Astronomical catalogs derived from wide-field imaging surveys are an
important tool for understanding the Universe. We construct an astronomical
catalog from 55 TB of imaging data using Celeste, a Bayesian variational
inference code written entirely in the high-productivity programming language
Julia. Using over 1.3 million threads on 650,000 Intel Xeon Phi cores of the
Cori Phase II supercomputer, Celeste achieves a peak rate of 1.54 DP PFLOP/s.
Celeste is able to jointly optimize parameters for 188M stars and galaxies,
loading and processing 178 TB across 8192 nodes in 14.6 minutes. To achieve
this, Celeste exploits parallelism at multiple levels (cluster, node, and
thread) and accelerates I/O through Cori's Burst Buffer. Julia's native
performance enables Celeste to employ high-level constructs without resorting
to hand-written or generated low-level code (C/C++/Fortran), and yet achieve
Celeste is a procedure for inferring astronomical catalogs that attains
state-of-the-art scientific results. To date, Celeste has been scaled to at
most hundreds of megabytes of astronomical images: Bayesian posterior inference
is notoriously demanding computationally. In this paper, we report on a
scalable, parallel version of Celeste, suitable for learning catalogs from
modern large-scale astronomical datasets. Our algorithmic innovations include a
fast numerical optimization routine for Bayesian posterior inference and a
statistically efficient scheme for decomposing astronomical optimization
problems into subproblems.
Our scalable implementation is written entirely in Julia, a new high-level
dynamic programming language designed for scientific and numerical computing.
We use Julia's high-level constructs for shared and distributed memory
parallelism, and demonstrate effective load balancing and efficient scaling on
up to 8192 Xeon cores on the NERSC Cori supercomputer.
Hubble Space Telescope spectroscopic observations of the nearby type Ia
supernova (SN Ia) SN2011fe, taken on 10 epochs from -13.1 to +40.8 days
relative to B-band maximum light, and spanning the far-ultraviolet (UV) to the
near-infrared (IR) are presented. This spectroscopic coverage makes SN2011fe
the best-studied local SN Ia to date. SN2011fe is a typical moderately-luminous
SN Ia with no evidence for dust extinction. Its near-UV spectral properties are
representative of a larger sample of local events (Maguire et al. 2012). The
near-UV to optical spectra of SN2011fe are modelled with a Monte Carlo
radiative transfer code using the technique of 'abundance tomography',
constraining the density structure and the abundance stratification in the SN
ejecta. SN2011fe was a relatively weak explosion, with moderate Fe-group
yields. The density structures of the classical model W7 and of a delayed
detonation model were tested. Both have shortcomings. An ad-hoc density
distribution was developed which yields improved fits and is characterised by a
high-velocity tail, which is absent in W7. However, this tail contains less
mass than delayed detonation models. This improved model has a lower energy
than one-dimensional explosion models matching typical SNe Ia (e.g. W7, WDD1).
The derived Fe abundance in the outermost layer is consistent with the
metallicity at the SN explosion site in M101 (~0.5 Zsolar). The spectroscopic
rise time (~19 days) is significantly longer than that measured from the early
optical light curve, implying a 'dark phase' of ~1 day. A longer rise time has
significant implications when deducing the properties of the white dwarf and
binary system from the early photometric behaviour.
We present ultraviolet (UV) spectroscopy and photometry of four Type Ia
supernovae (SNe 2004dt, 2004ef, 2005M, and 2005cf) obtained with the UV prism
of the Advanced Camera for Surveys on the Hubble Space Telescope. This dataset
provides unique spectral time series down to 2000 Angstrom. Significant
diversity is seen in the near maximum-light spectra (~ 2000--3500 Angstrom) for
this small sample. The corresponding photometric data, together with archival
data from Swift Ultraviolet/Optical Telescope observations, provide further
evidence of increased dispersion in the UV emission with respect to the
optical. The peak luminosities measured in uvw1/F250W are found to correlate
with the B-band light-curve shape parameter dm15(B), but with much larger
scatter relative to the correlation in the broad-band B band (e.g., ~0.4 mag
versus ~0.2 mag for those with 0.8 < dm15 < 1.7 mag). SN 2004dt is found as an
outlier of this correlation (at > 3 sigma), being brighter than normal SNe Ia
such as SN 2005cf by ~0.9 mag and ~2.0 mag in the uvw1/F250W and uvm2/F220W
filters, respectively. We show that different progenitor metallicity or
line-expansion velocities alone cannot explain such a large discrepancy.
Viewing-angle effects, such as due to an asymmetric explosion, may have a
significant influence on the flux emitted in the UV region. Detailed modeling
is needed to disentangle and quantify the above effects.
Detailed quantitative spectroscopy of Type Ia supernovae (SNe~Ia) provides
crucial information needed to minimize systematic effects in both ongoing SNe
Ia observational programs such as the Nearby Supernova Factory, ESSENCE, and
the SuperNova Legacy Survey (SNLS) and in proposed JDEM missions such as SNAP,
JEDI, and DESTINY.
Quantitative spectroscopy is mandatory to quantify and understand the
observational strategy of comparing ``like versus like''. It allows us to
explore evolutionary effects, from variations in progenitor metallicity to
variations in progenitor age, to variations in dust with cosmological epoch. It
also allows us to interpret and quantify the effects of asphericity, as well as
different amounts of mixing in the thermonuclear explosion.