1. Measurement of the Higgs boson mass and e+e-  ZH cross section using Zμ+μ- and Ze+e- at ILC √s = 250, 350, and 500 GeV European Linear Collider Workshop May 30 – Jun 5, 2016, Santander, Spain Jacqueline Yan , J. Tian , K.Fujii (KEK) and the ILC Physics Working Group
2013/07/20 1

2. Leptonic recoil mass study is finalized
@ECM = 250 GeV, 350 GeV, and 500 GeV
This study shows clearly the impact of ECM and polarization on the precise measurement of σZH and MH as well as demonstrates model independence to sub-% level
Higgs recoil against di-lepton system
Based on full ILD detector simulation
σZH is essential input to measurement of the pattern in the deviation of couplings in order to discriminate new physics models at the ILC
Paper submitted to P.R.D. : arxiv: undergoing review
Reference paper on model independence: arxiv:
Then I added all other BG and sig processes that hhave been generated as samples, even though I presumed them to be not important for Zmumu analysis at 250 GeV.
I wanted to check I didn’t msis anything.
I almost didn’t miss anything, except for WW_sl,
Some of the quarks may have decayed secondarily into muons, and I left it out.
There are about 130 events left after all cuts.
While other were mostly 0.
The results were only a little different after including this.
ILC sample used in analysis
channel MH
ECM L
polarization
e+eZh->μμh
e+eZh->eeh
125 GeV
250 GeV
350 GeV
500 GeV
250 fb-1
333 fb-1
500 fb-1
P(e-,e+) =
(-0.8,+0.3)
(+0.8,-0.3)

3. Data Selection Method Signal signature
a pair of isolated energetic leptons (μ/ e) with
invariant mass (Minv) close to Z mass
Dominant backgrounds Signatures
e+ e-  Z Z  l+ l- X : forward Z production angle
e+ e-  γ Z  γ l+ l- : energetic ISR γ which balance dilepton pt
e+ e-  W W  l+ l- ν ν : broad Minv distr.
cuts are designed to achieve high precision and at the same time maximize signal efficiency and minimize bias on Higgs decay modes
Final Selection
73 < GeV < M_inv < 120 GeV
10 GeV < pt_dl < (…) GeV
|cos(θ_missing)| < 0.98
TMVA cut
Evis cut
observe spectrum within signal region (depend on ECM)
Lepton Pair Candidate Selection
Prompt μ/e ID ,
isolated lepton finder (MVA based)
FSR and bremsstrahlung recovery
Lepton pairing : minimizeχ2(Minv, Mrec)
Here is the data selection I applied, with the numbers from last week Friday.
First I select muon pairs based on ratio of deposit in calorimeter wrt to total energy, opposite charge, and track quality.
Than I pick out the best muon pair candidates base on how close their inv mass is to real Z mass
Next , comes final selection for recoil mass analysis.
Here, in the order that these arei implemented
I require inv mass to be between 86 and 95 GeV,
Then recoil mass to be between 115 and 140 GeV
Then PT of dilepton system to be between 10 GeV and 70 geV
Then acoplanarity to be between 0.2 and 3
then cosine of Z production angle to be smaller than 0.91, to cut out very forward events characteristic of BG processes,
Then I calculate efficieny by counting events in signal region of 123 and 135 geV
Recoil mass is calculated taking into account beam x angle of 14 mrad.
Last week I received a comment from Kurata san about why I have this lower limicoplanarity
Ithe reason is that for some Bgprocesses, teher is a peak at very small cop values, although for the dominant 2f_Z_l, peak is at aroudn pai.
However I have doubts as to whether this cut is really necessary, when I already have the other cuts before it.
experimented with the outcome s when I take away the lower limit, and when I take away the cop cut altogether.
I also experimented with widening the inv maass cut window. 3

