The Supernova Legacy Survey

Astronomical Science

Mark Sullivan1 accreting carbon-oxygen white dwarf star minated in the late 1990s when two Christophe Balland 2 approaching the Chandrasekhar mass independent surveys for distant SNe Ia limit. As the white dwarf star gains mate- made the same remarkable discovery: rial from a binary companion, the core the high-redshift SNe Ia appeared about 1 Department of Physics, University of temperature of the star increases, leading 40 % fainter – more distant – than ex- Oxford, United Kingdom to a runaway fusion of the nuclei in the pected in a flat, matter-dominated Uni- 2 Laboratoire de Physique Nucléaire et white dwarf’s interior. The kinetic energy verse (Riess et al., 1998; Perlmutter et al., des Hautes Énergies (LPNHE), Centre release from this nuclear burning – some 1999), providing astonishing evidence National de la Recherche Scientifique 10 44 J – is sufficient to dramatically un- for an accelerating Universe. When these (CNRS) – Institut Nationale de Physique bind the star. The resulting violent explo- observations were combined with an- Nucléaire et de Physique des Particules sion and shock wave appears billions of alyses of the cosmic microwave back- (IN2P3), Universités Paris VI and Paris times brighter than our Sun, comfortably ground, a consistent picture emerged of VII, France* out-shining the galaxy in which the white a spatially flat Universe dominated by a dwarf resided. dark energy responsible for ~ 70–75 % of its energy, opposing the slowing effect The accelerating Universe was one of SN Ia explosions are observed to explode of gravity and accelerating the Universe’s the most surprising discoveries of 20th with approximately the same intrinsic rate of expansion. century science. The ‘dark energy’ luminosity to within a factor of two, pre- that drives it lacks a compelling theoret- sumably due to the similarity of the This incredible discovery sparked an in- ical explanation, and has sparked an triggering white dwarf mass and, conse- tense observational effort: at first to con- intense observational effort to under- quently, the amount of nuclear fuel avail- firm the seemingly bizarre and unpre- stand its nature. Over the past five able to burn. These raw luminosities dicted result, and later to place the tight- years, the Supernova Legacy Survey can be standardised further using simple est possible observational constraints on (SNLS) has made a concerted effort to empirical corrections between their lu- dark energy, in the hope that a theoretical gather 500 distant Type Ia Supernovae minosity, light-curve shape and colour – understanding could follow. Many hun- (SNe Ia), a sample of standard candles intrinsically brighter SNe Ia typically have dreds of SNe Ia have now been discov- with the power to make a 5 % statisti- wider (slower) light curves and a bluer ered out to a redshift of 1.5 in an effort to cal measurement of the dark energy’s optical colour than their fainter counter- map the Universe’s expansion history, equation of state. The SNLS sample parts (e.g. Phillips, 1993). The combi- and alternative cosmological probes have also provides one of the most uniform nation of extreme brightness, uniformity, been