Simons Array sets new small-scale limits on circular polarization in the cosmic microwave background

The POLARBEAR Collaboration’s preprint reports measurements made with the Polarbear-2b receiver of the Simons Array in Chile’s Atacama Desert. Observing at a nominal frequency of 90 gigahertz, the team searched two regions away from the Milky Way’s bright plane. Each region received about two and a half months of observations, with a combined effective sky area of roughly 12 square degrees.

The researchers found no evidence for circular polarization. Their main advance is extending measurements of its angular power spectrum—a description of how strongly a signal varies across different-sized sky features—to scales between 5.4 and 36 arcminutes. That is a search for spatial variations, rather than a measurement of one average polarization across the whole sky.

Why look for this signal? The standard cosmological model predicts measurable linear polarization in the microwave background, but no appreciable circular polarization. Several proposed processes could nevertheless generate it. Magnetized plasma can convert existing linear polarization into circular polarization; possible settings include galaxy clusters and remnants of the first stars’ supernovae. Other proposals involve interactions with the cosmic neutrino background or physics beyond the Standard Model. A non-detection at this sensitivity does not distinguish among these possibilities.

Making the signal measurable

The instrument’s key addition was a stationary quarter-wave plate placed in front of a continuously rotating half-wave plate. These optical elements change polarization in controlled ways. The quarter-wave plate converts incoming circular polarization into a linear-polarization component that the receiver can measure. The rotating plate then shifts that component into a regularly varying detector signal, helping separate it from slowly changing atmospheric brightness.

Conversion by the quarter-wave plate was close to 99% efficient, but the complete measurement chain was less efficient. After accounting for the rest of the system, the calibrated circular-polarization efficiency was about 53%. That distinction matters when translating detector measurements into limits on the sky.

The team also had to account for ordinary brightness and linear polarization leaking into the measurement channel. Jupiter observations, treating the planet as approximately unpolarized at this frequency, measured total intensity leakage of about 1.8%. Simulated polarized skies estimated contamination from linear polarization, including the effects of observing each field at different orientations.

Earth’s atmosphere presents another complication: oxygen can emit circularly polarized radiation because of its interaction with Earth’s magnetic field. The researchers filtered slowly varying signals and patterns that repeat with telescope pointing. They then passed artificial sky signals through the analysis to quantify how much genuine celestial structure those same filters would remove.

Checking a noisy measurement

Rather than deriving the spectrum from a single map, the analysis cross-correlated maps built from separate observations. With independent instrumental noise, this avoids the positive noise contribution that would appear when correlating a map with itself. Uncertainty estimates came from repeatedly resampling the resulting map-pair spectra.

The science spectra remained hidden while the team refined the analysis using consistency checks. These tests compared data split by detector properties, weather, observing time and other conditions. Eight observations used for initial development were excluded from the final science dataset. Both sky fields passed the final checks before the science spectra were revealed.

The combined 95% confidence upper limits run from 21.8 square microkelvin at the band centered on multipole 350 to 3,490 square microkelvin at multipole 1950. These values describe the conventionally weighted circular-polarization power spectrum, not temperature or polarization amplitudes directly. Higher multipoles correspond to smaller angular features, and the limits fluctuate between bands because the measurements are noise-dominated.

What the limits leave open

The study reaches a maximum multipole of 2000, almost three times the previous circular-polarization power-spectrum reach of about 700. Earlier Very Large Array observations did probe smaller angular scales, but constrained fractional polarization in individual fields rather than this angular power spectrum. The new result therefore extends a particular kind of measurement; it is not the first circular-polarization search at every small scale.

Instrument noise is the main limitation. Representative predicted signals shown in the paper lie roughly six orders of magnitude below these limits in power. The single observing band also prevents separation of potential sources through their frequency dependence. Residual diffuse foreground emission is included in the constrained sky signal, rather than independently identified.

The result is thus a conservative upper bound, not evidence that circular polarization is absent. Alongside that bound, the preprint demonstrates an observing and analysis method that other suitably equipped telescopes could adapt.