You are on page 1of 7

Could Sample Variance be Responsible for the Parity-Violating Signal Seen in the

BOSS Galaxy Survey?

Oliver H. E. Philcox1, 2, ∗ and Julia Ereza3, †


1
Department of Physics, Columbia University, New York, NY 10027, USA
2
Simons Society of Fellows, Simons Foundation, New York, NY 10010, USA
3
Instituto de Astrofı́sica de Andalucı́a (CSIC), Glorieta de la Astronomı́a, E-18080 Granada, Spain
Recent works have uncovered an excess signal in the parity-odd four-point correlation function
measured from the BOSS spectroscopic galaxy survey. If physical in origin, this could indicate
evidence for new parity-breaking processes in the scalar sector, most likely from inflation. At heart,
these studies compare the observed four-point correlator to the distribution obtained from parity-
conserving mock galaxy surveys; if the simulations underestimate the covariance of the data, noise
arXiv:2401.09523v1 [astro-ph.CO] 17 Jan 2024

fluctuations may be misinterpreted as a signal. To test this, we reanalyse the BOSS CMASS & LOWZ
parity-odd dataset with the noise distribution modeled using the newly developed GLAM-Uchuu
suite of mocks. These comprise full N -body simulations that follow the evolution of 20003 dark
matter particles in a ΛCDM universe, and represent a significant upgrade compared to the formerly
used MultiDark-Patchy mocks, which were based on an alternative (non N -body) gravity solver.
We find no significant evidence for parity-violation in the BOSS dataset (with a baseline detection
significance of 1.4σ), suggesting that the former signal (>3.5σ with our data cuts) could be caused
by an underestimation of the covariance in MultiDark-Patchy. The significant differences between
results obtained with the two sets of BOSS-calibrated galaxy catalogs showcases the heightened
sensitivity of beyond-two-point analyses to the treatment of non-linear effects and indicates that
previous constraints may suffer from large systematic uncertainties.

Is the large-scale Universe invariant under a parity trans- as a function of shape (which is a 4PCF in disguise).1 Effi-
formation? According to the standard model of cosmology, cient methods for computing this exist [e.g., 17–19] (using
the answer is yes, but a slew of recent observations have mathematical tools developed in [20]), which have facili-
challenged this notion. In particular, several works have tated parity-violation measurements from the SDSS-III
obtained evidence for birefringence in the Planck cosmic BOSS galaxy survey containing O(106 ) galaxies [21, 22].
microwave background (CMB) polarization [e.g., 1–7], In essence, the analyses of [9] and [10] proceed by com-
which could indicate novel tensorial physics such as an puting the parity-odd 4PCF of BOSS data and inter-
axion-photon coupling [8]. Moreover, analyses of the spec- preting this result in terms of the expected distribution
troscopic galaxy four-point correlation function (hereafter arising from cosmic variance and shot-noise. Both works
4PCF) have found an excess scalar parity-violating signal report significant deviations from the null distribution;
[9, 10] (using methods first discussed in [11]). Whilst it [9] find a 2.9σ excess in the CMASS sample, whilst [10]
seems unlikely that these effects are related, each could obtained a 7.1σ (3.1σ) deviation in the CMASS (LOWZ)
be an intriguing hint of physics beyond the standard sample. These analyses differ in their approach and as-
paradigm; however, one must keep in mind the possibility sumptions; [9] primarily makes use of non-parametric
that the signatures could be non-physical in origin. In this rank tests, whilst [10] utilize χ2 tests, facilitated either by
Letter, we focus on the galaxy parity-violation measure- rescaling a theoretical covariance or projected the data
ments, with the hope of disentangling novel phenomena into a low-dimensional subspace; furthermore, they find
from analysis systematics. larger signals when adopting finer radial binning.
To constrain parity-violation with the distribution of Analysis differences notwithstanding, both [9] and [10]
galaxies (which transforms as a scalar), one requires statis- find an intriguing excess in the 4PCF; if physical, this
tics based on the 4PCF or beyond. This occurs since would suggest parity-violating processes at work in the
tetrahedra (which describe the vertices of a quadruplet of Universe, which could be of relevance to both the obser-
galaxies) are the first chiral polyhedron, and thus the sim- vational and theoretical communities. For the former,
plest shape for which one can distinguish reflection and we note that searches for parity-violation have also been
rotation (as discussed in [11], see also [12–16]). Roughly performed using the CMB four-point function [23, 24]
speaking, parity-violation measurements proceed by iter- (and the tensorial three-point function [25]), and there
ating over all possible tetrahedra of galaxies in a survey remains the intriguing possibility that such effects could
and counting the number that are left- and right-handed be searched for in galaxy shape and spin statistics [e.g.,

1 This can be done uniquely; one considers the sign of the triple
∗ ohep2@cantab.ac.uk product r1 · r2 × r3 , where ri is the position vector from the
† jferrer@iaa.es zeroth vertex to the i-th and we fix r1 < r2 < r3 .
2

