Progress on G4 FDIRC Simulation. Doug Roberts University of Maryland

Transcription

1 Progress on G4 FDIRC Simulation Doug Roberts University of Maryland

2 Since SLAC Workshop Spent some time trying to streamline and speed-up the reconstruction technique Needed quicker turnaround on resolution measurements Looked at Side Reflection question Tried different Wedge Geometries Current BaBar wedge Micro-Wedge glued to bottom of BaBar wedge Replace wedge with block (bad, won t talk about) A few weeks ago, Jerry sent me a new design that I ve implemented This is the design that I ll present results for today

3 Jerry s New Design

4 Key New Features Micro-Wedge glued to bottom of existing wedge. Micro-Wedge extends ~mm below from edge of existing wedge Glue joint is.1 thick External Wedge between existing window and Focusing Block

5 Some Ring Images Color-coded by wavelength events in each plot, 4GeV π

6 Simulation Options Looked at three different geometry options 1. Default, as drawn (i.e. with micro-wedge and glue joint between micro-wedge and wedge. No Glue Joint Not really an option, just wanted to see if there was any visible effect due to Fresnel reflections from the glue-quartz interface Preliminary look shows ~1% of photons reflect from this interface for shallow dip angles (45 ~ 55, 1 GeV momentum π), smaller fraction as dip angle increases 3. No Micro-Wedge Everything else is external to the bar box, so would really need to justify making the modification Then simulated 4 GeV momentum π at dip angles of 45, 55, 65, 75, 85 and 9 Was previously looking at 1 GeV π, but MS has significant contribution to resolution at that momentum. More interesting to look at higher p. Recall: K-π Cerenkov angle difference at 4 GeV = 6.5 mrad, implies 3 σ separation requires resolution of. mrad

7 Single Photon Resolution Cerenkov Angle Reconstruction For each geometry, generate a dictionary of single photons Random over all 1 bars in a bar box Random direction Monochromatic (41 nm) Record position at focal plane, time of arrival after leaving quartz bar, and angles, θ x and θ y at the exit of the quartz bar For simulated tracks, for each photon hit generated: Look in dictionary for photons within same pixel as hit Currently using 3mm x 3mm pixels Grab t, θ x, and θ y and calculate θ C based on track s dip and phi angle, and Δt (hit time expected time from dictionary) (8-fold ambiguity from reflections in x, y, and z) (# photons from dictionary) Only look at hit solutions with Δt < 3 ns No simulation of PMT time resolution yet but this should be pretty loose for tubes being considered Including Quantum Efficiency based on H85, plus charge collection efficiency and packing efficiency (dead space)

8 θ X and θ Y Resolutions Because we focus in Y but not in X, we have different (> x4) resolutions in θ X and θ Y Would like to account for this in the reconstruction Central part of the ring is more focused than tails Most of the side reflections contribute to the tails Can also see contribution from multiple scattering and chromatic effect (bottom two plots) : about 3.5 mrad at 4 GeV

9 Red plot includes contribution from MS, E- loss, chromatic effects (~3.5 mrad) Red plots are removing side reflections from the Focusing Block. Expected Photon Resolution Can propagate θ X and θ Y resolutions into calculation of θ C to determine a solution-by-solution expected resolution Then, in θ C reconstruction, weight events by 1/σ (lower-right plot)

10 Summary of Single Photon Resolutions Geometry σ θ X (mrad) σ θ Y (mrad) σ θ C (mrad) Default No Glue No µwedge No Focusing Block Side Reflections Default No Glue No µwedge Gain from µwedge of ~.5 mrad σ θ X gets worse w/o side reflections: Lever-arm effect But, σ θ C gets better because we are primarily keeping photons in central part of the ring Compare σ θ C to BaBar s 9.6 mrad from dimuon events

11 Side Reflections Removing reflections from the side of the Focusing Block simplifies the images and seems to improve the single photon resolution on average But, we would be throwing away photons and would therefore possibly degrade the net θ C resolution Could end up with bar-dependent resolution For bars at the edges of the bar-box, image reflects on to itself and could degrade resolution Note: Still includes reflections from sides of External Wedge

