-
MNRAS 000, 1–12 (2017) Preprint 18 September 2017 Compiled using
MNRAS LATEX style file v3.0
Alignments of parity even/odd-only multipoles in CMB
Pavan K. Aluri,1? John P. Ralston,2† Amanda
Weltman1,3,4‡1Cosmology & Gravity Group, Department of
Mathematics and Applied Mathematics, University of Cape Town,
Rondebosch 7700, South Africa2Department of Physics and Astronomy,
University of Kansas, Lawrence, KS 66045, USA3Institute for
Advanced Study, Princeton, NJ 08540, USA4Center for Computational
Astrophysics, Flatiron Institute, 162 Fifth Avenue, New York, NY,
USA
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACTWe compare the statistics of parity even and odd
multipoles of the cosmic microwavebackground (CMB) sky from PLANCK
full mission temperature measurements. Anexcess power in odd
multipoles compared to even multipoles has previously been foundon
large angular scales. Motivated by this apparent parity asymmetry,
we evaluatedirectional statistics associated with even compared to
odd multipoles, along with theirsignificances. Primary tools are
the Power Tensor and Alignment Tensor statistics. Welimit our
analysis to the first sixty multipoles i.e., l = [2, 61]. We find
no evidence forstatistically unusual alignments of even parity
multipoles. More than one independentstatistic finds evidence for
alignments of anisotropy axes of odd multipoles, with asignificance
equivalent to ∼ 2σ or more. The robustness of alignment axes is
tested bymaking galactic cuts and varying the multipole range. Very
interestingly, the regionspanned by the (a)symmetry axes is found
to broadly contain other parity (a)symmetryaxes previously observed
in the literature.
Key words: methods: data analysis - cosmic background radiation
- submillimetre:diffuse background
1 INTRODUCTION
Many tests of symmetry of the cosmic microwave back-ground (CMB)
sky have revealed unexplained anomalies onlarge angular scales,
namely among low multipoles. Manylow multipoles are plagued with
anomalous features, associ-ated with a breakdown of isotropy, with
significances thatvaried between different data releases (de
Oliveira-Costa etal. 2004; Ralston & Jain 2004; Copi, Huterer
& Starkman2004; Schwarz et al. 2004; Eriksen et al. 2004;
Akrami etal. 2014; Vielva et al. 2004; Land & Magueijo 2005a;
Kim& Naselsky 2010; Finelli et al. 2012). In some cases
theanomalies have been attributed to statistical flukes. How-ever,
they have received significant interest from the cosmol-ogy
community by way of alternate or independent analysestowards
understanding these peculiarities (see for exampleHajian &
Souradeep (2003); Slosar & Seljak (2005); Land& Magueijo
(2005b); Bielewicz et al. (2005); de Oliveira-Costa & Tegmark
(2006); Copi et al. (2006); Wiaux et al.(2006); Abramo et al.
(2006); Bernui et al. (2007); Grup-puso & Burigana (2009);
Sarkar et al. (2011); Cruz et al.(2011); Rassat & Starck
(2013); Rassat et al. (2014); Polas-
? E-mail: [email protected]† Email: [email protected]‡ Email:
[email protected]
tri, Gruppuso & Natoli (2015); Copi et al. (2015); Schwarzet
al. (2016); Gruppuso et al. (2011); Hansen et al. (2011);Kim &
Naselsky (2011); Maris et al. (2011); Aluri & Jain(2012);
Naselsky et al. (2013); Eriksen et al. (2007); Bernui(2008); Lew
(2008); Hansen et al. (2009); Bernui (2009); Paciet al. (2010);
Santos, Villela & Wuensche (2012); Flender& Hotchkiss
(2013); Rath & Jain (2013); Bernui, Oliveira& Pereira
(2014); Quartin & Notari (2015); Aiola et al.(2015); Gurzadyan
et al. (2009); Naselsky, Hansen & Kim(2011); Ben-David, Kovetz
& Itzhaki (2012); Cruz et al.(2006, 2007); Nadathur et al.
(2014)). Regardless of interpre-tation, the large scale anomalies
have persisted from WMAPto PLANCK mission data, where the science
teams pursuedthem with no final conclusion (Bennett et al. 2011,
2013;Planck Collaboration XXIII 2014; Planck Collaboration
XVI2016).
In this paper we uncover yet another peculiarity associ-ated
with low multipole CMB data. We compare the align-ments of
parity-even and parity-odd multipoles separately toexplore any
preferred directions associated with each. Thesignificances of
these directions point to a particular paritypreference present in
the data, and possible clues about theirrelation to other large
angle CMB anomalies.
In Kim & Naselsky (2010), an anomalous point (inver-sion)
parity asymmetry was reported to be present in CMBdata at low−l.
Odd multipoles of the CMB were found to
c© 2017 The Authors
arX
iv:1
703.
0707
0v2
[as
tro-
ph.C
O]
15
Sep
2017
-
2 P. K. Aluri et al.
have significantly more power compared to the even multi-poles
in the angular power spectrum from WMAP seven yeardata, following
an earlier analysis that used WMAP firstyear data (Land &
Magueijo 2005a). Let P+ = 〈Dl〉even−land P− = 〈Dl〉odd−l denote mean
power in even and oddmultipoles, respectively, up to a chosen lmax
in the mul-tipole range l = [2, lmax]. Here Dl = l(l + 1)Cl/2π,
andCl is the CMB angular power spectrum. Since the powerl(l+1)Cl ∼
constant, at low multipoles, the ratio R(lmax) =P+/P− is expected
to fluctuate about ‘1’. However it wasfound to be significantly
lower than ‘1’ with a probability-to-exceed the observed value in
data reaching a minimum of∼ 3σ at lmax = 22.
This was followed by other studies confirming theanomalous
nature of this parity asymmetry (Gruppuso etal. 2011; Aluri &
Jain 2012). In the PLANCK 2015 anal-ysis (Planck Collaboration XVI
2016), the p−value of thisasymmetry was evaluated to be 0.2−0.3% at
lmax = 28, de-pending on the specific component separation method
usedto extract the CMB signal.
The directionality of this parity asymmetry was probedby Zhao
(2014), where the ratio R(lmax) and its variantswere computed in
different sky directions to obtain a mapof the even-odd power
asymmetry with a chosen lmax. Cu-riously, the minimum of the odd
parity excess statistic,R(lmax), was found to occur in the
direction of the CMBdipole.
Here we analyse the even and odd multipoles separatelyin a wider
multipole range, to explore any preferred direc-tions associated
with these point parity (a)symmetry modes.
2 POWER TENSOR, POWER ENTROPY, ANDALIGNMENT STATISTICS
The Power tensor is a robust diagnostic to test isotropy ofCMB
data (Ralston & Jain 2004; Samal et al. 2008, 2009).The CMB
temperature is conventionally expanded in termsof spherical
harmonics Ylm(n̂):
∆T (n̂) =
∞∑l=2
+l∑m=−l
almYlm(n̂) . (1)
Here alm are the spherical harmonic coefficients, ∆T (n̂)
de-notes the CMB temperature anisotropies after subtractingthe
monopole and dipole, and n̂ is the position vector on thedome of
the sky.
