Alternative look at Zb search data and MC Marj Corcoran, 28 Oct 2104 - - PowerPoint PPT Presentation

Alternative look at Zb search data and MC

1

Marj Corcoran, 28 Oct 2104

Types of background

There are three types of backgrounds: Sidebands: M(μ μ) for this background is outside of the upsilon signal region. It represents dimuon events which are not upsilons, and which have a third track associated with them. The sidebands in M(μ μ) should be used to get this background, and the normalization is determined by the area of this background under the upsilon peaks. Combinatorial: M(μ μ) for this background is in the upsilon signal region. There is a real upsilon, but it is associated with a random track. The original idea was to estimate this background with the mixed events —the dimuon from

• ne event and the third track from another event.

Feeddown: This background comes from three sources, , Y(2s )→ Y(1s)π π, Y(3s)→ Y(1s) π π, and Y(3s)→ Y(2s) π π. These decays have Delta-m distributions that fall in the 1 GeV mass region and overlap the Zb signals. We should be able to get an absolute normalization of these events from , Y(2s )→ μ μ and , Y(3s )→ μ μ Monte Carlo samples, along with the known branching ratios.

2

Sideband background vs. combinatorial background

The sideband and combinatorial backgrounds are not necessarily the same, here are plots of Delta-m for the two cases. Sideband background Combinatorial background from mixed events

3

Combinatorial background

The mixed event background was the original idea for getting the combinatorial background, but it is not a good representation, don’t quite know why. I show later that the sidebands actually model the combinatorial background well. Blue=data, green = mixed event background.

4

Some comments on the isolation cuts

We are making some very tight isolation cuts, I >0.9 for all three tracks. These cuts have a big impact on the signal acceptance, and I don’t think they are helping much. Zb1→ 1s π, with MC matching. Acceptance for all three isolation cuts is 12.8% 3s → 1s ππ, with MC matching. Acceptance for all three isolation cuts is 10.5%.

5

Isolation cuts

Zb1→ 1s π with pT(μμπ)>8 GeV cut Mixed-event background (old version, without iso cuts),, with pT(μμπ)>8 Gev . There’s a little to be gained in decreasing the combinatorial background, but we are losing 85% of any signal events. We could gain a factor of at least 2 in statistics.

6

Isolation impact on acceptance

There is one issue with the isolation cut as far as getting the acceptance of the Zb signal. We rely on the MC to give the acceptance, but we know from both Julie and Michelle’s analysis that the MC tends to overestimate the number of events in the I=1 spike. What we have done in the past is find a control data sample and reweight the MC so that data and MC agree for the control sample. We really should do this for this

• data. We can’t rely on the MC to get it right, and we are applying the isolation cut three times.

7

How to suppress the feeddown background—pT of third track

pT of third track, Y(2S)→Y(1S) ππ, reco-ed as Y(1S)π, with MC matching on the third track. pT of third track, Zb1 →Y(1S) π, with MC matching on the third track. There is a huge difference in the pT distributions of the third track. If we cut at pt>1, the feeddown background is pretty much killed, if we believe the Monte Carlo.

8

Suppress feeddown—pT of μμπ system

9

Zb1 →Y(1S) π, pt distribution of μμπ. Note the Zb MC treats the Zb as a spin 0 particle Y(2s) → Y(1s) ππ, pt distribution of μμπ A pt cut of 1 GeV on the third track, plus a pt cut of 10 GeV on the μμπ kills all of the Y(2s) → Y(1s) ππ background, and most of the Y(3s) → Y(1s) ππ background

Combinatorial background modeling

Another way to model the background is the MC sample Y(2s) → Y(1s) ππ. The green curve is an estimate of the combinatorial background from this MC, with pt cuts so that there is no signal. Turns out the sidebands (scaled to the right number of events) is also a good model of the combinatorics.

10

Data and feeddown backgrounds

11

RunIIb2 data, scaled sidebands, and Y(3s) → Y(2s) ππ MC. There is a cut of pt>1 on the third track, and pT(μμπ) >10. There is no actual Y(3s) → Y(2s) ππ signal here, since the πs are too soft to pass reco cuts.

Feeddown backgrounds

12

Here’s the Y(3s) → Y(1s) ππ feeddown MC. The cuts here are pt>1 for the third track and pt(μμπ )>10. Some

• f the feeddown signal is getting through, right where there is a peak in the data. These plots are the same,

just different ranges.

Delta-mass with Y(1S) removed

13

I can completely remove the Y(1S) contribution by cutting on the M(μμ). This is what I get, there is an excess at low mass, which can’t be feeddown background, at least not the channels we have

• considered. The combinatorial background from MC doesn’t work so well now, but the scaled

sidebands still seem to model the background. Magenta line shows all six Zb transitions.

Summary

14

The mixed-event sample does not give a good description of the combinatorial background, but the M(μμ) sidebands do. The tight cut on isolation for all three tracks presents some complications in determining the acceptance, since we know from past experience that the MC in general does not model the isolation spike at I=1 correctly. If we believe the MC samples that we have, the feeddown backgrounds can be mostly eliminated by cuts on th epT of the third track and pT(μμπ). The remaining the Y(3s) → Y(1s) ππ events sit right on the excess in the data. A cut on M(μμ) to exclude Y(1S) should eliminate all of the feeddown background. However an excess at low Delta-m still exists in the data. I think Brad saw the same thing. We need to address this excess—it is consistent with one of the Zb transitions, but it might also be an additional background we have not considered.