26–31]. To date, none of these studies have reported a statistics. In addition, the generation of these catalogs
significant detection. On the theory side, a wide variety of involves specifying a significant number of parameters
works have considered parity-violating models both in the (five), leading to a remarkable degree of parameter de-
scalar and tensor sector [14, 16, 32–51]. Applied to BOSS generacy. In contrast, GLAM-Uchuu has only one free
data, [45] drew the general conclusions: (a) late-time parameter. A further important difference is that, un-
solutions (i.e. non-inflationary) are heavily suppressed like MultiDark-Patchy, GLAM-Uchuu includes the
on large-scales, (b) none of the theoretical models yet redshift-dependent evolution of galaxy clustering. Whilst
analysed are consistent with the observational data. the MultiDark-Patchy catalogs were generated using
Perhaps the least exciting explanation for the BOSS a single 2.5 h−1 Gpc box, the GLAM-Uchuu catalogs
parity-odd-excess is a mischaracterization of the 4PCF were generated by joining boxes of different redshift into
noise distribution. Both previous analyses essentially spherical shells, reproducing the clustering evolution with
analyze the following quantity: redshift (see [53] for a detailed description of this method).
Moreover, [56] recently reported some evidence of model
χ̃2 = ζodd
T
C−1 ζodd , (1)
specification errors in MultiDark-Patchy, and [53]
where ζodd is the (possibly compressed) 4PCF and C is found differences in the two-point covariances.
some covariance matrix, calibrated to a suite of realis- In the below, we perform a similar analysis to [9], but
tic galaxy catalogs, namely the MultiDark-Patchy replacing MultiDark-Patchy with the GLAM-Uchuu
suite [52]. In the Gaussian limit, and assuming C is the galaxy catalogs. Assuming that they better characterize
true covariance of ζ, χ̃2 follows a χ2 distribution (with the null ζodd distribution (which is a fair assumption, viz
the null theory ζodd = 0), thus the value of χ̃2 obtained the above discussion), this will allow us to test whether the
from BOSS can be interpreted as a parity-violation de- formerly-reported BOSS 4PCF excess could be explained
tection probability. Of course, the actual analyses are by a simple underestimate of the data’s sample variance.
somewhat more nuanced, and take into account effects
such as the finite number of mocks used to estimate C
and non-Gaussianities in the distribution of χ̃2 (via a METHODS
simple invocation of simulation-based-inference) [9, 10].
At heart, all analyses make a key assumption: that the Our observational data-set is the twelfth data release
noise distribution of ζodd from BOSS matches the noise (DR12) of the Baryon Oscillation Spectroscopic Survey
distribution of ζodd from the MultiDark-Patchy mock (BOSS), part of SDSS-III [21, 22, 57]. As in [10] (but ex-
catalogs. tending beyond [9]) we use both the low-redshift LOWZ
In this Letter, we test the above assumption. This is sample (0.2 < z < 0.4) and the high-redshift CMASS
facilitated by a new suite of BOSS mocks: the GLAM- sample (0.43 < z < 0.7), which represent two physi-
Uchuu galaxy catalogs [53]. These are N -body cosmo- cally distinct galaxy populations, and split the data into
logical simulations that follow the evolution of 20003 the North and South galactic regions (denoted ‘N’ and
dark matter particles, each having a particle mass of ‘S’ hereafter). Adopting the standard BOSS systematic
1.06×1010 h−1 M⊙ . The cosmological parameters adopted weights [58, 59] (which we caution have been validated
are Ωm,0 = 0.309, Ωb,0 = 0.0486, ΩΛ,0 = 0.691, h = 0.677, only for the two-point functions), the effective number
ns = 0.9667, and σ8 = 0.816, representing the best fitting of galaxies in our sample is 606 000, 225 000, 196 000 and
ΛCDM parameters corresponding to the Planck 2015 cos- 91 103 for CMASS-N, CMASS-S, LOWZ-N and LOWZ-S
mology [54]. In contrast, the previously used MultiDark- respectively.
Patchy catalogs, rely on an approximate model for grav- We use two sets of mock catalogs in this study:
itational evolution, rather than employing N -body codes. MultiDark-Patchy and GLAM-Uchuu. Both model
The Patchy code generates fields for dark matter den- the full BOSS footprint (split across all four chunks) with
sity and peculiar velocity on a mesh, utilizing Gaussian similar redshift-distributions and observational masks.
fluctuations and implementing the augmented Lagrangian Here, we use Nsim = 2042 MultiDark-Patchy and
perturbation theory scheme [55]. Nsim = 627 GLAM-Uchuu mocks.2 Both sets of cata-
Whilst alternative methods to N -body simulations are logs include the observational veto mask and estimated
more computationally efficient, they necessarily sacrifice fiber collision weights, emulating the observational sam-
some accuracy, implying that the MultiDark-Patchy ple. We additionally utilize random catalogs to char-
simulations fail to generate a precise matter density field acterize the unperturbed galaxy distribution (from the
compared to a full N -body simulation such as GLAM- mask and radial distribution). These have been separately
Uchuu, which fully encodes nonlinear gravitational evolu-
tion (up to resolution and baryonic effects). The feasibility
of the resulting MultiDark-Patchy galaxy catalogs in
2
a real universe is uncertain, which introduces ambiguity in 6 MultiDark-Patchy and 3 GLAM-Uchuu mocks were cor-
their covariance error estimates for the BOSS clustering rupted in data transfer and are thus dropped from the analysis.
3