developed and matured: under- sets of SNe Ia available, with a photo- and a convenient month-long duration, standing dark energy has become a key metric and spectroscopic coverage makes SNe Ia observationally attractive goal of modern science. allowing new insights into the physical as calibrateable standard candles; ob- nature of SN Ia progenitors. With the jects to which a distance can be inferred survey recently completed, we report from only a measurement of their appar- The Supernova Legacy Survey on the latest science analysis, and the ent brightness on the sky. Applying the vital role that the ESO VLT has played various calibrating relationships to SN Ia The five-year Canada-France-Hawaii in measuring these distant cosmic ex- measurements provides distance esti- Telescope (CFHT) Supernova Legacy plosions. mates precise to ~ 7%, which can be Survey (SNLS) started in mid-2003 with used via the redshift-magnitude relation the ambitious goal of discovering, con- (or Hubble Diagram) to determine cos- firming and photometrically monitoring Type Ia Supernovae as cosmological tools mological models. around 500 SNe Ia to determine the nature of dark energy. The development Type Ia Supernovae (SNe Ia) are a violent For many years following the realisation of the square-degree imager MegaCam endpoint of stellar evolution, the result of the cosmological potential of SNe Ia, on the 3.6-m CFHT, and the efficiency of the thermonuclear destruction of an finding distant events in the numbers re- with which it could survey large volumes quired for meaningful constraints was of sky, meant that SNe Ia out to z = 1 a considerable logistical and technolo- could be discovered routinely and essen- * The full SNLS core collaboration is: Pierre Astier gical challenge. Years of searching were tially on demand. The multi-band opti- (LPNHE, CNRS-IN2P3), Dave Balam (University of Victoria), Christophe Balland (LPNHE, CNRS- required to discover only a handful of cal data (Figure 1) comes from the Deep IN2P3), Stephane Basa (LAM), Ray Carlberg (Uni- SNe Ia (e.g. Perlmutter et al., 1997). The component of the CFHT Legacy Survey versity of Toronto), Alex Conley (University of field only matured with the advent of (CFHT-LS), observing each of four fields Toronto), Dominique Fouchez (CPPM), Julien Guy large-format CCD cameras capable of every three or four days during dark time (LPNHE, CNRS-IN2P3), Delphin Hardin (LPNHE, CNRS-IN2P3), Isobel M. Hook (University of efficiently scanning large areas of sky, in a rolling search for around six luna- Oxford), Andy Howell (University of Toronto), and the simultaneous development of tions per year. As optical transient events Reynald Pain (LPNHE, CNRS-IN2P3), Kathy Perrett sophisticated image processing routines are discovered, the repeated imaging (University of Toronto), Chris J. Pritchet (University and powerful computers capable of rap- automatically builds up high-quality light of Vic toria), Nicolas Regnault (LPNHE, CNRS- IN2P3), Jim Rich (CEA-Saclay), Mark Sullivan (Uni- idly analysing the volume of data pro- curves which can be used to measure versity of Oxford). duced. The substantial search effort cul- the SN peak brightnesses, light-curve

