The 70th General Meeting of ILC Physics Subgroup

Asia/Tokyo
Keisuke Fujii (IPNS, KEK)
Description

Date/Time: 10:30~17:30 JST, March 9th, 2020

Place: Online Web Meeting

Official Language: English

 

Zoom Connection:

 https://cern.zoom.us/j/64597061826?pwd=N2lXMmQwWUxwVGQ1aTc0WnV4OVl0Zz09

TV meeting(via CERN Zoom):

1.  188.184.85.92 or 188.184.89.188

2.  meeting ID: 966 4733 5143

 

--------------------------------------------------------------------------------
Summary of The 70th General Meeting of the ILC Physics Subgroup
March 9, 2021 (Wednesday) held fully virtually because of COVID19
--------------------------------------------------------------------------------
0) Attendants:
 S.Tairafune (Tohoku)
 Y.Aoki      (KEK/Sokendai)
 K.Yumino    (KEK/Sokendai)
 K.Fujii     (KEK)
 D.Jeans     (KEK)
 A.Miyamoto  (KEK)
 T.Mizuno    (KEK/Sokendai)
 J.Nakajima  (KEK/Sokendai)
 J.Tian      (Tokyo)
 M.Kurata    (Tokyo)
 K.Hidaka    (Tokyo gakugei)
 T.Suehara   (Kyushu)

0) Opening comments (K.Fujii)


1) e+e- -> Z gamma: JES study (Takahiro Mizuno)
- JES study has never been presented at previous LCWS series.
  Photon selection:
     M2j < 125 GeV && Evis > 200 GeV
     cos(theta_jet.gamma)<0.95 for both jets
     -> correct photon fraction = 99.5%
  Kinematical reconstruction of JE:
     using theta, phi, Mj for 2 jets (8 variables) as inputs 
     to determine p_j, p_gamma, p_ISR (4 variables)
  JES calibration:
     Fit the kinematically reconstructed delta E distributions with a double gaussian 
     and decide the calibration factor to 10^-4 accuracy.


2) e+e- -> tau+ tau- (Keita Yumino)
Status:
- Tau polarization study using cone method and MC truth. 
  -> found reasonable agreement between MC truth polarimeter value and the one 
     from the cone method for both pi nu and rho nu decays.
Next Step:
- Repeat this at various levels of cheating.
- Think about new physics interactions that affect this process.


3) RHN (Jurina Nakajima)
Status:
- Signal channel
    e+e- -> NN -> eW eW (same sign leptons, BR~36%)
- Set 4 benchmark points consistent with the current limits and generated 
  RHN pair production events with Whizard.
- Study event selection 
    M_W problem solved by inserting a W-resonance as an intermediate state
    to pass to Whizard.
Next Step:
- Study potential backgrounds.


              -- Lunch Break --


4) e+e- -> gamma H @ 250GeV (Yumi Aoki)
Status:
    sigma(Hgamma) = 0.2fb in SM.
- Results for new SM sample (using proper 1-loop amplitudes)
- gamma H -> gamma (bb):
    Major Bkg: gamma Z -> gamma (bb)
    -> significance : 0.29 for 0.9ab^-1 (eLpR)
- gamma H -> gamma (WW*) -> gamma + 1 lepton + 2 jets:
    Major Bkg: WW (gamma)  
    -> significance : 0.26 for 0.9ab^-1 (eLpR)
  Checked Bkg statistics.
Next Step:
- Do eRpL case.
- Understand the role of this measurement in the BSM model tests and 
  in the global EFT analysis.


5) b mass study with e+e- -> bb (Seidai Tairafune)
Status:
- Motivation and method
  m_b(MSbar) runs differently depending on what is in the loop.
  Use m_b dependence of R_3 (3-jet fraction) to measure m_b.
- MC set
    new 250 GeV sample
- Event selection
  Bkg: radiative return to Z, WW, ZZ, ZH
    - veto energetic gamma (E_gamma > 50 GeV)
    - do forced 2 jet clustering and require Thrust > 0.85
- Plot R^{bl}_3(y)
    -> found R^{bl}_3(y_c) > 1, meaning that b quarks radiate a gluon more frequently
           than light quarks unlike naive expectation.
- Estimated m_b error
    for m_b = 2.75 GeV, delta R_3^{bl} = 0.996, 
        -> delta m_b ~ 1 GeV @ 250 GeV
           -> delta m_b ~ 0.1 GeV @ Giga-Z
    C: You should look at y_c < 0.001, where you you expect m_b-dependence.
       High y_c region has no sensitivity to m_b.


6) Deep Jet Clustering (Masakazu Kurata)
Status:
- Tool: Keras & Tensorflow backend
    Convolutional NN applied to 9 64x64-pixel images
         (E, Q, D0sig. Z0sig, Ecal, Hcal, direction vector)
    3 units for each of Encoder/Decoder 
  Application to ZHH->(qq)(bb)(bb) (q=u,d,s):
    For the moment ignore color singlet.
             (2jet clustering by Durham for each color singlet)
             cross entropy L = (1/n) sum t_i log y_j
        -> pseudo-labeling (reassignment of color to minimize L)
         Perfect Durham clustering is not always best for CNN.
        -> data cleansing 
        Use 230000 events for training (90% for training, 10% for validation)
    and compare confusion matrices for Durham and CNN
        -> 1st 3 jets start to improve, last 3 bit worse (correlation)
  New: Multi Center
    to recover splitting jets.
        -> now training so no result yet.
    C: Maybe it is difficult for AI to invent QCD.
Next Step:
- Try to improve further.
  C: look at performance for clean 6-jet events without hard gluon emission.


7) Discussions
7-1) JPS Meeting
    - e+e- -> Z gamma (A_LR): Takahiro Mizuno
    - e+e- -> tau+ tau-: Keita Yumino
    - e+e- -> N N: Jurina Nakajima
    - e+e- -> b b: Seidai Tairafune

7-2) LCWS2021
    - e+e- -> Z gamma (A_LR): Takahiro Mizuno
    - e+e- -> H gamma: Yumi Aoki
    - e+e- -> tau+ tau-: Keita Yumino
    - e+e- -> N N: Jurina Tairafune
    - e+e- -> b b: Seidai Tairafune
    - Kinematic fitting : Yu Kato
    - NN vertex finding : Kiichi Goto

We should discuss this next week at the regular weekly meeting with more participants.

----------------------------------------------------------
The slides shown at the meeting is available from
https://agenda.linearcollider.org/event/9134/
See them for details.
---------------------------------------------------------

*) Meeting Schedule:
  Next general meeting: 2021/05/12 (Wed) 10:30)


Working group web page:
http://www-jlc.kek.jp/jlc/en/subg/physics
 

There are minutes attached to this event. Show them.