CMASS-N CMASS-S LOWZ-N LOWZ-S 0.0030 GLAM-Uchuu 84.7%


1.6 MultiDark-Patchy 100.0%
0.0025 BOSS
)
GLAM) /var Patchy

1.4
0.0020
(

1.2

PDF
0.0015
1.0
0.0010
var(

0.8
0.0005
0 200 400 600 800 1000 1200 0.0000
Bin Index 5000 5500 6000
2
FIG. 1. Comparison of the parity-odd 4PCF variance es-
timated with two suites of galaxy mock catalogs: GLAM-
Uchuu (N -body) and MultiDark-Patchy (mesh, with aug- FIG. 2. Distribution of the parity-odd statistic χ̃2 (1) from
mented Lagrangian perturbation theory). We show results 2042 MultiDark-Patchy catalogs (used in previous analyses)
for the four BOSS samples, as indicated in the caption, and and 627 GLAM-Uchuu catalogs (new to this work), with the
collapse all 1288 4PCF bins into one dimension. In all samples BOSS result shown as a vertical red line. We show results
except LOWZ-S, the GLAM-Uchuu variance exceeds that of from the combined CMASS+LOWZ observational dataset,
MultiDark-Patchy. This gives rise to the differences seen with those from the four subsamples displayed in Fig. 3. The
in Figs. 2 & 3. rank-test probabilities (i.e. detection significances) are shown
in the upper right corner. Using the GLAM-Uchuu catalogs,
we find no evidence for parity-violation in BOSS, but a strong
preference when the 4PCF noise distribution is modeled with
generated for each dataset (BOSS, GLAM-Uchuu and the MultiDark-Patchy suite. Further, we find ≥3σ evidence
MultiDark-Patchy) and have size 40×, 30×, and 20× for parity-violation in 41% of the (parity-conserving) GLAM-
the data respectively. Uchuu mocks, when analyzed with MultiDark-Patchy.
Given the datasets, the 4PCFs are obtained using the Numerical constraints are given in Tab. I.
encore code [17, 60]. This computes the following quan-
tity (with a complexity quadratic in the number of galax-
ies): Computation of the 4PCFs was performed on a high-
performance computing cluster. Thanks to encore’s
Z efficient C++ and avx implementation, ≈ 90 minutes
ζℓ1 ℓ2 ℓ3 (r1 , r2 , r3 ) = dx dr̂1 dr̂2 dr̂3 Pℓ∗1 ℓ2 ℓ3 (r̂1 , r̂2 , r̂3 ) (2) was needed to process each mock (including all four data
chunks) on a 40-CPU machine, thus the full analysis
× δ(x)δ(x + r1 )δ(x + r2 )δ(x + r3 ),
required ≈ 16 000 CPU-hours.4 The full data-vector,
where δ is the galaxy overdensity,3 and Pℓ1 ℓ2 ℓ3 is an an- ζodd = {ζℓ1 ℓ2 ℓ3 (r1 , r2 , r3 ) ∀ ri , ℓi } contains Nbins = 1288
gular basis function relative to one vertex of the galaxy bins for each of the four (assumed independent) data
tetrahedron (as defined in [20]; see also the BiPoSH basis chunks.
[62]). The ri parameters give the (discretized) tetrahe- To analyze the data, we follow the methods of [9],
dron side lengths, whilst ℓi index the Fourier complement making use of a fiducial analytic covariance matrix C
of the internal angle, with odd ℓ1 + ℓ2 + ℓ3 encoding computed as described in [63]. This is estimated using
parity-violation. Our binning follows [9]; we use all ℓ various simplifying assumptions (in the Gaussian limit,
up to ℓ = 5, but drop ℓ = 5 due to residual leakage for example), thus we use it only as an approximate tool
from the window function, and use 10 radial bins with for dimensionality reduction (noting that former works
ri ∈ [20, 160)h−1 Mpc, additionally dropping any config- [9, 60, 63] have found that it well-represents the empirical
urations with internal radii (i.e. the three tetrahedron correlation structure of the 4PCF, but not the variance).
sides not specified by r1,2,3 ) below 14h−1 Mpc. On each of the data and mocks, we estimate a pseudo-χ2
statistic (hereafter χ̃2 ) via (1). If C is close to the true

