search of optimized cuts for pulsar detection

1Andrea Caliandro

Search of Optimized Cuts for Search of Optimized Cuts for

Pulsar DetectionPulsar DetectionAndrea Caliandro - INFN Bari

DC2 CloseOut

May 31 2006

Goddard Space Flight Center

2Andrea Caliandro

Summary Summary

Have been searched the best ROI & Energy band that maximize the probability for a point source to be periodic

– Parameter considered• Sky position• Flux• Observation Time

For this work a sample of 900 pulsars and diffuse emission have been simulated 15 times

– Each simulation is characterized by a value of flux attributed to all the pulsars and by the Observation time

• PSR flux: 1e-6, 0.3e-6, 1e-7, 0.3e-7, 1e-8 ph/cm2s• ObsTime: 1 Week, 2Week, 4Week

Finally this search has been tested on the DC2 sky– All the radio loud PSR in the pulsarDB have been analyzed

3Andrea Caliandro

Sky SimulationSky Simulation

AntiCenter Lateral Center Center Lateral

(2,-2) (6, 2) (10, 6)

• Is taken into account the strip of sky |b|<10

• This strip is subdivided in 9 regions– Center2-2, Center2_6, Center 6_10

– Lateral2-2, Lateral2_6, Lateral6_10

– Anticenter2-2, Anticenter2_6, Anticenter6_10

• In each region there are 100 psr uniformly distributed

– In this way the statistic is the same for each region

– This is not a simulation of a real pulsar population

• Diffuse emission simulation

– Galactic diffuse emission

– Extragalactic diffuse emission

• gtorbsim parameter– Pointing strategy: SLEWING

– Rocking angle = 35°

4Andrea Caliandro

Pulsar SimulationPulsar Simulation

• 2 peaks pulse shape– Phase position, width and height are random

• There is just the constrain for the peak distance to be more than 0.3 phase

• P = 1s, Pdot = 0s/s, Pddot = 0s/s2

– These choices are just to simplify the analysis, but do not affect the result

(Nel and De Jager,1995)• b = 1

• E0 = 10GeV

• g is a random number distributed as e gaussian N(1.4, 0.2)

– These choices are based on the data of the Gamma-ray Pulsars (Nel, De Jager 1995) and how I have chosen the Energy band to investigate

– See next slide

ScienceTools: CelestialSource/Pulsar

5Andrea Caliandro

Energy cutEnergy cut: Energy band from the selected

Emin value to the maximum observable Energy (200GeV)

(gtselect)

Binning: 20 MeV 50 MeV100 MeV200 MeV400 MeV600 MeV 1 GeV 10 GeV

Binning:0.25° 2°0.50° 3°0.75° 4°1° 6°1.5° 8°

ROIROI: Disk with Radius R

centerd on the source(gtselect)

Base AnalysisBase Analysis

Emin (MeV)

Rad

ius

(Deg

rees

)

Histogram by Cuts

Emin (keV)

• For each source are selected a series of ROI and Energy cut

• For each of these selection are calculated the Htest and the relative

probability (PHtest)

6Andrea Caliandro

Htest Algorithm Htest Algorithm

4312 pNpepb Np

n

1)( 312 HqeHqHw Hq

)44(max 2

201

mZH m

m

For N>=100 m=20

For N<100 m=N/5

bHeaPHtest 2eHdHecPHtest

HbneHwPHtest )(

Probability that H could be greater of the found value P(>H)

For H<=23

For 23<H<50

For N < 100

If H>=3030)30( nbewPHtest

For N < 20HbnePHtest 15.1

If H>=252515.1 nbePHtest

For N < 10: 1PHtest

84 ePHtest For H>=50

(De Jager et al. 1989)

7Andrea Caliandro

Single Pulsar Histograms

Histograms & Distributions Histograms & Distributions • From the Single Pulsar Histograms

– Htest & PHtest histograms

• Are obtained, for each value of PSR flux & Observation Time (i.e. for each simulation) and relative to each Region of the sky

– Mean Htest & PHtest Histograms– Simply the normalized sum of the Single

Pulsar Histogram

– Distribution of the Max Htest Values– Taken by each Htest Single Pulsar

Histogram» See next slides

– Distribution of the Minimum PHtest Values– Taken by each PHtest Single Pulsar

Histogram» See next slides

– Cut Distribution for which Htest is Max– From each Single Pulsar Histogram are

taken the cuts relative to the max value» See next slides

– Cut Distribution for which PHtest is Min– From each Single Pulsar Histogram are

