Overview of the Nearby Supernova Factory
G. Aldering
a
, G. Adam
b
, P. Antilogus
c
, P. Astier
d
, R. Bacon
b
, S. Bongard
c
, C. Bonnaud
b
,
Y. Copin
c
, D. Hardin
d
, F. Henault
b
, D. A. Howell
a
, J.-P. Lemonnier
b
, J.-M. Levy
d
,
S. Loken
a
, P. Nugent
a
, R. Pain
d
, A. Pecontal
b
, E. Pecontal
b
, S. Perlmutter
a
,
R. Quimby
a
, K. Schahmaneche
d
, G. Smadja
c
, and W.M. Wood-Vasey
a
,
the Nearby Supernova Factory collaboration
a
Lawrence Berkeley National Laboratory, Berkeley CA, USA
b
Centre de Recherche Astronomique, Universite Lyon I and Ecole Normale Superieure, Lyon, France
c
Institut de Physique Nucleaire, Universite Lyon I, Lyon, France
d
Laboratoire de Physique Nucleaire et de Hautes Energies, Universites Paris VI and VII, Paris, France
ABSTRACT
The Nearby Supernova Factory (SNfactory) is an international experiment designed to lay the foundation for the next
generation of cosmology experiments (such as CFHTLS, wP, SNAP and LSST) which will measure the expansion history
of the Universe using Type Ia supernovae. The SNfactory will discover and obtain frequent lightcurve spectrophotome-
try covering 3200-10000
˚
A for roughly 300 Type Ia supernovae at the low-redshift end of the smooth Hubble flow. The
quantity, quality, breadth of galactic environments, and homogeneous nature of the SNfactory dataset will make it the
premier source of calibration for the Type Ia supernova width-brightness relation and the intrinsic supernova colors used
for K-correction and correction for extinction by host-galaxy dust. This dataset will also allow an extensive investiga-
tion of additional parameters which possibly influence the quality of Type Ia supernovae as cosmological probes. The
SNfactory search capabilities and follow-up instrumentation include wide-field CCD imagers on two 1.2-m telescopes
(via collaboration with the Near Earth Asteroid Tracking team at JPL and the QUEST team at Yale), and a two-channel
integral-field-unit optical spectrograph/imager being fabricated for the University of Hawaii 2.2-m telescope. In addition
to ground-based follow-up, UV spectra for a subsample of these supernovae will be obtained with HST. The pipeline to
obtain, transfer via wireless and standard internet, and automatically process the search images is in operation. Software
and hardware development is now underway to enable the execution of follow-up spectroscopy of supernova candidates
at the Hawaii 2.2-m telescope via automated remote control of the telescope and the IFU spectrograph/imager.
Keywords: supernova, survey, cosmology, integral-field-unit, spectrograph
1. PROBING DARK ENERGY WITH SUPERNOVAE
A coherent view of the universe is emerging in which a mysterious form of “dark energy” accounts for about 2/3 of the total
energy density in the Universe. Direct evidence for this radical conclusion comes from distance measurements of Type Ia
supernovae (SNe Ia; see Fig. 1) which indicate the expansion of the Universe is not slowing down as would be expected
in a Universe filled with only matter and radiation.
1, 2
Further support for this result has come from recent measurements
of the CMB indicating a flat universe,
3
combined with determinations of Ω
M
∼ 0.3 from structure formation.
SNe Ia remain the most mature cosmological distance indicator, and therefore, offer the best current means of exper-
imentally probing the properties of the dark energy. Their cosmological use was developed in the early 1990’s, paving
the way for the discovery of dark energy.
1, 4–10
Now similar developmental efforts are needed so that the next order of
magnitude improvement of the experimental constraints on the properties of dark energy can be made.
Progress must be made on two fronts, at a level which cannot be pursued with existing programs alone. First a
large number of nearby SNe must be observed in an appropriate fashion since they provide the fulcrum of the lever-arm
needed to make cosmological inferences from high-redshift SNe observations. Furthermore, these SNe provide the critical
Correspondence: e-mail galdering@lbl.gov; telephone 510-495-2203
Ω
Μ
No Big Bang
1 2 0 1 2 3
ex
p
an
d
s
fo
re
v
er
Ω
Λ
Flat
Λ = 0
Universe
-1
0
1
2
3
2
3
closed
open
90
%
68
%
99
%
95
%
r
ec
ol
la
p
se
s
ev
e
nt
ua
ll
y
flat
Calan/Tololo
(Hamuy et al,
A.J. 1996)
Supernova
Cosmology
Project
effective m
B
(0.5,0.5)
(0, 0)
( 1, 0 ) (1, 0)
(1.5,–0.5)
(2, 0)
(Ω
Μ,
Ω
Λ
) =
( 0, 1 )
Flat
Λ = 0
redshift z
14
16
18
20
22
24
26
0.02 0.05 0.1 0.2 0.5 1.00.02 0.05 0.1 0.2 0.5 1.0
Figure 1. Cosmological constraints from Type Ia supernovae: Hubble diagram (brightness vs. redshift) and resulting constraints
on Ω
M
and Ω
Λ
from 42 high-redshift Type Ia supernovae.