12 Reconstructed θ C vs. Bar Number Side reflections add significant non-gaussian tails There s a noticeable broadening for bars 1 & 1 reflections map onto non-reflected image I haven t removed side reflections from the external wedge, so there s also a broadening w/o FBLOCK side reflections

13 Resolution Results, Dip = 9! σ =.76 mrad σ =.97 mrad Fit C Entries Mean.84 RMS.38 / ndf / 6 Constant 78.3 " 7.8 Mean.84 ".1 Sigma.761 ".45 Mean.87 RMS.399 / ndf / 3 Constant 57.5 " 7.1 Mean.87 ".1 Sigma.977 ".46 N /track Entries Mean.85 RMS 5.56 Mean 1.57 RMS (radians) C Photons per track Fit Mean vs. Bar Number Mean 6.5 RMS 3.45 Mean 6.5 RMS 3.45 # s/event vs. Bar Entries Mean Mean y.87 RMS 3.46 RMS y Mean Mean y 1.57 RMS 3.46 RMS y Fit vs. Bar Number Mean RMS Mean RMS Fit Weight/Event vs. Bar 3 # Entries Mean Mean y 3.454e+5 RMS 3.46 RMS y 8.55e+4 Mean Mean y 1.996e+5 RMS 3.46 RMS y 6.691e Event Fits for θ C On an event-by-event basis, I fit for the value of θ C, can look at mean and RMS of fit results vs. bar number Improved single photon resolution w/o side reflections is lost due to loss of photons - Better to keep Thought: Could we slope the sides so the images don t map onto themselves near the edges? (I think this was in one of Jerry s earlier designs)

14 Resolution vs. Dip Angle With and without side reflections Keeping side reflections is generally better

15 Final Comments Single photon resolution is comparable or better than BaBar DIRC Still some effects not included in the simulation However, θ C resolution doesn t seem to reach the target of. mrad for 3σ K-π separation at 4 GeV Lack of photons? QE, Charge Collection, Dead Space???? <N γ > 13~3, depending on dip angle Or could be non-optimal analysis But remember that goal is to compare geometry options, so even if it s non-optimal, it shouldn t be biased Resolution on θ C doesn t seem to go like 1/ N γ. Problem with weighting or handling ambiguities?

16 Backup Slides

17 Photon Dictionary, Central Pixel, Bar 6 θ Y (radians) θ X (radians)

18 Photon Dictionary, Central Pixel, Bar 1 θ Y (radians) θ X (radians)

19 Photon Dictionary, Edge Pixel, Bar 6 θ Y (radians) θ X (radians)

20 Photon Dictionary, Edge Pixel, Bar 1 θ Y (radians) θ X (radians)

21 45 degrees Resolution Results, Dip = 45 Fit C Entries Mean.81 RMS.337 / ndf / 6 Constant 79.1 ± 7.9 Mean.813 ±.1 Sigma.681 ±.44 Mean.815 RMS.3648 / ndf / 36 Constant 51.6 ± 7.1 Mean.816 ±.1 Sigma.917 ±.46 N /track Entries Mean 9.76 RMS Mean 1.1 RMS (radians) C Photons per track Fit Mean vs. Bar Number Mean 6.5 RMS 3.45 Mean 6.51 RMS 3.45 # s/event vs. Bar Entries Mean Mean y 3.31 RMS RMS y 7.11 Mean Mean y 1.1 RMS RMS y Fit vs. Bar Number Mean RMS Mean 6.6 RMS 3.45 Fit Weight/Event vs. Bar Entries Mean Mean y 4.519e+5 RMS RMS y 1.5e+5 Mean Mean y.39e+5 RMS RMS y 8.6e

