LiteBIRD's Foreground Problem Could Become a New Window on the Milky Way

To find the faintest swirl in the cosmic microwave background—a pattern called B-mode polarization—a telescope must first stare down the Milky Way. At microwave frequencies, interstellar dust and cosmic-ray electrons glow with polarized light. For cosmologists hunting the signature of inflation, that glow is a foreground to subtract. A new preprint argues it could also be a discovery in its own right.

A satellite built to find inflation—and the foregrounds in the way

LiteBIRD, a future space mission, is designed to map the cosmic microwave background (CMB) in 15 frequency bands from 40 to 402 GHz. Its main goal is to search for primordial B-modes—a divergence-free pattern in the CMB's polarization expected from tensor perturbations generated during inflation—and to tighten constraints on the tensor-to-scalar ratio, r, which tracks the energy scale of inflation. Inflation is the accelerated expansion phase thought to have occurred in the first fractions of a second of the universe.

The obstacle is the Milky Way. Thermal dust grains, aligned with Galactic magnetic fields, emit polarized microwaves. Cosmic-ray electrons spiraling around those fields produce synchrotron radiation. Both can be far brighter than the primordial B-mode signal. The question is whether they can be modeled well enough to remove them—and whether that same modeling can reveal the interstellar medium's physics.

What the team actually simulated

The study, led by S. Vinzl and colleagues in the LiteBIRD Collaboration, is a forecast, not an observation. The researchers built simulated LiteBIRD maps of intensity and polarization using three increasingly complex models of Galactic emission. They included thermal dust, synchrotron radiation, CMB, and instrument noise. They masked the bright Galactic plane and analyzed the diffuse sky at intermediate and high latitudes.

Rather than comparing individual pixels, they computed cross-frequency angular power spectra. This measures how the brightness of emission varies across angular scales—from large patches of sky to smaller ones—and how that pattern changes between frequencies. They then fit those spectra with two standard descriptions: a modified blackbody for dust, which depends on dust temperature and a spectral index, and a power law for synchrotron emission, which has its own spectral index. They also measured how strongly dust and synchrotron emission are correlated across the sky.

A simple power law that does not survive

Planck data suggested that, at intermediate and high Galactic latitudes, the angular power spectra of polarized foregrounds could be approximated by power laws. The new simulations indicate LiteBIRD is likely to rule out that approximation. In the simplest foreground model, the power-law fit produced an average chi-square per degree of freedom of 593, a sign of severe mismatch. At 402 GHz and a large angular scale (multipole ℓ=26.5), the difference between the simulated spectrum and the power-law fit reached about 70 microkelvin-squared in CMB temperature units—roughly six orders of magnitude above the expected CMB B-mode signal at that scale.

In other words, the foregrounds are not a smooth, simple foreground. Their brightness changes with frequency in a way that varies from place to place and from scale to scale. Future analyses will need more flexible models, such as moment expansions, which account for the averaging of different emission conditions along a line of sight, across the telescope beam, and in harmonic space.

Dust temperature and a magnetic correlation

The forecast is not only about problems. LiteBIRD should pin down key foreground parameters much more tightly than Planck. In the simulations, uncertainties at large angular scales reached about 0.2 kelvin for dust temperature, 0.006 for the dust spectral index, 0.04 for the synchrotron spectral index, and 0.01 for the dust-synchrotron correlation coefficient. Planck could not freely constrain dust temperature in these polarization fits; LiteBIRD could.

That correlation matters because both dust and synchrotron emission trace the Galaxy's magnetic field. Planck found them strongly correlated on large scales but ran out of sensitivity for synchrotron measurements at smaller scales. LiteBIRD should extend that measurement, giving a sharper view of how magnetic fields, cosmic rays, and dust grains are linked.

E, B, and intensity tell different stories

One of the most striking forecasts is that LiteBIRD will detect differences in the same spectral parameters between intensity, E-mode polarization, and B-mode polarization. E modes describe curl-free polarization patterns; B modes describe divergence-free, or swirl-like, patterns. In the most complex simulated sky, the researchers found an E-versus-B difference of about 4.2 sigma for dust temperature, 11 sigma for the dust spectral index, and 2.2 sigma for the synchrotron spectral index in a large-scale multipole bin centered near ℓ=16.5. Comparing intensity with B modes, the discrepancies reached 8.3 sigma for dust temperature and 15 sigma for the dust spectral index in a large-scale bin. The analysis did not quote synchrotron T/P discrepancies because the synchrotron spectral index remains unconstrained in intensity at LiteBIRD sensitivity.

These differences arise from mixing: a single telescope beam averages over regions with different temperatures, grain properties, and magnetic field structure. That averaging distorts the spectrum of the combined light, and it does so differently in intensity and polarization. The paper argues that moment expansion can model these distortions and help interpret them as signatures of the three-dimensional structure of the Galaxy.

The caveats are real

The work is based on simulations. It assumes idealized instrument properties—perfectly known frequency bandpasses, Gaussian beams, and a fixed CMB spectrum. It does not jointly solve for CMB and foregrounds, and it excludes some emission processes expected to be small at high Galactic latitudes. The likelihood used for the fits is Gaussian, a good approximation on intermediate and small scales but not at the very largest.

The moment expansion, while promising, is not a complete solution at first order. For the most complex foreground models, the average recovered spectral parameters remain biased. Adding second-order terms helped in tests, but those tests fixed some terms to their known input values, which would not be possible with real data. The paper also notes that the true dust spectrum may be more complex than a modified blackbody.

A foreground that becomes a laboratory

LiteBIRD's primary job is to probe inflation by measuring the tensor-to-scalar ratio. To do that, it must map the Milky Way's polarized glow with improved frequency coverage and sensitivity compared with Planck. The same effort should reveal how interstellar dust grains, cosmic rays, and magnetic fields interact across the Galaxy. The contamination that cosmologists dread may become a valuable scientific product—a new window on the interstellar medium, opened by the very light that once stood in the way.