Testing THE STATISTICAL ISOTROPY OF CMB maps
description
Transcript of Testing THE STATISTICAL ISOTROPY OF CMB maps
Testing Testing
THE THE STATISTICAL STATISTICAL ISOTROPYISOTROPY
OF OF CMB mapsCMB maps
Testing Testing
THE THE STATISTICAL STATISTICAL ISOTROPYISOTROPY
OF OF CMB mapsCMB maps
Tarun Souradeep
I.U.C.A.A, Pune
Amir Hajian, Tuhin Ghosh
MPA, Garching
(Aug 22 , 2005)
Open Question in Cosmology :
NASA/WMAP science team
Intriguing: Lack of power at large angular scales )60( o
Can imply more than just the
suppression of power in the
low multipoles !
Asymmetries in the CMB anisotropy
Broadly, statistical properties are not invariant under rotations
I.e., Breakdown of Statistical isotropy ?
N-S asymmetry Eriksen, et al. 2004, Hansen et al. 2004 (in local power)Larson & Wandelt 2004, Park 2004 (genus stat.)
Special directionsTegmark et al. 2004 (l=2,3 aligned)Copi et al. 2004 (multipole vectors)Land & Magueijo 2004 (cubic anomalies)Prunet et al., 2004 (mode coupling)
Underlying `template’ pattern Jaffe et al. 2005 (Bianchi VIIh:mild cosmic rotation?)..
High N-S asymmetry
Low N-S asymmetry
Fig: H. K. Eriksen, et al. 2003
WMAP first year data
Bianchi template
Difference map
smooth random function on a sphere (sky map).
General random CMB anisotropy: described by a
Probability Distribution Functional – Mean:
– Covariance(2-point correlation)
– ...
– N-point correlation
Statistics of CMB
)]ˆ([ nTP 0 iT
Nji TTT ...
)()(),( jijiij nTnTnnCC
)ˆ(nT
Gaussian Random CMB anisotropy
Completely specified by the covariance matrix ijC
Possibilities:• Statistically Isotropic, Gaussian models
• Statistically Isotropic, non-Gaussian models
• Statistically An-isotropic, Gaussian models
• Statistically An-isotropic, non-Gaussian models
Statistics of CMB
),(),(2
l
l
lmlmlmYaT
CMB Anisotropy Sky map => Spherical Harmonic decomposition
Statistics of CMB
Statistical isotropy
''*
'' mmlllmllm Caa
CMB anisotropy completely specified by the
angular power spectrumangular power spectrum
''*
'' mmlllmllm Caa
Single index n: (l,m) -> n
Diagonal
''*
'' : violationSI mmlllmllm Caa
Mild breakdown
**''''
*''
lmlmmlml
mllm
aaaa
aa
(Bond, Pogosyan & Souradeep 1998, 2002)
''*
'' : violationSI mmlllmllm Caa
Radical breakdown
**''''
*''
lmlmmlml
mllm
aaaa
aa
(Bond, Pogosyan & Souradeep 1998, 2002)
• Correlation is a two point function on a sphere
• Inverse-transform )()(
)}()({
21
21
2211
21
2121
21
nYnYCnYnY
mlmlmm
LMmmll
LMll
LMllLMll
LMll nYnYAnnC )}()({),( 2121 21
21
21
LM
mmmlml
LMllnnLMll
mlmlCaa
nYnYnnCddA
2211
21
2211
212121
*2121 )}()(){,(
BiPoSH
Linear combination of off-diagonal elements
Bipolar spherical harmonics.
Clebsch-Gordan
Bipolar Power spectrum (BiPS) :Bipolar Power spectrum (BiPS) :A Generic Measure of Statistical AnisotropyA Generic Measure of Statistical Anisotropy
0||21
21,,
2 llM
MllABiPS:BiPS:
rotationally invariantrotationally invariant
M
mmMlml
Mll mMlml
CaaA
1211
1
121121
*
BiPoSH BiPoSH coefficients :coefficients :
• Complete,Independent linear combinations of off-diagonal correlations.
• Encompasses other specific measures of off-diagonal terms, such as
- Durrer et al. ’98 :
- Prunet et al. ’04 :
M
Mmliml
Mllimllm
il CAaaD
1'1
)(
M
Mmlml
Mllmllml CAaaD
2'2
Recall: Coupling of angular momentum states
Mmlml |2211 0, 21121 Mmmlll
2
21221)ˆ,ˆ()(
8
1
nnCdndnd
A A weighted averageweighted average of the of the correlation function over all correlation function over all
rotationsrotations
)ˆ,ˆ(8
1)ˆˆ( :Recall 21221 nnCdnnC Bipolar multipole
index
Wigner Wigner rotation rotation matrixmatrix
)()(
mmmD
Characteristic Characteristic functionfunction
BiPS: real space constructBiPS: real space construct
''*
'' : violationSI mmlllmllm Caa
Radical breakdown
**''''
*''
lmlmmlml
mllm
aaaa
aa
(Bond, Pogosyan & Souradeep 1998, 2002)
MllA2
'
MllA4
'
Understanding BiPoSH coefficients
*' ' ' '
SI violation
:
lm l m l ll mma a C
' '''
'
Measure cross correlation in
lml m
LMlm lm
mm
LM
lm
ll a a CA
a
Bipolar spherical harmonics
Spherical Harmonic coefficents
BiPoSH coefficents
Angular power spectrum
BiPS
lma MllA'
lC
Spherical harmonics
Bipolar spherical harmonics
Spherical Harmonic Transforms
BipoSH Transforms
Angular power spectrum
BiPS
lma MllA'
lC
Spherical harmonics
Measure of Statistical IsotropyMeasure of Statistical Isotropy
00
isotropy Stat.
