🔭 astrophysics

Local bandpower diagnostics for the hemispherical power asymmetry in CMB E-mode polarization

This paper presents local bandpower diagnostics applied to Planck PR3 Commander E-mode polarization data, revealing a statistically significant hemispherical power asymmetry with a dipole direction consistent between independent half-mission maps and the temperature anomaly axis, thereby supporting a cosmic origin for the signal over noise or systematics.

Original authors: Robert A Lynch

Published 2026-08-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Robert A Lynch

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, glowing balloon that inflated rapidly in its earliest moments. This balloon is filled with the Cosmic Microwave Background (CMB), which is essentially the oldest light in existence, a faint afterglow from when the universe was just a baby. For decades, scientists have assumed this light is perfectly smooth and the same in every direction, like a perfectly uniform layer of frosting on a cake. This idea, called "statistical isotropy," is a cornerstone of our understanding of how the universe began. However, when astronomers look closely at this ancient light, they sometimes spot strange bumps and wiggles that don't quite fit the perfect cake model. One of the most persistent oddities is the "Hemispherical Power Asymmetry," where one half of the sky seems to have more energy (or "power") than the other half. It's as if the frosting is thicker on the left side of the cake than the right, and nobody is quite sure if that's a real feature of the universe or just a smudge on the camera lens.

To solve this mystery, scientists usually look at the temperature of this light, but they are now trying to look at its "polarization." Think of polarization like the direction in which light waves wiggle. While the temperature tells us how hot the light is, polarization tells us how it vibrates. If the weird "thick frosting" on one side of the sky is a real cosmic feature, it should show up in the polarization data too, not just the temperature. The problem is that measuring this polarization is incredibly hard because the signal is tiny and easily drowned out by noise, like trying to hear a whisper in a hurricane. This paper is a clever attempt to build a new kind of "ear" to listen for that whisper, specifically to figure out if the weird asymmetry is a real signal from the universe or just a random fluctuation of the noise.

The researcher, Robert A. Lynch, developed a new way to test this using data from the Planck satellite. Instead of looking at the whole sky at once, they chopped the sky into hundreds of overlapping circular patches, like looking at a mosaic through a magnifying glass. In each patch, they measured the power of the polarization. But here's the tricky part: because they couldn't see the whole sky (some parts were blocked by our own galaxy), the measurements got messy, with different types of signals leaking into each other. To fix this, the team used a mathematical "deconvolution" technique, which is like using a sophisticated filter to untangle the messy signals and separate the true cosmic wiggles from the noise.

The real genius of their method is how they checked if the signal was real. They split the satellite's data into two independent halves, HM1 and HM2. These two halves look at the exact same patch of sky but were taken at different times, meaning they have different noise patterns. If the "thick frosting" asymmetry is a real cosmic feature, both halves should point to the same direction, even if the noise tries to push them apart. If the asymmetry is just a random glitch or noise, the two halves would point in completely different directions.

When they applied this test to the polarization data, they found something fascinating. In the "Commander" data processing method, the two halves of the data pointed in directions that were only 8.7 degrees apart. This is a very tight agreement. To see if this was special, they ran 300 simulations of a perfectly normal, random universe. In none of those 300 simulations did the two halves agree this closely; the chance of this happening by random luck is less than 0.33%. Furthermore, when they looked at other types of polarization signals (which shouldn't show this asymmetry if it's real), the agreement vanished, suggesting the signal is specific to the type of light they expected.

However, the author is very careful not to claim they have found a new law of physics. They explain that while the two halves agree so well that the signal must be coming from the cosmic light itself and not just random noise, they cannot yet prove it isn't just a rare, lucky fluke of the universe's natural randomness. It's like hearing a whisper in two different rooms and realizing the voice is definitely real, but not knowing if the speaker is actually saying something important or just humming a tune. The signal is consistent with the "thick frosting" asymmetry seen in temperature data, but the current data isn't sensitive enough to say for sure. The paper concludes that this new diagnostic tool works beautifully and that future, more sensitive satellites like LiteBIRD will be needed to finally confirm whether this cosmic asymmetry is a real feature of our universe or just a cosmic coincidence.

Technical Summary: Local Bandpower Diagnostics for the Hemispherical Power Asymmetry in CMB E-mode Polarization

Problem Statement
The hemispherical power asymmetry (HPA) is a persistent large-scale anomaly in the Cosmic Microwave Background (CMB) temperature field, characterized by a dipolar modulation where one hemisphere exhibits systematically more power than the other. While confirmed in temperature data with statistical significance exceeding $3\sigma$ , the origin of this anomaly remains debated. The CMB E-mode polarization offers an independent probe of the same primordial fluctuations; however, at the sensitivity of the Planck satellite, the polarized sky is noise-dominated. A critical challenge in analyzing polarization on incomplete skies is E-to-B leakage, where the masking of the sky causes power to leak between E-mode and B-mode components, potentially mimicking or obscuring cosmological signals. Previous analyses have struggled to distinguish whether candidate signals in E-mode polarization originate from the CMB, instrumental noise, or systematic artifacts.

