Download Gone in a flash: supernovae in the survey era

The transient universe • Sullivan: Supernovae
1: The spiral galaxy M101, home to supernova SN 2011fe, the brightest Type Ia supernova in the sky for more than 25 years. The left-hand image shows the
galaxy before the explosion, and the right-hand image shows the supernova near peak brightness. (BJ Fulton [IfA, Univ. Hawaii]/LCOGT/PTF/STScI)
Gone in a flash:
supernovae in the survey era
T
he study of supernovae – stellar explosions, the dramatic final moments of
a star’s life – is one of the richest and
longest running fields of astronomy. Supernovae
have been observed, recorded and studied by
humans for nearly 2000 years, since the first
descriptions of “guest stars” appeared in the
records of Chinese astronomers. These relatively nearby galactic supernovae, appearing
brighter in the sky than Venus – perhaps bright
enough to cast shadows – must have astonished
and perplexed observers of the time. Those of
1572 (noted by Tycho Brahe) and 1604 (Kepler’s
supernova) were used as arguments against the
then commonly held Aristotelian world-view
that the heavens were immutable. Even today,
the idea of an entire star self-destructing in an
explosion five billion times brighter than the
Sun is both compelling and awe-inspiring.
Milky Way supernovae
There are around seven recorded supernovae
in the Milky Way for which plausible remnants have been identified (depending on how
one counts), including SN 1572, SN 1604 and
SN 1054, the explosion that gave birth to the
A&G • December 2013 • Vol. 54 The transient universe
Mark Sullivan outlines the scientific
heritage of supernovae and the
potential for future discoveries.
Together with Walter Baade, he coined the term
“super-novae” in the early 1930s, and over the
course of the next 50 years found 120 events –
a record not bettered until recently by Scottish
amateur astronomer Tom Boles (see page 6.9).
The development of dedicated robotic searches
Crab pulsar. Some 20 further ancient galactic in the 1960s and 70s saw the discovery rate
supernova candidates exist in Chinese, Islamic pick up to a few tens of events per year; CCD
and other astronomical records.
searches in the 1980s on larger telThe first detected extragalactic
escopes provided the first events
The
supernova, in Andromeda
at cosmological distances.
idea of a star self(S Andromedae), was not
Around this time came the
destructing in an
discovered until the end of
realization of the idea, first
the 19th century, although
postulated in the 1960s,
explosion five billion
that Type Ia supernovae
it was not recognized for
times brighter than
what it was at the time.
can act as particularly highthe Sun is an aweFor the next few decades
quality standard candles.
inspiring field of
This sparked a significant
the discovery rate remained
study
low – a “blind luck” era, when
observational effort in the early
finding supernovae was the result
1990s to find larger samples of disof chance rather than any concerted
tant events, pushing the boundaries of
plan (see figure 2). The first dedicated “super- what was then achievable with the available
nova patrol” did not commence until the 1930s, telescopes, instruments and computing. The
when the Caltech astronomer Fritz Zwicky used culmination was the announcement by two
the Mount Wilson telescopes and an arduous teams in the late 1990s that high-redshift Type
manual comparison of photographic plates. Ia supernovae appeared about 40% fainter – or,
‘‘
’’
6.17
The transient universe • Sullivan: Supernovae
equivalently, more distant – than expected in
a flat, matter-dominated universe. This indicated that the rate of expansion of the universe
was speeding up, and that the substance of the
universe was dominated by a “dark energy”
responsible for 70–75% of its energy (Riess et
al. 1998, Perlmutter et al. 1999). This result
later won the 2011 Nobel Prize in Physics, and
has forever changed our understanding of the
universe.
New sky surveys continue to find supernovae
at an extraordinary rate, partly motivated by
trying to understand dark energy and partly
motivated by an innate curiosity as to how stars
explode. Cheaper, larger CCDs and the birth
of large-scale computing have made the task of
surveying large areas of sky significantly simpler
and quicker. Even as recently as the turn of the
last century, supernovae were only discovered
at the rate of a few tens per year. Modern surveys are enabling events to be located with a
frequency that is almost an order of magnitude
larger – 2012 saw more than 2000 spectroscopically confirmed supernovae announced, and
2013 is on track to better this (figure 2).