In Dirac notation, the coefficients of the spherical har-monic
expansion are
alm = 〈l,m|a(l)〉 , (2)
where |l,m〉 represent eigenstates of the angular momen-tum
operators J2 and Jz. Under a small rotation, the alm’schange to
|a(l)〉′ = |a(l)〉+ |δa(l)〉 , (3)
where the infinitesimal change is given by |δa(l)〉 = −iJ
·Θ|a(l)〉. Here Ji (i = 1 · · · 3) are the angular momentummatrices
in spin−l representation, and Θi are the angles ofrotation. To find
the axes along which the maximum changeis achieved, compute the
Hessian, which is
∂
∂Θi∂Θj〈δa(l)|δa(l)〉 = 〈a(l)|J iJj |a(l)〉 ≡ Aij . (4)
The eigenvectors of Aij define the frame to which maxi-mal
change is developed under rotations. The correspondingstatistic Aij
that we call Power tensor, is defined as
Aij(l) =1
l(l + 1)(2l + 1)
∑mm′m′′
almJimm′J
jm′m′′a
∗lm′′ . (5)
Under the assumption of statistical isotropy, different
spher-ical harmonic coefficients are uncorrelated i.e., 〈alma∗l′m′〉
=Clδll′δmm′ , and hence 〈Aij〉 = (Cl/3)δij . Thus, in an en-semble
realizations of an uncorrelated, statistically isotropicCMB sky,
the eigenvalues of the Power tensor are randomlydistributed about
the mean value of Cl/3. The Power tensoreigenvectors are also
distributed uniformly over the sky.
Thus, Power tensor maps the complicated pattern ofeach multipole
on the sky onto an ellipsoid whose axeslengths are given by its
eigenvalues, and the three ellip-soid axes by its eigenvectors.
Hence, Power tensor can beused to characterise axiality, planarity,
as well as consis-tency with isotropy of each multipole by
comparing the ra-tio of its eigenvalues (ie., shape of the
ellipsoid), with thecorresponding eigenvector denoting the
direction of isotropybreakdown.
In any given realization, the eigenvalues of the Powertensor
will not be equal. Let the eigenvalues and eigenvec-tors
corresponding to a multipole be denoted Λα and e
iα,
where ‘α’ denotes the three eigen-indices and ‘i’ denotesthe
components of each eigenvector eα. We also define theprincipal
eigenvector (PEV) as the eigenvector associatedwith the largest
eigenvalue. Each PEV is then taken as theanisotropy axis
corresponding to a multipole l.
The significance of anisotropy/axiality represented by aPEV can
be quantified using Power entropy, defined as
S = −3∑
α=1
λα ln(λα) , (6)
where λα = Λα/∑β Λβ are the normalized eigenvalues of
the Power tensor. In the limit that a multipole is
highlyanisotropic, one normalized eigenvalue will tend to being‘1’.
Correspondingly, the Power entropy S → 0. If statisti-cal isotropy
holds, then each normalized eigenvalue is equalto 1/3, and S →
ln(3) ≈ 1.0986, which is the maximumpossible value.
The PEV’s make it possible to compare the orientationsof
different multipoles, which a priori contain information,that is
independent of the power. A typical statistic is
thedot-product-squared of PEV’s from two distinct multipoles land
l′. Squaring the dot-product removes the arbitrary signconvention
of eigenvectors.
To quantify correlations in a set of PEV’s from a rangelmin ≤ l
≤ lmax, we use the Alignment tensor X, which isdefined as
Xij(lmin, lmax) =
lmax∑l=lmin
ẽil ẽjl , (7)
where ẽl is the principal eigenvector of a multipole l. Let
ζαand fα denote the normalized eigenvalues and eigenvectorsof this
Alignment tensor. The eigenvalues are normalized toremove the
trivial effect of the l-range. One then computesthe Alignment
entropy, SX , which is a rotationally invariant
MNRAS 000, 1–12 (2017)
-
Parity wise alignments of CMB multipoles 3
summary of the ratios of ζα, that is given by
SX = −3∑
α=1
ζα ln(ζα) . (8)
When the PEV’s over the range are uncorrelated, Xij ∼δij and all
ζα are equal. In the extreme opposite case whenthe PEV’s over the
set of multipoles are all parallel to a sin-gle eigenvector, then
all but one ζα → 0. That leads to themaximal range of Alignment
entropy as 0 ≤ SX ≤ ln(3).The lower limit SX → 0 represents the
maximum pos-sible correlation. The upper limit SX → ln(3)
representsthe completely uncorrelated hypothesis of the standard
BigBang. We define the collective alignment vector of a set
ofmultipoles as the principal eigenvector of the correspond-ing
Alignment tensor (f̃α). It’s significance is assessed us-ing
Alignment entropy. The reader may refer to Ralston &Jain
(2004); Samal et al. (2008) for more details about thePower tensor
method, as well as it’s relation to axes inferredfrom other
statistics viz. the angular momentum dispersionmaximization (de
Oliveira-Costa et al. 2004) and Maxwell’smultipole vectors (Schwarz
et al. 2004).
3 DESCRIPTION OF PROCEDURE ANDDATA SETS
3.1 Analysis procedure
The Power tensor and Alignment tensor allow us toprobe any
underlying anisotropy axis associated with CMBanisotropies from a
desired multipole range or a set of mul-tipoles.
Under point inversion, alm → (−1)lalm, and so theeven(odd)
multipoles are symmetric(antisymmetric) undersuch operation. In the
present work, we apply the Align-ment tensor statistic to even and
odd multipoles separately.Thus we can explore any common preferred
axes underlyingthese modes separately.
We first compute the principal eigenvector (PEV) cor-responding
to each multipole in a chosen multipole rangel = [lmin, lmax]. The
PEVs are separated between even andodd multipoles to construct the
Alignment tensor for eachparity set separately. The PEV of the
Alignment tensor willprovide the common anisotropy axis
corresponding to eachset of parity even/odd multipoles under study.
The signifi-cance of anisotropy represented by this axis is
measured us-ing Alignment entropy. This is done by computing the
lowertail probability deduced from simulations in comparison tothe
observed entropy value from data. We also study align-ments in
cumulative multipole bins, by varying the upperand lower end of the
l−range being considered.
3.2 Real and mock data used
For this study we use the full sky Commander CMB map,derived
from PLANCK 2015 data, that is made publiclyavailable1. It is a
maximum likelihood estimate of the CMBmap, along with various
astrophysical components such as
1 http://irsa.ipac.caltech.edu/data/Planck/release_2/
all-sky-maps/matrix_cmb.html
galactic synchrotron, thermal dust, their spectral indices,etc.,
that uses multi-frequency CMB observations and ex-ternal
observations/templates for various galactic emissiontypes (Eriksen
et al. 2004, 2008; Planck Collaboration IX2016; Planck
Collaboration X 2016).
The Commander map is available at a resolution ofHEALPix2 Nside
= 2048. However, we downgrade the mapto a lower resolution of Nside
= 256, and smooth it to havea Gaussian beam FWHM = 1◦ (degrees).
Since we areinterested in large angular scales, this is sufficient
for ourpurposes.
We also prepare the mock data accordingly. ThePLANCK
collaboration has also provided sets of CMB real-izations that have
the appropriate instrument effects such asbeam smoothing, as well
as noise realizations for public use3.These are referred to as Full
Focal Plane (FFP) simulations.We use the FFP8 and FFP8.1 simulation
sets for our pur-pose. The set FFP8 was an initial release that
complementsthe PLANCK 2015 full mission data release. However dueto
a slight mismatch in the theoretical power spectrum ofCMB used to
generate these realizations, with the angularpower spectrum
consistent with final PLANCK 2015 cos-mological parameters, the CMB
realizations were updatedwith a new set denoted as FFP8.1 that
match PLANCK2015 cosmology (Planck Collaboration XII 2016). Hence
weuse simulated CMB skies from the set FFP8.1, but will usethe FFP8
realizations for noise.
The FFP simulations of CMB and noise that are pro-vided,
correspond to a specific frequency channel, and arenot readily
usable. These simulation sets do not constituteindividual component
separated maps corresponding to var-ious cleaning algorithms used
by PLANCK such as Comman-der, SMICA, etc., to obtain clean CMB maps
from the rawsatellite data (Planck Collaboration IX 2016). Thus, to
ob-tain a set of realistic CMB maps, we process this ensembleof
multi-channel maps as follows.
