ü ü ü Data analysis Systematics Results ü Motivations Outline

(1)

(2)

Outline

ü  Motivations

ü  Data analysis

ü  Systematics

ü  Results

(3)

Motivations (1)

ü 

Electrons/Positrons are “special” CRs:

- 

Large energy losses à“local” sources

- 

Only e.m. interactions in the ISM à complementary info on CR propagations

- 

e

⁺

Sensitive to exotic sources à DM indirect searches

ü 

Most of the High Energy measurements are not distinguishing the charge sign:

- 

Direct measurements: eg. BETS, PPB-BETS, ATIC, FERMI, CALET, DAMPE

- 

Indirect measurement: Hess,Hess-LE, CTA

3

B.Bertucci AMS days, April 16, 2015

(4)

(e

⁺

+ e

^-

) measurements before AMS

Chang, 2008

Aahronian, 2008, 2009

Abdo,2009

Ackerman,2010

(5)

(e

⁺

+ e

^-

) measurements before AMS Large spread in the measurements results

5

AMS-02 “charge insensitive” measurement can be directly

compared to previous measurements

(6)

Motivations (2)

Charge insensitive measurement offers unique features:

v  No specific selection related to charge sign

ü  more statistics

ü  reduce systematics in the evaluation of the selection efficiency

v  No systematic uncertainties related to charge sign

•  Reach higher energies

•  Smaller systematic uncertainties

(7)

Data Analysis

7

N(E,E+ΔE): number of e

⁺

+e

^-

A(E) : Acceptance

ε

_trig

: trigger efficiency T(E) : time exposure ΔE : energy interval

(E, E + E) = N(E, E + E)

A(E)✏

_trig

T(E) E

(8)

Data Analysis

N(E,E+ΔE): number of e

⁺

+e

^-

A(E) : Acceptance

ε

_trig

: trigger efficiency T(E) : time exposure ΔE : energy interval

Event Selection &

Measurement Technique

(E, E + E) = N(E, E + E)

A(E)✏

_trig

T(E) E

(9)

The Data Sample

9

41 billion events up to Nov. 2013 (e⁺+e^- measurement)

> 60 billion events collected now

Events collected

Events reconstructed

(10)

Event Selection

600 GeV electron DAQ:

- efficient data periods (no SAA, TRD gas refills, AMS z-axis more than 40° w.r.t. local zenith)

Geomagnetic effects:

E>1.20 max geomagnetic cutoff

TRD:

-  Minimum 8 hits used for e/p identification -  |Z| = 1

TOF:

-  relativistic down-going particle (β>0.83) TRACKER:

-  |Z| = 1

-  track/ECAL matching to define fiducial volume

ECAL:

-  Shower axis within the fiducial volume -  Not MIP in the first 5X₀

-  Electromagnetic shape of the shower (ECAL estimator)

(11)

Measurement strategy

a) Define a clean sample of electrons/protons based on Tracker/ECAL detectors in order to study the TRD signals for electrons/protons b) Define a clean sample of electron/protons based on TRD/Tracker

detectors in order to study the ECAL signals for electrons/protons c) Efficiently select a sample of ISS data enriched in (e

⁺

+e

^-

) signal

based on ECAL

d) Measure the number of (e

⁺

+e

^-

) by a fit of the TRD classifier

distribution of the selected sample to the reference distributions in TRD for signal and background.

B.Bertucci

N

_e

, N

_p

evaluated only from Data

(12)

e/p separation with ECAL

electrons and protons behave diﬀerently when entering the ECAL

1)  Matching measured momentum in tracker with the

deposited energy in ECAL [ not used for event selecAon, but to select control samples ]

2)  3D imaging of the energy shower allows to discriminate electron or proton iniAated showers [ ECAL classiﬁer, used to preselect events for further analysis]

E ≈ 200 MeV E ≈ 20-30 GeV E ≈ 20GeV

Two complementary techniques can exploit electron/proton diﬀerences in ECAL

B.Bertucci

(13)

e/p separation with ECAL

1) Deﬁne a set of variables related to the shower:

NHITS: number of hits in the ECAL Shower Mean: weighted average of the longitudinal energy deposit

2) Combine them sta9s9cally by means of a mul9variate analysis based on Boosted Decision Tree (BDT) technique

13

0 50 100 150 200 250 300 350 400 450 !

! Nhits in ECAL ! Occurrency ! 0 0.02 0.04 0.06 0.08 0.1 0.12!