4. Data Selection Performance
Example of cut table for 250 GeV, Zμμ
Major residual BG: f_zz_sl , 4f_zz_l, 2f_z_l
ECM=250GeV Zμμ
ECM=250GeV Zee
Here is the data selection I applied, with the numbers from last week Friday.
First I select muon pairs based on ratio of deposit in calorimeter wrt to total energy, opposite charge, and track quality.
Than I pick out the best muon pair candidates base on how close their inv mass is to real Z mass
Next , comes final selection for recoil mass analysis.
Here, in the order that these arei implemented
I require inv mass to be between 86 and 95 GeV,
Then recoil mass to be between 115 and 140 GeV
Then PT of dilepton system to be between 10 GeV and 70 geV
Then acoplanarity to be between 0.2 and 3
then cosine of Z production angle to be smaller than 0.91, to cut out very forward events characteristic of BG processes,
Then I calculate efficieny by counting events in signal region of 123 and 135 geV
Recoil mass is calculated taking into account beam x angle of 14 mrad.
Last week I received a comment from Kurata san about why I have this lower limicoplanarity
Ithe reason is that for some Bgprocesses, teher is a peak at very small cop values, although for the dominant 2f_Z_l, peak is at aroudn pai.
However I have doubts as to whether this cut is really necessary, when I already have the other cuts before it.
experimented with the outcome s when I take away the lower limit, and when I take away the cop cut altogether.
I also experimented with widening the inv maass cut window.

5. Reconstructed Recoil Mass Spectrum Sig + BG BG Signal 250 GeV:
Zμμ channel Pol: (- 0.8, +0.3)
Sig + BG BG
Signal
500 GeV
350 GeV
Now here is the plot for the fitted recoil mass including all signal and BG samples listed earlier, and after implementing all the cuts.
Ifor fitting, I used GPET for signal and 3rd order polynomial for BG.
The fitted recoil mass came out to be GeV, and the error of this was about 70 MeV.
at present stage this fitted mass precision has much meaning , until AFTER I evaluate the validity of my analysis thouroughly using pseudo experiments, by generating Toy MC events.

6. Reconstructed Recoil Mass Spectrum Sig + BG BG Signal 250 GeV:
Zee channel Pol: (- 0.8, +0.3)
Sig + BG BG
Signal
500 GeV
350 GeV
Now here is the plot for the fitted recoil mass including all signal and BG samples listed earlier, and after implementing all the cuts.
Ifor fitting, I used GPET for signal and 3rd order polynomial for BG.
The fitted recoil mass came out to be GeV, and the error of this was about 70 MeV.
at present stage this fitted mass precision has much meaning , until AFTER I evaluate the validity of my analysis thouroughly using pseudo experiments, by generating Toy MC events.

7. Statistical Uncertainties at ECM=250GeV, left pol, assuming 250 fb-1 ,
ZH xsec relative error
at ECM=250GeV, left pol, assuming 250 fb-1 ,
we can achieve
xsec : 2.5%
ΔMH : 37 MeV
ΔMH (MeV)
assumed Lumi:
250
(for each beam pol)
Precision clearly worse at higher energies
(at 350 GeV: ΔσZH 1.3 x , ΔMH 2.7 x , at 500 GeV: ΔσZH 2.1 x, ΔMH 14 x)
Right pol: BG is suppressed, but precision 〜10% worse due to signal xsec smaller (2/3 x)
Then I added all other BG and sig processes that hhave been generated as samples, even though I presumed them to be not important for Zmumu analysis at 250 GeV.
I wanted to check I didn’t msis anything.
I almost didn’t miss anything, except for WW_sl,
Some of the quarks may have decayed secondarily into muons, and I left it out.
There are about 130 events left after all cuts.
While other were mostly 0.
The results were only a little different after including this.
ZH xsec: include contribution from invisible Higgs decay
T. Junping,
A. Ishikawa,

8. After full H20 run: gZZH: 0.4%, ΔMH : 14 MeV
Scaled to H20 Scenario
ZH xsec relative error
ΔMH (MeV)
Then I added all other BG and sig processes that hhave been generated as samples, even though I presumed them to be not important for Zmumu analysis at 250 GeV.
I wanted to check I didn’t msis anything.
I almost didn’t miss anything, except for WW_sl,
Some of the quarks may have decayed secondarily into muons, and I left it out.
There are about 130 events left after all cuts.
While other were mostly 0.
The results were only a little different after including this.
After full H20 run: gZZH: 0.4%, ΔMH : 14 MeV

