ABSTRACT
We use the measurement of the cosmic microwave background taken during the MAXIMA-1 flight to estimate the bispectrum of cosmological perturbations. We propose an estimator for the bispectrum that is appropriate in the flat sky approximation, apply it to the MAXIMA-1 data, and evaluate errors using bootstrap methods. We compare the estimated value with what would be expected if the sky signal were Gaussian and find that it is indeed consistent, with a chi(2) per degree of freedom of approximately unity. This measurement places constraints on models of inflation.
ABSTRACT
We show that the supernovae results, which imply that there is evidence for an accelerating universe, may be closely related to the recent discovery of redshift dependence in the fine-structure constant alpha. The link is a class of varying speed-of-light (VSL) theories that contain cosmological solutions that are similar to quintessence. During the radiation-dominated epoch, the cosmological constant Lambda is prevented from dominating the universe by the usual VSL mechanism. In the matter-dominated epoch, the varying-c effects switch off, allowing Lambda to eventually surface and lead to an accelerating universe. By the time this happens, the residual variations in c imply a changing alpha at a rate that is in agreement with observations.
ABSTRACT
We extend a previous bispectrum analysis of the cosmic microwave background temperature anisotropy, allowing for the presence of correlations between different angular scales. We find a strong non-Gaussian signal in the "interscale" components of the bispectrum: their observed values concentrate close to zero instead of displaying the scatter expected from Gaussian maps. This signal is present over the range of multipoles l = 6-18, in contrast with previous detections. We attempt to attribute this effect to galactic foreground contamination, pixelization effects, possible anomalies in the noise, documented systematic errors studied by the COBE team, and the effect of assumptions used in our Monte Carlo simulations. Within this class of systematic errors, the confidence level for rejecting Gaussianity varies between 97% and 99.8%.
