Improvements to Bremsstrahlung energy loss correction for electron momentum ... · 2015-04-10 · Improvements to Bremsstrahlung energy loss correction for electron momentum reconstruction

Improvements to Bremsstrahlung energy loss correctionfor electron momentum reconstruction

PANDA Collaboration Meeting

Ermias ATOMSSA

Institut de Physique Nucleaire d’Orsay

March 17, 2015

Ermias ATOMSSA (IPNO) PCM, Bremsstrahlung Correction Mar. 17, 2015 1 / 17

Effect of Bremsstrahlung radiation

Photon emission by electrons before exiting the tracking system

Some or all of the tracking points are created after the track loses momentum

Leads to mis-reconstruction of the momentum that shows as a tail in resolutiondistributions and invariant mass spectra


General approach

Thesis work by Binsong Ma, 23 Sept 2014, Universite Paris SudEnergy distribution highly non Gaussian: Can not be corrected easily with Kalman FilterSearch for clusters in the EMC generated by the Bremsstrahlung photons associated witheach track and add the total energy to the track’s momentum

Relies on the observation that90% of Bremsstrahlung photons are emitted within a cone of 2 mradThe angular separation of the reconstructed electron momentum and photon clusterposition is a reliable predictor of the radius at which the photon was emitted


Cluster merging

Depending on the momentum and point of emission, Bremsstrahlung photon clusters canbe reconstructed separately from or merged with that of the electron

Higher momentum tracks and late emission of photons tend to result in merging, whereaslower momenta and early emission result in separate clusters

The method has to handle both cases for every track


Separate bumps

Dominant at low momentum and late emissionSearch for bumps in the EMC with Ebump > 1%Etrack

Apply angular selection on ∆φ = ±(φbump − φe∓ ) and ∆θ = ±(θbump − θe∓ )∆θ: |∆θ| < 2◦ (no bending assumed in theta)∆φ (Barrel): −1◦ < ∆φ < 2 arcsin((0.3 · B · Rext

TRK )/precT )∆φ (Forward): −1◦ < ∆φ < (0.3 · B · ZSTT+MVD+GEM · tan θrec )/Prec

T

Corrected momentum: pCOR = pKF + ETOTγ,sep , where ETOT

γ,sep =∑

EBrem cutγ,sep. bumps


Merged bumps

Project the energy deposits in the cluster associated to the electron in φ direction

Split the phi projection along local minima into “φ-bumps”

The right (left) most φ-bump with sufficient energy is considered as originating from thee+ (e−) and the total energy of the phi bumps to the left (right) is added to themomentum of the track depending on charge

pCOR = pKF + ETOTγ,sep + ETOT

γ,mrg , where ETOTγ,mrg =

∑Eφ−bumps


Performance of the method

MC)/p

COR - p

MC(p

0.1− 0.05− 0 0.05 0.1 0.15 0.2 0.25 0.30

200

400

600

800

1000

1200

1400

UncorrectedSep. bumps onlyFull corr.

= 0.5 GeV/cT

p < 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.1− 0.05− 0 0.05 0.1 0.15 0.2 0.25 0.30

200

400

600

800

1000

1200

1400 = 1.0 GeV/cT

p

< 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.1− 0.05− 0 0.05 0.1 0.15 0.2 0.25 0.30

200

400

600

800

1000 = 2.0 GeV/cT

p < 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.1− 0.05− 0 0.05 0.1 0.15 0.2 0.25 0.30

200

400

600

800

1000

1200

1400

1600

= 0.5 GeV/cT

p

< 20θ 10 <

FWD. ENDCAP

MC)/p

COR - p

MC(p

0.1− 0.05− 0 0.05 0.1 0.15 0.2 0.25 0.30

200

400

600

800

1000

1200

1400

= 1.0 GeV/cT

p

< 20θ 10 <

FWD. ENDCAP

MC)/p

COR - p

MC(p

0.1− 0.05− 0 0.05 0.1 0.15 0.2 0.25 0.30

200

400

600

800

1000 = 2.0 GeV/cT

p

< 20θ 10 <

FWD. ENDCAP

Significantly improved resolution at all angles (L:Barrel, R:FWD) and momenta

Correction with separated only and both separated and merged photons

Contribution from merged increases with momentum


Neutral candidates vs. bumps

Initial code used neutral candidates for separated bump correction to avoid adding bumpsalready associated with the electron

Results in a dependence on track-cluster association criteriaIssue first noticed in recent PANDAroot releases (≈ scrut14 and above)Possibly due to bumps being assigned to secondary electrons/fake tracks?