We downgrade all the CMB and noise realizations toa common
resolution of HEALPix Nside = 256, and smoothto have a uniform beam
resolution of FWHM = 1◦ (de-grees) Gaussian beam. We apply the
HEALPix facilitiesanafast, alteralm and synfast in that order to
bring themto the afore mentioned common HEALPix resolution andbeam
smoothing. We used the circularized beam transferfunctions
corresponding to each PLANCK frequency chan-nel, that are provided
with the second public release ofPLANCK data. We then compute the
noise rms correspond-ing to each channel using these
smoothed/downgraded re-alizations. These noise rms maps are used to
combine thesmoothed/downgraded individual frequency specific CMBand
noise realizations through inverse noise variance weight-ing. Thus
we are considering only the diagonal part of thefull covariance
matrix that results from beam smoothing.However since we are
interested in studying large angularscales, the coadded CMB and
effective noise maps thus ob-tained would sufficiently represent
the observed sky.
A set of 1000 CMB and noise Monte Carlo realizationsare provided
with appropriate instrument and noise char-acteristics through
PLANCK public release 2. Correspond-
2 http://healpix.jpl.nasa.gov/3 http://crd.lbl.gov/cmb-data
MNRAS 000, 1–12 (2017)
http://irsa.ipac.caltech.edu/data/Planck/release_2/all-sky-maps/matrix_cmb.htmlhttp://irsa.ipac.caltech.edu/data/Planck/release_2/all-sky-maps/matrix_cmb.htmlhttp://healpix.jpl.nasa.gov/http://crd.lbl.gov/cmb-data
-
4 P. K. Aluri et al.
l (`, b) (degrees)
2 (239.8◦,57.2◦)3 (244.3◦,63.0◦)
Table 1. The directions corresponding to the PEVs of l = 2,
3
modes obtained from Power tensor statistic are listed here.
Theseaxes are headless and the quoted direction is from the upper
galac-
tic hemisphere. These broadly point towards the CMB
kineticdipole direction (`, b) = (264◦, 48◦) as shown in subsequent
plots.They are aligned at a mere separation of ≈ 6.2◦.
ingly we generate 1000 co-added CMB maps with noise fromthe FFP
realizations following this procedure.
4 RESULTS
We are interested in any preferred directional
correlationsassociated with even versus odd multipoles
correspondingto large angular scales of the CMB sky. We use the
multi-pole range l = [2, 61] for this study. Before proceeding
wediscuss the anomalous alignment of quadrupole and octopolemodes
of the CMB seen in WMAP as well as PLANCK data,that have received
considerable attention (see Bennett et al.(2013); Planck
Collaboration XXIII (2014) for the assess-ment of the WMAP and
PLANCK collaborations).
4.1 Quadrupole-Octopole alignment
The alignment of the quadrupole (l = 2) and octopole(l = 3)
anisotropy axes as seen in the PLANCK full missionCommander map
deserves comment. The directions inferredfrom the principal
eigenvector (PEV) corresponding to l = 2and 3 multipoles are listed
in Table 1. Since eigenvectors ofthe Power tensor are headless
vectors, we report the direc-tion of these axes from only one of
the hemispheres. We findthat these two modes are well aligned with
a separation ofonly ≈ 6.2◦ (degrees). This corresponds to a random
chanceoccurrence probability of 1 − cos(6.2◦) ≈ 0.0058, which
isclose to a 3σ significance. Together with the CMB dipole,the
quadrupole and octopole modes point towards the Virgocluster
(Ralston & Jain 2004). These axes are shown in sub-sequent
plots as some of the reference anisotropy directionsseen in the CMB
sky. Note that the alignment of CMB tem-perature quadrupole and
octopole modes was found to im-prove by appying any additional
corrections such as resid-ual galactic bias correction (Aluri et
al. 2011) or kineticquadrupole correction, frequency independent
(Schwarz etal. 2004) or frequency dependent (Notari & Quartin
2015).The PLANCK 2015 foreground cleaned maps have the fre-quency
independent kinetic Doppler boost contribution sub-tracted (Planck
Collaboration IX 2016). Here we used thePLANCK’s Commander 2015 CMB
map as provided.
4.2 Parity alignments
Using the PEVs computed for each ‘l’ from the multipolerange of
our interest, we construct the Alignment tensordefined in Eq. [7]
for even and odd multipoles separately.
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=2, lmax=[7,61]
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
8
16
24
32
40
48
56
lmax
Figure 1. Collective alignment vectors i.e., principal
eigenvec-tors of the Alignment tensor X(lmin, lmax) (Eq. 7) for
even and
odd multipoles, obtained from PLANCK 2015 Commander full sky
CMB map are shown here in Galactic co-ordinates. The +’s de-note
even−l and and the •’s correspond to odd−l alignment axes.Other
prominent anisotropy axes seen in the CMB sky are alsolabeled.
0.01
0.1
1
10 20 30 40 50 60
Fre
qu
ency
/PT
E
lmax
2σ CL
lmin=2, lmax=[7,61]
Even-l
Odd-l
Figure 2. Significances of the collective alignment axes,
shownin Fig. [1], as measured using Alignment entropy, SX , are
plot-
ted here as a function of lmax. The lower end of the
multipolebin is fixed to lmin = 2. The upper end of the multipole
bin is
varied as lmax = [7, 61]. The probability to exceed (PTE)
the
observed value of SX from data in comparison to simulations
isplotted in black and red solid curves for even and odd
multipoles
respectively. We see that the odd multipole alignment axes
are
significantly directional at ∼ 2σ level on large angular
scales.
First we present results for the case of varying lmax,meaning,
we fix lmin = 2 and vary lmax = [7, 61]. So, thesmallest range
considered is l = 2 · · · 7, and the Alignmenttensor is computed
separately for even and odd multipolesusing l = 2, 4, 6 and l = 3,
5, 7 PEVs respectively. Then wekeep increasing the multipole range
up to lmax = 61 by twomultipoles each time (so that there are an
equal numberof even and odd multipoles for computing the
Alignmenttensor), and obtain the common anisotropy axis for the
setof even/odd multipoles in the current range every time.
Theresults are shown in Fig. [1].
There seems to be an apparent clustering of even mul-tipoles,
denoted by +’s, broadly oriented along the CMB ki-netic dipole (l =
1) direction. By progressively adding moremultipoles to the
Alignment tensor, the derived PEV movescloser to the CMB dipole
direction. On the other hand, thecommon alignment axis of odd
multipole PEVs, plotted in
MNRAS 000, 1–12 (2017)
-
Parity wise alignments of CMB multipoles 5
the same figure using • point types, steadily drifts from be-ing
close to the southern galactic pole towards the galacticplane.
We assess the significance of these collective alignmentaxes of
even/odd multipoles using the Alignment entropy(SX) defined in Eq.
[8]. The value of the Alignment entropyobtained from the data is
compared with the same quantitycomputed from simulations. The
p−value plot for the ob-served value of SX as a function of lmax is
shown in Fig. [2].We find that the apparent clustering indicated by
the com-mon alignment axes of even multipoles (black curve) is
notsignificant, as the p−value curve is always within 2σ in
themultipole range considered. However, it could be an indica-tion
of a remnant anisotropy (or a leakage) that is resultingin the
apparent clustering of these axes towards CMB dipole.
In the same plot, Fig. [2], we also show the significancesof odd
multipole alignment axes as a function of lmax (redcurve). We find
that these axes are highly directional, despitethe change in their
orientation steadily with the addition ofmore multipoles. The
significance fluctuates about the 2σconfidence level up to lmax =
27, and becomes insignificantthereafter. So, by adding more
multipoles, the directionalityof common alignment axis of odd
multipoles seen at low−lis weakened.