3 In practice, we use the data and random catalogs to create a 4 Though the encore GPU implementation described in [17] is
Landy-Szalay-type estimator [61] to remove the window function, considerably faster, we do not use it here due to the far larger
as described in [17, 60]. number of CPUs available on a typical cluster.
4

CMASS-N CMASS-S LOWZ-N LOWZ-S


1.4% 65.4% 64.0% GLAM-Uchuu 100.0%
0.007 98.8% 97.3% 96.6% MultiDark-Patchy 95.0%
BOSS
0.006
0.005
PDF

0.004
0.003
0.002
0.001
0.000
1200 1400 1600 1800 1200 1400 1600 1200 1400 1600 1000 1200 1400 1600
2 2 2 2

FIG. 3. As Fig. 2, but assessing the probability of parity-violation in each of the four BOSS data chunks (indicated by the title).
The PDFs of GLAM-Uchuu and MultiDark-Patchy are significantly shifted in all except LOWZ-N, as expected from Fig. 1.
We find no detection of parity-violation in three of the four samples (see Tab. I), though the LOWZ-S result is anomalous,
suggesting possible underestimated covariances or observational systematics.

covariance, this will provide an optimal noise weighting in which is closely related to (1) but uses the empirical
the Gaussian limit. Given the set of χ̃2 values obtained covariance Csim = (Πζodd )(Πζodd )T sim estimated from
from a suite of (parity-conserving) mock catalogs, we con- Nsim mocks. Assuming a Gaussian distribution for Πζodd ,
struct an empirical distribution psim (χ̃2 ) and assess the (3) follows a T 2 distribution, as discussed in [65]. When
probability-to-exceed of BOSS, i.e. psim (χ̃2BOSS ). This is combining samples, we add the T 2 values, with the com-
practically implemented as a rank test, which is a simple bined theoretical distribution obtained via a convolution
form of simulation-based inference. Since we take the dis- as in [9], or from the empirical mock-based distribution,
tribution of χ̃2 directly from the catalogs, we do not need as above. Assuming that the fiducial and empirical co-
to assume Gaussianity nor that C well-approximates the variances differ, this test will provide a different weighting
true covariance; however, we must assume that psim (χ̃2 ) than the rank-based test described above, and could thus
is accurate, i.e. that the simulated noise matches the ob- differently highlight parity-violating signatures, though
servational noise. When analyzing multiple data chunks, we caution that it assumes Gaussianity.
we sum the χ̃2 values of each, weighted by the ratio of
the fiducial covariance C to the empirical one, as esti-
mated from 10% of the MultiDark-Patchy catalogs RESULTS
using a Kullback-Leibler minimization (i.e. inverse noise
weighting). Once again, we note that this cannot induce We begin by discussing the empirical 4PCF covariances
bias. obtained from the GLAM-Uchuu and MultiDark-
We additionally perform a second test for parity- Patchy suites. Whilst the correlation structures are
violation, used in both [9] and [10] (following [60, 64]). very similar (see Fig. 3 of [9]), the variances differ con-
Given the high-dimensionality of ζodd , we cannot compute siderably, as shown in Fig. 1. For all data chunks except
T LOWZ-S, the GLAM-Uchuu covariance significantly
the empirical covariance Csim = ζodd ζodd sim
directly
from the mock catalogs. Instead, we use the fiducial exceeds that of MultiDark-Patchy, with a roughly
covariance C to define a minimum-variance eigenbasis uniform rescaling found across the radial and angular
containing Neig = 250 elements, into which the data and 4PCF components. In LOWZ-S, the conclusion reverses;
mock 4PCFs are projected via some Neig × Nbins operator the MultiDark-Patchy variance is larger than that
Π.5 This is then analyzed with the following statistic: of GLAM-Uchuu. These results are not unsurprising,
given that: (a) [53] found a significant excess in the two-
T 2 = (Πζodd )T C−1
sim (Πζodd ), (3) point correlation function variances of GLAM-Uchuu
compared to MultiDark-Patchy for the CMASS-N
data-chunk, and (b) the 4PCF covariance contains contri-
5 The choice of Neig is somewhat arbitrary, but [9] found Neig = 250 butions from the 2PCF, 3PCF, 4PCF, 5PCF and 8PCF.
to yield large detection significances with MultiDark-Patchy The above observation is central to the remainder of the
catalogs whilst limiting finite-mock effects. paper. If the MultiDark-Patchy 4PCF covariance was
5

