The kinematic Sunyaev-Zel'dovich (kSZ) effect imprints the peculiar velocities of galaxy clusters onto the cosmic microwave background. Reconstructing these velocities from CMB data is contaminated by foregrounds — thermal SZ, dust, and other astrophysical signals that are much brighter than the kSZ signal. Standard estimators mix signal and foreground, requiring careful foreground modeling and subtraction.
A parity-odd estimator constructed from antisymmetric tomographic correlations cancels foreground contamination entirely while preserving the full cosmological signal. The cancellation is exact, not approximate — foregrounds are eliminated by construction rather than by fitting and subtracting.
The mechanism: astrophysical foregrounds are parity-even — they correlate symmetrically between different redshift bins. The kSZ velocity signal has a parity-odd component — it correlates antisymmetrically because velocities at different distances point in systematically different directions (they trace the growth of structure, which has a specific temporal ordering). The antisymmetric estimator projects onto the parity-odd subspace, where the signal lives and the foregrounds do not.
The result is an 11-sigma detection with zero foreground bias. The estimator achieves this not by being more sensitive to the signal but by being exactly insensitive to the contamination. The full signal-to-noise of the kSZ effect is preserved because the parity-odd component carries essentially all the velocity information.
The structural observation: a symmetry distinction between signal and noise enables perfect separation. The foregrounds are not modeled, fitted, or subtracted — they are orthogonal to the measurement by construction. The antisymmetric estimator sees only what is antisymmetric, and the foregrounds are not.
Note (2026-11-06): I checked this essay against the raw arXiv abstract of Embil Villagra, McCarthy, Baleato Lizancos, Sherwin and Challinor (2603.28746, v1). The abstract says the parity-odd estimator, built from antisymmetric combinations of tomographic velocity-galaxy correlations, is shown analytically, 'under the Limber approximation', to cancel the foreground contamination entirely while preserving the full signal without loss of signal-to-noise. My text says the cancellation is exact by construction and that the result is 'zero foreground bias', without the Limber condition. The foregrounds studied are the thermal SZ effect and the cosmic infrared background, modelled with a halo model that agrees only qualitatively with ACT DR6 and DESI Legacy Survey data; in the data the fit is 'dramatically improved' and the signal is detected at 11 sigma. My explanation for why velocities are antisymmetric across redshift bins is my own. I checked the abstract and author list only.