For reference, we also plot other interesting
anisotropydirections seen in the CMB data with different point
typesin black. The quadrupole and octopole axes listed in Ta-ble 1
of the present analysis are denoted by up and invertedtriangles
respectively. The CMB dipole direction, and thelow−l hemispherical
power asymmetry axis - that is ob-tained from the analysis of
PLANCK 2015 data using theBipoSH framework (Planck Collaboration
XVI 2016), arehighlighted using a black circle and a cross
respectively. Aset of interesting anisotropy axes corresponding to
a mirrorparity (a)symmetry are also found in the CMB data
(PlanckCollaboration XVI 2016). However, only the mirror asym-metry
axis is found to be anomalous. The maximum mirrorsymmetry axis is
labeled max(S+), and the maximum mir-ror asymmetry axis is labeled
as max(S−). These two axesare highlighted using a black diamond and
a square respec-tively in Fig. [2].
It is interesting to note that the even/odd multipoles’common
axes span two broad regions of the sky in an appar-ently
non-random/non-overlapping manner. One can readilysee that the
common alignment axes of even multipole PEVsfound here and the
(insignificant) even mirror parity direc-tion - max(S+), are
broadly aligned with the CMB kineticdipole direction. The region
spanned by the odd multipolecommon alignment axes contain the odd
mirror parity axis -max(S−), and the odd parity low−l dipole
modulation axis.
Aluri & Jain (2012) found that the significance ofthe
even-odd multipole power asymmetry in CMB angularpower spectrum
significantly decreases when the first fewmultipoles are omitted.
We now test for low multipole con-tributions to the ∼ 2σ
significance seen for the directionalityof odd multipole alignment
axes. We repeat the calculations,while choosing different lmin
values i.e., lmin = 4, 6 and 8.The results are shown in Fig. [3] in
the left column. We findthat the distribution of common alignment
axes still persistsfor different low−l cuts i.e., using different
lmin, but varyingthe other end of the multipole window upto lmax =
61.
However, similar to what was observed by Aluri & Jain
(2012), we find that the significance of odd multipole
PEValignments quickly disappears when lmin of the multipolewindow
is varied. The p−value plots corresponding to choos-ing different
lmin are shown in the right column of Fig. [3].The even-multipole
alignments remain insignificant in thiscase as well.
To study the alignment preferences of high−l in themultipole
range under consideration, we fix lmax and varylmin. In Fig. [4],
we show the collective alignment axes ob-tained by varying lmin in
the range l = [2, 56], with fixedlmax = 61. The significance of
these axes as a function oflmin are plotted in Fig. [5]. This study
suggests a possi-bility of two distinct populations for l . 30
compared to30 . l ≤ 61 when contrasted with varying lmax case.
Wefind a ∼ 2σ significance upto lmax ∼ 30 in the varying lmaxcase.
However in the varying lmin case, the significance keepsbuilding up
upto lmin ∼ 30 which indicates two distinct pop-ulations of
anisotropy axes.
We observe the alignment axis of even multipole PEVsdrifting
towards the galactic plane as more and more low−lare discarded. In
comparison, the odd multipole PEVs’alignment axis now seem to have
settled at the galacticplane. The significance of the common
alignment axis be-comes acute for lmin ∼ 28. A residual foreground
bias mayexplain the clustering of these axes in the galactic
plane,and also the corresponding anomalous significance. We pur-sue
this aspect later in the paper.
Now we probe the observed clustering of common align-ment axes
of even multipole PEVs further. The absolutescalar product of the
common axes obtained from the small-est and largest subset of
multipole bins of even/odd ‘l’ PEVsfrom the whole multipole range l
= [2, 61] is computed. Thisproduct denoted by cos(α) is taken as
representative of theseaxes being closer or scattered away from
each other. The fre-quency plots of cos(α) corresponding to even
and odd mul-tipoles, as obtained from simulations, are shown in
Fig. [6].The two cases of varying lmax and lmin while fixing the
otherend of the multipole window are shown in that figure, in
theleft and right panels respectively. The observed value of
theinner product of the same axes from the data are denotedby
vertical dashed lines in respective colours. From the his-togram
plot, we see that the clustering of even multipolecommon axes is
not statistically significant in both casesof varying lmax and
lmin. In contrast with this, the scalarproduct of odd multipoles’
common axes from the smallestand largest subsets is statistically
significant in comparisonto simulations.
The simulations suggest that the collective alignmentaxes,
computed using Alignment tensor, from the smallestand largest
multipole bin windows, tend to be closer to eachother. This could
be because the small multipole bin win-dow is a subset of the
larger multipole window, and thuscorrelated, leading to this
preference. Upon extending themultipole window range (lmax), we
observe that the distri-butions tend towards being uniform, as
expected.
We tested the stability of alignment axes by applyinggalactic
masks with different sky fractions, and inpaintingthe masked CMB
maps using iSAP software4 (Starck, Ras-sat & Fadili 2013). The
publicly available PLANCK HFI
4 http://www.cosmostat.org/software/isap/
MNRAS 000, 1–12 (2017)
http://www.cosmostat.org/software/isap/
-
6 P. K. Aluri et al.
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=4, lmax=[9,61]
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
16
24
32
40
48
56
lmax
0.01
0.1
1
10 20 30 40 50 60
Fre
qu
ency
/PT
E
lmax
2σ CL
lmin=4, lmax=[9,61]
Even-l
Odd-l
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=6, lmax=[11,61]
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
16
24
32
40
48
56
lmax
0.01
0.1
1
10 20 30 40 50 60
Fre
qu
ency
/PT
E
lmax
2σ CL
lmin=6, lmax=[11,61]
Even-l
Odd-l
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=8, lmax=[13,61]
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
16
24
32
40
48
56
lmax
0.01
0.1
1
10 20 30 40 50 60
Fre
qu
ency
/PT
E
lmax
2σ CL
lmin=8, lmax=[13,61]
Even-l
Odd-l
Figure 3. Same as Fig. [1] and [2], but for different lmin.
Although the broad orientation of the axes persists by
progressively excludingthe first few multipoles in these plots, we
find that their significances however fall (below 2σ) as seen from
the p−value plots shown inright column.
masks were used which exclude 1%, 3%, 10%, 20% and 30%of the sky
fraction5. We find that the odd multipole align-ment axes are
stable up to an exclusion of 10% of the skyin the galactic plane.
However, the even multipole commonaxes are found to be sensitive to
galactic cuts. They pro-gressively move towards or away from the
galactic plane inthe varying lmax and lmin cases respectively,
while remain-ing broadly clustered. Applying a galactic mask with
80% orless sky fraction is found to destroy the collective
orientationof these axes. This analysis is presented in Appendix
A.
We then tested the effect of including more multipolesby
extending the multipole range to lmax = 71, 81, 91 and101. Any
significant alignments seen in studying the multi-
5 http://irsa.ipac.caltech.edu/data/Planck/release_2/
ancillary-data/
pole window l = [2, 61] vanish. This is not unexpected, as
itcould be a simple consequence of diluting the signal.
Finally, we analysed clean CMB maps obtained usingother cleaning
procedures and data sets. We find a simi-lar behaviour for the
even/odd multipole common axes inWMAP provided nine year Internal
Linear Combination6
(ILC) map (Bennett et al. 2013), and the
Local-generalizedMorphological Component Analysis (LGMCA) map
thatwas produced using both the WMAP and PLANCK fullmission
observations7 (Bobin, Sureau & Starck 2016).
We also checked collective alignment axes in multipoleblocks of
∆l = 6 from the same range l = [2, 61], with threeeven/odd
multipoles in each block. The alignment axes thus
6 https://lambda.gsfc.nasa.gov/product/map/current/7
http://www.cosmostat.org/product/lgmca_cmb/
MNRAS 000, 1–12 (2017)
http://irsa.ipac.caltech.edu/data/Planck/release_2/ancillary-data/http://irsa.ipac.caltech.edu/data/Planck/release_2/ancillary-data/https://lambda.gsfc.nasa.gov/product/map/current/http://www.cosmostat.org/product/lgmca_cmb/
-
Parity wise alignments of CMB multipoles 7
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=[2,56], lmax=61
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
8
16
24
32
40
48
56
lmin
Figure 4. The alignment axis of even and odd multipole
PEVs(denoted by + and • respectively), in Galactic co-ordinates,
forfixed lmax = 61 and varying lmin in the range l = [2, 56].