22 55 Degrees Resolution Results, Dip = 55 Fit C Entries Mean.817 RMS.3538 / ndf 35. / 3 Constant 4.9 ± 6.9 Mean.816 ±.1 Sigma.3163 ±.54 Mean.817 RMS.364 / ndf 49. / 9 Constant 7.9 ± 6.7 Mean.818 ±.1 Sigma.336 ±.61 N /track Entries Mean.5 RMS Mean RMS (radians) C Photons per track Fit Mean vs. Bar Number Mean 6.5 RMS 3.45 Mean RMS # s/event vs. Bar Entries Mean Mean y.7 RMS RMS y Mean Mean y RMS RMS y Fit vs. Bar Number Mean RMS 3.51 Mean 3.61 RMS 3. Fit Weight/Event vs. Bar Entries Mean Mean y 3.35e+5 RMS RMS y 8.665e+4 Mean Mean y 1.749e+5 RMS RMS y 7.46e

23 65 Degrees Resolution Results, Dip = 65 Fit C Entries Mean.818 RMS.317 / ndf 7.66 / 6 Constant 67.1 ± 7.5 Mean.819 ±.1 Sigma.877 ±.48 Mean.8 RMS.383 / ndf / 33 Constant 36 ± 6.9 Mean.819 ±.1 Sigma.36 ±.6 N /track Entries Mean RMS 4.64 Mean 6.88 RMS (radians) C Photons per track Fit Mean vs. Bar Number Mean 6.5 RMS 3.45 Mean 6.5 RMS 3.45 # s/event vs. Bar Entries Mean 6.53 Mean y RMS RMS y 4.64 Mean 6.53 Mean y 6.88 RMS RMS y Fit vs. Bar Number Mean RMS 3.49 Mean 6.49 RMS Fit Weight/Event vs. Bar Entries Mean 6.53 Mean y.698e+5 RMS RMS y 7.59e+4 Mean 6.53 Mean y 1.475e+5 RMS RMS y 6.66e

24 75 Degrees Resolution Results, Dip = 75 Fit C Entries Mean.8 RMS.339 / ndf 5. / 8 Constant 41.9 ± 7. Mean.81 ±.1 Sigma.3166 ±.57 Entries 1984 Mean.81 RMS.371 / ndf / 34 Constant 7.9 ± 6.5 Mean.8 ±. Sigma.3365 ±.59 N /track Entries Mean 1.91 RMS Entries 1984 Mean RMS (radians) C Photons per track Fit Mean vs. Bar Number Mean 6.5 RMS 3.45 Mean 6.5 RMS 3.45 # s/event vs. Bar Entries Mean 6.53 Mean y 1.91 RMS RMS y Entries 1984 Mean 6.54 Mean y RMS RMS y Fit vs. Bar Number Mean 6.54 RMS 3.46 Mean RMS 3.58 Fit Weight/Event vs. Bar Entries Mean 6.53 Mean y.33e+5 RMS RMS y 6.396e+4 Entries 1984 Mean 6.54 Mean y 1.141e+5 RMS RMS y 4.939e

25 85 Degrees Resolution Results, Dip = 85 Fit C Entries Mean.89 RMS.365 / ndf 4.7 / 8 Constant 6.8 ± 6.6 Mean.88 ±.1 Sigma.33 ±.6 Entries 1973 Mean.88 RMS.431 / ndf / 3 Constant 188 ± 5.9 Mean.86 ±.1 Sigma.393 ±.8 N /track Entries Mean RMS 4.93 Entries 1973 Mean 5.56 RMS (radians) C Photons per track Fit Mean vs. Bar Number Mean 6.5 RMS 3.45 Mean 6.5 RMS 3.45 # s/event vs. Bar Entries Mean Mean y RMS RMS y 4.93 Entries 1973 Mean 6.4 Mean y 5.56 RMS RMS y Fit vs. Bar Number Mean RMS Mean RMS 3.49 Fit Weight/Event vs. Bar Entries Mean Mean y.33e+5 RMS RMS y 5.966e+4 Entries 1973 Mean 6.4 Mean y 9.577e+4 RMS RMS y 4.593e