Much of this survey work has been performed
on dedicated smaller telescopes used with large
format cameras. Prime examples include:
● T he Palomar Transient Factory (PTF), using
the Palomar 48-inch Schmidt telescope refurbished with a large-format CCD camera
● Pan-STARRS, using a purpose-built 1.8 m
telescope
● T he Catalina Real-Time Transient Survey
(CRTS), using three dedicated telescopes in the
northern and southern hemispheres.
Other surveys, such as the La Silla Quest Variability Survey on the ESO 1 m Schmidt telescope
on La Silla, and the SkyMapper survey at Siding
Spring, are also now coming online.
This rapid discovery rate has dramatically
improved our understanding of supernova physics, but, as might be expected, has also raised
more questions to be answered. Here we highlight recent progress on two fronts: finding new
ways in which stars can explode, and understanding the progenitor stars that make the
cosmologically important Type Ia supernovae.
An explosion of supernova types
In 1941, Minkowski proposed the first observational classification of supernova into the now
familiar Type I (no hydrogen lines in the spectra)
and Type II (with hydrogen lines) events. This
basic classification scheme survives today, but
has acquired ever more complex subdivisions
(Ia, Ib, Ic, IIb, IIn, Ia-CSM, Iax, etc) to accommodate the large variety of supernova types
now known. This means that the classification
of supernovae sometimes seems to require the
attention of a black-belt expert – and even then,
similar objects can be classified in subtly different ways by experienced observers.
6.18
2: The discovery rate of approximately 14 000 extragalactic supernovae located over the last
century, complete up to 31 August 2013. This includes all supernovae announced via IAU circulars
and Astronomer’s Telegrams, as well as those discovered by the Palomar Transient Factory, the
Supernova Legacy Survey, and Pan-STARRS. Several key dates and observations are highlighted
in the figures. The top panel shows an overview of the last 130 years, while the bottom panel
focuses on the last 20 years and the future.
Historically, supernovae were believed to originate in two physically distinct ways. The first
group results from the thermonuclear destruction of a carbon–oxygen white dwarf star, as
it accretes or gains material from a companion
until carbon burning is ignited at or near its core.
These are the Type Ia supernovae (SNe Ia), with
distinctive features of silicon (a product of carbon burning) in the spectra. The second group
forms through the gravitational core-collapse of
a massive star, more than eight times the mass of
the Sun, triggered by iron photo-disintegration
and the consequent loss of support in the star’s
core. This mechanism is believed to produce
nearly all the other spectral types.
Once the progenitor star has exploded, there
must also be some way of generating electromagnetic radiation so the supernova can be
detected. There are three basic contributors to
the radiation, as presently understood. The first,
important in nearly all supernova types, is from
the radioactive decay of unstable elements synthesized in the explosive nucleosynthesis. The
most important is 56 Ni, which then decays into
56
Co and eventually into stable 56Fe. This decay
generates gamma-rays, which are trapped in the
ejecta and thermalize it so that it glows. This
mechanism is the only power source for SNe Ia –
indeed, without radioactive 56 Ni, SNe Ia would
never be seen. A second source is the release
of internal energy deposited by the explosion
via photon diffusion. A final contributor is the
A&G • December 2013 • Vol. 54
The transient universe • Sullivan: Supernovae
expected from radioactive 56 Ni. A third class,
SLSN-R, may actually be closely related to
SLSN-I, and are consistent with being powered
by the radioactive decay of several solar masses
of 56 Ni (here, “R” stands for radioactive).