0.001
0.01
0.1
1
2 10 20 30 40 50 60
Fre
qu
ency
/PT
E
lmin
2σ CL
3σ CLlmin=[2,56], lmax=61
Even-l
Odd-l
Figure 5. The lower tail probabilities or the probability to
ex-
ceed (PTE) the observed Alignment entropy, SX , of the
collectivealignment axes from data in comparison to 1000
simulations as
a function of lmin = [2, 56], while fixing lmax = 61 are
shown
here. The significances of observed SX of even and odd
multipolecommon anisotropy axes are plotted in black and red solid
curves
respectively.
inferred for even/odd multipoles accordingly span the
sameregion, from lowest multipole bin to the highest, as seen
invarying lmax and lmin cases. However the cumulative statis-tics
are better suited for our purpose i.e., to probe the widestpossible
correlations across (even/odd) multipoles.
4.3 Dissecting cumulative statistics
The cumulative statistics do not give much information onwhich
regions of the data dominate the analysis. The Align-ment entropy
is also just a single-number summary that can-not completely
identify the source of this anomaly. To gleanmore information about
the observed alignments, we lookinside the cumulative statistics in
this section, while alsointroducing an independent statistic for
testing isotropy.
To make a more informative statistic from PEVs |ẽl〉,we first
observe that normalized eigenvectors are equivalentto rank-1
projection operators Πl = |ẽl〉〈ẽl|. We can then de-fine a
Hilbert-Schmidt inner product (HSIP) (Reed & Simon1972) as
Bll′ = Tr{Π†lΠl′} = 〈ẽl|ẽl′〉2. (9)
For a set of ‘n’ unit vectors, there will be a total of‘n(n−
1)/2’ such independent inner products possible. Thedistribution of
these independent HSIPs treated as a ran-dom variable, Bll′ → x
(for all l, and l′ < l), has an analyticform given by f(x) =
1/(2
√x) for 0 ≤ x ≤ 1 (see Appendix
B for details). Correspondingly, its cumulative
distributionfunction is given by F (x) =
√x. We refer to the analytic
isotropic null distribution function as aPDF, and the
corre-sponding cumulative distribution function as aCDF.
Analo-gously, we refer to the empirical counterparts as ePDF
andeCDF, respectively. The aPDF in this form is normalized tohave
unit area under the curve.
Before proceeding further we first check that f(x) =1/(2√x) is
the true PDF of Hilbert-Schmidt inner products
of isotropically distributed unit vectors. We generate 1000sets
of n = 30 units vectors. All possible HSIPs among theseunit vectors
are computed for each set of 30 normalized vec-tors which will be a
total of 30 × 29/2 = 435. Then themean empirical distribution
function is built by taking theaverage of individual ePDF
histograms of 1000 sets of 30isotropic unit vectors to compare with
the analytic distri-bution function. The mean and analytic PDFs are
shownin Fig. [7]. The 435 independent HSIPs for each set of
30isotropic unit vectors are sorted into 50 bins to compare theaPDF
and ePDF. We find excellent agreement between thetwo distribution
functions.
Now we evaluate the ePDF and eCDF of HSIPs from thedata (PLANCK
2015 Commander map) and compare themwith their analytic forms. We
illustrate the distributionsfor three representative multipole
ranges l = [2, 25], [2, 61]and [26, 61]. There are a total of 12,
30 and 18 even or oddmultipole PEVs in these three sets. Thus 12 ×
11/2 = 66,30×29/2 = 435 and 18×17/2 = 153 independent HSIPs
arepossible, respectively, in each set of multipoles among evenor
odd multipole PEVs. Recall that, in the cumulative statis-tics, we
chose the multipole range such that there are equalnumber of
even/odd multipoles available in the l−range be-ing considered.
These are then sorted into 20 bins to buildthe ePDF and eCDF. The
results are shown in Fig. [8] and[9], for the three multipole
ranges mentioned above. In de-scribing these plots below, we only
highlight a visual discrep-ancy. Later, we use Anderson-Darling
(AD) test statistic tofind whether the data conforms with the
isotropic null dis-tribution function or not, and also quantify
it’s significanceusing simulations.
The eCDF plots highlight the peculiarity of odd multi-pole PEV
alignments rather more dramatically than ePDFplots. One notices
that there is a mild deficit at low HSIPbin values, and a mild
excess at intermediate HSIP bin val-ues in the empirical PDF of odd
multipole PEV alignmentsfor the range l = [2, 25] in Fig. [8]. The
discrepancy withthe isotropic hypothesis is more striking in the
empirical cu-mulative distribution function of odd multipole PEV
HSIPsfor the same range compared to the analytic distribution
inFig. [9]. With larger lmax = 61, the discrepancy nearly
van-ishes. The diagonal dashed line is the reference curve
aboutwhich the data statistic coming from the null distributionis
expected to fluctuate. The empirical CDF of even multi-pole PEV
HSIPs essentially criss-crosses this reference curvein Fig. [9], in
agreement with our findings from previoussections. However, as
noted above, the odd multipole align-ments deviate significantly.
Our earlier observation on the
MNRAS 000, 1–12 (2017)
-
8 P. K. Aluri et al.
0
0.04
0.08
0.12
0.16
0.2
0.24
0 0.2 0.4 0.6 0.8 1
Fra
cti
onal
counts
cos(α)
lmin=2, lmax=[7,61]
Even-l
Odd-l
0
0.04
0.08
0.12
0.16
0.2
0.24
0 0.2 0.4 0.6 0.8 1
Fra
cti
onal
counts
cos(α)
lmin=[2,56], lmax=61
Even-l
Odd-l
Figure 6. Distribution of the observed clustering of even or odd
multipole PEV common axes computed as dot products of
collective
alignment axis from the smallest and largest multipole bin sets
from the range l = [2, 61]. For the varying lmax case (left plot),
the innerproduct is taken for the axes obtained from the multipole
bins l = [2, 7] and l = [2, 61]. The varying lmin case (right
plot), uses common
alignment axes obtained from the Alignment tensor for the bins l
= [2, 61] and l = [56, 61]. The scalar product of collective
alignment
axes corresponding to smallest and largest bins of even/odd
multipoles are shown in black and red solid curves.
0
1
2
3
4
5
6
7
8
9
0 0.2 0.4 0.6 0.8 1
PD
F
x = < e~
i | e~
j >2
Emperical Dist. Func.Analytic Dist. Func. = 1/(2*sqrt(x))
Figure 7. Test of agreement between empirical PDF of
Hilbert-Schmidt inner product of isotropically distributed unit
vectors on
a sphere, and their analytic distribution function. The
simulationused 1000 random sets of 30 isotropically distributed
unit vec-tors recording the ePDF histogram each time. The mean
ePDF
obtained from averaging individual ePDFs is shown here in
bars.
The analytic PDF is shown as a solid (blue) line.
presence of two populations of anisotropy axes is also
cor-roborated by the eCDF curves for l = [2, 25] and l = [26,
61]that are non-overlapping in multipole range.
The Anderson-Darling (AD) test (Anderson & Dar-ling 1954;
Bohm & Zech 2010) quantifies the agreement ofthe data with the
isotropic null distribution. The Anderson-Darling statistic is
defined as
AD = −N −N∑i=1
2i− 1N
[ln(F (xi))+
ln(1− F (xN−i+1))] , (10)
where ‘N ’ is the number of sample points, and F (xi) is
theanalytic cumulative distribution function evaluated for thedata
sample point xi. For our specific case of HSIPs, F (xi) =√xi, and
for a set of ‘n’ even/odd multipole PEVs, there
are N = n(n − 1)/2 number of independent inner productspossible.