Methodology
This paper introduces a set of "local bandpower diagnostics" designed to test whether a candidate dipolar modulation in E-mode polarization data is intrinsic to the CMB rather than driven by noise or systematics. The methodology relies on the following components:

Local Bandpower Estimation (LBE): Instead of using a global E-mode map (which requires inpainting to handle masking), the author applies the standard pseudo- $C_\ell$ framework (implemented via NaMaster) directly to overlapping circular patches ( $R=15^\circ$ ) on the masked sky. This approach computes pseudo-power spectra from the Stokes $Q$ and $U$ maps within each patch and uses an analytically computed coupling matrix to deconvolve the mask-induced mixing, recovering true E-mode power without constructing an intermediate global map.
Half-Mission Directional Consistency Check: The core diagnostic involves fitting a dipole independently to the bandpower maps of the two Planck half-mission splits (HM1 and HM2). These splits share the same CMB sky but possess independent noise realizations. The test measures the angular separation ( $\theta$ $θ$ ) between the dipole directions recovered from HM1 and HM2.
- Logic: If a signal is noise-driven or systematic, the directions in the two splits should be uncorrelated. If the signal originates in the shared CMB (whether a real modulation or a chance fluctuation), the directions should align more tightly than expected from isotropic simulations.
- Limitation: The test cannot distinguish between a real primordial modulation and a chance dipolar fluctuation in the shared CMB component; it only diagnoses whether the signal resides in the CMB rather than the noise.
Pipeline Selection and Validation: The author compares four Planck component-separation pipelines (Commander, SEVEM, SMICA, NILC) using signal injection tests. They inject known dipolar modulations into simulations to determine which pipeline is most responsive to the signal. Commander is selected as the fiducial pipeline based on its proportional response to injected modulations.
Spectral and Cross-Spectrum Checks: The analysis separates EE, BB, and EB spectra to verify that coherence is confined to EE (as expected for scalar modulation). Additionally, a cross-spectrum (HM1 $\times$ HM2) check is performed, which cancels independent noise terms by construction, providing a noise-bias-free directional estimate to corroborate the auto-spectrum results.

Key Contributions

Dense Spatial Sampling: The application of pseudo- $C_\ell$ deconvolution to 600 overlapping patches, enabling dense spatial sampling of the sky, which facilitates the half-mission directional consistency check.
Leakage Correction via Deconvolution: Demonstrating that standard pseudo- $C_\ell$ deconvolution is essential for E-mode HPA analysis. The author shows that without this correction (using a Local Variance Estimator on a global map), recovered directions are dominated by mask geometry and leakage, yielding no coherence.
Diagnostic Framework: Establishing a diagnostic protocol that uses directional agreement between independent noise splits to validate the CMB origin of a signal, even when the signal-to-noise ratio is too low to detect the modulation amplitude directly.

Results

Temperature Validation: Applied to Planck PR3 Commander temperature data ( $\ell=15-40$ ), the pipeline recovers the known HPA direction with an angular separation of $0.1^\circ$ between half-missions, validating the method on a known signal.
E-mode Polarization (Commander):
- With deconvolution, the Commander pipeline yields dipole directions for HM1 and HM2 separated by $8.7^\circ$ .
- This separation is smaller than in any of the 300 isotropic FFP10 simulations, resulting in a $p$ -value of $<0.33\%$ .
- The recovered directions lie near the temperature HPA axis ( $\sim 11.6^\circ$ away).
- Without deconvolution (LVE), the directions are separated by $101.6^\circ$ ( $p=75.0\%$ ), confirming the necessity of leakage correction.
Spectral Coherence: The directional coherence is present only in the EE spectrum. The BB and EB spectra show no directional agreement between half-missions ( $p > 76\%$ ), consistent with theoretical expectations for scalar modulation.
Pipeline Gradient: The significance of the directional consistency varies across pipelines: Commander ( $p < 0.33\%$ ), SEVEM ( $3.0\%$ ), SMICA ( $5.7\%$ ), and NILC ( $25.3\%$ ). This gradient correlates with the pipelines' responsiveness to injected signals, with NILC showing the weakest response, consistent with its known excess large-scale noise.
Cross-Spectrum: The cross-spectrum check, free of noise bias, corroborates the directional coherence found in the auto-spectra, though the specific tightness of the Commander auto-spectrum result may be partially influenced by noise subtraction effects.

Significance and Claims
The author explicitly states that they do not interpret these results as a detection of a primordial modulation. The paper emphasizes the modest nature of the claim:

Diagnostic, Not Probative: The directional consistency check confirms that the candidate signal originates in the shared CMB component rather than in independent noise or systematics. However, it cannot distinguish between a real physical modulation and a chance dipolar fluctuation (cosmic variance) in the specific CMB realization observed.
Noise Limitation: Signal injection tests reveal that at the physical amplitude of the HPA ( $A \approx 0.07$ ), the test is noise-limited. A 7% modulation is indistinguishable from the null hypothesis in individual realisations at Planck's sensitivity.
Conditional Significance: The quoted $p$ -values are conditional on the accuracy of the Planck Full Focal Plane 10 (FFP10) noise model. If the simulations underestimate the noise, the significance could be overstated.
Future Outlook: The convergence of diagnostics (half-mission consistency, alignment with temperature HPA, cross-spectrum corroboration, and EE-only coherence) suggests a CMB-level signal worth further investigation. The author concludes that future experiments with higher polarization sensitivity, specifically LiteBIRD, are required to move E-mode measurements into a regime where the directional consistency check can definitively distinguish between a real modulation and a chance fluctuation.

More like this