1
At left, the 68%, 90%, and 99% confidence regions for an unconstrained
fit for Ω
M
and Ω
Λ
are shown. For a flat universe only a cosmological constant or other form of dark energy can explain the data.
Even without assuming a flat universe a cosmological constant is hard to avoid for any reasonable choice of Ω
M
. At right, the Hubble
diagram of the current Supernova Cosmology Project dataset of Type Ia supernovae is shown. The low-redshift SNe Ia shown are most
of those currently available that are suitable for analysis using the same methods as used for the high-redshift SNe Ia.
empirical calibration of the SN lightcurve brightness-width relation, as well as providing the intrinsic SN colors needed
to correct for dust extinction. At present, half the statistical uncertainty in the dark energy measurements arises from the
limited pool of low-redshift SNe; we propose to increase this pool to many hundreds of well-observed nearby SNe located
in the smooth Hubble flow.
Second, our understanding of the physics of the SNe Ia must be pushed to a deeper level. Presently SNe Ia appear to
be excellent standardized candles, but we do not understand the details of why this is so. The chief remaining loophole
in the interpretation of the SN Ia results is the possibility of a conspiratorial evolutionary effect in the explosions them-
selves.
11
There is already some empirical data that constrains such “conspiracies,” but as we move to the next generation
of experiments much tighter constraints will be needed. Low-redshift supernova data provide both the necessary empiri-
cal constraints, and the matching deeper physical understanding of the SNe Ia, and are therefore a necessary complement
to the ongoing high-redshift work and future projects like the SN program of the CFHT Legacy Survey (CFHTLS), the
w Project (wP), the Large Synoptic Survey Telescope (LSST) and the SuperNova/Acceleration Probe (SNAP). This work
to constrain evolutionary effects requires that the low-redshift SNe Ia be observed with spectroscopy over their entire
lightcurves and across a wide range of galactic environments (spanning stellar age and metal content).
2. THE NEED FOR A LARGE NEARBY SN SAMPLE
In recognition of the importance and urgency of this work, we have begun a new experiment – the Nearby Supernova
Factory (SNfactory) – designed to exploit as fully as possible the potential of low-redshift SNe. The SNfactory will
concentrate on the discovery and spectrophotometric follow-up of ∼300–600 SNe Ia. This is an order of magnitude
more SNe than currently available samples and will yield two orders of magnitude more spectrophotometry. Moreover,
our goal is produce a sample which is homogeneous, well-calibrated, and recalibratable. Such a large sample of nearby
Hubble-flow SNe Ia can make significant contributions to both the statistical and systematic aspects of using SNe Ia for
cosmology, as we now describe.
2.1. Anchoring the zero-point of the Hubble diagram
Roughly 50% of the statistical uncertainty in the current cosmological constraints from SNe Ia result stems from the
small number of low-redshift SNe Ia which are suitable to serve as the zero-point for the SNe Ia Hubble diagram. This
zero-point is the product L
SN
H
2
o
, where H
o
is the Hubble constant and L
SN
is the luminosity of a standardized Type Ia
supernova. For purposes of cosmology this zero-point is a “nuisance” parameter, containing no useful information while
contributing to the statistical uncertainty.
The minimal criteria for low-redshift SNe Ia to be cosmologically useful is that they be in the smooth Hubble flow
— so that their radial velocities reflect cosmological redshift rather than galaxy peculiar velocities; that there exists good
lightcurve photometry beginning no later than 5 days after maximum light — so that determination of the peak magnitude
involves little extrapolation; and that the SNe Ia are discovered in the blind-search mode used for the high-redshift SNe Ia
— so that any subtle selection effects and the range of host galaxy properties are as similar as possible at both high and
low redshift. The largest sample of SNe Ia satisfying these minimal criteria are from Ref 7, and are shown in Fig. 1. One
can see that the cosmologically useful nearby SNe Ia are actually outnumbered by the high-redshift SNe Ia!