Naturally, several theories as to their origin
and power source have emerged to explain this
confusing observational picture. At the time of
writing, two of these theories seem to be gaining
most traction. The first is that SLSN-R result
from the death of stars of greater than ~100 M⊙
via the pair instability mechanism (Gal-Yam
et al. 2009): an extremely massive star will go
through a phase of electron–positron “pair”
generation in its core, rendering the star unstable and making it explode, long before it evolves
to the point that iron core collapse can occur
as happens in lower-mass objects. The second is that SLSN-I (and perhaps, in part, the
other SLSN types) may be powered by energy
deposition from the spin-down of a magnetar
(Kasen and Bildsten 2010). However, this magnetar theory can only be consistent with all the
observations with significant tuning – and new
3: The light curves of common supernova types compared to superluminous supernovae. In
observations are emerging that may challenge
all cases, day zero represents the day of peak brightness. Typical supernova explosions reach
the pair instability origin for SLSN-R (Nicholl
absolute magnitudes of more than –19.5 and decay over a few weeks, whereas superluminous
et al. 2013).
supernovae can reach up to –22.5, more than a factor of 10 brighter, and can remain bright for
several months. (Reprinted from Gal-Yam 2012)
As has happened over several decades with
SNe Ia, understanding the detailed physics
of SLSNe (and even testing them as standard
interaction of the supernova ejecta with any and unrecognizable because its high redshift candles) requires populations of events, rather
dense circumstellar medium (CSM) around the meant that it probed the rest-frame ultraviolet than a handful of relatively poorly observed
progenitor star, with the kinetic energy of the – a region poorly explored in supernova science. examples. Determining the rates of SLSNe as
ejecta essentially converted into electromagnetic
This high redshift also implied an extreme a function of redshift is essential to understand
energy. This mechanism is particularly impor- ultraviolet luminosity, making them interest- their progenitors, because both star formation
tant in Type IIn supernovae, which also show ing candidates for a new generation of
and metallicity evolve over cosmic hisnarrow emission lines in their spectra that indi- cosmological candle. Their bright
tory. Meanwhile, improved spectral
The
apparent magnitudes allow
cate interaction with the CSM.
and photometric data will allow
for studies of the interstellar
studies of their chemistry and
new surveys are
Superluminous supernovae
medium of their host galaxenergetics. Their intrinsic rarrevealing a much
ity – they occur at rates about
The new surveys described above are revealing ies via absorption line speclarger diversity of
a much larger diversity of rare explosive tran- troscopy, or perhaps even as
1/1000th of the SN Ia rate –
rare explosive
sients than this simple picture suggests. Super- standard candles in the verymeans that very large volumes
transients
luminous supernovae (SLSNe) are a new class of high-redshift universe – beyond
of space must be searched in
order to find significant numextreme supernova that have emerged over the the reach of our current best tool,
past five years (for a review, see Gal-Yam 2012), SNe Ia.
bers. But with very large imagers on
reaching absolute magnitudes of –23 (50 times
To date there are three identified classes
4 m-class facilities (e.g. Megacam on the
brighter than a Type Ia supernova). The first of SLSN (Gal-Yam 2012), commonly defined as CFHT, or DECam on the Cerro Tololo Interevent, SCP 06F6, was identified in 2009 and had being brighter than –21 in absolute magnitude American Observatory 4 m Blanco telescope),
broad, unexplained spectral absorption lines (figure 3). By analogy to the classification of these objects can be found up to a redshift of 4
and a very slowly evolving light curve. The lack normal SNe, the SLSNe-I have hydrogen-free (Cooke et al. 2012). It should also be possible to
of an apparent host galaxy meant that even the optical spectra, and may possibly be related study them in large numbers using projects such
redshift was unknown. Consequently, an imagi- to SNe Ic, whereas SLSNe-II show narrow as the Dark Energy Survey (DES).
native range of possibilities were postulated to hydrogen lines. The luminosity of these latter
explain the object, ranging from a tidally dis- events is probably related to interaction with Progenitors of Type Ia supernovae
rupted white dwarf at moderate redshift, to an CSM around the progenitor star, which then Type Ia supernovae are routinely used to
asteroid colliding with a white dwarf in our own also causes the observed hydrogen lines. But measure cosmological parameters, such as the
galaxy. During the operation of the Palomar SLSNe-I are more puzzling. The normal meth- equation of state of dark energy, w (figure 4).