Sample GLAM-Uchuu MultiDark-Patchy 0.008 GLAM-Uchuu 96.0%


Combined 84.7% 1.4σ 100.0% >3.5σ
0.007 MultiDark-Patchy 99.4%
CMASS-N 1.4% 0.0σ 98.8% 2.5σ BOSS
CMASS-S 65.4% 0.9σ 97.3% 2.2σ 0.006
LOWZ-N 64.0% 0.9σ 96.6% 2.1σ 0.005
LOWZ-S 100.0% >3.2σ 95.0% 2.0σ

PDF
0.004
TABLE I. Probability of a detection of parity-violation from
the BOSS dataset, obtained using the rank test shown in 0.003
Figs. 2 & 3. We give the one-tailed probability of BOSS com-
pared to the distribution of GLAM-Uchuu (left columns,
0.002
new to this work) and MultiDark-Patchy (right columns, 0.001
matching previous work) simulations, and present also the
effective Gaussian significances. Where the BOSS result is 0.000
larger than all values in the empirical histogram, we give the 1000 1200 1400 1600 1800 2000
minimum detection significance. T2
FIG. 4. Alternative test for parity-violation, projecting the
underestimated, could this explain the excess parity-odd parity-odd statistic onto the 250 lowest-noise eigenmodes of
signal? a fiducial covariance matrix, then computing the distribution
In Fig. 2 we present the main result of this work: a com- of the T 2 statistic (3) using the empirical GLAM-Uchuu or
MultiDark-Patchy covariance. PDFs are obtained from
parison of the parity-violation parameter χ̃2 from BOSS,
jackknifing (histograms) and analytic Gaussian theory (lines)
GLAM-Uchuu and MultiDark-Patchy. This is anal- and we show results for all samples combined, with the num-
ogous to Fig. 4 of [9] but now includes the LOWZ dataset, bers on the top-right giving the detection probabilities relative
as well as the empirical distribution from the GLAM- to the mocks. The full (dashed) red lines show the BOSS
Uchuu catalogs. The results are striking: we find that the results when analyzed using the GLAM-Uchuu (MultiDark-
BOSS data has larger χ̃2 than every MultiDark-Patchy Patchy) mocks. We find no significant detection of parity-
catalog but only 84.7% of the GLAM-Uchuu suite. Con- violation when using GLAM-Uchuu but a detection when
using MultiDark-Patchy.
verted to one-tailed Gaussian significances (cf. Tab. I), this
corresponds to a >3.5σ excess in BOSS when analyzed
with MultiDark-Patchy (consistent with [9, 10]) or
MultiDark-Patchy and/or GLAM-Uchuu catalogs
a 1.4σ deviation using GLAM-Uchuu. If we regard
for this chunk (which has a more complex selection func-
MultiDark-Patchy as the ‘true’ noise distribution we
tion than CMASS). We note, however, that this region
thus obtain a strong detection of parity-violation from
contains the fewest galaxies, and thus the least statistical
BOSS, but this disappears if we assume GLAM-Uchuu
weight.
is a more realistic representation of the BOSS survey.
Furthermore, we can treat each GLAM-Uchuu mock Finally, we show results from the alternative Gaussian
as a ‘validation’ dataset for the MultiDark-Patchy analysis in Fig. 4. When projecting the data onto 250
analysis: in this framework, we find ≥ 3σ detections of eigenmodes and analyzing via the analytic T 2 distribu-
parity-violation in 41% of the GLAM-Uchuu mocks, tion, we find a 3.0σ excess in BOSS when analyzed with
indicating significant bias in the MultiDark-Patchy MultiDark-Patchy or a 2.4σ detection with GLAM-
pipeline. Uchuu. From the figure it is clear that the T 2 distribution
is somewhat broader than the analytic theory, which is a
To understand these results further, we show the χ̃2
signature of likelihood non-Gaussianity and leads to in-
distributions from each BOSS chunk in turn in Fig. 3,
flated detection significances. If one uses the empirical T 2
with associated numerical results given in Tab. I. In all
distribution rather than the theoretical, these detection
cases, we find significant differences between the GLAM-
significances reduce to 2.7σ (MultiDark-Patchy) and
Uchuu and MultiDark-Patchy distributions, which
2.1σ (GLAM-Uchuu), indicating no significant evidence
occurs primarily due to the different 4PCF covariances
for parity-violation in the GLAM-Uchuu analysis.
(cf. Fig. 1). In all datasets except LOWZ-S, χ̃2BOSS falls in
the center or the left of the GLAM-Uchuu distribution
but in the tails of the MultiDark-Patchy distribu-
tion, matching the findings of Fig. 2. For LOWZ-S, we DISCUSSION
find the opposite results: the GLAM-Uchuu distribu-
tion is shifted to lower χ̃2 , thus its detection significance What do the above results imply for the cosmic parity-
is larger than that obtained with MultiDark-Patchy. violation study of [9]? The precise conclusions depend sig-
This may indicate inaccuracies in the construction of the nificantly on our assumptions about the GLAM-Uchuu
6

