1 of 25

Shower energy systematics in hdgeant/4, revision 3.0

Richard Jones, University of Connecticut

March 9, 2019

updated March 23, 2019

2 of 25

Purpose

  • Claims are made that 2𝛾 mass spectra in the calorimeters are different in hdgeant4 simulations than they are with hdgeant.
  • May be due to inconsistent energy scales in the two simulations.
  • May originate in either fcal, bcal or both

2

  1. Compare reconstructed shower energy with generated.
  2. Same conditions (control.in) in hdgeant and hdgeant4 sims.
  3. Single-photon simulation -- gets rid of position resolution, track matching efficiency, other possible reasons for bias in a 2𝛾 mass spectrum.
  4. Separate BCal and FCal showers, check for systematics with polar angle.

Plan:

3 of 25

Part 1: FCal showers

  • uniformly generated in energy [0, 8] GeV
  • uniformly generated in polar angle [0, 12°], azimuthal angle [0, 360°]
  • sliced into bins in polar angle generated
  • plot and fit reconstructed shower energy vs generated.
    • standard control.in physics settings
    • all photons shot from along the axis of the target
    • material interactions (ie. conversions) allowed to happen
    • run number 51500 (with magnetic field, no DIRC radiators)
    • 1M photons generated, 92% are reconstructed as single shower
    • only first (most energetic) reconstructed shower is included in plots
    • Gaussian fit to spectrum in each bin in E𝛾 generated, plot Gaussian mean from fit

3

4 of 25

Fcal: 3° - 5°, cubic fits

4

5 of 25

Fcal: 5° - 7°, cubic fits

5

6 of 25

Fcal: 7° - 9°, cubic fits

6

7 of 25

Fcal: 9° - 11°, cubic fits

7

8 of 25

Fcal: 3° - 5°, cluster time comparison (relative to RF)

8

9 of 25

Part 2: BCal showers

  • uniformly generated in energy [0, 8] GeV
  • uniformly generated in polar angle [20°, 90°], azimuthal angle [0, 360°]
  • sliced into bins in polar angle generated
  • plot and fit reconstructed shower energy vs generated.
    • standard control.in physics settings
    • all photons shot from along the axis of the target
    • material interactions (ie. conversions) allowed to happen
    • run number 51500 (with magnetic field, no DIRC radiators)
    • 1M photons generated, ~90% are reconstructed as single shower
    • only first reconstructed shower is included in plots
    • Gaussian fit to spectrum in each bin in E𝛾 generated, plot Gaussian mean from fit

9

10 of 25

BCal: 20°-30°, cubic fits

10

11 of 25

BCal: 30°-40°, cubic fits

11

12 of 25

BCal: 40°-50°, cubic fits

12

13 of 25

BCal: 50°-60°, cubic fits

13

14 of 25

BCal: 60°-70°, cubic fits

14

15 of 25

BCal: 70°-80°, cubic fits

15

16 of 25

BCal: 80°-90°, cubic fits

16

17 of 25

Conclusions

  1. significant systematic bias is seen as a function of polar angle
  2. small but significant y-offset is visible in both simulations
  3. agreement between hdgeant and hdgeant4 is quite good
  4. nonlinear correction is needed
  5. nonlinearity changes sign with angle
    1. positive at small polar angle (low-E showers leak energy out the front surface)
    2. negative at close to 90o (high-E showers leak energy out the back of the modules)
  6. stated errors on fit parameters are too small, peaks are non-gaussian.

​

17

18 of 25

Part 3: study of BCal timing

  • There is a divergence between hdgeant and hdgeant4 in the handling of shower timing in the BCal.
    • hdgeant uses energy-weighted average shower times in the bcal
    • hdgeant4 uses the earliest hit per sector for times in the bcal
  • What is the best algorithm?
  • Look at both and see if there is any visible difference.
  • Largest differences expected for most forward showers, if any.
  • Currently “earliest hit” does not take propagation delay down the length of the modules into account.

18

19 of 25

BCal shower timing: earliest hit time

19

20 of 25

BCal shower timing: earliest hit time

20

21 of 25

BCal shower timing: energy-weighted time

21

22 of 25

BCal shower timing: energy-weighted time

22

23 of 25

Conclusions on bcal timing

  • no significant differences found between earliest time and energy-weighted
  • important weakness in current bcal time algorithm (both sims)
    • light propagation down module mixes z and t
    • we do not measure <t> or tmin but <t - z/v> or (t - z/v)min and <t + z/v> or (t + z/v)min
    • deprecated tags bcalSiPMUpHit and bcalSiPMDownHit exist in hddm_s
  • proposal: rehabilitate these end-tags and fill them in parallel with bcalTruthCell
  • differences in bcal timing seen in rho simulations remain unexplained

23

24 of 25

Appendix: new bcalSiPMUpHits/bcalSiPMDownHits

24

25 of 25

bcalSiPMUpHits/bcalSiPMDownHits timing

25