1. BSM search using Higgs to invisible decay at the ILC
Yu Kato,
Jacqueline Yan(KEK), Junping Tian, Satoru Yamashita
The University of Tokyo
Hello everyone, this is Yu Kato, from the University of tokyo, and I will be talking about BSM Search using Higgs to Invisible Decay at the ILC.
The 50th general Feb. 4, 2017

2. BSM search using Higgs to invisible decay
Status
Reproduced all root file without using kt algorithm, and reanalyzed
Evaluated UL based on Yamashita san’s advice
Made presentation in LCWS2016
Pause analysis for TPC experiment
Plans
Come back around June
Improve sensitivity through further optimization of analysis methods and jet energy resolution
Analysis at E cm = 350, 500 GeV, compare between different scenarios > contribute to optimization of ILC run scenario
2017/2/7 50th
BSM search using Higgs to invisible decay

3. Setting & Flow of Analysis
Generator: WHIZARD 1.95
Samples: DBD sample + Dirac sample ( e + e − →qqH, H→ ZZ ∗ →4𝜈)
Detector: ILD full simulation
Ecm = 250 GeV, ∫𝐿𝑑𝑡 = 250 fb-1 , 𝑃 𝑒 − , 𝑃 𝑒 + = −0.8,+0.3 , (+0.8,−0.3)
Flow of analysis
Isolated lepton tagging → veto
Jet clustering
forced 2-jet reconstruction using LCFIPlus
Event selection (next page)
assume BR(H→invisible)=10%
Fit & Toy Monte Carlo to set upper limit
Next, I shall explain the setup and flow of analysis.
First the set up for simulation. The samples are generated using WHIZARD1.95. 
The study is based on full ILD detector simulation, conducted at center of mass energy 250 GeV, and assumes 250 fb-1 for each polarization.
The beam parameters in the ILD is taken into account. 
The flow of analysis is shown below.
First, I conduct isolated lepton tagging in order to veto leptons from PFOs.
At the stage of jet clustering, LCFIPlus is used for 2-jet clustering.
After that, at the stage of event selection, I designed cuts assuming a BR(H→invisible) of 10%.
Then finally I fit the spectrum of signal and bkg, and used Toy MC to set the UL of the BR.
2017/2/7 50th
BSM search using Higgs to invisible decay

4. BSM search using Higgs to invisible decay
Event Selection
isolated lepton veto
loose restriction of transverse di-jet momentum, di-jet invariant mass, and recoil mass from di-jet
number of PFOs and charged tracks: Npfo, Ntrack
di-jet (Z) pt: PtZ
di-jet mass: MZ
di-jet polar angle: θZ
recoil mass: Mrecoil
multi-variate analysis: Boosted Decision Tree(BDT) method
Here are the details of the event selection.
I will not go through everything, but in general, I apply cuts on the number of PFOs and charged tracks,
also on the Pt, mass and angle of the di-jet system.
Last of all, I apply a MVA analysis.
2017/2/7 50th
BSM search using Higgs to invisible decay

5. Recoil Mass Plots [Ecm = 250 GeV, 250 fb-1,BR(H->inv.)=10%]
Preliminary
Recoil Mass Plots [Ecm = 250 GeV, 250 fb-1,BR(H->inv.)=10%]
Left
Right
significance: 15.5
efficiency: 63.5%
significance: 19.7
efficiency: 65.9%
This is the recoil mass plot after the event selection.
The left plot is the left polarization case, and the right plot is the right polarization case.
The table on the bottom right is the cut table.
As you can see, W bkg is much reduced for the right polarization, and the peak is obviously sharper.
The significance is 15.5 for left, and 19.7 for right　and signal efficiency is …
No.
Cut 1
Isolated lepton veto 5
80 < Mz < 100 2
Loose Cut (Ptz,Mz,Mrecoil) 6
|cosθZ|<0.9 3
Npfo>15 & Ntrack>6 & Ntrackji>1 7
100 < Mrecoil < 160 4
20 < Ptz < 80 8
BDT cut
MVA input variables mz
cosj1
cosz
cosj2
TMVA v-4.2.0
2017/2/7 50th
BSM search using Higgs to invisible decay

6. Fit Signal & Background
Preliminary
Fit Signal & Background
signal: GPET (Gaussian + Exponential) bkg : 5th order polynomial
Left
These are the results for the fitting of the signal and bkg on the previous page.
I used GPET for the signal and 5th order polynomial for the bkg.
Right
2017/2/7 50th
BSM search using Higgs to invisible decay

7. Toy Monte Carlo to Set Upper Limit
Preliminary
Toy Monte Carlo to Set Upper Limit
Left polarization
Right polarization
UL=1.03% (95%CL)
UL=0.69% (95%CL)
Then, I generated Toy Mc events using the fitted results on the previous page, and used that to determine the UL.
I set the signal yield as 0, bkg shape fixed and the each yields floated, then generated pseudo experiments.
The BR is derived from the number of signal of each pseudo experiment.
Here, SM Higgs to invisible decay is not included in the # of signals.
Assuming luminosity of 250 fb^-1 for each polarization, the result for the 95%CL is 1.03% for left and 0.69% for right.
bkg : shape fixed, yields floated
∫𝐿𝑑𝑡 = 250 fb-1
bkg yields floated → times pseudo experiment
not include SM Higgs to invisible decay
2017/2/7 50th
BSM search using Higgs to invisible decay

