Can we detect anisotropy in the mass compositionfrom the Telescope Array Surface Detector data?

Yana Zhezher for the Telescope Array Collaboration

Institute for Nuclear Research RAS

ICRC’19, 30 July 2019

Introduction: mass composition anisotropy state-of-art

I The anisotropy of cosmic ray mass composition is predicted inmultiple astrophysical models (B. R. d’Orfeuil et al., Astron.Astrophys. 567, A81 (2014)).

I Due to large shower to shower statistical fluctuations, primaryparticle type can’t be assigned for each event. Mass compositionobtained by averaging over large number of events.

I We are in need of mass composition indicator, as discriminatingas possible, to study it’s spatial distribution.

I Is the TA SD BDT ξ parameter a plausible candidate?I Test of ξ parameter sensitivity to mass composition anisotropies

with the use of the Monte-Carlo.

2 / 18

Introduction: the Telescope Array experiment

The largest UHECR experimentin the Northern Hemisphere

I Utah, USAI 507 surface detectors,

S = 3 m2, distance1.2 km

I 3 fluorescense stationsI > 11 years of constant

data acquisition3 / 18

Mass composition study with the TA SDBoosted Decision Trees:

ROOT::TMVA

(a, AoP, . . . ) → ξ

SD detector array: > 90 % dutycycle, larger data statistics

compared to FD

Comparison of ξ distributions fordata with Monte-Carlo modelling

〈lnA〉 (E)

TA, Phys. Rev. D 99, 022002 (2019)4 / 18

ξ parameter distributionh_pm

Entries 1244

ξ-0.3 -0.2 -0.1 0 0.1 0.2 0.30

50

100

150

200

250

300

h_pmEntries 1244

h_fmEntries 2604

h_fmEntries 2604

18.0<log(E)<18.2h_dt

Entries 741h_pm

Entries 10383

ξ-0.3 -0.2 -0.1 0 0.1 0.2 0.30

200

400

600

800

1000

1200

1400h_pm

Entries 10383h_fm

Entries 16257h_fm

Entries 16257

18.4<log(E)<18.6h_dt

Entries 4707

h_pmEntries 2982

ξ-0.3 -0.2 -0.1 0 0.1 0.2 0.30

100

200

300

400

500

600

700h_pm

Entries 2982h_fm

Entries 3578h_fm

Entries 3578

19.0<log(E)<19.2h_dt

Entries 1393h_pm

Entries 297

ξ-0.3 -0.2 -0.1 0 0.1 0.2 0.30

5

10

15

20

25

30

35

h_pmEntries 297

h_fmEntries 441

h_fmEntries 441

19.6<log(E)<20.0h_dt

Entries 167

proton, iron, dataTA, Phys. Rev. D 99, 022002 (2019)

5 / 18

Monte-Carlo setup

Three MC sets are constructed for testing potentialcomposition anisotropy:1. Isotropic set with data composition (p + Fe mixture).2. Set with data composition and light “hotspot” at energies

logE > 19.2.3. Set with data composition and heavy “hotspot” at energies

logE > 19.2.

I Two energy ranges: full energy range logE > 18.0 and “hotspot”energy range logE > 19.2.

I Hotspot position: R.A. = 146.7◦, Dec. = 43.2◦ (TA, ApJ 790L21 (2014)).

6 / 18

Monte-Carlo setup

p and Fe Monte-Carlo spectral sets with QGSJETII-03 – 9 years.Statistics: 78487 proton events, 117010 iron events.

Quality cuts:1. event includes 7 or more triggered stations;2. zenith angle is below 45◦;3. reconstructed core position inside the array with the distance of

at least 1200 m from the edge of the array;4. χ2/d .o.f . doesn’t exceed 4 for both the geometry and the LDF

fits;5. χ2/d .o.f . doesn’t exceed 5 for the joint geometry and LDF fit.6. an arrival direction is reconstructed with accuracy less than 5◦;7. fractional uncertainty of the S800 is less than 25 %.

7 / 18

Data composition and statistics

0

1

2

3

4

5

6

18 18.5 19 19.5 20

p

He

N

Si

Fe

<ln

A>

log10 E, eV

TA SD, QGSJET II-03

logE εp N. of events18.0-18.2 0.76 74418.2-18.4 0.46 348218.4-18.6 0.52 470718.6-18.8 0.53 383418.8-19.0 0.58 276619.0-19.2 0.47 139719.2-19.4 0.83 69119.4-19.6 0.54 28019.6-20.0 0.72 167

TA, Phys. Rev. D 99, 022002 (2019)

8 / 18

Isotropic “composition” MC oversampling, 20◦ spherical caps

E > 1018.0 eV E > 1019.2 eV

“Spots” may appear as statistical fluctuations,task for a future analysis.

red line: Galactic PlaneN. Hayashida et al. 1999, Astropart. Phys., 10, 303

9 / 18

Light “hotspot” MC oversampling, 20◦ spherical caps

E > 1018.0 eV E > 1019.2 eV

red line: Galactic Plane

10 / 18

Heavy “hotspot” MC oversampling, 20◦ spherical caps

E > 1018.0 eV E > 1019.2 eV

red line: Galactic Plane

11 / 18

1D test: ξ distribution over R.A.

I r.a. bins with 20◦ width.I Average ξ in each bin.I χ2-test for different MC sets.

Light “hotspot”: no significant sensitivity to anisotropies

-0.001

0

0.001

0.002

0.003

0.004

0.005

0.006

0.007

0.008

0 50 100 150 200 250 300 350

ξ

R.A. (degrees)

log(E) > 18.0

isotropic MClight hotspot MC

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0 50 100 150 200 250 300 350

ξ

R.A. (degrees)

log(E) > 19.2

isotropic MClight hotspot MC

χ2/d .o.f . = 0.66, p = 0.85 χ2/d .o.f . = 0.99, p = 0.4612 / 18

1D test: ξ distribution over R.A.