2
''
'''
' ''
BA
CaaA
Mll
Mll
Mmllm
mm
Mll mlml
• Averaging over l,l’& M beats down Cosmic variance .
• Fast: Advantage of fast SH transform. (1 min. /alpha 1.25 GHz proc.: Healpix 512, BiPS upto 20 )
• Orientation independent.
bias
SH transform of the map
Cosmic Bias
• Analytically calculate multi-D integrals over
– Gaussian statistics => express as products of covariance.
For SI correlation
)()()()( 4321 nTnTnTnT
))1(1()12(212
1
1
12
12
||
||lll
l
l
lll CCB
~B
(A. Hajian and Souradeep, ApJ Lett. 2003)
“True” Cl
Cosmic Variance
• Analytically calculate multi-D integrals over
– Gaussian statistics => express as products of covariance.
Tedious exercise: 105 terms, 96 connected terms.
''
', , ,
2
,
2
''
',
4
,,
2
1
2,
3
1
22
,
2
1
24
312211424231
1
131
2
242
22112
21
1
311141214131
1
1
2111
1
132
321
1
2
21
12
21
1
1
1
4
))1(2(412
)12(8
12
)12()1(15)12(4))12()1(4
12
)12(9()var(
Mmlml
Mmlml
Mmlml
MM
l
lmm
l
lmm
Mmlmll
lll
Mmlml
Mmlml
Mmlml
MM
l
lm
Mmlml
llll
llll
lll
llll
ll
l
CCCCCC
CCCCCCCCl
CCl
CCl
C
i
222 ~~)(
)()()()()()()()( 87654321 nTnTnTnTnTnTnTnT
(A. Hajian and Souradeep, ApJ Lett. 2003)“True” underlying theory
Bias corrected BiPS measurement
Analytic estimate for bias and cosmic variance match numerical measurements on simulated statistically isotropic maps !
~B
(A. Hajian and Souradeep, ApJ Lett. 2003)
222 ~~)(
Bias
Cosmic Variance
1
Testing Statistical Isotropy of WMAP
Circles search (Cornish, Starkman, Spergel, Komatsu 2004)
Foreground cleaned map
(Tegmark et al. 2003)
ILCNASA/WMAP science team
(for WMAP best fit model)
(Hajian, TS, Cornish, ApJLett in press )
Angular power spectra of the maps
• `Spergel’ Circles search map
• `Tegmark’ Foreground cleaned map
• ILC: WMAP internal combination map
(compared to the WMAP best fit model)
• WMAP best fit curve
• Average of 1000 realizations
Scanning the l-space with different windows•Maps can be filtered by isotropic window to retain power on certain angular scales, (eg., l~30 to 70)
lmllm aWa
•Cosmic variance >> Noise
•Tegmark’s map is ‘foreground’ free
Testing Statistical Isotropy of WMAPLow pass Gaussian filter at l= 40
(assuming WMAP best fit model)
Low pass Gaussian filter at l= 40
Statistically isotropic!
Probability Distribution of BiPS
Obtained from measurements of 1000 simulated SI CMB maps.
Can compute a Bayesian probability of map being SI for each BiPS multipole
(Given theory Cl)
Probability of a Map being SI
Low pass Gaussian filter at l= 40
Bayesian probability
Probability of a Map being SI
Band pass filter between multipoles 20-30
Bayesian probability
BiPS imply
WMAP is Statistically
Isotropic !!
What does the null BiPS
meaurement of CMB maps imply
Sources of Statistical AnisotropySources of Statistical Anisotropy• Ultra large scale structure and cosmic topology.
• Anisotropic cosmology• Primordial magnetic fields (based on Durrer et al. 98, Chen et al. 04)
• Observational artifacts: – Anisotropic noise – Non-circular beam – Incomplete/unequal sky coverage – Residuals from foreground removal
Cosmic topologyCosmic topology
Multiply connected universe ?Multiply connected universe ?
Simple Torus (Euclidean)
Compact hyperbolic space
MC spherical space (“soccer ball”)
Eg., Zeldovich & Starobinsky 1972
Eg., Gott ’70, Cornish et al. 1996, Linde ‘04
BiPS signature of a “soccer ball” universe
(Hajian, Pogosyan, TS, Contaldi, Bond : in progress.)