All bumps used now with a condition that EmcIndex(bump) 6= EmcIndex(track)

MC - p)/p

MC(p

0.2− 0.1− 0 0.1 0.2 0.30

1000

2000

3000

4000

5000

6000

7000UncorrectedNeut. candsBumps

= 0.5T

p

< 45θ 30 <


Over-correction

Visible asymmetry of corrected resolution, too much energy is being added back

For tracks that emit late, the momentum reconstruction is mostly based on “good” hits

Adding the full energy of the photon for such tracks over-corrects momentum


Emission radius dependence of resolution degradation

Photons emitted in later half of tracking system do not affect the reconstructedmomentum as much as photons emitted early

Reconstructed momentum resolution improves with increasing MC radius (RTrue) ofemission (ie. fraction of tracking points before emission)

MC)/p

KF - p

MC(p

0.2− 0 0.2 0.4 0.6 0.8 1

4−10

3−10

2−10

1−10

Resolution of reconstructed momentum

< 45θ 30 <

= 0.5T

p

(cm) < 7True0 < R






MC)/p

KF - p

MC(p

0.2− 0 0.2 0.4 0.6 0.8 1

4−10

3−10

2−10

1−10


< 45θ 30 <

= 0.5T

p

(cm) < 14True7 < R






MC)/p

KF - p

MC(p

0.2− 0 0.2 0.4 0.6 0.8 1

4−10

3−10

2−10

1−10


< 45θ 30 <

= 0.5T

p

(cm) < 21True14 < R






MC)/p

KF - p

MC(p

0.2− 0 0.2 0.4 0.6 0.8 1

3−10

2−10

1−10


< 45θ 30 <

= 0.5T

p

(cm) < 28True21 < R






MC)/p

KF - p

MC(p

0.2− 0 0.2 0.4 0.6 0.8 1

3−10

2−10

1−10


< 45θ 30 <

= 0.5T

p

(cm) < 35True28 < R






MC)/p

KF - p

MC(p

0.2− 0 0.2 0.4 0.6 0.8 1

3−10

2−10

1−10


< 45θ 30 <

= 0.5T

p

(cm) < 42True35 < R



Emission radius weighted correction

Solution: Add a smaller fraction of the energy for photons emitted later:ETOTγ,sep =

∑W (R)× EBrem cut

γ,sep. bumps

Constraints on weight function: W (0) = 1 and W (RextTRK ) = 0.

Estimator of R based on the relation between RTrue ,∆φ, pT : RCalc = K× 2pT sin (∆φ/2)0.3B

Correlation plot: Events with single Bremsstrahlung photon (Remission < RTRK ) and singleidentified separate bump (cut on ∆φ, ∆θ), MC matching requiredWeight: Fermi function with inflection point at the middle of tracking system

0

50

100

150

200

250

300

350

Calculated radius vs. true radius

True R0 5 10 15 20 25 30 35 40

Cal

c R

0

5

10

15

20

25

30

35

40

= 0.5 GeV/cT

p

Calculated radius vs. true radius

CalcR0 5 10 15 20 25 30 35 40

)C

alc

W(R

0

0.2

0.4

0.6

0.8

1(R-R0)/R11+e1W(R) =

/2)TRKextR0 = 21 cm (= R

/10)TRKext R≈R1 = 5.0 cm (


Results with weighted correction (Barrel Region)

MC)/p

COR - p

MC(p

0.2− 0.1− 0 0.1 0.2 0.30

1000

2000

3000

4000

5000

6000

7000

8000Full corr.

Full corr. (wtd)

= 0.5T

p

< 45θ 30 <

Resolution of fully corrected (with weight) momentum

MC)/p

COR - p

MC(p

0.2− 0.1− 0 0.1 0.2 0.30

1000

2000

3000

4000

5000

6000

7000

8000Full corr.

Full corr. (wtd)

= 1.0T

p

< 45θ 30 <

Resolution of fully corrected (with weight) momentum

Weighting fixes over-correction at all momenta and angles (more plots in backup)

The same function is used everywhere

Particularly useful for low momentum tracks where search window can be wide


Full event simulation

Presence of other particles that generate hits in the EMC can bias the correctionPotential issue especially at low momenta, where ∆φ search window is largeTo test the stability in presence of other particles: π0J/ψ and π0π+π−

Significantly improved J/ψ mass resolution with minimal effect on π+π− spectrumModifications (discussed on PANDAroot forum, msg:17933) to the track-cluster matching algorithms were usedReason: A failed association can result in adding the energy of the electron’s cluster back to the momentum

Allows for much tighter mass cut while improving efficiency

-e+eM0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0

200

400

600

800

1000

1200

1400

1600

inv. massψJ/ inv. massψJ/

-π+πM0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0

50

100

150

200

250

300

inv. mass-π+π

Raw (eid)

Corrected (eid)

inv. mass-π+π


https://forum.gsi.de/index.php?t=tree&th=4573&start=0&

Implementation in PANDAroot (1/3)

φ-bumps are created for all clusters and stored separately in PndEmcPhiBumpSpiltter.cxx(called from XYZ..). No change required in “official” simulation macros for this step

Corrected momentum calculated for all reconstructed tracks and stored separately in aTCA by PndPidBremCorrector.cxx. Add the following to pid complete.C:

// after all the Pid modules...

