesting more MC is better and/or easier than re-implementing run1's solution

[hadjichris]

@afernez Thanks Alex for this summary. Give me a few more days to digest the material so that I can verify what you are proposing and we can talk early next week. I will also contact Scott to get more information about the $\Lambda_b^0$ background. Eventually, we might want to include that in our MC request.

[afernez]

In our analysis meetings I guess we've agreed to make a new .dec file with just the missing decays, and we'll make a MC request with that. To put together/finalize what we've discussed for stats/BFs: I think we can use the same relative ratios as for the analogous decays, ie.

Decay B0sig (old)                  -->          Decay B0sig (new)
0.0020 MyD*- MyOtherD0 K+          -->          0.0020 MyOtherD*- MyD0 K+  
0.0018 MyD*- MyOtherD0 K*+         -->          0.0018 MyOtherD*- MyD0 K*+
0.0033 MyD*- MyOtherD*0 K*+        -->          0.0033 MyOtherD*- MyD*0 K*+  

and then for overall stats multiply by 1.5 to account for the fact that MyOtherD*- can decay as $D^0 \pi^-$ (isospin argument->2/3; this is the only decay mode for MyD*-) or as $D^- \pi^0(/\gamma)$ (isospin argument->1/3). This is similar to what was done in the 11894600 dec file for MyD*- MyOtherD*0 K+ (BF=0.0079) vs MyOtherD*- MyD*0 K+ (BF=0.0118). Overall, this will mean 1.5*0.066*547.4M=54.1M simulated events (the 0.066 comes from the sum of these decays' rel BFs over the sum of all the B0sig rel BFs in the 11894600 dec file). Broken down by year (referencing table):

Decay	Nsim [M] 2015	Nsim [M] 2016	Nsim [M] 2017	Nsim [M] 2018	Nsim [M]	% Prev Tot Req
B0->D0(Xc->munuX')X (delta)	2.65	15.26	15.91	20.29	54.10	0.7%

[afernez]

We've requested the missing DDX MC, it's been produced, and we've integrated it into our tuples, so I'll close this.