Kun's weekly group discussion

Friday 15 May 2026, 14:30 → 15:50 Asia/Shanghai

Lab Room #S244 (Tsung-Dao Lee Institute)

Kun Liu (TDLI / SJTU)

会议链接/Zoom Link: 
https://cern.zoom.us/j/66708976616?pwd=wIvRGTja7ooamUftTYtz7brdc6ynl9.1

会议号/Meeting ID: 667 0897 6616

会议密码/Passcode: 949845

Tracker Team: Hardware, Reconstruction, Simulation and Analysis

1 Tracker R&D

Hardware: DS Tracker Module based on Baby_2 and AC-LGAD sensors

Speakers: Dian YU (Tsung-Dao Lee Institute), Lin Yang (TDLI), ML Huang (TDLI), xiang li (tdli)

Comparison and Discussion of Momentum Resolution Extraction Methods: Impact Analysis of STD and FWHM Definitions

• Sensitivity of STD to extreme values: The conventional Standard Deviation (STD) includes all data points in the distribution during statistical calculation. Due to the amplification effect of its squared terms, the calculated STD is highly sensitive to outliers and non-Gaussian tailing effects present in the error distribution.

• Faithful representation of measured values by FWHM: The definition of Full Width at Half Maximum (FWHM) inherently focuses on the core peak distribution near the measured momentum value. This approach effectively isolates interference from extreme values and tails, faithfully reflecting the true resolution of the main signal.

2 Track Reconstruction

Dark SHINE and CEPC

Speakers: Dian YU (Tsung-Dao Lee Institute), Lin Yang (TDLI)

3 Simulation and Analysis

Dark SHINE, CEPC and ITk Test Beam

Speaker: Dian YU (Tsung-Dao Lee Institute)

Test-Beam Analysis

current status of interposed modules

Prospective goal before RAL engagement

• finish all analysis components where the progress is within my control：
  • R3_20USEM30000292
  • R3_20USEM30000054
  • SS_20USBMS0000406
  • SS_20USBMS0000381

 

 

ATLAS: Analysis, Performance and Physics

4 Egamma performance

Run2, Run3

Speakers: Lin Yang (TDLI), xiang li (tdli)

ATLAS_Energy_Calibration.pdf

Photon ID

Re-Weighted Electron v.s. Photon

absolute yield

normalized distribution

Normalizing Flow Framework

bottlenecks

1. restricted and crowded LXPLUS computing resources
2. limited flexibility of the inherited Photon ID measurement framework

implemented improvements

1. Repeated full data loading per trial -> load once
2. Multiple data copies during split -> repeated usage
3. Model checkpoint output is excessive -> only necessary
4. non-intermediate results -> update after each trial

ID measurement implementation strategy

NF train:
 MC Zee electrons  ->  MC Single Photon target
 produces model + standardization + diagnostics

NF apply:
 MC Zee electrons  -> corrected MC photon-candidate h5
 Data Zee electrons -> corrected data photon-candidate h5

One-time photon reference conversion:
 original MC Single Photon ROOT/h5 -> canonical photon-reference h5

ID measurement:
 canonical MC Single Photon h5
 canonical corrected MC Zee h5
 canonical corrected Data Zee h5
 -> efficiencies, closure ratios, scale factors

 

 

5 Higgs analysis

Speakers: Chonghao Wu (SJTU, TDLI), Dian YU (Tsung-Dao Lee Institute), Lennart Baumgartel, Lin Yang (TDLI), xiang li (tdli)

515.pdf

6 General issues

Feel free to continue the discussion and share your opinions with the group.

Speakers: Chonghao Wu (SJTU, TDLI), Dian YU (Tsung-Dao Lee Institute), Jialin Li, Lin Yang (TDLI), ML Huang (TDLI), xiang li (tdli)