8. Toy Monte Carlo to Set Upper Limit
Preliminary
Toy Monte Carlo to Set Upper Limit
Left polarization
Right polarization
250 fb-1
250 fb-1
250 fb-1
UL=1.03% (95%CL)
UL=0.69% (95%CL)
H20 (800 fb-1 )
800 fb-1
800 fb-1
Here are the results assuming the H20 scenario, where we have 800 fb^-1 for each pol., using the same methods.
UL=0.58% (95%CL)
UL=0.39% (95%CL)
2017/2/7 50th
BSM search using Higgs to invisible decay

9. How to set UL [Statistical method]
Template
Assume BR(H→invisible)=[1,2,…,10]% -> Event selection
Get # of events (NS+B) in window range (Mrecoil∈[120,140] GeV)
Generate Poisson distribution of NS+B -> Get 95% CL limit (NUL)
Repeat for each BR(H→invisible)=[1,2,…,10]% -> Get calibration line between NUL and UL
Toy MC
Fit template bkg -> Generate pseudo experiment by fluctuated bkg function
Translate NS+B into UL of BR(H→invisible) using calibration line
Repeat times -> Obtain UL distribution
Up to here we have used the conventional method, but from now on we will reanalyze the UL statistical fluctuation by the following method.
ad lib UL
NS+B
NS+B
NUL
2017/2/7 50th
BSM search using Higgs to invisible decay

10. Statistical fluctuation of Upper Limit
Preliminary
Statistical fluctuation of Upper Limit
Left polarization
Right polarization
250 fb-1
250 fb-1
250 fb-1
UL=0.95±0.45%
UL=0.53±0.31%
H20 (800 fb-1 )
Here are the results.
ad lib
800 fb-1
800 fb-1
UL=0.55±0.25%
UL=0.25±0.17%
2017/2/7 50th
BSM search using Higgs to invisible decay

11. BSM search using Higgs to invisible decay
Summary
Use measurement of BR(H→𝑖𝑛𝑣.) as a means for indirect BSM search
Motivation
set upper limit (UL) on BR(H→𝑖𝑛𝑣.)
develop analysis method to achieve high sensitivity
compare between alternative polarization
study is based on full ILD detector simulation, at E cm = 250 GeV, assuming 250 fb −1
Status
optimized data selection methods
set UL using toy MC for both left and right scenario
Left polarization
Right polarization
significance [250 fb −1 ]
15.5
19.7
Here is the summary.
We use measurement of BR(H→𝑖𝑛𝑣.) as a means for indirect BSM search.
Regarding motivation, we want to set upper limit (UL) on BR(H→𝑖𝑛𝑣.).
So as to set UL, we have to develop analysis method to achieve high sensitivity.
And we also compare between alternative polarization, left and right mode.
This study is based on full ILD detector simulation, at Ecm = 250 GeV, assuming 250 fb^−1.
As for our status, we optimized data selection methods and significance is 15.5 for left pol. and 19.7 for rightl.
After that, the UL we obtained ,assuming 250 fb^-1 for each pol. is 1.03% for left and 0.69% for right.
And assuming H20 scenario, the UL is 0.58% for left and 0.39% for right.
These results are consistent within 1σ.
UL of BR [%] (95%CL)
Left polarization
Right polarization
Statistical
Left
Right
250 fb −1
1.03
0.69
0.95±0.45
0.53±0.31
H20 scenario (800 fb −1 )
0.58
0.39
0.55±0.25
0.25±0.17
Previous study(250 fb −1 )
0.95 -
2017/2/7 50th
BSM search using Higgs to invisible decay

12. BSM search using Higgs to invisible decay
Back Up
2017/2/7 50th
BSM search using Higgs to invisible decay

13. BSM search using Higgs to invisible decay
Motivation
In SM, Higgs decays invisibly through H→Z Z ∗ →4𝜈 (BR(H→𝑖𝑛𝑣.)~0.1%)
If BR(H→𝑖𝑛𝑣.) exceeds SM prediction , It signifies new physics beyond SM (BSM)
In this study, we estimate upper limit of BR(H→𝑖𝑛𝑣.)
Compare between left & right polarization at the ILC q
BSM X
invisible
Dark Matter…
SUSY…
visible
𝐵𝑅 H→XX ~???% q SM Z ν
𝐵𝑅 H→ ZZ ∗ →4𝜈 ~0.1%
invisible
visible
First, I shall explain the motivation of this study. 
In the SM, this is the only process in which the Higgs decays invisibly. The BR is about 0.1%.
On the other hand, if the Higgs decay in non-SM process, because the invisible decay products cannot be seen in the detector, the BR of H to inv. would exceed the SM prediction.
In other words, this is an indication of new physics beyond the standard model.
I aim to establish the upper limit of the BR for invisible Higgs decay and to compare results between different polarizations.
𝑁 𝑆 ∝ 𝜎 ZH ・𝐵𝑅 H→𝑖𝑛𝑣𝑖𝑠𝑖𝑏𝑙𝑒
A. Ishikawa (Tohoku Univ.),
”Search for Invisible Higgs Decays at the ILC”
M. A. Thomson (Univ. of Cambridge), arXiv: (2015)
2017/2/7 50th
BSM search using Higgs to invisible decay