Similar to the case of varying lmax discussed in theprevious
section, the AD statistic is obtained as a functionof lmax from the
multipole range l = [2, 61]. At each lmax,the AD statistic is
computed from the even/odd multipolePEV sub sets of the current
l−range separately. Likewise,we also show the results for varying
lmin case.
The AD statistic values as a function of lmax are shownin the
left panel of Fig. [10], and as a function of lmin inthe right
panel in the same figure. The expected value ofthe AD statistic is
denoted by a (blue) dashed line. It iscomputed from 1000 ILC-like
noisy CMB maps obtainedfrom FFP simulations described in Sec.
[3.2]. The mean ADstatistic from simulations is evaluated in both
cases for evenand odd multipoles separately. Since the two curves
are in-distinguishable, as expected, only one of them is shown
toavoid redundancy. From Fig. [10], one can readily see thatthe AD
statistic for lmax = 9, 19, 23 and 25 acquires veryhigh values,
hinting at the origin of the 2σ level significanceseen for the
common alignment axes of odd multipole PEVson large angular scales.
From Fig. [4], we see that many ofthe collective alignment axes in
the case of varying lmin set-tled in the galactic plane.
Correspondingly, in the right-handpanel of Fig. [10], we see that
the distribution of the HSIPsquantified by the AD statistic is very
high compared to itsexpectation in the same multipole range, in the
varying lmincase.
The p−values of the AD statistic for the PLANCK 2015Commander
map derived HSIPs as a function of lmax areshown in the left-hand
panel of Fig. [11]. The significancesof the AD statistic for even
and odd multipole PEV HSIPs
MNRAS 000, 1–12 (2017)
-
Parity wise alignments of CMB multipoles 9
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
No
rma
lize
d
PD
F
x = < e~
i | e~
j >2
Even-l
Emperical PDF : l=[2,25]l=[2,61]
l=[26,61]Analytic PDF = 1/(2*sqrt(x))
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
x = < e~
i | e~
j >2
Odd-l
Emperical PDF : l=[2,25]l=[2,61]
l=[26,61]Analytic PDF = 1/(2*sqrt(x))
Figure 8. The empirical distribution functions of
Hilbert-Schmidt inner products (HSIPs) of PEVs from data computed
separately foreven and odd multipoles, from three representative
multipole ranges l = [2, 25], [2, 61] and [26, 61] are shown here.
The even and odd
HSIP ePDFs are shown in left and right panels respectively. The
analytic distribution function is shown by a dashed (blue)
line.
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Em
pir
ica
l C
DF
Analytic CDF
Even-l
l=[2,25]l=[2,61]
l=[26,61]
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Em
pir
ica
l C
DF
Analytic CDF
Odd-l
l=[2,25]l=[2,61]
l=[26,61]
Figure 9. Same as Fig. [8], but shown here are the empirical
cumulative distribution functions built from data HSIPs. See text
for
details.
0
1
2
3
4
5
6
2 10 20 30 40 50 60
An
der
son
-Da
rlin
g s
tati
stic
lmax
PLANCK15 Commander map
Even-lOdd-l
Expectation of AD
0
1
2
3
4
5
6
7
8
2 10 20 30 40 50 60
An
der
son
-Da
rlin
g s
tati
stic
lmin
Figure 10. The Anderson-Darling statistic computed separately
for even and odd multipole PEV HSIPs from PLANCK 2015 Commnadermap
are shown here. The case of varying lmax(lmin) are shown in
left(right) panel. The even and odd multipole statistic values
are
shown in black and red solid lines with square and filled circle
point types respectively. The dashed (blue) curve denotes the
expected
statistic value, obtained from an ensemble of 1000 mock observed
CMB maps, that is same for even or odd multipole PEVs.
MNRAS 000, 1–12 (2017)
-
10 P. K. Aluri et al.
are computed separately, and are shown in black and redsolid
lines with square and circle point types respectively.
The Anderson-Darling statistic gives independent con-firmation
that the odd multipole PEV alignments areanomalous on large angular
scales. Significance exceeding 2σconfidence level is found for lmax
= 9, 19, 23 and 25 whichare found to have high values for AD
statistic from the leftplot of Fig. [10]. The even multipole PEV
HSIPs show nosignificant signal of differing from the isotropic
null distri-bution in this analysis, consistent with the finding
from pre-ceding section. Thus there are some anomalous
alignmentsamong odd multipole anisotropy axes on large angular
scalesrepresented by their principal eigenvectors that are
result-ing in the high significance of our test statistic. Owing
tothe highly deviant AD statistic in the varying lmin case, theAD
statistic is found to be anomalous for the same rangeof multipoles.
The p−value plot for the same is shown inright panel of Fig. [11],
which follows a trend similar to thesignificances found in Fig.
[5].
5 CONCLUSIONS
We have compared alignment statistics of parity even andodd
multipoles with several independent methods. We usedthe clean CMB
signal estimate from PLANCK 2015 dataobtained using the Commander
algorithm. Analysis was re-stricted to the first sixty multipoles
i.e., l = [2, 61]. Powertensor and Alignment tensor statistics were
used to probethe alignments of even and odd parity multipoles,
separately.
We studied the data in several ways. The collectivealignment
axes of even and odd multipoles show different be-haviors. The
anisotropy axes of even-parity multipoles fromlarge angular scales
are broadly clustered near the directionof the CMB dipole. The
anisotropy axes of odd multipolesare much less concentrated, but
are significantly directionalas quantified by Alignment
entropy.
We constructed cumulative statistical measures thatfixed the
lower limit lmin = 2, while varying the upperlimit to reach lmax =
61. The Alignment entropy, SX , ofeven-parity multipoles was as
expected from an uncorre-lated isotropic distribution. The
odd–parity multipole SXwas unusually small on large angular scales
with signifi-cance exceeding 2σ magnitude. As lmax was increased
abovelmax ∼ 27 the significance disappeared, apparently by
dilu-tion in the larger set. This significance nevertheless
disap-pears by ignoring the first few multipoles. A similar
effectwas seen in studying even-odd multipole power asymmetry,using
the WMAP seven year temperature power spectrum(Aluri & Jain
2012). To understand the alignment prefer-ences of small angular
scales in the range being studied, wefixed the upper limit at lmax
= 61 while varying the lowerlimit lmin. A regime of multipoles with
small SX at 2σ ormore significance for odd-parity multipoles was
observed,with lowest p−value for SX occurring at lmin & 26.
Thetwo different effects from varying lmin and lmax analysis ina
single data set pose a puzzle. The resolution may involvestwo
different populations separated by a middle range ofl ∼ 27, with
each population diluting a distinctive signal ofthe other when
populations are mixed. The observation thatthe axes of the l >
27 set settled at the galactic plane maybe an indication of a
residual galactic bias in this subset.
These results are further tested against potential resid-ual
contamination in the full sky map by excising differentfractions of
the sky, and then inpainting the masked region.The odd multipoles’
common axes are stable against galac-tic cuts up to excluding (and
then inpainting) 10% of thesky, whereas the even multipoles are
found to be sensitiveto galactic cuts.
An independent statistic was used to dissect the cumu-lative
statistical studies. The Hilbert-Schmidt inner prod-ucts (HSIP) are
rotationally invariant statistics with an an-alytic isotropic null
distribution. The distribution of the datacompared to the HSIP null
was computed using Anderson-Darling (AD) test statistic. For the
odd multipole PEVs, theAD statistic for the data HSIPs shows a
significance simi-lar to that found using the Alignment entropy
method. TheAD method pinpoints lmax = 9, 19, 23 and 25 as
contain-ing unusual alignments that are rendering the AD
statisticanomalous at a significance of 2σ or more.
Interestingly, we find that the even mirror parity axisfrom the
PLANCK 2015 results, and the even multipoles’common axes from large
angular scales computed here,broadly point in the CMB dipole
direction. Likewise, theodd mirror parity axis from the PLANCK 2015
analysis,and the odd parity low−l hemispherical power asymmetryaxis
fall in the region spanned by the odd multipole align-ment axes.