6 8 10 12 14 16 18 !

!!

Shower Mean (Layer Number)!

0.07!

0.06!

0.05!

0.04!

0.03!

0.02!

0.01!

0!

!

Occurrency !

B.Bertucci

83-100 GeV

(14)

ECAL Classifier

protons electrons

ISS data: 73–140 GeV

ECAL classifier

Fraction of events

(15)

e/p separation with TRD

15

TRD-Classifier = -Log₁₀(P_e)-2 ! !

hEle_25__1__1 Entries 19229 Mean 376.6 RMS 369.1

ADC counts

500 1000 1500

Normalized Entries

10-5

10-4

10-3

10-2

10-1

hEle_25__1__1 Entries 19229 Mean 376.6 RMS 369.1

Electrons Protons

TRD - Single tube spectrum

15

Combined Probability to be electron :!

Electrons

[50,100] GeV [25,35] GeV

Protons

[330,500] GeV [24,26] GeV

(16)

AMS data on ISS:

redundancy and complementarity

e

^-

e

⁺

p p

spillover p

N o rm al ize d En tr ie s

Control sample e^-

Control Sample p

E C A L c la ss if ie r

**sign(R)*TRD classifier**

Anti-p, spillover protons,

ISS Data: 73-140 GeV!

(17)

1D fit to measure N

_e

and N

_p

17

Reference spectra for the signal and the background are fitted to data as a function of the TRD classifier for different cuts on the ECAL BDT estimator

electrons + positrons

protons

Measurement is performed for the cut on the ECAL classifier that minimizes the overall statistical + systematic uncertainty ( à ε_BDT)

(18)

1D fit to measure N

_e

and N

_p

Reference spectra for the signal and the background are fitted to data as a function of the TRD classifier for different cuts on the ECAL BDT estimator

protons

ECAL selection ^protons

electrons + positrons electrons + positrons

Measurement is performed for the cut on the ECAL classifier that minimizes the overall statistical + systematic uncertainty ( à ε_BDT)

(19)

ECAL classifier cut efficiency

19

The efficiency of the ECAL classifier is evaluated on the negative sample (R<0), selected by means of the Tracker

à the Signal/Background in the sample is naturally enhanced and the evaluation is reliable up to highest energies

(20)

Raw (e

⁺

+e

^-

) counts

10.6 million (e

⁺

+ e

^-

) events

(21)

Systematic errors:

Stability of the signal

21

ECAL

∈ 0.78 0.8 0.82 0.84 0.86

EN

220 240 260 280 300 320

0 2 4 6 8 10 12 14 16 18 a) [500 - 700] GeV 20

Trials

EN

220 240 260 280 300

320 RMS = 4% b)

50 100 150 200

Dominating systematic uncertainties on N^e⁺^+e^-

-  Knowledge of the TRD reference distributions -  Stability of the fit result for different

background levels, e.g. ECAL classifier cuts

The analysis was repeated 2000 times in each energy bin varying the ECAL classifier cut and different values of selection cuts used to construct the

templates and the stability of the results verified within a 5% window in ECAL classifier cut efficiency

(22)

Systematic errors:

Stability of the signal

The RMS of the N_e as been used as systematics uncertainty, the effect of purely statistical contributions were taken into account and subtracted estimated from a dedicated simulation.

Negligible contribution to the measurement error below ≈ 200 GeV Dominant source of systematic error at higher energies (> 500 GeV)

ECAL

∈ 0.78 0.8 0.82 0.84 0.86

EN

220 240 260 280 300 320

0 2 4 6 8 10 12 14 16 18 a) [500 - 700] GeV 20

Trials

EN

220 240 260 280 300

320 RMS = 4% b)

50 100 150 200

(23)

Data Analysis

23

N(E,E+ΔE): number of e

⁺

+e

^-

A(E) : Acceptance

ε

_trig

: trigger efficiency T(E) : time exposure Δ(E) : energy interval

Geometrical Acceptance Selection efficiency

& Data/MC comparison

(E, E + E) = N(E, E + E)

A(E)✏

_trig

T(E) E

(24)

Detector Acceptance

Calculated with MC (Geant 4)

e-!