and MultiDark-Patchy simulation suites. Two choices We thank Giovanni Cabass and Colin Hill for insightful comments.
are apparent: OHEP is a Junior Fellow of the Simons Society of Fellows and thanks
the night markets of Taipei for their pig-related delicacies. JE
1. We assume that the noise distribution of BOSS is acknowledges financial support from the Spanish MICINN funding
well-modeled by the new GLAM-Uchuu catalogs grant PGC2018-101931-B-I00. The GLAM-Uchuu catalogs will
but not by the older MultiDark-Patchy catalogs be made publicly available at http://www.skiesanduniverses.org/
(which are less realistic, as, for example, they use an Simulations/Uchuu/.
alternative method to N -body and do not account
for redshift evolution of the galaxy samples). In this
case, we find no evidence for parity-violation, with
our main analysis yielding only 1.4σ.
[1] J. R. Eskilt, Astron. Astrophys. 662, A10 (2022),
2. We assume that neither the GLAM-Uchuu nor arXiv:2201.13347 [astro-ph.CO].
MultiDark-Patchy suites accurately represent [2] J. R. Eskilt and E. Komatsu, (2022), arXiv:2205.13962
the BOSS noise distribution (seen, for example, [astro-ph.CO].
by the anomalous results with the LOWZ-S data [3] P. Diego-Palazuelos et al., Phys. Rev. Lett. 128, 091302
chunk). In this case, the significant variation in (2022), arXiv:2201.07682 [astro-ph.CO].
detection significances from two differently produced [4] E. Komatsu, Nature Rev. Phys. 4, 452 (2022),
arXiv:2202.13919 [astro-ph.CO].
galaxy catalogs (both of which are calibrated to [5] Y. Minami and E. Komatsu, Phys. Rev. Lett. 125, 221301
BOSS) indicates that our pipeline may have large (2020), arXiv:2011.11254 [astro-ph.CO].
systematic errors and cannot be used to claim any [6] P. Diego-Palazuelos et al., JCAP 01, 044 (2023),
strong detection of parity-violation. arXiv:2210.07655 [astro-ph.CO].
[7] S. E. Clark, C.-G. Kim, J. C. Hill, and B. S. Hensley,
In either case, we do not find significant evidence for Astrophys. J. 919, 53 (2021), arXiv:2105.00120 [astro-
parity-violation from the BOSS survey (using the ap- ph.GA].
proach of [9]), and thus no observational motivation for [8] M. S. Turner and L. M. Widrow, Phys. Rev. D 37, 2743
novel inflationary or late-time physics. (1988).
Although the statistical methods differ, this result also [9] O. H. E. Philcox, Phys. Rev. D 106, 063501 (2022),
arXiv:2206.04227 [astro-ph.CO].
has implications for the analysis of [10]. In this work,
[10] J. Hou, Z. Slepian, and R. N. Cahn, Mon. Not. Roy.
the main constraints come from an unprojected Gaussian Astron. Soc. 522, 5701 (2023), arXiv:2206.03625 [astro-
analysis using the theoretical covariance matrices of [63], ph.CO].
with the effective volume and shot-noise fitted to the [11] R. N. Cahn, Z. Slepian, and J. Hou, Phys. Rev. Lett.
empirical MultiDark-Patchy covariance. Although 130, 201002 (2023), arXiv:2110.12004 [astro-ph.CO].
this introduces mock-based information only through a [12] A. Lue, L.-M. Wang, and M. Kamionkowski, Phys. Rev.
pair of parameters, one still affects significant variation Lett. 83, 1506 (1999), arXiv:astro-ph/9812088.
[13] D. Jeong and M. Kamionkowski, Phys. Rev. Lett. 108,
in the constraints if the GLAM-Uchuu catalogs are 251301 (2012), arXiv:1203.0302 [astro-ph.CO].
replaced with those of MultiDark-Patchy (from Fig. 1, [14] M. Shiraishi, Phys. Rev. D 94, 083503 (2016),
given the roughly constant variance ratio). The parity- arXiv:1608.00368 [astro-ph.CO].
odd excess of [10] is also enhanced if more bins are used [15] V. Gluscevic and M. Kamionkowski, Phys. Rev. D 81,
in the data-vector; this would be interesting to investigate 123529 (2010), arXiv:1002.1308 [astro-ph.CO].
further with the GLAM-Uchuu catalogs, though we [16] T. Liu, X. Tong, Y. Wang, and Z.-Z. Xianyu, JHEP 04,
caution that larger dimensionality is often associated with 189 (2020), arXiv:1909.01819 [hep-ph].
[17] O. H. E. Philcox, Z. Slepian, J. Hou, C. Warner,
likelihood non-Gaussianity. R. N. Cahn, and D. J. Eisenstein, arXiv e-prints ,
Finally, the analysis described in this Letter has ram- arXiv:2105.08722 (2021), arXiv:2105.08722 [astro-ph.IM].
ifications for future surveys such as DESI and Euclid [18] O. H. E. Philcox and Z. Slepian, Proc. Nat. Acad. Sci.
[66, 67]. These experiments will lead to an explosion in 119, e2111366119 (2022), arXiv:2106.10278 [astro-ph.IM].
the data volume, facilitating much more precise probes of [19] O. H. E. Philcox and Z. Slepian, arXiv e-prints ,
the parity-violating sector (amongst many other topics). arXiv:2106.10278 (2021), arXiv:2106.10278 [astro-ph.IM].
Enhanced precision does not preclude bias however, and [20] R. N. Cahn and Z. Slepian, J. Phys. A 56, 325204 (2023),
arXiv:2010.14418 [astro-ph.CO].
it remains possible that model-independent analyses like [21] K. S. Dawson, D. J. Schlegel, C. P. Ahn, S. F. Anderson,
the above will be limited by the accuracy of our cata- É. Aubourg, S. Bailey, R. H. Barkhouser, J. E. Bautista,
logs, in particular their ability to reproduce the statistical A. Beifiori, et al., AJ 145, 10 (2013), arXiv:1208.0022
properties of high-dimensional observables. If these hur- [astro-ph.CO].
dles can be surmounted, however, the parity-odd dataset [22] S. Alam, M. Ata, S. Bailey, F. Beutler, D. Bizyaev,
could open doors to a wide variety of exciting inflationary J. A. Blazek, A. S. Bolton, J. R. Brownstein, A. Bur-
studies. den, C.-H. Chuang, et al., MNRAS 470, 2617 (2017),
arXiv:1607.03155 [astro-ph.CO].
7