PndPidBremCorrector *bremCorr = new PndPidBremCorrector();

fRun->AddTask(bremCorr);

...

In analysis macro, request RhoCandidate objects to be filled with corrected momenta byadding “Brem” to the selection string

PndAnalysis* theAnalysis = new PndAnalysis();

RhoCandList eplus;

while (theAnalysis->GetEvent() && i++<nevts) {

...

theAnalysis->FillList(eplus, "BremElectronVeryTightMinus");

...

}

...



Alternatively, working directly with charged candidates

// MyTask.h:

TClonesArray* fBremCorr;

...

// MyTask.cxx

#include "PndPidBremCorrected4Mom.h"

// in MyTask::Init() method

fBremCorr =

dynamic_cast<TClonesArray *> (ioman->GetObject("BremCorrected4Mom"));

...

// in MyTask::Exec() method

for (int j=0; j<fChargedCandList.GetLength(); ++j) {

int trk_id = fChargedCandList[j]->GetTrackNumber();

PndPidBremCorrected4Mom *bremCorr =

(PndPidBremCorrected4Mom*) fBremCorr->At(trk_id);

TVector3 mom_corrected = bremCorr->GetMomentum());

double energy_corrected = bremCorr->GetEnergy());

...

}

...



Does not affect any analysis that doesn’t request it explicitly (separate storage)

PID consideration:

All PID probabilities are still calculated with uncorrected momentumNot advisable to use corrected momenta as input for PID algorithms at this pointunless PID probabilities are recomputedPotential to use the corrected momentum as additional input to PID algorithms

Will be available soon in trunk release (pending approval)

In the meantime all the relevant source files can be found here:https://github.com/atomssa/brempatch


https://github.com/atomssa/brempatch

Conclusion

Bremsstrahlung correction method for electrons fully debugged and implemented inPANDAroot with some improvements

Offers improved spectra for signals with electrons in the exit channel and improved masscut efficiency for resonances

Simple usage instructions for simulations are provided


Backup


Performance with varying momentum (Barrel Region)

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

100

200

300

400

500

No weight

W/ weight

= 0.15 GeV/cT

p

< 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400 = 0.25 GeV/cT

p

< 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400

1600

1800

2000 = 0.50 GeV/cT

p

< 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400

1600

1800

2000 = 1.00 GeV/cT

p

< 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400

1600

1800= 1.50 GeV/c

Tp

< 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400 = 2.00 GeV/cT

p

< 45θ 30 <

BARREL


Performance with varying energy (FWD Region)

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400

No weight

W/ weight

= 0.15 GeV/cT

p

< 20θ 10 <

FWD.ENDCAP

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400= 0.25 GeV/c

Tp

< 20θ 10 <

FWD.ENDCAP

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400

1600

1800

2000 = 0.50 GeV/cT

p

< 20θ 10 <

FWD.ENDCAP

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400

1600

1800 = 1.00 GeV/cT

p

< 20θ 10 <

FWD.ENDCAP

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400

1600 = 1.50 GeV/cT

p

< 20θ 10 <

FWD.ENDCAP

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400= 2.00 GeV/c

Tp

< 20θ 10 <

FWD.ENDCAP


Performance with varying angles (Barrel Region)

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

200

400

600

800

1000

1200

1400

1600

1800

2000

No weight

W/ weight

= 0.50 GeV/cT

p

< 45θ 30 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

500

1000

1500

2000

2500 = 0.50 GeV/cT

p

< 60θ 45 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

500

1000

1500

2000

2500= 0.50 GeV/c

Tp

< 75θ 60 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

500

1000

1500

2000

2500

= 0.50 GeV/cT

p

< 90θ 75 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

500

1000

1500

2000

2500= 0.50 GeV/c

Tp

< 105θ 90 <

BARREL

MC)/p

COR - p

MC(p

0.2− 0.15− 0.1− 0.05− 0 0.05 0.1 0.15 0.20

500

1000

1500

2000

2500 = 0.50 GeV/cT

p

< 120θ105 <

BARREL


Documents

Improvements to Bremsstrahlung energy loss correction for electron momentum ... · 2015-04-10 · Improvements to Bremsstrahlung energy loss correction for electron momentum reconstruction