42 The Messenger 133 – September 2008 Figure 1. The point of light marked on the right image is a distant Type Ia Supernova, nearly four billion light-years away at a redshift of 0.31. This false-colour image is generated from g, r and i data taken using MegaCam at the 3.6-m Canada-France-Hawaii Telescope on Mauna Kea. Once these transient events have been located, they can be spectroscopi- cally conﬁrmed by 8-m-class telescopes such as the ESO VLT.

20 Figure 2. The light curves of more than 150 SNe Ia,

e 21 discovered and photometrically monitored by CFHT. d

u Each point represents a single MegaCam observa-

t 22 i

n tion (several SNe are observed simultaneously due

g 23 a to that instrument’s wide ﬁeld of view). The solid

M 24

i curves are light-curve template ﬁts to each SN and 25 are used to interpolate the brightness at maximum 20 light for the subsequent cosmological analyses (e.g.

e 21

d Astier et al., 2006). The three panels show data taken

u 22 t i in the i ﬁlter (upper), r (middle) and g (lower). z data n 23 g is also taken but is not shown. The multi-band data a 24

M is essential for both accurate k-corrections to the

r 25 rest-frame, and for measurement of the optical col- 26 our of the SN at maximum light. 21 e d

u 22 t i

n 23 g

a 24 M

25 g 26 27 MayJul SepNov JanMar MayJul SepNov Jan 2005 2006 Time widths, and colours required for the cos- date remains optically bright. Our ESO/ in Figure 3). When the analysis is com- mological analysis (Figure 2). In addition, VLT real-time follow-up (Basa et al., in plete, this number is expected to rise a vast database of deep and accurate prep.) has used ToO mode with FORS1 to more than 200, representing the larg- photometry yielding well-sampled multi- and FORS2 (Appenzeller et al., 1998), the est number of SNe Ia confirmed with colour light curves for all classes of opti- latter for the higher-redshift candidates a single telescope. This will be a dataset cal transients is available. where the sensitive red response be- with considerable legacy value, not only comes more critical. In general, FORS1 for studying dark energy, but also for was operated in MOS mode with the learning about the physics of the SN ex- The role of the VLT moveable slits, observing not only the plosions themselves. principal transient target, but the host gal- A critical component of any SN survey is axies of several other old variable events, VLT spectra represent a large fraction of spectroscopic follow-up of candidate the light from which has since faded. the SNLS SNe Ia spectra, and considera- events, confirming their nature and meas- This multiplexing has resulted in a large ble work has been done to produce a uring the redshifts essential for place- number of redshifts of transients as well clean identification of their types and red- ment on a Hubble Diagram. The SNLS is as spectra of the SNe Ia. shifts, necessary for their subsequent no exception. Being optically faint – cosmological use. In particular, two new fainter than 24th magnitude at a redshift Over the duration of two ESO large pro- techniques have been developed for our of one – distant SN spectroscopy re- grammes, we have followed up nearly VLT spectra. The first is a dedicated pipe- quires the light-collecting power of 8-m- 320 optically transient events, and with line that makes use of photometric in- class telescopes, such as the ESO VLT. the last six months of data still being ana- formation during the spectral extraction As with all transient events a rapid re- lysed, have confirmed 200 as SNe, and phase (Balland et al., in prep.). Distant sponse is essential while the SN Ia candi- more than 160 as SNe Ia (see examples SNe Ia are often buried in their host gal-

The Messenger 133 – September 2008 43 Astronomical Science Sullivan M., Balland C., The Supernova Legacy Survey

Rest-frame Wavelength (Å) Rest-frame Wavelength (Å) Figure 3. Example spectra of SNe Ia from the VLT/ 3000 3500 4000 4500 5000 5500 6000 6500 2500 3000 3500 4000 4500 5000 5500 6000 FORS follow-up campaign (Balland et al., in prep). 7 3 Each panel shows a different SN Ia distributed over z = 0.415 z = 0.537 , , ) 04D2fp ) 05D4ek z 0.4 to z 1. In each case the blue line is the

Å 6 Å / / 2.5 2 2 observed FORS spectrum, and the red the model m 5 m c c template ﬁt. The characteristic Ia features allow /

/ 2 s s / 4 / robust SN classiﬁcations, and in the spectra with a g g r r 1.5 e e

3 higher signal-to-noise, the chemical features can 8 8 1 1 – – 1 also be used to study the redshift evolution of SN Ia 0 0 1 2 1 ( (

properties. x x 0.5 u u

l 1 l F F 0 0 –0.5 4000 5000 6000 7000 8000 9000 4000 5000 6000 7000 8000 9000 Observed Wavelength (Å) Observed Wavelength (Å)

Rest-frame Wavelength (Å) Rest-frame Wavelength (Å) 2500 3000 3500 4000 4500 5000 5500 2500 3000 3500 4000 4500 5000

3.5 z = 0.690 1.5 z = 0.769 ) 3 05D1ke ) 04D4id Å Å / / 2 2.5 2 m m 1 c c / / s 2 s / / g g r 1.5 r e e 0.5

8 8 1 1 – 1 – 0 0 1 1 ( (

0.5 0 x x u u l 0 l F F –0.5 –0.5

4000 5000 6000 7000 8000 9000 4000 5000 6000 7000 8000 9000 Observed Wavelength (Å) Observed Wavelength (Å)

Rest-frame Wavelength (Å) Rest-frame Wavelength (Å) 2500 3000 3500 4000 4500 2 000 2500 3000 3500 4000 4500

1.2 z = 0.915 4 z = 1.031 ) ) 04D1ow 04D4dw Å Å / / 1 2 2 3 m m 0.8 c c / / s s 2 / / 0.6 g g r r e e 0.4