14. BSM search using Higgs to invisible decay
Recoil Mass Method
We can measure Higgs without directly looking at it → model independent
𝑀 𝑟𝑒𝑐 2 = 𝑠 − 𝐸 𝑍 2 − 𝑝 𝑍 2
Higgs-strahlung cross section can be obtained using leptonic decay of Z
𝜎 ZH = 𝑁 𝑆 𝐵𝑅 Z→ 𝑙 + 𝑙 − ε 𝑆 𝐿
ε 𝑆 : signal efficiency, N S : # of signal, 𝐿 : integrated luminosity
J.Yan(KEK) et al. Phys. Rev. D (arXiv: ) (2016)
For BSM search, this study uses hadronic channel 𝑍→𝑞𝑞 and 𝐸 𝑐𝑚 =250 GeV
This study uses the  Higgsstrahlung process. This method allows us to measure the Higgs without directly looking at the Higgs.
This model independent measurement can only be realized at the ILC.
The total ZH cross section, which is crucial for the BR measurement,  is given in this paper by Jacqueline.Yan et al, using the leptonic decay of the Z.
On the other hand, this study will be conducted at center of mass energy 250 GeV, using the hadronic decay of the Z
2017/2/7 50th
BSM search using Higgs to invisible decay

15. Cut table [Ecm=250GeV,250fb-1,Left]
Hadronic Channel Analysis
Cut table [Ecm=250GeV,250fb-1,Left]
2017/2/7 50th
BSM search using Higgs to invisible decay

16. Cut table [Ecm=250GeV,250fb-1,Right]
Hadronic Channel Analysis
Cut table [Ecm=250GeV,250fb-1,Right]
2017/2/7 50th
BSM search using Higgs to invisible decay

17. Toy Monte Carlo to set upper limit
Preliminary
Toy Monte Carlo to set upper limit
Left
Right
UL=0.78% (95%CL)
UL=0.52% (95%CL)
bkg : shape & yields fixed
∫𝐿𝑑𝑡 = 250 fb-1
bkg fixed → times pseudo experiment
not include SM Higgs to invisible decay
2017/2/7 50th
BSM search using Higgs to invisible decay

18. BSM search using Higgs to invisible decay
Previous Study
A. Ishikawa (Tohoku Unv.),
”Search for Invisible Higgs Decays at the ILC”
Smaller upper limit of BR(H→𝑖𝑛𝑣.) →　higher sensitivity to BSM
We aim to improve this result
∫𝐿𝑑𝑡=250,350,500 fb-1 for E cm =250,350,500 GeV
こちらは東北大学の石川さんによる、先行研究のデータです。
ULはご覧のような結果が得られています。
我々はこの結果を向上することを目標にします。
2017/2/7 50th
BSM search using Higgs to invisible decay

19. Signal feature [hadronic channel]q X
invisible
𝐵𝑅 H→XX ~???%
Two-Jet
Invisible particle exist
𝑀 𝑞𝑞 ≈ 𝑀 𝑍 (𝐵𝑅 Z→𝑞𝑞 ~70%)
𝑀 𝑟𝑒𝑐𝑜𝑖𝑙 ≈ 𝑀 𝐻𝑖𝑔𝑔𝑠
Major Backgrounds [hadronic channel]
The major backgrounds have the final states qqll,qqlν,qqνν.
ZZ semileptonic
WW semileptonic
Z 𝜈 𝑒 𝜈 𝑒 , Z→qq
W 𝜈 𝑒 𝜈 𝑒 , W→qq
ννH,H→ZZ,Z→𝑞𝑞
qqH,H→SM decay
2017/2/7 50th
BSM search using Higgs to invisible decay

20. Signal feature [leptonic channel]
𝑙 + X
invisible
𝐵𝑅 H→XX ~???%
Two-leptons
Invisible particles without neutrino
𝑀 𝑙𝑙 ≈ 𝑀 𝑍 (𝐵𝑅 Z→𝑙𝑙 ~3.4%)
𝑀 𝑟𝑒𝑐𝑜𝑖𝑙 ≈ 𝑀 𝐻𝑖𝑔𝑔𝑠
𝑙 −
Major Backgrounds [leptonic channel]
The major backgrounds have the final states with di-lepton & missing energy .
ZZ leptonic
WW leptonic
single Z & 𝜈 𝑒 𝜈 𝑒 , Z→𝑙𝑙
single W & e 𝜈 𝑒 , W→e ν e
ννH,H→ZZ,Z→𝑙𝑙
𝑙𝑙H,H→SM decay
2017/2/7 50th
BSM search using Higgs to invisible decay