Do Type 1A Supernovae prove Lambda > 0?

This paper by Michael Rowan-Robinson re-examines the analysis of supernova data that has been used to claim a positive cosmological constant.

Type 1a supernovae are white dwarf stars in binary systems which accrete enough mass to push them over the Chandrasekhar limit and explode. They are (more or less) standard candles that can be used out to relatively large distances. For this reason, supernova data can be used to measure cosmological parameters, in particular Omega_M (the matter density of the universe) and Omega_Lambda (the vacuum energy, or cosmological constant). Two groups focused on making these measurements in the 1990s: the Supernova Cosmology Project (SCP) and the High-Z Supernova Search Team (HZT). Around 1998-2000 the consensus agreement was that Omega_M = 0.3 and Omega_Lambda = 0.7.

This paper points out several potential problems with the supernova analysis. First, there may be a sample bias. The author shows that the distribution of all supernovae discovered after 1956 (excluding data points that are not well-measured) is different from the distribution used by the SCP. The SCP's data tends toward brighter supernovae at lower redshift. Since the relative dimness of more distant supernovae is what is being measured, this is an important problem. The SCP uses supernovae whose light curves were measured both at maximum brightness and 15 days later, as this allows them to adjust the absolute magnitude estimate. The author of this paper suggests that such the supernovae which were measured again 15 days later may be brighter than average. There is also a suggestion that there may be a systematic error in pre-1990 photographic photometry.

The 15-day measurement is made in order to adjust the estimate of the absolute magnitude of the supernova. It is believed that there is a correlation between the absolute magnitude of the supernova and the decay time of the light curve. There are several methods which are used to make this correction. The first is to estimate dM/dm(15) (the ratio of maximum brightness to the change in brightness over 15 days) for supernovae which have an independent distance measurement, and apply that correction to other data points. The second is to use multi-color light curve shapes (MLCS) to fit the light curves. The third is to apply a "stretch" factor to adjust for the decay times of the light curves. The author of this paper is skeptical of the way these adjustments have been done. He argues that any supernova that was only detected after the period of maximum brightness should be excluded from the analysis, because otherwise you have to use the same brightness-decay time relationship to extrapolated backwards, and this artificially reduces the scatter in the results. He also shows that the "stretch" factor method, used by the SCP, gives systematically different results from the other approaches.

The author of this paper criticizes the supernova teams for neglecting the effects of extinction within the galaxy which hosts the supernova. Both teams claim the extinction is negligible. The author believes this may not be valid because 1) the majority of supernovae take place in spiral galaxies, which have more extinction than ellipticals, and 2) star forming systems are more common as one looks backwards to z=1, so one would expect more extinction in galaxies hosting high-redshift supernovae. The HZT group give their estimates of extinction, which the author shows to be 0.22 magnitudes less than found by other methods.

Finally the author choses his own sample of supernova data, applies what he believes to be more consistent corrections for decay time and extinction, and finds that the significance of the dimming of distant supernova (which is the evidence for a positive cosmological constant) is less significant than other people have found. Making a plot of the Hubble diagram, including gravitational lensing and S-Z clusters, he finds a best-fit value of Omega_M = 0.81 +/- 0.12. His analysis does not rule out a cosmological constant, but he concludes that the evidence for it is relatively weak. He argues that since there is more motivation for an Omega_M = 1 universe, we should not be so quick to rule it out.