1 8 8 1 1 – – 0 0 0.2 1 1 0 ( (

x x 0 u u l l –1 F F –0.2 –0.4 –2 4000 5000 6000 7000 8000 9000 4000 5000 6000 7000 8000 9000 Observed Wavelength (Å) Observed Wavelength (Å) axies, with light from the continuum of Figure 4 shows an example of such a fit, The resulting clean, host-subtracted the galaxy drowning out signal from the with the spectrum of this distant SN well SN Ia spectra can be used to analyse any SN, making the task of SN identification measured despite its location in the core evolution in the strength of the SN difficult (Figure 4). The spatial profile of the of its host. chem ical features with redshift, placing host galaxy is measured from MegaCam constraints on the degree to which images in several photometric bands pro- The second technique concerns the the SNe themselves change with cosmic jected along the slit and then matched spectral identification. This uses a spec- time. This is one of the most direct meth- to the spectral profiles from FORS at the trophotometric model of SNe Ia con- ods available for probing any chang- corresponding wavelengths. This tech- structed from a sample of both nearby ing composition of the SN Ia progenitors. nique allows a precise estimate of the and distant SNe covering a wide range of host contamination at the SN position, op- epochs. Each new SN candidate spec- timally recovering the spectra of both trum is fit to this model, and the best-fit Cosmological measurements the SN and its host. If the SN is too close parameters are compared on a case-by- to its host galaxy centre for a separate case basis to the average properties of The key measurement made by the SNLS extraction, the combined spectrum is the SN Ia model sample. Differences are is the determination of the equation of extracted and fit to a two-component interpreted as the signature of peculiar state of the dark energy, w, the ratio of its model comprising a spectral model of the or non SNe Ia spectra. Although the final pressure to energy density. Dark energy SN Ia and a galaxy model drawn from identification relies on human judgement, must have a strong negative pressure to a large set of template spectra spanning this procedure limits the subjectivity usu- explain the observed cosmic acceleration the Hubble sequence. The left panel of ally entering SN classification. and hence have a negative w. The sim-

44 The Messenger 133 – September 2008 Rest-frame Wavelength (Å) Rest-frame Wavelength (Å) Figure 4. An example of 3 000 3500 4000 4500 5000 5500 3000 3500 4000 4500 5000 5500 host galaxy subtraction 5 techniques developed 9 z = 0.437 z = 0.437 for analysing VLT/FORS 06D4co + Host 06D4co spectra of SNLS SNe Ia. ) ) 4 The left panel shows the Å Å 8 / / 2 2 raw spectrum (blue) and m m 7 c c model ﬁt (red), together / / 3 s s / / 6 with the best-ﬁtting host g g r r galaxy spectrum (green). e e

8 8 5 1 1 2 Once the host galaxy is – – 0 0 Host model subtracted (right panel), 1 1 ( ( 4

