8.3. Supernova 1997ff - serendipity and careful planning

In the spirit of "extraordinary claims require extraordinary proof," the magnitude of the discovery of an accelerating universe requires an extremely careful elimination of potential contaminating astrophysical effects. A pervasive screen of "gray" dust, for example, that can dim the light while leaving little imprint on the spectral energy distribution, has been suggested as an alternative explanation for the unexpected faintness of high-z SNe Ia (Aguirre 1999). While gray dust, e.g., quasispherical grains larger then 0.1 µm, could be made to reproduce some of the observations, some measurements do disfavor a 30% visual opacity due to gray dust at the ~ 2.5 level (Riess et al. 2000). Nevertheless, both the possibility of gray dust and, in particular, the ever-existing challenge of luminosity evolution, did leave the case for an accelerating universe somewhat short of compelling by 2001. The strongest evidence so far against alternatives to an accelerating universe came from SN 1997ff.

In 1997, Ron Gilliland and Mark Phillips reobserved the Hubble Deep Field North with HST, with the goal of detecting high-z supernovae (Gilliland, Nugent and Phillips 1999). They discovered two supernovae: SN 1997fg, at a redshift of z = 0.95 was found in a late-type galaxy, while SN 1997ff was found to be hosted by an elliptical galaxy. The initial, photometrically determined redshift, placed the latter galaxy at z ~ 0.95-1.32 with considerable uncertainty. Serendipitously, the Guaranteed Time Observer (GTO) program of Thompson et al. (1999) included imaging the host of SN 1977ff with the near infrared camera, NICMOS, on board HST. One of these exposures, amazingly enough, was actually taken within hours of the discovery of the SN. Furthermore, six months after the GTO program, Dickinson (2003, in preparation; GO 7817) reobserved the field with NICMOS. The combination of all the HST and ground-based observations (in the U, B, V, I, J, H, and K bands) allowed for an improved photometric redshift determination for the host of z = 1.65 ± 0.15 (Budavári et al. 2000), making SN 1997ff the farthest known supernova to date. A probability density function based on the partial supernovae light curve in three bands also gave z 1.7, as did a tentative determination of the redshift based on the spectrum of the host taken with the Keck telescopes (see Riess et al. 2001, for a full description).

The importance of detecting a SN Ia at such a high redshift cannot be overemphasized. Even with the suggested value of the vacuum energy density of 0.7, at z 1 attractive gravity would still have dominated over the cosmic repulsion, resulting in a decelerated expansion. Consequently, SNe Ia in the redshift range z ~ 1-2 should appear brighter relative to SNe in a coasting universe. On the other hand, any effect that increases monotonically with redshift, as would be expected from dust extinction and simple evolutionary effects, would predict that the SNe Ia would be dimmer with increasing redshift. SN 1997ff was found to be brighter by ~ 1.1 mag (at the > 99.99% confidence level) than expected for an alternative source of dimming (e.g., dust or evolution) beyond z ~ 0.5. Figure 29 shows a redshift-distance relation in which the points are redshift-binned data from Perlmutter et al. (1999) and Riess et al. (1998), together with a family of curves for flat cosmological models. The transition from a decelerating to an accelerating phase occurs (if the "dark energy" is represented by a cosmological constant; see Section VIII D) at z_tr = (2 / _M)^1/3 - 1. As can be seen from the figure, the only cosmological model that is consistent with all the data (including SN 1997ff) is one in which _M 0.35, 0.65.

Figure 29. Hubble diagram of SNe Ia minus an empty (i.e., M + = 0) universe compared to various cosmological and astrophysical models. The points are redshift-binned data from Riess et al. (1998) and Perlmutter et al. (1999). The observations of SN 1997ff are inconsistent with monotonic evolutionary or dust effects that could mimic an accelerating universe. Adapted from Riess et al. (2001).

Measurements of the power spectrum of the cosmic microwave background (e.g., de Bernardiset al. 2002, Abroe et al. 2002, Hu et al. 2001, Netterfield et al. 2002) provided strong evidence for a geometrically flat (_M + 1) universe. When combined with several estimates of _M from mass to light ratios, x-ray temperature of intracluster gas, numbers, and dynamics of clusters of galaxies, all giving _M 0.2-0.3 (e.g., Carlberg et al. 1996, Bahcall, Fan and Cen 1997, Bahcall et al. 2000, Strauss and Willick 1995), the inescapable conclusion was that there is a dark energy component of 0.7 - consistent with the value found from high-redshift supernovae. Most recently, the best fit cosmological model has been determined by the Wilkinson Microwave Anisotropy Probe (WMAP). In combination with other large-scale structure work, the WMAP results imply _tot = 1.02 ± 0.02, 0.73 (Bennett et al. 2003, Spergel et al. 2003).