Abstract
We present a method for comparing and combining maps with different resolutions and beam shapes, and apply it to the Saskatoon, QMAP, and COBE-DMR data sets. Although the Saskatoon and QMAP maps detect signals at the 21$σ$ and 40$σ$ levels, respectively, their difference is consistent with pure noise, placing strong limits on possible systematic errors. In particular, we obtain quantitative upper limits on relative calibration and pointing errors. Splitting the combined data by frequency shows similar consistency between the Ka and Q bands, placing limits on foreground contamination. The visual agreement between the maps is equally striking. Our combined QMAP+Saskatoon map, nicknamed QMASK, is publicly available on the web together with its 6495$×$6495 noise covariance matrix. This thoroughly tested data set covers a large enough area (648 square degrees—currently the largest degree-scale map available) to allow a statistical comparison with COBE-DMR, showing good agreement.
Introduction
Published in 2001, this research represents a landmark effort in the "Golden Age" of Cosmic Microwave Background (CMB) cosmology. During this era, multiple experiments were mapping the early universe, but comparing their results was notoriously difficult due to differences in resolution, beam shapes, and noise properties. This paper provides a rigorous framework for merging these disparate datasets into a unified view.
Methodology: Beyond Power Spectra
In the late 1990s, researchers typically compared experiments using "power spectra"—statistical averages of fluctuations. However, this approach discards "spatial phase information"—the actual locations of hot and cold spots. To hunt for systematic errors like pointing offsets or calibration drifts, a direct map-to-map comparison is required.
The authors developed a novel matrix-based deconvolution method. Unlike standard approaches that "smooth" high-resolution data to match lower-resolution maps (which destroys information), they used a "Virtual Map" technique. By treating the problem as a combination of real data and a "virtual map" with infinite noise, they could mathematically "unsmooth" the data into a common resolution without making up information. This was implemented using a heavy linear algebra approach ($y = Ax + n$), where all experimental complexities (beam shapes, correlated noise) are encapsulated in a single, massive noise covariance matrix ($Σ$).
Datasets and the QMASK Map
The study integrated three major datasets:
- Saskatoon: Known for its complex, "caterpillar-like" weight functions and 0.68° resolution.
- QMAP: Comprising two flights and two frequency bands (Ka at 0.89° and Q at 0.68°).
- COBE/DMR: The low-resolution (7.08°) gold standard for large-scale features.
The resulting combined map, dubbed QMASK, covers 648 square degrees, making it the largest degree-scale CMB map available at the time of publication.
Key Insights and Results
The most significant finding was the remarkable consistency between independent experiments. The difference between the Saskatoon (21$σ$ signal) and QMAP (40$σ$ signal) maps was found to be consistent with pure detector noise. This result strongly validates that the observed features are real physical signals from the early universe rather than experimental artifacts.
Calibration and Foregrounds
The analysis placed a quantitative limit on relative calibration errors between experiments at less than 20% (at 3$σ$). Furthermore, by comparing the Ka and Q bands of QMAP, the authors found a spectral index ($β$) near zero, as expected for thermal CMB radiation. This effectively ruled out significant contamination from galactic dust or synchrotron radiation in the observed region.
The "Cold Spot"
The maps revealed an unusually cold spot at RA=3h20m, DEC=84°55'. At -4.1$σ$, this feature is difficult to explain via standard Gaussian fluctuations. While detected by both QMAP and Saskatoon, its frequency dependence was slightly anomalous, not perfectly matching the Sunyaev-Zeldovich effect or standard foregrounds.
Limitations and Future Outlook
While groundbreaking, the study faced computational hurdles. Processing the 6495 $×$ 6495 noise covariance matrix was a significant undertaking in 2001. Scaling these methods to the millions of pixels provided by modern satellites like Planck requires even more optimized algorithms.
The authors pointed toward the then-upcoming MAP satellite (later renamed WMAP) as the definitive tool to determine if features like the "Cold Spot" were rare statistical flukes or indicators of new physics. By releasing the QMASK map and its covariance matrix to the public, they provided a high-fidelity benchmark that helped pave the way for the precision cosmology era.
Reference: Xu, Y., Tegmark, M., et al. (2001). Comparing and combining the Saskatoon, QMAP and COBE CMB maps. Phys. Rev. D.
Summary updated