the spectrum is ready x x u u 1 l l 3 for both classification F F and science analysis. 2 0 1 4 000 4500 5000 5500 6000 6500 7000 7500 8000 8500 4000 4500 5000 5500 6000 6500 7000 7500 8000 Observed Wavelength (Å) Observed Wavelength (Å) plest explanation is a Cosmological Con- photometrically calibrating the physical precisely and cannot be used a priori. stant, an intrinsic and non-evolving SN fluxes, as well as empirically control- The SN Ia method critically relies on sets property of empty space with a negative ling the various light-curve width and col- of local SNe at 0.015 < z < 0.10, where pressure equal to its energy density our relations, is therefore considerable. the effect of varying the cosmological pa- such that w = –1. Other ideas include the Furthermore, the values of the other cos- rameters is small, and which essentially broad family of quintessence models, mological parameters that enter the lumi- anchor the analysis and allow relative dis- which predict a dynamic and varying nosity distance calculation, such as the tances to the more distant events to be form of dark energy field generally with matter density or amount of curvature in measured. w ≠ –1, and phantom energy, a form the Universe, are not perfectly known. of dark energy with w < –1 that would Other complementary observations must The cosmological analysis of the first ultimately tear apart all gravitationally be used in conjunction with SNe Ia (see year SNLS dataset (SNLS1) is published bound structures in a ‘big rip’ (for a de- Figure 5) which place constraints, or pri- in Astier et al., 2006; the key results tailed review of the different possibilities ors, on the matter density (e.g, observa- are shown in Figure 5. The result, w¯ = see Copeland et al., 2006). An alterna - tions of large-scale structure) or spatial –1.023 ± 0.090 (statistical error), is con- tive considered by some theorists is that flatness (e.g., observations of the Cosmic sistent with a cosmological constant (i.e., the cosmologist’s fundamental tool, Microwave Background). Finally, the ab- w = –1) to a better than 9 % precision. General Relativity, may simply fail on very solute luminosity of a SN Ia is not known Analyses of SNLS3, the third-year sam- large scales. 2 –0.4 9 SNe Ia are used to measure cosmologi- 9,7% g 9 n –0.6 5 % + = 1 cal parameters by comparing their stand- a M X 1.8 B 6 8 % ard candle distances (derived from their ig SNLS 1st Year –0.8 B apparent brightnesses and a knowledge o 1.6 N S % –1 N of the SN Ia absolute luminosities) with B L ,7 S A 9 1 luminosity distances calculated from their O 9 s w –1.2 t 1.4 Y ( S e redshifts together with a set of cosmo- % a D 5 r S 9 –1.4 logical parameters and the equations of ) S

S 1.2 ) General Relativity. As the cosmologi- S % –1.6 D 8 S ( cal parameters are, in principle, the only 6

1 A unknowns in this analysis, constraints –1.8 B can be placed on their values with a sufﬁ- –2 cient number of SNe Ia. 0.8 00.1 0.20.3 0.40.5 0.6

M This apparently simple concept has sev- ng 0.6 ati C ler Figure 5. Cosmological constraints from SNLS1. On eral non-apparent difficulties. Measuring lo ce F s c g la e A in the left are the cosmological constraints in 7 ver- O t d at M departures in dark energy from w = –1 p ler 0.4 e ce sus 7 , (assuming w¯ = –1), and on the right the requires an extremely precise experiment: n De constraints in 7 M versus w¯ (assuming a flat Uni- a 10 % difference in w from –1 is equiv- verse and a constant equation of state w). The best- 0.2 alent to a change in SN Ia brightness at fitting result was 7 M = 0.263 ± 0.042 for a flat , CDM z = 0.6 of only 0.04 magnitudes, an ab- model, and w¯ = –1.023 ± 0.090 for a flat cosmol- 0 ogy with a constant w. The statistical error in w¯ of solute precision perhaps not routinely 00.2 0.40.6 0.811.2 9 % will improve to better than 6 % in the upcoming achieved in astronomy. The challenge of SNLS3 papers. M

The Messenger 133 – September 2008 45 Astronomical Science Sullivan M., Balland C., The Supernova Legacy Survey

Figure 6. The prelimi- not only precise SN light curves, but also 26 nary Hubble Diagram SNLS 3rd year; preliminary extremely deep image stacks from which from the SNLS3 analysis. Each black ﬁlled cir- SN Ia host galaxy information can be 24 cle represents a SN obtained (Sullivan et al., 2006). Analyses detected and monitored of these data allow the measurement of at the CFHT, and spec- galaxy properties such as stellar mass, 22 troscopically conﬁrmed star formation activity and mean age, and r