A

_geom

(E) = A

_gen

⇥ N

_sel

(E) N

_gen

(E)

A

_eff

(E) = A

_geom

⇥ ✏

^sel

⇥ (1 + )

A

_gen = acceptance of the generation surface

N

_sel = events passing through TRD,TOF,TRK,ECAL

ε

_sel

=

selection efficiency

δ =

data driven correction

(25)

Detector Acceptance

25

Estimated with MC (Geant 4)

3.9 m!

A

_geom

(E) = A

_gen

⇥ N

_sel

(E) N

_gen

(E)

A

_eff

(E) = A

_geom

⇥ ✏

^sel

⇥ (1 + )

A

_gen = acceptance of the generation surface

N

_sel = events passing through TRD,TOF,TRK,ECAL

ε

_sel

=

selection efficiency (70-90% above GeV)

δ =

data driven correction (-4%@ 2 GeV, -3%@1TeV)

Preselection Selected

In AMS

e-!

(26)

Detector Acceptance

Data driven correction evaluated from the comparison of each selection cut efficiency on ISS data and MC sample

Energy (GeV)

10 10

²

Efficiency

0.9 0.92 0.94 0.96 0.98 1

Energy (GeV) 5 6 7 8 910 20 30 40 50 10²

Data/MC

0.96 0.98 1 1.02

1.04 Data/MC comparison

Energy (GeV)

10 10²

Efficiency

0.9 0.92 0.94 0.96 0.98 1

Energy (GeV)

5 6 7 8 910 20 30 40 50 10²

Data/MC

0.96 0.98 1 1.02

1.04 Data/MC comparison

Example : TRD acceptance + quality cut

• 

Data/MC ratio on all cuts used to evaluate δ

• 

Deviation from unity used to assess systematic uncertainty

(27)

Data Analysis

27

N(E,E+ΔE): number of e

⁺

+e

^-

A(E) : Acceptance

ε

_trig

: trigger efficiency T(E) : time exposure Δ(E) : energy interval

(E, E + E) = N(E, E + E)

A(E)✏

_trig

T(E) E

(28)

Trigger Efficiency

B.Bertucci

100% efficiency at E>3 GeV Determined with ISS data using with unbias trigger (pre-scaled by 1/100)

Electromagnetic Trigger:

4/4 TOF + ECAL energy deposit Z=1 Trigger

4/4 TOF + No Veto

(29)

Data Analysis

29

N(E,E+ΔE): number of e

⁺

+e

^-

A(E) : Acceptance

ε

_trig

: trigger efficiency T(E) : time exposure Δ(E) : energy interval

Data taking &

Geomagnetic effects

(E, E + E) = N(E, E + E)

A(E)✏

_trig

T(E) E

(30)

Exposure time:

geomagnetic effects

<Acquisition rate> ≈ 500 Hz

CR (R<R_cutoﬀ)

CR (R>R_cutoﬀ)

AMS

CR

Latitude (°) Latitude (°)

<Livetime> ≈ 89%

Longitude (°) Longitude (°)

Effect on data taking:

-  Reduced livetime: in South

Atlantic Anomaly region and close to geomagnetic poles.

(31)

Exposure time

31

The exposure time to a given energy along the orbit is performed only considering the time spent in the regions where the rigidity cutoff used in the event selection is lower than the energy.

Primary

Secondary

|GeoLat| < 20°

|GeoLat| > 40°

20°<|GeoLat| < 40°

dN /dE (H z G eV

-1

)

≈ 80% efficiency

Geomagnetic cutoff & orbit

(32)

Data Analysis

N(E,E+ΔE): number of e

⁺

+e

^-

A(E) : Acceptance

ε

_trig

: trigger efficiency T(E) : time exposure Δ(E) : energy interval

Energy resolution

& Calibration

(E, E + E) = N(E, E + E)

A(E)✏

_trig

T(E) E

(33)

33

Chapter 2. The AMS-02 detector 73

information used to have a redundant signal in case of anode breakdowns and also to build up

1661

the ECAL standalone trigger (see Section det:daq_trg

2.8).

1662 1663

ECAL PMT response is equalized setting the PMT gain to a common value and correcting

1664

the residual response of each cell to hadronic MIP particles o✏ine [BasaraICRC2013

202].

1665 1666

Electrons, positrons and photons reaching ECAL interact starting an electromagnetic shower.