9. Model Independence Taking Into Account Exotic Decay Modes
final efficiency in last row
(stat uncertainty = 0.16%)
no visible bias beyond 1 sigma
observe deviation from avg efficiency
Known modes
unknown modes : Bx 10%
any exotic decay modes should resemble these wide kinematic range of SM modes
assign 10% of “unknown mode” BR
(2) fluctuate known SM modes by largest BR uncertainty predicted from HL-LHC (Snowmass report arXiv: ) and ILC for H cc and gg (“ILC Higgs White Paper”, arXiv: )
Here is the data selection I applied, with the numbers from last week Friday.
First I select muon pairs based on ratio of deposit in calorimeter wrt to total energy, opposite charge, and track quality.
Than I pick out the best muon pair candidates base on how close their inv mass is to real Z mass
Next , comes final selection for recoil mass analysis.
Here, in the order that these arei implemented
I require inv mass to be between 86 and 95 GeV,
Then recoil mass to be between 115 and 140 GeV
Then PT of dilepton system to be between 10 GeV and 70 geV
Then acoplanarity to be between 0.2 and 3
then cosine of Z production angle to be smaller than 0.91, to cut out very forward events characteristic of BG processes,
Then I calculate efficieny by counting events in signal region of 123 and 135 geV
Recoil mass is calculated taking into account beam x angle of 14 mrad.
Last week I received a comment from Kurata san about why I have this lower limicoplanarity
Ithe reason is that for some Bgprocesses, teher is a peak at very small cop values, although for the dominant 2f_Z_l, peak is at aroudn pai.
However I have doubts as to whether this cut is really necessary, when I already have the other cuts before it.
experimented with the outcome s when I take away the lower limit, and when I take away the cop cut altogether.
I also experimented with widening the inv maass cut window.
relative bias on σZH
< % for Zmm
< 0.19% for Zee
Bias get smaller for higher ECM
conclusion: current bias is more than 5 times smaller than even the best statistical uncertainty expected from full H20 run

10. Summary
Completed full simulation study which demonstrates measurement precision of the e+e-  ZH cross section (σZH) and the Higgs boson mass (MH) at the ILC
clear comparison established between three ECMs = 250, 350, and 500 GeV, and two beam polarizations (Pe-,Pe+) =(−80%, +30%) and (+80%, -30%)
 results may contribute to further optimization of ILC run scenario
Assuming Lumi = 250 fb-1 for left pol at ECM = 250 GeV : σZH = 2.5%, ΔMH=37 MeV
scaling all results to H20 run: ΔgHZZ/gHZZ = 0.4%, ΔMH = 14 MeV
model independence demonstrated to the sub-% level
data selection methods designed to minimize bias on σZH to more than 5 times smaller than even the smallest σZH statistical uncertainties
Paper submitted to P.R.D. : arxiv: undergoing review
Reference paper on model independence: arxiv:
Next steps: Study systematic uncertainties due to beam-beam effects
Here is the data selection I applied, with the numbers from last week Friday.
First I select muon pairs based on ratio of deposit in calorimeter wrt to total energy, opposite charge, and track quality.
Than I pick out the best muon pair candidates base on how close their inv mass is to real Z mass
Next , comes final selection for recoil mass analysis.
Here, in the order that these arei implemented
I require inv mass to be between 86 and 95 GeV,
Then recoil mass to be between 115 and 140 GeV
Then PT of dilepton system to be between 10 GeV and 70 geV
Then acoplanarity to be between 0.2 and 3
then cosine of Z production angle to be smaller than 0.91, to cut out very forward events characteristic of BG processes,
Then I calculate efficieny by counting events in signal region of 123 and 135 geV
Recoil mass is calculated taking into account beam x angle of 14 mrad.
Last week I received a comment from Kurata san about why I have this lower limicoplanarity
Ithe reason is that for some Bgprocesses, teher is a peak at very small cop values, although for the dominant 2f_Z_l, peak is at aroudn pai.
However I have doubts as to whether this cut is really necessary, when I already have the other cuts before it.
experimented with the outcome s when I take away the lower limit, and when I take away the cop cut altogether.
I also experimented with widening the inv maass cut window.