[23] O. H. E. Philcox, (2023), arXiv:2303.12106 [astro-ph.CO]. Phys. 126, 937 (2011), arXiv:1108.0175 [astro-ph.CO].
[24] O. H. E. Philcox and M. Shiraishi, (2023), [49] N. Bartolo, S. Matarrese, M. Peloso, and M. Shiraishi,
arXiv:2308.03831 [astro-ph.CO]. JCAP 01, 027 (2015), arXiv:1411.2521 [astro-ph.CO].
[25] O. H. E. Philcox and M. Shiraishi, (2023), [50] M. Shiraishi, A. Ricciardone, and S. Saga, JCAP 11, 051
arXiv:2312.12498 [astro-ph.CO]. (2013), arXiv:1308.6769 [astro-ph.CO].
[26] O. H. E. Philcox, M. J. König, S. Alexander, and D. N. [51] K. W. Masui, U.-L. Pen, and N. Turok, Phys. Rev. Lett.
Spergel, (2023), arXiv:2309.08653 [astro-ph.CO]. 118, 221301 (2017), arXiv:1702.06552 [astro-ph.CO].
[27] W. R. Coulton, O. H. E. Philcox, and F. Villaescusa- [52] F.-S. Kitaura, S. Rodrı́guez-Torres, C.-H. Chuang,
Navarro, (2023), arXiv:2306.11782 [astro-ph.CO]. C. Zhao, F. Prada, H. Gil-Marı́n, H. Guo, G. Yepes,
[28] H. Jia, H.-M. Zhu, and U.-L. Pen, Astrophys. J. 943, 32 A. Klypin, C. G. Scóccola, et al., MNRAS 456, 4156
(2023), arXiv:2210.04168 [astro-ph.CO]. (2016), arXiv:1509.06400 [astro-ph.CO].
[29] H.-R. Yu, P. Motloch, U.-L. Pen, Y. Yu, H. Wang, H. Mo, [53] J. Ereza, F. Prada, A. Klypin, T. Ishiyama, A. Smith,
X. Yang, and Y. Jing, Phys. Rev. Lett. 124, 101302 C. M. Baugh, B. Li, C. Hernández-Aguayo, and J. Ruedas,
(2020), arXiv:1904.01029 [astro-ph.CO]. (2023), arXiv:2311.14456 [astro-ph.CO].
[30] P. Motloch, U.-L. Pen, and H.-R. Yu, Phys. Rev. D 105, [54] P. A. R. Ade et al. (Planck), Astron. Astrophys. 594, A13
083512 (2022), arXiv:2111.12590 [astro-ph.CO]. (2016), arXiv:1502.01589 [astro-ph.CO].
[31] K. Kogai, K. Akitsu, F. Schmidt, and Y. Urakawa, JCAP [55] F. S. Kitaura, G. Yepes, and F. Prada, MNRAS 439,
03, 060 (2021), arXiv:2009.05517 [astro-ph.CO]. L21 (2014), arXiv:1307.3285 [astro-ph.CO].
[32] S. Alexander and N. Yunes, Phys. Rept. 480, 1 (2009), [56] B. Yu, U. Seljak, Y. Li, and S. Singh, JCAP 2023, 057
arXiv:0907.2562 [hep-th]. (2023), arXiv:2211.16794 [astro-ph.CO].
[33] S. Alexander, A. Marciano, and D. Spergel, JCAP 04, [57] D. J. Eisenstein, D. H. Weinberg, E. Agol, H. Ai-
046 (2013), arXiv:1107.0318 [hep-th]. hara, C. Allende Prieto, S. F. Anderson, J. A. Arns,
[34] S. H. S. Alexander, Int. J. Mod. Phys. D 25, 1640013 É. Aubourg, S. Bailey, E. Balbinot, et al., AJ 142, 72
(2016), arXiv:1604.00703 [hep-th]. (2011), arXiv:1101.1529 [astro-ph.IM].
[35] C. Creque-Sarbinowski, S. Alexander, M. Kamionkowski, [58] M. Vargas-Magaña, S. Ho, A. J. Cuesta, R. O’Connell,