1667

The mean longitudinal profile of the energy deposit by an electromagnetic shower is usually

1668

described by a gamma distribution [Grindhammer1993kw

203]:

1669

h 1 E

dE(t)

dt i = ( t)^{↵ 1}e ^t

(↵) (2.2) eq:show_long_prof

where t = x/X₀ is the shower depth in units of radiation length, ⇠ 0.5 is the scaling parameter

1670

and ↵ the shape parameter. The total thickness of the ECAL (⇠ 17 X⁰) allows the containment

1671

of 75% of the shower energy deposit for 1 TeV electrons.

1672

The energy of the incoming particle is measured applying corrections for the rear and lateral

1673

energy leakage, and for the anode efficiency, to the deposited energy. These corrections ensure

1674

the energy linearity to be under control to less than 1% up to 300 GeV. The calorimeter energy

1675

resolution (E)/E has been measured during the test beams [DiFalcoICRC2013

204] (see Figure fig:det:ecal:ecal_eneres

2.16) and can be

1676

parametrized as a function of the particle energy E by:

1677

(E)

E = 10.4± 0.2

pE(GeV)% (1.4± 0.1)% (2.3) eq:ecal_ene_res

fig:det:ecal:ecal_eneres

Figure 2.16: ECAL energy resolution measured using e^± test beams for

perpendicularly incident particles [DiFalcoICRC2013 204].

The fine ECAL 3D readout granularity allows to reconstruct the shower axis and direction

1678

with high precision. The ECAL pointing accuracy is an extremely important parameter for

1679

gamma ray astrophysics. The ECAL angular resolution has been measured to be better than 1

1680

for energies above 50 GeV [VecchiICRC2013

205]. The ECAL standalone trigger, whose efficiency is better than

1681

99% at energies above 5 GeV, allows to measure photons inside the AMS field of view and which

1682

did not interact before the calorimeter. Given the amount of radiation length X₀ in front of

1683

Chapter 2. The AMS-02 detector 73

information used to have a redundant signal in case of anode breakdowns and also to build up

1661

the ECAL standalone trigger (see Section det:daq_trg

2.8).

1662

1663

ECAL PMT response is equalized setting the PMT gain to a common value and correcting

1664

the residual response of each cell to hadronic MIP particles o✏ine [BasaraICRC2013

202].

1665

1666

Electrons, positrons and photons reaching ECAL interact starting an electromagnetic shower.

1667

The mean longitudinal profile of the energy deposit by an electromagnetic shower is usually

1668

described by a gamma distribution [Grindhammer1993kw

203]:

1669

h1 E

dE(t)

dt i = ( t)^{↵ 1}e ^t

(↵) (2.2) eq:show_long_prof

where t = x/X0 is the shower depth in units of radiation length, ⇠ 0.5 is the scaling parameter

1670

and ↵ the shape parameter. The total thickness of the ECAL (⇠ 17 X⁰) allows the containment

1671

of 75% of the shower energy deposit for 1 TeV electrons.

1672

The energy of the incoming particle is measured applying corrections for the rear and lateral

1673

energy leakage, and for the anode efficiency, to the deposited energy. These corrections ensure

1674

the energy linearity to be under control to less than 1% up to 300 GeV. The calorimeter energy

1675

resolution (E)/E has been measured during the test beams [DiFalcoICRC2013

204] (see Figure fig:det:ecal:ecal_eneres

2.16) and can be

1676

parametrized as a function of the particle energy E by:

1677

(E)

E = 10.4± 0.2

pE(GeV)% (1.4± 0.1)% (2.3) eq:ecal_ene_res

fig:det:ecal:ecal_eneres

Figure 2.16: ECAL energy resolution measured using e^± test beams for

perpendicularly incident particles [DiFalcoICRC2013 204].

The fine ECAL 3D readout granularity allows to reconstruct the shower axis and direction

1678

with high precision. The ECAL pointing accuracy is an extremely important parameter for

1679

gamma ray astrophysics. The ECAL angular resolution has been measured to be better than 1

1680

for energies above 50 GeV [VecchiICRC2013

205]. The ECAL standalone trigger, whose efficiency is better than

1681

99% at energies above 5 GeV, allows to measure photons inside the AMS field of view and which

1682

did not interact before the calorimeter. Given the amount of radiation length X₀ in front of

1683

Energy Resolution

The finite energy resolution could affect the flux measurement due to bin-to-bin migration effects.

Energy (GeV)

1 10 10² 10³

Relative error

0 0.01 0.02 0.03 0.04

Bin to bin migration error

The excellent energy resolution of ECAL results in a negligible effect of on the measurement error above few GeVs.