Heavy “hotspot”: no significant sensitivity to anisotropies

-0.004

-0.002

0

0.002

0.004

0.006

0.008

0 50 100 150 200 250 300 350

ξ

R.A. (degrees)

log(E) > 18.0

isotropic MCheavy hotspot MC

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0 50 100 150 200 250 300 350

ξ

R.A. (degrees)

log(E) > 19.2

isotropic MCheavy hotspot MC

χ2/d .o.f . = 1.24, p = 0.22 χ2/d .o.f . = 0.89, p = 0.56

13 / 18

2D test: HEALPix maps comparison

I How to take into account large-scale gradients of ξ? Spatialdistribution of average ξ may be pixelized (i.e. HEALPix,K.M, Gorski et. al., Astrophys.J, 622:759-771, 2005).

I HEALPix maps are one-dimensional arrays, each pixel refers to aspecific position on the sky.

I HEALPix maps are suitable for χ2-test.I Resolution: nside= 8 to make the statistics sufficient enough in

each pixel.

14 / 18

2D test: HEALPix maps comparison

I Light “hotspot” vs isotropic MC:

logE > 18.0 logE > 19.2χ2/d .o.f . = 1.04, p = 0.25 χ2/d .o.f . = 1.47, p = 6.9× 10−5

3.8 σ

I Heavy “hotspot” vs isotropic MC:

logE > 18.0 logE > 19.2χ2/d .o.f . = 1.01, p = 0.43 χ2/d .o.f . = 1.63, p = 6.3× 10−7

4.8 σ

Dicrimination between isotropic and anisotropic ξ distributions.

15 / 18

2D test: two-point correlation function

I Correlation functions in cosmology: distribution of galaxies andclusters of galaxies, non-Gaussianity in the CMB (Totsuji &Kihara (1969), Peebles (1973)).

I C2 (θ1, θ2) = 〈(F (θ1)− 〈F 〉) (F (θ2)− 〈F 〉)〉

Light “hotspot”: no significant sensitivity to anisotropies

-6x10-5-4x10-5-2x10-5

0 2x10-5 4x10-5 6x10-5 8x10-5 0.0001

0.00012

0.00014

0 20 40 60 80 100 120 140 160 180

C2

(θ)

θ (degrees)

log(E) > 18.0

isotropic MClight hotspot MC

-0.0004

-0.0002

0

0.0002

0.0004

0.0006

0.0008

0 20 40 60 80 100 120 140 160 180

C2

(θ)

θ (degrees)

log(E) > 19.2

isotropic MClight hotspot MC

16 / 18

2D test: two-point correlation function

Heavy “hotspot”: no significant sensitivity to anisotropies

-0.0001

-5x10-5

0

5x10-5

0.0001

0.00015

0.0002

0 20 40 60 80 100 120 140 160 180

C2

(θ)

θ (degrees)

log(E) > 18.0

isotropic MCheavy hotspot MC

-0.0008

-0.0006

-0.0004

-0.0002

0

0.0002

0.0004

0.0006

0 20 40 60 80 100 120 140 160 180

C2

(θ)

θ (degrees)

log(E) > 19.2

isotropic MCheavy hotspot MC

17 / 18

Summary

1. Potential capability to distinguish the mass compositionanisotropy in the experimental data by using the BDT ξparameter for both light and heavy “hotspot” scenario.

2. HEALPix pixelization is the strongest approach at the moment.3. Better discrimination always needed: neural network-based

composition studies (see poster by O. Kalashev & M. KuznetsovCRI172, Tue & Wed).

Supported by Russian Science Foundation

18 / 18

Backup

Observables, sensitive to the primary composition

Shower frontq Linsley curvature parameterq Area-over-peakq Area-over-peak slopeq Number of detectors,excluded from the showerfront approximation duringevent reconstruction

LDFq Sbq Sum of signals from alldetectors of the eventq Number of detectors hitqχ2/d.o.f . for LDFapproximations

Muonsq Full number of peaks in allFADC traces of the eventq Number of peaks in thedetector with the largestsignalq Assymmetry between upperand lower layers of thedetectorq Number of peaks, presentonly in the upper layer of thedetectorsq Number of peaks, presentonly in the lower layer of thedetectors

+ zenith angle, energy of the event

Linsley front curvature parameter

Deeper showermaximum means morecurved shower front.

Shower front is fit with the following function:

t0(r) = t0 + tplane+

+ a× 0.67 (1 + r/RL)1.5LDF (r)−0.5

LDF (r) = S(r)/S(800 m)

S(r) =(

rRm

)−1.2(1 +

rRm

)−(η−1.2)(1 +

r2

R21

)−0.6

Rm = 90.0 m, R1 = 1000 m, RL = 30 mη = 3.97− 1.79(sec(θ)− 1)

t iplane =

1c~n(~Ri − ~Rcore

)tplane – plane front arribal timea – Linsley front curvature parameterLDF – Lateral distribution function

Area-over-peak (AoP) and area-over-peak slope

I Time-resolved signal from a station

I AoP(r) fit with the following function:I AoP(r) = α− β(r/r0 − 1.0)I r0 = 1200m, α - AoP at 1200 m, β - slope parameter

Sb parameter

Sb =N∑

i=1

[Si ×

(ri

r0

)b],

where Si – i-th detector signal, ri – distance to the shower core in m,r0 = 1200 м – characteristic distance. b = 3 и b = 4.5 are chosen asproviding the best separation between primaries.

G. Ros, A. D. Supanitsky, G. A. Medina-Tanco et al., Astropart. Phys., 2001

“Hotspot” MC set

proton, iron