r using 8–10-m class o c B facilities such as the subsequent studies of how SN Ia proper- m 20 ESO/VLT. The blue cir- ties relate to these different variables. cles are the lower-redshift comparison sam- 18 ple which anchor the In particular, the classical view that most Hubble diagram analy- SNe Ia result from old, evolved stellar SNLS sis. populations appears incorrect. Although 16 Low-z some SNe Ia do occur in passive systems with little or no recent star formation 0.0 0.20.4 0.60.8 1.0 activity, consistent with a long delay time SN Redshift from stellar birth to SN explosion, most seem to occur in actively star-forming ple, are now nearing completion (a pre- their fluxes can be performed. This may, galaxies, suggesting a short delay time liminary Hubble Diagram can be found in however, be an over-simplification, and (Figure 7; see also Sullivan et al., 2006). Figure 6). With a sample size three times would ignore the considerable uncer- These prompt and delayed SNe Ia pos- larger than SNLS1, the analysis provides tainty that exists over the underlying phys- sess different light curves: prompt SNe Ia not only a step forward in the statisti - ics governing SN Ia explosions. For ex- appear brighter with broader light curves, cal precision, but in the understanding of ample, the configuration of the progenitor while the delayed component SNe are SNe Ia as astrophysical events, and will system prior to explosion is very uncer- fainter with fast light curves. By virtue lead to a better than 6 % constraint on tain. Both single degenerate systems of the evolving mix of quiescent and star- w¯ . A ~ 5 % measurement of w¯ is ex- (a white dwarf star together with a main- forming galaxies with redshift, a subtle pected from the final SNLS sample, as sequence or red-giant companion) or redshift evolution in SN Ia population well as the first detailed measurements double degenerate systems (two white demographics is predicted (Figure 7) and of the degree to which w changes out to dwarf stars) could theoretically result in a has now been observed in SNLS data z = 1. SN Ia explosion. There are also open (Howell et al., 2007). Although such shifts questions as to how the metallicity or age do not affect cosmological conclusions if of the progenitor star may influence the the SN Ia calibrating relationships remain Supernova astrophysics observed properties of the SN explosion, universally applicable, further analysis of leading to possible biases as the dem- the SNLS dataset is required to test this Taken at face value, the simplicity of the ographics of the SN Ia population shifts assumption. SN Ia technique – comparing the relative slightly with look-back time. brightnesses of events at different dis- The most straightforward interpretation tances – suggests the ultimate accuracy The SNLS has provided some new in- of this environmentally-dependent SN Ia of their use may only be limited by the sight into these issues. The homogene- rate is a wide range of delay times, but extent to which relative calibrations of ous nature of the CFHT-LS data pro vides the exact physical implications are un-

r Passive Star-forming Burst a e y

r e p

s s –12 a 10 m

r a l l e t s

Figure 7. SNLS has provided evidence that SN Ia t i

n properties are dependent on the age of the progeni- u

r tor system. The SN Ia rate per unit stellar mass ver- e

p sus host star-formation rate per unit stellar mass –13 e 10 t (Sullivan et al., 2006). Blue points refer to SNLS data, a

R orange points to local estimates. The red area indi-

N cates the SNLS SN Ia rate in passive, or zero star- S SNLS formation rate, galaxies. A SN Ia population with a Mannucci et al. wide range of delay times is supported: simplistically, a delayed population in quiescent galaxies together –12–11 –10–9 –8 with a prompt population whose rate correlates with LOG Specific SFR (M yr –1 per unit stellar mass) recent star formation.