11. BACKUP
Here is the data selection I applied, with the numbers from last week Friday.
First I select muon pairs based on ratio of deposit in calorimeter wrt to total energy, opposite charge, and track quality.
Than I pick out the best muon pair candidates base on how close their inv mass is to real Z mass
Next , comes final selection for recoil mass analysis.
Here, in the order that these arei implemented
I require inv mass to be between 86 and 95 GeV,
Then recoil mass to be between 115 and 140 GeV
Then PT of dilepton system to be between 10 GeV and 70 geV
Then acoplanarity to be between 0.2 and 3
then cosine of Z production angle to be smaller than 0.91, to cut out very forward events characteristic of BG processes,
Then I calculate efficieny by counting events in signal region of 123 and 135 geV
Recoil mass is calculated taking into account beam x angle of 14 mrad.
Last week I received a comment from Kurata san about why I have this lower limicoplanarity
Ithe reason is that for some Bgprocesses, teher is a peak at very small cop values, although for the dominant 2f_Z_l, peak is at aroudn pai.
However I have doubts as to whether this cut is really necessary, when I already have the other cuts before it.
experimented with the outcome s when I take away the lower limit, and when I take away the cop cut altogether.
I also experimented with widening the inv maass cut window.

12. Comparison between ECMs and polarizations
assume Lumi nearly scales with ECM
assumed Lumi:
250
(for each beam pol)
Precision clearly worse at higher energies (esp for MH) ΔσZH is not so bad at ECM=350 GeV
Right pol: BG is suppressed, but precision 〜10% worse due to signal xsec smaller x 1.5
scaled to H20 scenario
Then I added all other BG and sig processes that hhave been generated as samples, even though I presumed them to be not important for Zmumu analysis at 250 GeV.
I wanted to check I didn’t msis anything.
I almost didn’t miss anything, except for WW_sl,
Some of the quarks may have decayed secondarily into muons, and I left it out.
There are about 130 events left after all cuts.
While other were mostly 0.
The results were only a little different after including this.
After full H20 run: gZZH: 0.4%, ΔMH : 14 MeV

13. Also tried MH measurement using Hbb mode
B likelihood: Red: 2f BG Blue : signal
MH measurement does not need to be model-independent,
If use Hbb mode, the only major residual BG is 4f_zz_sl
Zmm 250GeV
250GeV
ΔmH
[MeV]
ΔmH [MeV]
(H-->bb)
Zmm
left 39 38
right 43
Zee
121
106
149
124
350 GeV
103
114
120
125
450
382
502
464
500 GeV
592
530
660
566
1160
1030
1190
1130
B likeliness
improvement (10-15%) for Zee at all ECM
and both Zmm and Zee ECM=500 GeV
(2f BG is more serious for Zee and higher ECM)
Here is the data selection I applied, with the numbers from last week Friday.
First I select muon pairs based on ratio of deposit in calorimeter wrt to total energy, opposite charge, and track quality.
Than I pick out the best muon pair candidates base on how close their inv mass is to real Z mass
Next , comes final selection for recoil mass analysis.
Here, in the order that these arei implemented
I require inv mass to be between 86 and 95 GeV,
Then recoil mass to be between 115 and 140 GeV
Then PT of dilepton system to be between 10 GeV and 70 geV
Then acoplanarity to be between 0.2 and 3
then cosine of Z production angle to be smaller than 0.91, to cut out very forward events characteristic of BG processes,
Then I calculate efficieny by counting events in signal region of 123 and 135 geV
Recoil mass is calculated taking into account beam x angle of 14 mrad.
Last week I received a comment from Kurata san about why I have this lower limicoplanarity
Ithe reason is that for some Bgprocesses, teher is a peak at very small cop values, although for the dominant 2f_Z_l, peak is at aroudn pai.
However I have doubts as to whether this cut is really necessary, when I already have the other cuts before it.
experimented with the outcome s when I take away the lower limit, and when I take away the cop cut altogether.
I also experimented with widening the inv maass cut window.
2f BG is greatly reduced, but mass precisions limited by signal statistics
merit of this method is not totally clear yet (?)