From these observations, we speculate that theseanomalous axes may
have a common origin in their peculiarparity (a)symmetry
properties.
We plan to investigate these speculations more in a
laterwork.
ACKNOWLEDGEMENTS
We acknowledge the use of freely available HEALPix8 (Gorskiet
al. 2005) package and iSAP software9 in this work. Partof the
results presented here are based on observations ob-tained with
PLANCK10, an ESA science mission with in-struments and
contributions directly funded by ESA Mem-ber States, NASA, and
Canada. We also acknowledge theuse of WMAP data made available from
Legacy Archivefor Microwave Background Data Analysis11 (LAMBDA)
sitethat is a part of NASA’s High Energy Astrophysics
ScienceArchive Research Center (HEASARC). This research
usedresources of the National Energy Research Scientific Com-puting
(NERSC) Center, a DOE Office of Science User Fa-cility supported by
the Office of Science of the U.S. Depart-ment of Energy under
Contract No. DE-AC02-05CH11231.PKA is funded by the post-doctoral
fellowship program ofthe Claude Leon Foundation, South Africa at
UCT. Thiswork is based on the research supported by the
SouthAfrican Research Chairs Initiative of the Department of
Sci-ence and Technology and the National Research Foundationof
South Africa as well as the Competitive Programme forRated
Researchers (Grant Number 91552) (AW). Any opin-ion, finding and
conclusion or recommendation expressed in
8 http://healpix.jpl.nasa.gov/9
http://www.cosmostat.org/software/isap/10
http://www.esa.int/Planck11
https://lambda.gsfc.nasa.gov/product/map/dr5/
MNRAS 000, 1–12 (2017)
http://healpix.jpl.nasa.gov/http://www.cosmostat.org/software/isap/http://www.esa.int/Planckhttps://lambda.gsfc.nasa.gov/product/map/dr5/
-
Parity wise alignments of CMB multipoles 11
0.01
0.1
1
2 10 20 30 40 50 60
And
erso
n-D
arl
ing
sta
t. p
-va
lue
lmax
95% CL
Even-lOdd-l
0.01
0.1
1
2 10 20 30 40 50 60
And
erso
n-D
arl
ing
sta
t. p
-va
lue
lmin
95% CL
Figure 11. p−values of AD statistics of data shown in Fig. [10],
are plotted here as a function of lmax and lmin. The
significancesshow a similar trend for even or odd multipoles as
seen with the Alignment tensor method in Fig. [2] and [5]. The 95%
confidence level
is also shown for reference, as a dashed grey line in the plot.
The statistic shows higher significances for lmax = 9, 19, 23, 25
indicating
the possible source of the ∼ 2σ significances seen earlier, with
odd multipole PEV alignments on large angular scales, in the
Alignmententropy analysis.
this material is that of the authors and the National Re-search
Foundation (NRF) of South Africa does not acceptany liability in
this regard.PKA also thanks Pankaj Jain for helpful exchanges on
anearlier version of the paper. AW would like to thank DavidSpergel
for helpful discussions on this work.We thank the anonymous referee
for a careful reading andhelpful comments on our paper.
REFERENCES
Abramo L. R., Bernui A., Ferreira I. S., Villela T., and
Wuensche
C. A., 2006, Phys. Rev. D, 74, 063506
Aiola S., Wang B., Kosowsky A., Kahniashvili T., and
FirouzjahiH., 2015, Phys. Rev. D, 92, 063008
Akrami Y. et al., 2014, ApJ, 784, L42
Aluri P. K., Samal P. K., Jain P., and Ralston J. P., 2011,
MNRAS414, 1032
Aluri P. K., and Jain, P, 2012, MNRAS, 419, 3378
Anderson T. W., and Darling D. A., 1954, Journal of the
Ameri-can Statistical Association, 49, 765
Ben-David A., Kovetz E. D., Itzhaki N., 2012, ApJ, 748, 39
Bennett C. et al., 2011, ApJS, 192, 17
Bennett C. L. et al., 2013, ApJS, 208, 20
Bernui A., Mota B., Reboucas M. J., and Tavakol R., 2007, A
&A, 464, 479
Bernui A., 2008, Phys. Rev. D, 78, 063531
Bernui A., 2009, Phys. Rev. D, 80, 123010
Bernui A., Oliveira A. F., and Pereira T. S., 2014, JCAP, 10,
041
Bielewicz P., Eriksen H. K., Banday A. J., Gorski K. M., and
LiljeP. B., 2005, ApJ, 635, 750
Bobin J., Sureau F., and Starck J.-L., 2016, A & A, 591,
A50
Bohm G. and Zech G., Introduction to Statistics and Data
Analy-sis for Physicists (Verlag Deutsches
Elektronen-Synchrotron,
Hamburg, Germany, 2010)
Copi C. J., Huterer D., and Starkman G. D., 2004, Phys. Rev.
D,
70, 043515
Copi C. J., Huterer D., Schwarz D. J., and Starkman G. D.,
2006,MNRAS, 367, 79
Copi C. J., Huterer D., Schwarz D. J., and Starkman G. D.,
2015,MNRAS, 449, 3458
Cruz M., Tucci M., Martinez-Gonzalez E., and Vielva P.,
2006,
MNRAS, 369, 57
Cruz M., Cayon L., Martinez-Gonzalez E., Vielva P., and Jin
J.,
2007, ApJ, 655, 11
Cruz M., Vielva P., Martinez-Gonzalez E., and Barreiro R.
B.,
2011, MNRAS, 412, 2383
de Oliveira-Costa A., Tegmark M., Zaldarriaga M., and Hamil-ton
A., 2004, Phys. Rev. D, 69, 063516
de Oliveira-Costa A., and Tegmark M., 2006, Phys. Rev. D,
74,023005
Eriksen H. K., Hansen F. K., Banday A. J., Gorski K. M., and
Lilje P. B., 2004, ApJ, 609, 1198
Eriksen H. K. et al., 2004, ApJS, 155, 227
Eriksen H. K., Banday A. J., Gorski K. M., Hansen F. K., and
Lilje P. B., 2007, ApJ, 660, L81
Eriksen, H. K. et al., 2008, ApJ, 676, 10
Finelli F., Gruppuso A., Paci F., and Starobinsky A. A.,
2012,
JCAP, 07, 049
Flender S., and Hotchkiss S., 2013, JCAP, 09, 033
Gorski K. M. et al., 2005, ApJ, 622, 759
Gruppuso A., and Burigana C., 2009, JCAP, 08, 004
Gruppuso A. et al., 2011, MNRAS, 411, 1445
Gurzadyan V. G. et al., 2009, A & A, 498, L1
Hajian A., and Souradeep T., 2003, ApJ, 597, L5
Hansen F. K., Banday A. J., Gorski K. M., Eriksen H. K.,
andLilje P. B., 2009, ApJ, 704, 1448
Hansen M., Frejsel A. M., Kim J., Naselsky P., and Nesti F.,
2011,Phys. Rev. D, 83, 103508
Kim J., and Naselsky P., 2010, ApJL, 714, L265
Kim J., and Naselsky P., 2011, ApJ, 739, 79
Land K., and Magueijo, J., 2005a, Phys. Rev. D, 72,
101302(R)
Land K., and Magueijo J., 2005b, Phys. Rev. Lett., 95,
071301
Lew B., 2008, JCAP, 09, 023
Maris M., Burigana C., Gruppuso A., Finelli F., and Diego J.