Transient Factory, several similar events were ods of powering supernovae discussed above – The pursuit of w is a goal of many wide-field
located (Quimby et al. 2011), this time with host radioactive decay of 56 Ni, or the interaction of surveys, for example: the Supernova Legacy
galaxy redshifts. This, in turn, allowed a red- the supernova ejecta with dense CSM – cannot Survey, the Sloan Digital Sky Survey Supershift determination for SCP 06F6 to be made, at reproduce both their extreme brightness and nova Search, the Lick Observatory Supernova
z = 1.19. The SCP 06F6 spectrum was so unusual their decay-rates, which are faster than that Search, PTF, SkyMapper, Pan-STARRS, DES,
‘‘
’’
A&G • December 2013 • Vol. 54 6.19
The transient universe • Sullivan: Supernovae
the Large Synoptic Survey Telescope, and
Euclid. This represents a significant allocation
of observing resources. The technique itself is
simple. The SN Ia “standard candle” distances
(derived from their apparent brightness and a
knowledge of the SN Ia absolute luminosities)
can be compared with luminosity distances, calculated from their redshifts together with a set
of cosmological parameters and the equations of
General Relativity. As the cosmological parameters are, in principle, the only unknowns in
this analysis, with a sufficient number of SNe Ia,
constraints can be placed on their values.
Although SNe Ia provided the first direct evidence for dark energy (and still provide the most
mature and constraining measurements), there
are several potential drawbacks to the technique – the apparently simple standard candle
concept has several less apparent difficulties,
fundamentally related to the precision required.
Detecting departures in dark energy from
w = –1, mathematically equivalent to the cosmological constant, requires an extremely sensitive experiment. A 10% difference in w from
–1 is equivalent to a change in SN Ia brightness
at z = 0.6 of only 0.04 magnitudes, an absolute
precision perhaps not historically or routinely
achieved in astronomy. The challenge is even
more complex when fitting variable w models.
Therefore, the SN Ia experiments are not simply
about obtaining more data, but about obtaining higher-quality data in which these experimental systematics can be controlled. While
these challenges are considerable, it is at least
a well-defined and tractable problem on which
substantial progress is being made.
Coupled with these experimental challenges
is the relative confusion over the nature and
astrophysics of the SN Ia progenitor system.
Although the nature of the exploding star as
a carbon–oxygen white dwarf is not seriously
challenged, and has recently been confirmed in
a nearby SN Ia (SN 2011fe; Nugent et al. 2011;
figure 5), this white dwarf must accrete material
from a companion until the density and temperature are high enough to ignite a thermonuclear
explosion. The identity of this second star is not
known and has puzzled astronomers for decades: is it non-degenerate such as a giant or main
sequence star donating mass (as in a nova or
similar system), or is it a second white dwarf in
some form of merger? These two broad classes,
“single degenerate” and “double degenerate”,
probably trigger the thermonuclear explosion in
different ways: different delays between system
formation and explosion, different accretion
rates and ignition conditions, and perhaps different white dwarf compositions and ages.
There has been a growing realization over the
last few years that, in fact, it is very likely that
both (and possibly even other) channels operate. The remarkable SN Ia PTF 11kx (Dilday et
al. 2012) showed significant evidence for inter­
6.20
4: Typical confidence
contours in the
cosmological
parameters w (the
equation of state of
dark energy) and ΩM
(the matter density of
the universe). Shown
are the SN Ia results
(blue), constraints from
the Hubble constant
(red), and those from
galaxy redshift surveys
(green). Results from
observations of the
cosmic microwave
background (the
WMAP satellite) are
also used. The data
are consistent with
w = –1 (the cosmological
constant) with a 7%
uncertainty, including
all identified systematic
uncertainties. (Adapted
from Sullivan et al. 2011)
5: Constraints on
the mass, effective
temperature, radius,
and average density
of the primary star of
the Type Ia supernova
SN 2011fe. The
shaded red region is
excluded from the
non-detection of an
optical counterpart in
Hubble Space Telescope
(HST) imaging (see
also figure 6 for the
constraints this places
on the companion star).
