2.2. Type Ia Supernovae as Standardized Candles
SNIa have been used as extragalactic distance indicators since Kowal  first published his Hubble diagram ( = 0.6 mag) for type I SNe. We now recognize that the old type I SNe spectroscopic class is comprised of two distinct physical entities: SNIb/c which are massive stars that undergo core collapse (or in some rare cases might undergo a thermonuclear detonation in their cores) after losing their hydrogen atmospheres, and SNIa which are most likely thermonuclear explosions of white dwarfs. In the mid-1980s it was recognized that studies of the type I SN sample had been confused by these similar appearing SNe, which were henceforth classified as type Ib [59, 94, 102] and type Ic . By the late 1980s/early 1990s, a strong case was being made that the vast majority of the true type Ia SNe had strikingly similar light curve shapes [11, 46, 47, 48], spectral time series [6, 18, 28, 62], and absolute magnitudes [47, 54]. There were a small minority of clearly peculiar type Ia SNe (e.g., SN1986G , SN1991bg [19, 49], and SN1991T [19, 78]), but these could be identified and removed by their unusual spectral features. A 1992 review by Branch and Tammann  of a variety of studies in the literature concluded that the intrinsic dispersion in B and V maximum for type Ia SNe must be < 0.25 mag, making them "the best standard candles known so far."
In fact, the Branch and Tammann review indicated that the magnitude dispersion was probably even smaller, but the measurement uncertainties in the available datasets were too large to tell. The Calan/Tololo Supernova Search (CTSS), a program begun by Hamuy et al.  in 1990, took the field a dramatic step forward by obtaining a crucial set of high quality SN light curves and spectra. By targeting a magnitude range that would discover type Ia SNe in the redshift range z = 0.01 - 0.1, the CTSS was able to compare the peak magnitudes of SNe whose relative distance could be deduced from their Hubble velocities.
The CTSS observed some 25 fields (out of a total sample of 45 fields) twice a month for over three and one half years with photographic plates or film at the Cerro Tololo Inter-American Observatory (CTIO) Curtis Schmidt telescope, and then organized extensive follow-up photometry campaigns primarily on the CTIO 0.9 m telescope, and spectroscopic observation on either the CTIO 4 m or 1.5 m telescope. Toward the end of this search, Hamuy et al.  pointed out the difficulty of this comprehensive project: "Unfortunately, the appearance of a SN is not predictable. As a consequence of this we cannot schedule the followup observations a priori, and we generally have to rely on someone else's telescope time. This makes the execution of this project somewhat difficult." Despite these challenges, the search was a major success; with the cooperation of many visiting CTIO astronomers and CTIO staff, it contributed 30 new type Ia SN light curves to the pool  with an almost unprecedented control of measurement uncertainties.
As the CTSS data began to become available, several methods were presented that could select for the "most standard" subset of the type Ia standard candles, a subset which remained the dominant majority of the ever-growing sample . For example, Vaughan et al.  presented a cut on the B-V color at maximum that would select what were later called the "Branch Normal" SNIa, with an observed dispersion of less than 0.25 mag.
Phillips  found a tight correlation between the rate at which a type Ia SN's luminosity declines and its absolute magnitude, a relation which apparently applied not only to the Branch Normal type Ia SNe, but also to the peculiar type Ia SNe. Phillips plotted the absolute magnitude of the existing set of nearby SNIa, which had dense photoelectric or CCD coverage, versus the parameter m15(B), the amount the SN decreased in brightness in the B-band over the 15 days following maximum light. The sample showed a strong correlation which, if removed, dramatically improved the predictive power of SNIa. Hamuy et al.  used this empirical relation to reduce the scatter in the Hubble diagram to < 0.2 mag in V for a sample of nearly 30 SNIa from the CTSS search.
Impressed by the success of the m15(B) parameter, Riess et al.  developed the multi-color light curve shape method (MLCS), which parameterized the shape of SN light curves as a function of their absolute magnitude at maximum. This method also included a sophisticated error model and fitted observations in all colors simultaneously, allowing a color excess to be included. This color excess, which we attribute to intervening dust, enabled the extinction to be measured. Another method that has been used widely in cosmological measurements with SNIa is the "stretch" method described in Perlmutter et al. [74, 77]. This method is based on the observation that the entire range of SNIa light curves, at least in the B and V-bands, can be represented with a simple time stretching (or shrinking) of a canonical light curve. The coupled stretched B and V light curves serve as a parameterized set of light curve shapes , providing many of the benefits of the MLCS method but as a much simpler (and constrained) set. This method, as well as recent implementations of m15(B) [24, 65], also allows extinction to be directly incorporated into the SNIa distance measurements. Other methods that correct for intrinsic luminosity differences or limit the input sample by various criteria have also been proposed to increase the precision of type Ia SNe as distance indicators [9, 17, 93, 95]. While these latter techniques are not as developed as the m15(B), MLCS, and stretch methods, they all provide distances that are comparable in precision, roughly = 0.18 mag about the inverse square law, equating to a fundamental precision of SNIa distances of ~ 6% (0.12 mag), once photometric uncertainties and peculiar velocities are removed. Finally, a "poor man's" distance indicator, the snapshot method , combines information contained in one or more SN spectra with as little as one night's multi-color photometry. This method's accuracy depends critically on how much information is available.