(34)

Energy scale

ECAL energy absolute scale

ü  measured on ground with test beams

ü  Minimum ionization signal from p/He used to cross-calibrate the energy scale in flight

Comparison of the reconstructed energy in the ECAL with the Momentum (P) measured in the Tracker is used to verify the stability over time

10

20 80 100 120 180 290

(35)

35

Chapter 6. The e⁺ + e flux measurement 187

Energy (GeV)

1 10 10² 10³

(E) / Eσ

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

Energy resolution

Absolute scale uncertainty

fig:flux:ecal:scale:enerr_param

Figure 6.2: Parametrization of the ECAL energy resolution (in blue) and the energy measurement absolute scale systematics (in red). Both curves have been parametrized between 10 GeV and 290 GeV using test beam e^±. The ECAL resolution energy dependence has been extrapolated continuously outside this range. The uncertainty on the absolute scale grows towards low and high energies in order to cover further discrepancies with the MC simulation.

dure explained in Section fluxintro:MCnorm

3.3 assuming an injection flux ^⇤_gen(E_gen) representative of the incoming

3916

e⁺+ e flux. The analysis selection chain has been applied on MC simulation e events. The bias

3917

introduced in the energy measurement by every cut on the analysis is in this way taken into ac-

3918

count. The selected e events are then accumulated in histograms according to the reconstructed

3919

energy value E_rec and the reconstructed flux ^⇤_rec(E_rec) is then computed, as done for data, ac-

3920

cording to the prescription of Equation ^eq:flux3.11. Finally ^⇤_rec(E_rec) is compared to the known input

3921

flux ^⇤_gen(E_gen). Any discrepancy between the two values has to be considered a consequence of

3922

bin-to-bin event migrations due to the finite ECAL energy measurement resolution and to any

3923

non-linearity in the energy scale.

3924

The uncertainty of the flux measurement due to the bin-to-bin event migration induced by the

3925

finite ECAL energy measurement resolution has been studied using this approach. For each MC

3926

simulation event, the value of E_rec has been determined by applying a smearing to the value of

3927

E_gen according to the energy resolution observed in the ECAL migration matrix. Any deviation

3928

from the linearity, observed in Figure fig:flux:ecal:scale:resomatrix

6.3, has been neglected in this approach to disentagle the

3929

bin-to-bin event migrations from the event migrations due to a miscalibration of the energy scale.

3930

This last point will be discussed later in this section. The result of the comparison of ^⇤_rec(E_rec)

3931

with the known input flux ^⇤_gen(E_gen) is shown in Figure fig:flux:ecal:scale:bintobinerror

6.4. The red points quantify the amount

3932

of systematic uncertainty to the flux measurement due to the discrepancy observed in ^⇤_rec(Erec) if

3933

compared to ^⇤_gen(Egen). This e↵ect has been studied independently also by other analysis groups

3934

within the AMS collaboration. The parametrization that has been chosen by the collaboration to

3935

•  For each energy bin, the flux measurement is reported to a representative value Ē of the energy in the bin for a flux E⁻³

•  the uncertainty on the energy scale is associated as an error to the choosen Ē

ECAL energy scale known at 2% level in [10.0 – 290.0] GeV

Energy scale

(36)

Measurement Error

Dominated by acceptance

systematics below ≈100 GeV Dominated by statistics above 130 GeV.

Finite Statistics of reference distributions in the fit are the major source of systematics.

With more data both errors will decrease.

(37)

B.Bertucci

Results: PRL 113, 221102 (2014)

(38)

Results: the flux before AMS

(39)

Results: the flux after AMS

39

(40)

Solar effects on fluxes

At low energies the flux is modulated by solar activity :

further studies from AMS-02 data will allow to accurately deconvolve the Local Interstellar Spectrum

Oct. 2013 Stat err. only Oct. 2011

Stat err. only

AMS-02 Data Neutron monitor

Φ(e+ +e- ) relative flux Neutron monitor relative counts

Flux on PRL

(41)

Results: the spectral index

41

γ = d log (Φ)/ d log (E)

γ

d log (Φ)/ d log (E) = -3.170 ±0.008 (stat+sys) ±0.008 (E scale)

A single power law describes the spectrum for E>30.2 GeV

(42)

Results: the spectral index

d log (Φ)/ d log (E) = -3.170 ±0.008 (stat+sys) ±0.008 (E scale)

A single power law describes the spectrum for E>30.2 GeV Energy (GeV)

10 10

²

Spectral Index

-3.6 -3.4 -3.2 -3 -2.8 -2.6 -2.4 -2.2

Energy (GeV)

102 10³

)-1 sr sec ]2 [ m2 (GeVΦ × 3 E

102

Data

Eγ

Fit to C ×

AMS-02 ATIC01&02 BETS9798 BETS04 Fermi-LAT HEAT94 HEAT94&95 HEAT95 H.E.S.S.