The shaded green
region is excluded
from considerations of
the non-detection of a
shock breakout/heating
at early times, and the
blue region is excluded by the non-detection of a counterpart in Chandra X-ray data. The location
of the H, He and C main sequence is shown, with the symbol size scaled for different primary star
masses. Several observed white dwarf and neutron stars are shown. The primary radius in units
of R⊙ is shown for a primary star mass of 1.4 M⊙. (Reprinted from Bloom et al. 2012)
action between the SN Ia ejecta and a dense
CSM, with hydrogen emission lines and calcium
features switching from absorption to emission
during the supernova evolution. This would naturally be expected in a recurrent nova (i.e. single
degenerate) system where the supernova ejecta
encounters dense shells of material from earlier
nova eruptions, although questions remain over
the large CSM mass implied by the data.
On the other hand, observations of SN Ia remnants many hundreds of years old have failed to
locate any surviving companion stars (Schaefer
and Pagnotta 2012). These would be expected
in SN Ia remnants which originated from sin-
gle degenerate systems, as the star from which
material is accreted onto the white dwarf should
survive the explosion. Furthermore, historical
pre-supernova Hubble Space Telescope imaging
of M101, the host galaxy of SN 2011fe (see figure 1), also rules out a luminous giant companion star at the site of the supernova explosion (Li
et al. 2011; figure 6). Thus, a diversity in SN Ia
progenitors seems very likely.
Challenges for future surveys
A critical component of any supernova survey is
the spectroscopic follow-up of candidate events,
confirming their nature and measuring the
A&G • December 2013 • Vol. 54
The transient universe • Sullivan: Supernovae
6: Constraints on the
companion star to
the white dwarf that
exploded as the Type Ia
supernova SN 2011fe.
This Hertzprung–Russell
diagram is constructed
from pre-explosion
Hubble Space Telescope
imaging of the supernova
location. The parameter
space above the yellow
line is ruled out by the
imaging data, excluding
most red giants and
many known recurrent
nova systems (e.g.
RS Oph). Other popular
potential SN Ia progenitor
systems, such as
V445 Pup (a helium nova)
and U Sco (a recurrent
nova with a main
sequence companion),
are shown as grey
shaded areas. The stellar
main sequence is shown
as a black line, with giant
branches for different
mass stars shown as
coloured dots. (Reprinted
from Li et al. 2011)
redshifts essential for either their detailed study, confirmed samples, although with the clear posor placement on a Hubble diagram. Perhaps sibility of contamination in the samples from
obviously, a larger telescope is usually required events that are not normal SNe Ia. Such potenfor spectroscopy than for discovery – for exam- tial contamination must be carefully accounted
ple, events discovered on 1 m class telescopes are for in cosmological studies.
However, photometric typing does not allow
usually followed up on 3–4 m class facilities,
and those fainter objects discovered on 3–4 m studies of the detailed physics of the supertelescopes require 8 m-class telescopes. This nova explosions, for which spectroscopy is
quickly generates a bottleneck, with insufficient required to give chemical composition, nor
telescope time available to observe all
does it allow the identification of odd,
the events discovered. As an exampeculiar or extreme events within
A
ple, the high-redshift Supernova
the mainstream of a populasecond approach
tion. Such events can often
Legacy Survey (SNLS), part of
the Canada–France–Hawaii
have an important impact in
is also needed:
Telescope Legacy Survey, ambitious programmes the field. For example, two
used some 1600 hours of
extreme SNe Ia (the overdedicated to
8 m telescope time to clasluminous SN 1991T and the
transient
sify around 500 supernova
under-lumious SN 1991bg)
follow-up
helped to establish the nowevents. This situation will be
even worse for DES, which will
famous brighter-slower/fainterfind 3500 SNe Ia alone. For future
faster relation for SNe Ia used in
projects finding even more events, the
cosmological applications (the “Phillips
situation will become impossible. Similar scal- relation”), while the apparently super-Chaning arguments apply to future local searches.