K
Ideal, noise free maps predictions
BiPS signature of a “soccer ball” universe
(Hajian, Pogosyan, TS, Contaldi, Bond : in progress.)
013.1tot
Ideal, noise free maps predictions
Measured BiPS for a “soccer ball” universe
(Hajian, Pogosyan, TS, Contaldi, Bond : in progress.)
013.1tot
0.5
2.5
2.0
1.0
1.51000 simulated full
sky maps with WMAP noise
Is there a hidden pattern?
Jaffe et al. 2005
WMAP first year data
Rotating Universe Template
Subtraction of above two maps
Bianchi CMB Map & Power spectrum
Template of a rotating universe looking along the axis of rotation.
BiPS of Bianchi plus random map
BiPS of Bianchi plus random map
=1e-3
=4e-4
BiPS of Bianchi plus random map
Target specific l-space with different windows•Maps can be filtered by isotropic window to retain power on certain angular scales,
lmllm aWa
Low pass Gaussian filters
Band pass filters
Null BiPS Probability for Bianchi VIIh ( =0.5)
Gaussian filter
Band pass filter
Null BiPS Probability for Bianchi VIIh ( =0.5)
Band pass filterCl filter
(/H)
Gaussian filter
Band passfilter
Null BiPS Probability for Bianchi VIIh ( =0.99)
Summary • Propose BiPS as a generic measure for detecting
and quantifying Statistical isotropy violations. BiPS is insensitive to the overall orientation of SI breakdown (e.g., orientation of preferred axes). Hence constraints are not orientation specific. Computationally fast method
• Null results on some WMAP full sky maps. SI improves for a theory that predicts low power on low multipoles.
• Can constrain/detect cosmic topology.. BiPS constrains Dodecahedron universe strongly.
• Constrain anisotropic cosmological models BianchiVIIh claim (Jaffe et al. 2005), (/H)0 < 2.5*10-10 (99%CL)
• Diagnostic tool for observational artifacts in CMB maps• BipoSH & BiPS of CMB Polarization maps ?
Thank you !!!Thank you !!!
Testing maps for observational artifacts:
• Anisotropic noise
• Non-circular beam
• Incomplete/unequal sky coverage
• Residuals from foreground removal
Anisotropic Noise• Simplest case:
ijiijC 2
lmlmlm
N
mlmL
nYfnnC
f
)ˆ()ˆ,ˆ(
|| 2
Anisotropic Noise• Simplest case:
ijiijC 2
Effect of the Mask(AH, Souradeep astro-ph/0501001)
Wiener Filtered Map
Foregrounds ….
Testing Statistical Isotropy of WMAP
low pass filter to retain the lowest multipoles
Probability of a Map being SI
Lowest multipoless l=2-4
Bayesian probability
(assuming WMAP bf
theory Cl )
Statistical Isotropy of WMAPProbability depend on the ‘true’ model
WMAP best fit theory spectrum over-predicts power on low
multipoles
Statistical Isotropy of WMAP
WMAP maps are SI if the model fits the power on low multipoles !!!!
Probability depend on the ‘true’ model
Statistical Isotropy of WMAPProbability depend on the ‘true’ model
WMAP best fit theory spectrum over-predicts power on low
multipoles
CSSK
TOH
(Shafeiloo & Souradeep )
Improved Error sensitive iterative Richardson-Lucy deconvolution method
)()( kGkPk
dkC ll
Primordial power spectrum from Early universe can deconvolved from CMB anisotropyspectrum
Recovered spectrum shows an infra-red cut-off on Horizon scale !!!
Is it cosmic topology ? Signature of pre-inflationary phase ? Trans-Planckian physics ? ….
Recovering the primordial power spectrum Recovering the primordial power spectrum
Horizon scale
Angular power spectrum from the recovered P(k)(Shafieloo & Souradeep 2003)
Probability of a Map being SI
Lowest multipoles l=2-4
Bayesian probability
(assuming theory Cl for an ‘optimal’ primordial power
spectrum )
BiPS imply
WMAP is Statistically
Isotropic !!
Testing Statistical Isotropy of WMAP
Band pass filter between multipoles 20-30
Testing Statistical Isotropy of WMAPGalactic mask!
(assuming WMAP best fit model)
Ultra Large scale structure of the universe
BiPS of Primordial Magnetic Fields
B=0 B=30nG
BiPS of Primordial Magnetic Fields
Hajian, Chen, Souradeep, Kahniashvili, Ratra: in progress
I.U.C.A.A., Pune, India I.U.C.A.A., Pune, India
T= TR+5*10-3 TB
T= TR+3*10-3 TB
T= TR+2*10-3 TB
T= TR+1*10-3 TB
+
+ =
=
ClT=Cl
Random+2ClBianchi
2
• Where max 2
22
2
( )
l
l T CDMl l
C
C C
ClT=Cl
Random+2ClBianchi
2 P(2)
22