taken the cuts relative to the min value» See next slides

PHtestHtest

PHtest

Htest

This two histograms are

relative to PSRflux=1e-7

ObsTime=4Weeks

Region=Lateral2-2

8Andrea Caliandro

AntiCenter6_10AntiCenter2_6AntiCenter2-2

Lateral6_10Lateral2_6Lateral2-2

Center6_10

Center2_6Center2-2



Center6_10Center2_6Center2-2

HtestHtest

• Cut Distribution for which Htest is Maximum

• Htest Max Value Distributions

AllRegionsconsider the Regions

all together - all 900 psr

AllRegions

The plots showed in this slide are


ObsTime=4Weeks

9Andrea Caliandro



Center6_10

Center2_6Center2-2

PHtestPHtest

AllRegions

• Cut Distribution for which PHtest is Minimum

The plots showed in this slide are


ObsTime=4Weeks



Center6_10Center2_6Center2-2

• PHtest Min Value Distributions

AllRegionsconsider the Regions

all together - all 900 psr

10Andrea Caliandro

Htest

Max Value

Distribution

Cut Distribution for which Htest is Maximum

Overview on All RegionsOverview on All RegionsHtest DistributionsHtest Distributions

Flux 1e-6 ph/cm^2 sFlux 1e-6 ph/cm^2 s Flux 1e-7 ph/cm^2 sFlux 1e-7 ph/cm^2 s Flux 1e-8 ph/cm^2 sFlux 1e-8 ph/cm^2 s

(ObsTime = 4Weeks)

• Cuts Distribution:– Decreasing the Flux

• The distribution become more broad

• The peak is displaced toward high Energy

• Max Value Distribution:– Decreasing the Flux

• The distribution is less broad

• The peak is at lower values

11Andrea Caliandro

vs Fluencevs Fluence

Fluence (ph/cm^2)

Fluence (ph/cm^2)

Fluence (ph/cm^2)

Hte

stE

min

(M

eV)

Rad

ius

(deg

ree)

Overview on All RegionsOverview on All RegionsHtest PlotsHtest Plots

vs Fluxvs Flux

Flux (ph/cm^2s)

Flux (ph/cm^2s)

Flux (ph/cm^2s)

Hte

stE

min

(M

eV)

Rad

ius

(deg

ree)

1Week2Weeks4Weeks

By Max value Distribution

•Htest vs Flux and Fluence• Mean value of distribution

• Errors bars: right and left RMS

–Htest is proportional to the flux and the fluence

By Cuts Distribution

•Energy Cut Plots• Energy coordinate of the Peak of the distribution

• Errors bars: right and left RMS respect to the Peak

–Seems to converge towards Emin=200MeV

•Radius Plots• Radius coordinate of the Peak of the distribution


–The peaks increase toward 1.5°

12Andrea Caliandro

PHtest

Min Value

Distribution

Cut Distribution for which PHtest is Minimum

Overview on All RegionsOverview on All RegionsPHtest DistributionsPHtest Distributions

Flux 1e-6 ph/cm^2 sFlux 1e-6 ph/cm^2 s Flux 1e-7 ph/cm^2 sFlux 1e-7 ph/cm^2 s Flux 1e-8 ph/cm^2 sFlux 1e-8 ph/cm^2 s

(ObsTime = 4Weeks)

This peak is due to the case N<100 & H>30

( see slide 6 )

• Cuts Distribution:– The behavior is the same of Htest Cuts

Distribution

– But the positions of the peak aren’t the same of Htest Cuts Distribution

This peak correspond to the case N>100 & H>50

(the best one)

13Andrea Caliandro

vs Fluencevs Fluence

Fluence (ph/cm^2)

Fluence (ph/cm^2)

Fluence (ph/cm^2)

PH

test

(l

og

10)

Em

in (

MeV

)R

adiu

s (d

egre

e)

Overview on All RegionsOverview on All RegionsPHtest PlotsPHtest Plots

vs Fluxvs Flux

Flux (ph/cm^2s)

Flux (ph/cm^2s)

Flux (ph/cm^2s)

PH

test

(l

og

10)

Em

in (

MeV

)R

adiu

s (d

egre

e)

1Week2Weeks4Weeks

By Min value Distribution

•PHtest vs Flux and Fluence• Mean value of distribution

• Errors bars: standard RMS

–There is a smooth step starting at 2E-2ph/cm2 ending at 3E-1ph/cm2

By Cuts Distribution

•Energy Cut Plots• Energy coordinate of the Peak of the distribution


–For low fluence is convenient to cut at Energy greater of 200MeV

–For high fluence cut aren’t needed

•Radius Plots• Radius coordinate of the Peak of the distribution


–Radius decrease until 0.25°

14Andrea Caliandro

Diffuse HtestDiffuse Htest

Htest Htest PHtestPHtest

Relative CalculationRelative Calculation

• The described analysis has been performed on the Diffuse Data file alone

– Related to each psr sky position, the Htest & PHtest histograms are builded

– The Distributions are calculated in two different way

• Absolute Calculation– The max value is taken from

each Htest histogram

– The min value is taken from each PHtest histogram

• Relative Calculation– From the Htest and PHtest

histograms is taken the value relative to the Best Cut for the corresponding pulsar