drasekhar SN Ia SN 2003fg showed that not all
The solution appears to be two-fold. For cos- SNe Ia need explode near the Chandrasekhar
mological applications, photometric typing mass (Howell et al. 2006). Finally, of course,
techniques have been developed which, with an superluminous supernovae would not be known
after-the-event host galaxy spectroscopic red- were it not for spectroscopy.
shift obtained with multi-object spectrographs,
Thus a second, complementary approach
such as AAOmega at the Anglo-Australian is also needed: ambitious programmes (or
Telescope, can provide reliable classifications facilities) dedicated to transient follow-up. An
(Campbell et al. 2013). This allows Hubble example is a new programme at the European
diagrams to extend beyond spectroscopically Southern Observatory 3.6 m New Technology
‘‘
’’
A&G • December 2013 • Vol. 54 Telescope: the Public ESO Spectroscopic Survey
for Transient Objects (PESSTO). This public
survey will attempt to address the lack of spectroscopic follow-up, at least for local supernova
surveys, classifying some 2000 SNe over the
next five years (>10% of all supernovae ever
classified), and conducting detailed time-series
studies of 150 events.
Supernovae have a profound influence upon
many diverse areas of astrophysics, including
chemical enrichment, feedback in galaxy formation, cosmology and stellar evolution, and
we are fortunate to be living through a golden
age for their study. The combination of the discovery of the mysterious dark energy and the
uncovering of new types of explosive optical
transients, has motivated a diverse set of discovery and follow-up programmes; yet significant questions remain about many aspects of
supernova rates, their light curves and spectra,
their demographics, and the dependence of all of
these properties on their environment and progenitor composition and configuration. Indeed,
as we have discussed, in many cases even the
nature of the star that explodes, or the configuration of the progenitor system, is still unclear.
This picture can only improve over the next
decade, with new facilities planned that will
dwarf existing programmes. In particular,
planned upgrades to the PTF to an approximately 40 square-degree imager, tentatively
called the Zwicky Transient Factory, will allow
thousands of square degrees to be surveyed
every night to 21st magnitude. This will provide measurements of supernova evolution on
very short, intra-night timescales, including the
first hours after explosion probing the epoch of
shock break-out. In the more distant future, the
Large Synoptic Survey Telescope, an 8 m-class
facility, will survey the entire visible sky to 24th
magnitude every few nights. Surveys like these
will undoubtedly uncover new explosive transient types, and provide a rich dataset to finally
answer the outstanding questions in the field. ●
Mark Sullivan, School of Physics and Astronomy,
University of Southampton, UK; m.sullivan@
soton.ac.uk
References
Bloom J et al. 2012 ApJ 744 L17.
Campbell H et al. 2013 ApJ 763 88.
Cooke J et al. 2012 Nature 491 228.
Dilday B et al. 2012 Science 337 942.
Gal-Yam A 2012 Science 337 927.
Gal-Yam A et al. 2009 Nature 462 624.
Howell et al. 2006 Nature 443 308.
Kasen D and Bildsten L 2010 ApJ 717 245.
Li W et al. 2011 Nature in press arXiv:1310.4446.
Nicholl M et al. 2011 Nature 480 348.
Nugent P et al. 2011 Nature 480 344.
Perlmutter S et al. 1999 ApJ 517 565.
Quimby R et al. 2011 Nature 474 487.
Riess A et al. 1998 AJ 116 1009.
Schaefer B and Pagnotta A 2012 Nature 481 16.
Sullivan M et al. 2011 ApJ 737 102.
6.21