and O. Philcox, (2023), arXiv:2303.04815 [astro-ph.CO]. A. J. Ross, D. J. Eisenstein, W. J. Percival, J. N.
[36] N. Bartolo and G. Orlando, JCAP 07, 034 (2017), Grieb, A. G. Sánchez, et al., MNRAS 477, 1153 (2018),
arXiv:1706.04627 [astro-ph.CO]. arXiv:1610.03506 [astro-ph.CO].
[37] N. Bartolo, S. Matarrese, M. Peloso, and M. Shiraishi, [59] A. J. Ross et al. (BOSS), Mon. Not. Roy. Astron. Soc.
JCAP 07, 039 (2015), arXiv:1505.02193 [astro-ph.CO]. 464, 1168 (2017), arXiv:1607.03145 [astro-ph.CO].
[38] L. Sorbo, JCAP 06, 003 (2011), arXiv:1101.1525 [astro- [60] O. H. E. Philcox, J. Hou, and Z. Slepian, (2021),
ph.CO]. arXiv:2108.01670 [astro-ph.CO].
[39] M. Shiraishi, M. Liguori, and J. R. Fergusson, JCAP 01, [61] S. D. Landy and A. S. Szalay, ApJ 412, 64 (1993).
007 (2015), arXiv:1409.0265 [astro-ph.CO]. [62] D. A. Varshalovich, A. N. Moskalev, and V. K. Kherson-
[40] M. Shiraishi, M. Liguori, and J. R. Fergusson, JCAP 05, skii, Quantum Theory of Angular Momentum (1988).
008 (2014), arXiv:1403.4222 [astro-ph.CO]. [63] J. Hou, R. N. Cahn, O. H. E. Philcox, and Z. Slepian,
[41] S. H. S. Alexander, Phys. Lett. B 660, 444 (2008), Phys. Rev. D 106, 043515 (2022), arXiv:2108.01714 [astro-
arXiv:hep-th/0601034. ph.CO].
[42] J.-O. Gong, M. Mylova, and M. Sasaki, JHEP 10, 140 [64] R. Scoccimarro, H. M. P. Couchman, and J. A. Frieman,
(2023), arXiv:2303.05178 [hep-th]. ApJ 517, 531 (1999), arXiv:astro-ph/9808305 [astro-ph].
[43] J. Soda, H. Kodama, and M. Nozawa, JHEP 08, 067 [65] E. Sellentin and A. F. Heavens, MNRAS 464, 4658 (2017),
(2011), arXiv:1106.3228 [hep-th]. arXiv:1609.00504 [astro-ph.CO].
[44] L. Bordin and G. Cabass, JCAP 07, 014 (2020), [66] DESI Collaboration, A. Aghamousa, J. Aguilar, S. Ahlen,
arXiv:2004.00619 [astro-ph.CO]. S. Alam, L. E. Allen, C. Allende Prieto, J. Annis, S. Bai-
[45] G. Cabass, M. M. Ivanov, and O. H. E. Philcox, Phys. Rev. ley, C. Balland, et al., arXiv e-prints , arXiv:1611.00036
D 107, 023523 (2023), arXiv:2210.16320 [astro-ph.CO]. (2016), arXiv:1611.00036 [astro-ph.IM].
[46] G. Cabass, S. Jazayeri, E. Pajer, and D. Stefanyszyn, [67] R. Laureijs, J. Amiaux, S. Arduini, J. L. Auguères,
(2022), arXiv:2210.02907 [hep-th]. J. Brinchmann, R. Cole, M. Cropper, C. Dabin, L. Duvet,
[47] M. Shiraishi, JCAP 06, 015 (2012), arXiv:1202.2847 A. Ealet, et al., arXiv e-prints , arXiv:1110.3193 (2011),
[astro-ph.CO]. arXiv:1110.3193 [astro-ph.CO].
[48] M. Shiraishi, D. Nitta, and S. Yokoyama, Prog. Theor.

You might also like