M.,
2011, MNRAS, 415, 2546
Nadathur S., Lavinto M., Hotchkiss S., and Rasanen S.,
2014,Phys. Rev. D, 90, 103510
Naselsky P., Hansen M., and Kim J., 2011, JCAP, 09, 012
Naselsky P., Zhao W., Kim J., and Chen S., 2012, ApJ, 749,
31
Notari A., and Quartin M., 2015, JCAP, 06, 047
Paci F. et al., 2010, MNRAS, 407, 399
Planck Collaboration XXIII : P. A. R. Ade et al., 2014, A &
A,
571, A23
Planck Collaboration XVI: P. A. R. Ade et al., 2016a, A &
A,
MNRAS 000, 1–12 (2017)
-
12 P. K. Aluri et al.
594, A16
Planck Collaboration IX: Adam R. et al., 2016a, A & A, 594,
A9
Planck Collaboration X: Adam R. et al., 2016b, A & A, 594,
A10
Planck Collaboration XII : Ade P. A. R. et al., 2016b, A & A
594,
A12
Polastri L., Gruppuso A., and Natoli P., 2015, JCAP, 04, 018
Quartin M., and Notari A., 2015, JCAP, 01, 008
Ralston J. P., and Jain P., 2004, IJMPD, 13, 1857
Rassat A., and Starck J.-L., 2013, A & A, 557, L1
Rassat A., Starck, J.-L., Paykari P., Sureau F., and Bobin
J.,
2014, JCAP, 08, 006
Rath P. K., and Jain P., 2013, JCAP, 12, 014
Reed M., and Simon B., Methods of modern mathematical
physics
- 1. Functional analysis, Acad. Press (1972)
Santos L., Villela T., and Wuensche C. A., 2012, A & A,
544,
A121
Samal P. K., Saha R., Jain P., and Ralston J. P., 2008,
MNRAS,
385, 1718
Samal P. K., Saha R., Jain P., and Ralston J. P., 2009,
MNRAS,396, 511
Sarkar D., Huterer D., Copi C. J., Starkman G. D., and SchwarzD.
J., 2011, Astropart. Phys., 34, 591
Schwarz D. J., Starkman G. D., Huterer D., and Copi C. J.,
2004,
Phys. Rev. Lett., 93, 221301
Schwarz D. J., Copi C. J., Huterer D., and Starkman G. D.,
2016,
CQG, 33, 184001
Slosar A., and Seljak U., 2005, Phys. Rev. D, 70, 083002
Starck J.-L., Rassat A., and Fadili M. J., 2013, A & A, 550,
A15
Vielva P., Martinez-Gonzalez E., Barreiro R. B., Sanz J. L.,
Cayon
L., 2004, ApJ, 609, 22
Wiaux Y., Vielva P., Martinez-Gonzalez E., and VandergheynstP.,
2006, Phys. Rev. Lett., 96, 151303
Zhao W., 2014, Phys. Rev. D, 89, 023010
APPENDIX A: STABILITY OF ALIGNMENTAXES
Here we probe the stability of the even/odd multipole align-ment
axes using different foreground exclusion masks. Weused PLANCK 2015
HFI masks with varying sky fractions,that are provided along with
the second public release ofPLANCK data12. The respective sky
fractions of the masksused are 1%, 3%, 10%, 20% and 30%. The
excluded regionscorresponding to these masks are shown in Fig.
[A1].
We used these masks at their native resolution ofHEALPix Nside =
2048 on the PLANCK 2015 CommanderCMB temperature map which is also
made available at thesame resolution. The masked CMB map is then
inpaintedusing the freely available iSAP software13 (see Starck,
Ras-sat & Fadili (2013)). We used the default settings of
themrs_alm_inpainting facility of iSAP to inpaint the CMBsky.
Following the same procedure as described in the mainanalysis,
the inpainted CMB map is then downgraded toNside = 256 and
simultaneously smoothed to have a beambeam resolution of FWHM = 1◦
(degree) Gaussian beam.
The common alignment axes of even and odd multipolePEVs from
masking and inpainting 3%, 10% and 20% ofthe CMB sky are shown in
Fig. [A2]. Here we performed a
12 http://irsa.ipac.caltech.edu/data/Planck/release_2/
ancillary-data/13 http://www.cosmostat.org/software/isap/
Figure A1. Foreground exclusion masks that are applied to
testthe stability of the collective alignment axes. From red to
deep
blue they progressive exclude 1%, 3%, 10%, 20% and 30% of
the
sky. We used the freely available iSAP software to inpaint
themasked region.
qualitative analysis only. By visual inspection we see thatthe
odd multipole alignment axes are broadly stable up to10% of the sky
being masked and inpainted. However theeven multipole PEV alignment
axes steadily drift towardsgalactic plane in the varying lmax, and
move towards thepoles in the case of varying lmin. Applying
galactic cutswith 20% or more masking fraction (followed by
inpaintingthe masked sky) is found to destroy the alignment
patternsseen otherwise.
APPENDIX B: THE ISOTROPIC NULL HSIPDISTRIBUTION
Let |ẽl〉 be a random eigenvector from an isotropic
distri-bution. Since eigenvectors have no magnitude and no
sign,|ẽl〉 is equivalent to the rank-one projector Πl =
|ẽl〉〈ẽl|.Consider the distribution of x = Tr{Π†lΠl′} =
〈ẽl|ẽl′〉
2
(for all l 6= l′). Choose coordinates where the first
instance|ẽ1〉 is along the z axis, so that 〈ẽl|ẽ1〉 = cos θl. In
anisotropic ensemble the distribution of cos θl is constant overthe
range −1 ≤ cos θ ≤ 1 as shown by the solid angle mea-sure dΩl = d
cos θldφl. Averaging over all cases we can dropthe index l. For
each x = cos2 θ there are two signs of cos θ.The distribution of x
over the range 0 ≤ x ≤ 1 is then
f(x) =dN
dx= 2
dN
d cos θ|d cos θdx
| = 22
d√x
dx=
1
2√x.
The same result comes from f(x) =(2π/4π)
∫ 1−1 d cos θ δ(cos
2 θ − x), accounting for twosolutions of the delta function.
This paper has been typeset from a TEX/LATEX file prepared
by
the author.
MNRAS 000, 1–12 (2017)
http://irsa.ipac.caltech.edu/data/Planck/release_2/ancillary-data/http://irsa.ipac.caltech.edu/data/Planck/release_2/ancillary-data/http://www.cosmostat.org/software/isap/
-
Parity wise alignments of CMB multipoles 13
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=2, lmax=[7,61]
- inpainted 3%
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
8
16
24
32
40
48
56
lmax
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=[2,56], lmax=61
- inpainted 3%
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
8
16
24
32
40
48
56
lmin
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=2, lmax=[7,61]
- inpainted 10%
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
8
16
24
32
40
48
56
lmax
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=[2,56], lmax=61
- inpainted 10%
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
8
16
24
32
40
48
56
lmin
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=2, lmax=[7,61]
- inpainted 20%
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
8
16
24
32
40
48
56
lmax
-π/2
-π/3
-π/6
0
π/6
π/3
π/2
π5π/43π/27π/40π/4π/23π/4π
PLANCK15 Commander : lmin=[2,56], lmax=61
- inpainted 20%
CMB dipole
PT-PEV l=2
l=3
low-l Hemi. Power Asym.max(S-)
max(S+)
even
odd
8
16
24
32
40
48
56
lmin
Figure A2. Common alignment axes obtained after applying
galactic masks with different sky fraction and inpainting using
iSAP. Thevarying lmax and lmin cases are shown in the left and
right columns, respectively, for masking fractions of 3%, 10% and
20% of the sky.By excluding 20% or more sky fraction (and then
inpainting), the broad orientations of the common alignment axes
disappears.
MNRAS 000, 1–12 (2017)
1 Introduction2 Power tensor, Power entropy, and Alignment
Statistics3 Description of procedure and data sets3.1 Analysis
procedure3.2 Real and mock data used
4 Results4.1 Quadrupole-Octopole alignment4.2 Parity
alignments4.3 Dissecting cumulative statistics
5 ConclusionsA Stability of alignment axesB The Isotropic Null
HSIP Distribution