H.E.S.S. (LE)

(43)

Minimal model Fit

Minimal model combined fit to (e

⁺

+e

^-

) and positron fraction

common e^+/-source with cut-off

- primary e^-from SNR - secondary e⁺from

ISM interactions

Secondary e⁺

Primary e^- from SNR

Common e^+/-source with cut-off

43

B.Bertucci

C_SE _se ^E/E^S

e

= C

_e

E

^e

+

e⁺

= C

_e⁺

E

^e+

+

^CSE _se ^E/E^S

(44)

..next step

(45)

Conclusion

45

ü 

The statistics and the resolution of AMS provide a precision measurement of the (e

⁺

+e

^-

) flux.

ü 

ü ü ü Data analysis Systematics Results ü Motivations Outline

Outline

ü Motivations

ü Data analysis

ü Systematics

ü Results

Motivations (1)

Electrons/Positrons are “special” CRs:

Large energy losses à“local” sources

Only e.m. interactions in the ISM à complementary info on CR propagations

e

Sensitive to exotic sources à DM indirect searches

Most of the High Energy measurements are not distinguishing the charge sign:

Direct measurements: eg. BETS, PPB-BETS, ATIC, FERMI, CALET, DAMPE

Indirect measurement: Hess,Hess-LE, CTA

(e

+ e

) measurements before AMS

(e

+ e

) measurements before AMS Large spread in the measurements results

AMS-02 “charge insensitive” measurement can be directly

compared to previous measurements

Motivations (2)

Charge insensitive measurement offers unique features:

• Reach higher energies

• Smaller systematic uncertainties

Data Analysis

N(E,E+ΔE): number of e

+e

A(E) : Acceptance

ε

: trigger efficiency T(E) : time exposure ΔE : energy interval

(E, E + E) = N(E, E + E)

A(E)✏

T(E) E

Data Analysis

N(E,E+ΔE): number of e

+e

A(E) : Acceptance

ε

: trigger efficiency T(E) : time exposure ΔE : energy interval

Event Selection &

Measurement Technique

(E, E + E) = N(E, E + E)

A(E)✏

T(E) E

The Data Sample

Event Selection

Measurement strategy

a) Define a clean sample of electrons/protons based on Tracker/ECAL detectors in order to study the TRD signals for electrons/protons b) Define a clean sample of electron/protons based on TRD/Tracker

detectors in order to study the ECAL signals for electrons/protons c) Efficiently select a sample of ISS data enriched in (e

+e

) signal

based on ECAL

d) Measure the number of (e

+e

) by a fit of the TRD classifier

distribution of the selected sample to the reference distributions in TRD for signal and background.

N

, N

evaluated only from Data

e/p separation with ECAL

electrons and protons behave diﬀerently when entering the ECAL

1) Matching measured momentum in tracker with the

deposited energy in ECAL [ not used for event selecAon, but to select control samples ]

2) 3D imaging of the energy shower allows to discriminate electron or proton iniAated showers [ ECAL classiﬁer, used to preselect events for further analysis]

Two complementary techniques can exploit electron/proton diﬀerences in ECAL

e/p separation with ECAL

1) Deﬁne a set of variables related to the shower:

NHITS: number of hits in the ECAL Shower Mean: weighted average of the longitudinal energy deposit

2) Combine them sta9s9cally by means of a mul9variate analysis based on Boosted Decision Tree (BDT) technique

ECAL Classifier

ISS data: 73–140 GeV

e/p separation with TRD

AMS data on ISS:

redundancy and complementarity

e

e

p p

ü  Motivations

ü  Data analysis

ü  Systematics

ü  Results

•  Reach higher energies

•  Smaller systematic uncertainties

1)  Matching measured momentum in tracker with the

2)  3D imaging of the energy shower allows to discriminate electron or proton iniAated showers [ ECAL classiﬁer, used to preselect events for further analysis]

**sign(R)*TRD classifier**