A number of groups are planning much larger, more comprehensive, experiments using high-redshift SNe Ia to probe
the nature of dark energy. CFHTLS expects to discover and follow-up ∼300–600 high-redshift (0.3 < z < 0.9) SNe Ia
over five years beginning in 2003. The wP search at the CTIO expects to find 200 0.15 < z < 0.75 SNe Ia over the next
five years . These experiments have the potential to measure the effective time-averaged equation of state, w = p/ρ, of the
dark energy, with a level of precision that will begin to test whether dark energy is due to something other than Einstein’s
cosmological constant (for which w = −1, independent of time). Further down the road, the SuperNova/Acceleration
Probe will begin the measurement of time variations in the equation of state — the next generation of tests for dark-energy
that differs from the simple cosmological constant. These experiments will rely very heavily on the a dataset such as the
SNfactory’s in order to calibrate L
SN
H
o
.
This is dramatically illustrated in Fig. 2 for the case of 300 CFHTLS SNe. CFHTLS will attempt to rule out a
cosmological constant by examining the case of a constant effective equation of state, w(z) = constant = w
0
. In the
example shown, for a nominal w
0
= −0.8 the high-redshift SNe Ia alone still allow w
0
= −1 at the 68% confidence
level (projection of solid contour). However, addition of 300 SNe Ia from the SNfactory allows a clear rejection of
w
0
= −1 (projection of dashed contour). This is the result of the SNfactory constraint on L
SN
H
2
o
. The SNfactory is also
essential for getting the best statistical results for a SNAP measurement of w(z). w(z) is often expressed to linear order
as w(z) = w
0
+ w
1
z. The projection of the solid contours of Fig. 3 give the 68% confidence region for w
0
or w
1
for
a SNAP-like dataset of 2000 SNe Ia with 0.3 < z < 1.7. In this example, SNAP does not rule out w
0
= −1, w
1
= 0,
that is, a cosmological constant is still allowed. However, addition of the SNfactory SNe Ia allows a clear rejection of
a cosmological constant and significantly improves the measurement of w
0
. In the general case, the SNfactory dataset
typically halves the uncertainty in w
0
by measuring L
SN
H
2
o
to 1%. Thus, is it clear that the SNfactory is a necessity if
the high-redshift supernova cosmology experiments are to realize their full potential.
2.2. Calibration of the Luminosity–Lightcurve Width Relation
The slope, α, of the relation between SN Ia intrinsic luminosity and lightcurve width has been determined from only a
relatively small (∼ 30) number of Hubble-flow SNe Ia. Each of these SNe Ia has an intrinsic peak-brightness uncertainty
of about 10% and measurement errors which are comparable after host-galaxy extinction correction. Moreover, the
population of SNe Ia with narrow or wide lightcurves is small, thus limiting the lever-arm available to measure α. As a
result, α is determined to only about 25%.
12
This doesn’t effect individual SNe too greatly because most SNe are clustered
around the typical lightcurve width. However, for the future large high-redshift SNe samples, where the probative value
comes from averaging, there exists the potential for σ
α
to become an important source of statistical uncertainty since it
is a correlated uncertainty for all the SNe. The large SNfactory dataset should reduce the uncertainty on α by at least a
factor of three.
2.3. Calibration of Intrinsic Colors for Dust Extinction Correction
Correction of SN brightnesses for host-galaxy dust extinction involves a comparison of the measured color (usually at
maximum light) of a new SN with colors of SNe Ia which are extinction-free (e.g., those in elliptical galaxies, which are
-1 -0.9 -0.8 -0.7
-0.6 -0.5
w
0
0.26
0.28
0.3
0.32
0.34
Ω
M
CFHLS
CFHLS+SNF
Figure 2. The importance of low-redshift SNe Ia for near-future dark energy probes: The impact of adding 300 SNfactory SNe Ia
to a representative sample of 300 high-redshift (0.3 < z < 0.7) SNe Ia, such as the CFHTLS or w P might obtain over the next 5 years
is demonstrated. The projection of the solid contour is the 68% confidence region in the Ω
M
–w
0
plane from the high-redshift survey
alone, while the projection of the dashed contour is the 68% confidence region when the SNfactory SNe Ia are added. In this example
the addition of the SNfactory SNe Ia allows elimination of a cosmological constant model (w
0
= −1). A flat universe and a prior of
±0.04 on Ω
M
has been assumed. When comparing to Fig. 3 note that here w
1
is set to zero and therefore contributes no uncertainty;
if w
1
were allowed to float the uncertainty in w
0
would be much larger. Courtesy E. Linder, R. Miquel, and D. Huterer.
mostly free of dust). The change in color must be multiplied by 4.1 to obtain the extinction-corrected brightness. These
intrinsic colors are a function of lightcurve epoch and depend on whether a SN is intrinsically over- or under-luminous.