46 The Messenger 133 – September 2008 clear. The SNLS relation between SN Ia ing possibility is the existence of more 0.01–0.015 magnitudes; future, planned rate and star-formation rate (Figure 8) than one progenitor mechanism (e.g. experiments will require a calibration implies that around 1% of all white dwarfs Mannucci et al., 2006). The key to making of better than 1% in both the distant and end their lives as SNe Ia (Pritchet et al., progress is to pinpoint any fundamental nearby sample – as much effort is re- 2008), independent of their initial mass. environmental differences between de- quired for the local sample as was As the single degenerate model typically layed and prompt events. For example, needed for the higher-redshift SNLS has lower conversion efficiencies at lower metallicity is predicted to affect SN Ia lu- dataset. The second challenge is under- masses, this suggests that some other minosities and rates. Timmes et al. (2003) standing the limitations of SN Ia by in - mechanism is responsible for the produc- predict that higher metallicity progenitors vestigating their astrophysical properties tion of at least some SNe Ia. However, produce white dwarfs richer in 22Ne, with and controlling any subtle evolutionary the precise implication for the progenitor an increased neutronisation during nu- effects. SNLS is providing the essential systems must await the construction clear burning producing stable 58Ni at the stepping stone for both efforts. of a more detailed delay-time distribution. expense of the 56Ni that powers the light curves. As a result, a ≤ 25 % difference in luminosity is expected between high References Future perspectives and low metallicity environments. Recent Appenzeller, I., et al. 1998, The Messenger, 94, 1 observational results hint at these effects Astier, P., et al. 2006, A&A, 447, 31 The upcoming analysis of the SNLS third (Gallagher et al., 2008), but urgently need Copeland, E. J., Sami, M. & Tsujikawa, S. 2006, year dataset will provide the most precise confirmation with detailed spectroscopy International Journal of Modern Physics D, measurement yet of the nature of the of the host galaxies of larger, complete 15, 1753 Gallagher, J. S., et al. 2008, arXiv:0805.4360 dark energy driving the accelerating cos- and homogeneous samples, such as the Howell, D. A., et al. 2007, ApJL, 667, L37 mic expansion. While these cosmologi- SNLS. Such a programme will soon com- Mannucci, F., Della Valle, M. & Panagia, N. 2006, cal results will inevitably draw most atten- mence using the VLT. MNRAS, 370, 773 tion, SNLS has also allowed new insights Perlmutter, S., et al. 1997, ApJ, 483, 565 Perlmutter, S., et al. 1999, ApJ, 517, 565 into the astrophysics governing SN Ia As with any experiment, the final preci- Phillips, M. M. 1993, ApJL, 413, L105 progenitors and their explosions. To date, sion of the SNLS results is governed by Pritchet, C. J., Howell, D. A. & Sullivan, M. 2008, no effect has been uncovered that chal- both statistical and systematic uncer- arXiv:0806.3729 lenges the conclusions that have been tainties. As more SNe Ia are used in the Riess, A. G., et al. 1998, AJ, 116, 1009 Sullivan, M., et al. 2006, ApJ, 648, 868 drawn from using SNe Ia in cosmological analysis and the statistical error de- Timmes, F. X., Brown, E. F. & Truran, J. W. 2003, applications, but some open questions creases, the contribution of systematic ApJL, 590, L83 remain. Why are the brightest SNe Ia as- errors becomes increasingly important. sociated with short delay times and the Ultimately, the challenge of controlling youngest galaxies? How well do SNe Ia systematics in SN cosmology is two-fold. from different environments inter-calibrate The first is photometric calibration. The in a cosmological analysis? A tantalis- SNLS calibration is accurate to about

Redshift 0.0 0.5 1.01.5 2.02.5 3.0

−3.8 SNLS Other )

3 Combined rate evolution – −4.0 c p M

1 –

r −4.2 y

s t

n Fainter Brighter e v −4.4 15 e (

) 10 e r t

e 5

a Young ‘B’ SNe Ia b

R 0 m −4.6 (star-forming galaxies) u a I N

N 5 Figure 8. The volumetric SN Ia rate redshift evolution S

( derived from the SNLS SN Ia properties. The relative −4.8

G 0 mix of the two components will evolve with redshift 0.70.8 0.91.0 1.11.2 O (main panel). As SN light curve width (‘stretch’) cor- L Stretch relates with star-formation activity in the host (inset −5.0 Old ‘A’ SNe Ia histograms), a mild evolution in mean SN Ia light- (passive galaxies) curve width with redshift is implied as the relative mix of the two components changes. This effect has been detected in SNLS data (Howell et al., 2007), 13 11 97 54 3 2 and must be carefully controlled in Hubble diagram Age of Universe (billions of years) analyses.

The Messenger 133 – September 2008 47