Absolute CalculationAbsolute Calculation

Diffuse Distribution example for the simulation with ObsTime = 4Weeks

and psr Flux = 1e-6

15Andrea Caliandro

PHtest with Diffuse linePHtest with Diffuse lineAllRegionsAllRegions

Diffuse Relative CalculationDiffuse Relative CalculationDiffuse Absolute CalculationDiffuse Absolute Calculation

• Absolute Calculation– The PHtest on the Diffuse is

independent from the fluence

– We can say that a source is a pulsar with good significance if log10(PHtest)<-2.5

• Relative Calculation– The PHtest on the Diffuse is not

constant but increase with the fluence

– PHtest values under the dotted line means pulsar with good significance

– above ~6E-2ph/cm2 the PHtest value of pulsar are all good

16Andrea Caliandro

PHtest Center Regions PHtest Center Regions Center2_6Center2_6Center2-2Center2-2 Center6_10Center6_10

over ~1E-1ph/cm2 the PHtest value of pulsar are all good

17Andrea Caliandro

PHtest Lateral Regions PHtest Lateral Regions Lateral2_6Lateral2_6Lateral2-2Lateral2-2 Lateral6_10Lateral6_10


18Andrea Caliandro

PHtest Anticenter Regions PHtest Anticenter Regions AntiCenter2_6AntiCenter2_6AntiCenter2-2AntiCenter2-2 AntiCenter6_10AntiCenter6_10


19Andrea Caliandro

First ConclusionsFirst Conclusions

Max Htest value in Cut Histogram not always correspond to Minimum PHtest value– Infact the distribution are different

• See slide 10 & 12

Cuts Distribution are too broad to determine a precice value for Energy cut and Radius of the ROI as function of the fluence and the position in the sky region– For low fluence are necessary Energy cut above 100 or

200MeV

– For high fluence to have a good PHtest value is enough to take radius not greater than 1.5° or 2°

By the PHtest performed on the diffuse we can say that in first approximation a source is a pulsar with good significance if log10(PHtest)<-2.5 – We have also seen that the application of the best cut

increase this value in function of the fluence

20Andrea Caliandro

DC2 Pulsar AnalysisDC2 Pulsar Analysis

C

en

ter2

-2

C

en

ter2

_6

Ce

nte

r6_

10

An

tiC

en

ter2

-2

An

tiC

en

ter2

_6

An

tiC

en

ter6

_1

0

L

ate

ral2

-2

La

tera

l2_

6

La

tera

l6_

10

E

xte

rna

l

• Best cuts have been searched for all the pulsar in the DC2pulsarDB

• DC2pulsarDB and LATsourceCatalog have been compared

– 40 pulsars have been found in the LATsourceCatalog

– In this way we also obtain a flux estimation for these pulsars

• A different analysis has been performed for pulsars with and without flux estimation

• Distribution of pulsar population in the sky

– About 90% of the pulsar population is in the Central and Lateral Region with |b|<6

21Andrea Caliandro

DC2 psr with Flux estimation DC2 psr with Flux estimation

The best ROI radius for all pulsars but 3 is less than 2°

Below 1ph/cm2 the Energy cut are broadly distributed from Emin=50MeV to Emin=1GeV

Above 1ph/cm2 there is a concentration of pulsars without Energy cut

Considering the threshold at log10(PHtest)=-2.5 we lost 6 pulsars

22Andrea Caliandro

Pulsar not in LAT catalogPulsar not in LAT catalog

• Distribution of PHtest

– There are 23 pulsars below the threshold of log10(PHtest)=-2.5

• 15 have the best possible value of the PHtest

• Distribution of Cuts

– Most of the pulsars have the best ROI radius <2°

– and No Energy cut

23Andrea Caliandro

Some Light CurveSome Light Curve

Rad=0.75Emin=200Log(P)=-6.8Nevt=2098


Rad=8Emin=200Log(P)=-4.4Nevt=30413



Rad=0.75Emin=100Log(P)=-6.6Nevt=2223Rad=0.5

Emin=400Log(P)=-2.6Nevt=632


Maybe a trick of the

simulators?

24Andrea Caliandro

ConclusionsConclusions

Appling the search of the Energy cut and the ROI radius that minimize the PHtest value we have found 23 pulsars that aren’t in the LATsourceCatalog– The cut found are not standard cut

• Most of the psr have a best radius <-2°

• Most of the psr have not Energy cut

– Maybe could be useful to build a Science Tools that make this search

search of optimized cuts for pulsar detection

Documents