The current uncertainty in the intrinsic (dust-free) colors of SNe Ia is not negligible. Only about 10% of all host
galaxies are ellipticals, so the number of calibrating SNe Ia is small. Moreover, few of those SNe Ia are in the smooth
Hubble-flow, where the effects between SN color and brightness due to dust and intrinsic luminosity can be separated.
As a result, the uncertainty in the intrinsic SN Ia colors is one of the dominant uncertainties in the current cosmology
measurements. Note that this is a correlated uncertainty in the calibration of the intrinsic colors of all SNe Ia, so it does not
average out as larger samples of high-redshift SNe Ia are obtained, unlike the color measurement errors of each individual
high-redshift SN Ia. The SNfactory spectrophotometric lightcurve measurements are designed to greatly improve this
calibration of intrinsic colors, making it possible to take advantage of the large statistics from the next generations of
high-redshift SN Ia projects.
2.4. K-corrections
Because SNe Ia are observed over a range of redshifts, in the restframe of the SN any filter used to obtain an image will not
exactly match the standard B-band filter. As a result, the brightness of a SN Ia will be affected by spectral features which
are either included or excluded due to filter mismatch and must be corrected. This “K-correction” requires knowledge
of the SN spectrum and the photon response of the instrument.
13, 14
For high-redshift SNe Ia the spectrum is usually
only available from around the time of maximum light, whereas each photometry point along the SN lightcurve requires
its own K-correction. Thus, the appropriate spectrum to be used for K-corrections at other epochs on the lightcurve
must be inferred from the spectra of low-redshift analogs. The choice of the best analog relies on comparison of the
-1
-0.95
-0.9
-0.85
-0.8
-0.75
w
0
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
w
1
SNAP
SNAP+SNF
Figure 3. The importance of low-redshift SNe Ia for next-generation dark energy probes: The impact of adding 300 SNfactory
SNe Ia to a more ambitious high-redshift SNe Ia experiment such as SNAP is demonstrated. The projection of the solid contour
gives the 68% confidence region from SNAP alone, while the projection of the dashed contour is the 68% confidence region when the
SNfactory SNe Ia are added. In this example the addition of the SNfactory SNe Ia allows elimination of a cosmological constant model
(w
0
= −1, w
1
= 0). A flat universe has been assumed and a prior of ±0.04 on Ω
M
has been imposed. Courtesy E. Linder, R. Miquel,
and D. Huterer.
maximum-light spectra and the colors of the low- and high-redshift SNe (it could also depend on additional inputs, such
as the lightcurve shape). The better the analog, the better the accuracy of the K-correction.
The SNfactory’s spectral timeseries will allow synthetic photometry, thereby eliminating errors in the K-corrections
for SNfactory SNe. This will allow excellent calibration of SNe Ia standardization relations. Moreover, the large SNfactory
sample will vastly increase the number of analogs available for the K-correction of high-redshift SNe Ia. Thus, one can
see that these gains — although merely technical — have the power to significantly improve the results from supernova
cosmology experiments.
All the above steps are needed for the proper application of the current standardization methods used by all the groups
who do cosmology with Type Ia supernovae. Furthermore, the reader will have noted that these various calibration and
correction steps are not independent: K-corrections must be applied to get the SN color, which is then used to determine
the dust extinction, after which α and L
SN
H
2
o
can be determined. Therefore a large homogeneous dataset is required to
separate-out the contributions of these various effects. In particular, a large-scale search is necessary in order to find the
rarer (elliptical host galaxies, large/small lightcurve width, etc.) events which provide leverage for the calibration.
2.5. Converting Systematic Uncertainties into Statistical Uncertainties
Improvement of systematic uncertainties is as — or even more — important than improving the statistical uncertainties
just discussed. In particular, we now want to scrutinize the SNe Ia closely enough that we can find any existing second-
order differences that are not already parameterized by the lightcurve width vs. luminosity relation. Well-observed nearby
SNe Ia, especially in host galaxies spanning a wide range in star-formation histories, are essential for hunting for such
possible second-order systematic trends and the observables that could constrain them.
By measuring key spectral
15, 16
and lightcurve features for each SN the physical conditions of the explosion can be
tightly constrained, making it possible to recognize sets of SNe with matching initial conditions. The current theoretical