►
From YouTube: Modeling Sub-Chandrasekhar White Dwarf Explosions as Type Ia Supernovae, Abigail Polin, Carnegie Obs
Description
In this seminar series, the recipients of NERSC Early Career Awards for using HPC at NERSC, describe their research and how HPC was an important aspect of it.
A
So
I
want
to
welcome
you
to
the
next
in
our
seminars,
by
the
recipients
of
the
2021
nurse
early
queer
achievement
awards,
and
just
for
those
of
you
who
maybe
don't
know,
we
have
these
achievement
awards
we
give
out
every
year
and
the
presented,
and
we
should
recognize
extraordinary
contributions
from
early
career
scientists
to
use
in
our
student
research.
So
we're
really
happy
to
be
able
to
support
very
career
scientists
using
nurses.
We
have
two
one
for
high
impact,
scientific
achievement
and
another
category
for
innovative
use
of
high
performance
computing.
A
We're
we're
really
pleased
today
to
be
able
to
hear
from
abby
I'm
abigail
abbott.
I
think
people
call
call
you
from
caltech
we're
talking
about
her
work
with
supernova
and
abby,
just
in
recognition.
Here's
a
virtual
certificate
that
we
can
get
a
real,
a
real
mail
address
from
you
we'll
send
you
a
real
real
hard
copy
of
it,
and
with
that
I'm
going
to
stop
sharing
and
turn
it
over
to.
B
I
always
forget:
what's
the
best
way
to
do
this,
but
can
you
see
my
talk
now.
B
Great
I'm
realizing.
I
already
forgot
to
change
the
title,
but
thank
you.
First
of
all
for
the
award
and
for
having
me
speak
today,
I'm
abigail
or
abby
poland,
I'm
currently
a
postdoc
at
carnegie
observatory,
observatories
in
caltech
a
joint
position,
but
most
of
this
work
was
done
when
I
was
a
grad
student
at
berkeley
and
an
affiliate
at
lbl.
B
But
I
forget
the
title
I
gave
you
guys
on
the
email,
but
this
one
is
how
to
how
to
blow
up
sub-chandra
sacraments
white
doors
and
how
to
find
them.
I
do
realize
that
not
everybody
in
the
audience
today
is
going
to
be
an
astronomer,
though
so
I
thought
I
would
start
with
a
little
bit
of
background
on
the
general
problem
before
getting
into
the
nitty-gritty
of
what
I
do
with
nurse.
B
Can
I
advance
slides?
Yes,
I
can
great
so
for
a
supernova
physicist.
A
really
good
night
is
when
a
galaxy
goes
from.
Looking
like
this
to
looking
like
this,
that
one
little
speck
of
light
is
what
we
go
after
each
and
every
night,
and
there
are
two
main
ways
that
we
can
get
information
out
of
that
light.
The
first
one,
the
most
basic
and
the
easiest
data
to
take
is
a
light
curve.
B
B
B
We
can
also
where
this
takes
more
telescope
time.
We
can
take
a
spectrum
so
at
any
given
point
in
time.
B
These
missing
chunks
are
attributed
to
a
specific
element
that
we
know
is
present
in
that
ejecta
absorbing
photons,
and
if
we
come
back
night
after
night
and
take
more
and
more
spectrum
spectra
as
the
ejecta
expands
as
it
expands,
it
becomes
more
diffuse.
We
can
see
through
more
of
it
and
we
can
see
further
and
further
into
the
core
of
the
supernova.
B
So
if
this
will
play
over
time
as
we
trace
the
photosphere
retreating
into
the
center
of
the
ejecta,
we
can
measure
the
composition
of
various
radii
in
this
ejecta
over
time
and
eventually
you'll
see
these
absorption
chunks.
These
absorption
troughs
change
into.
Oh,
I
guess
the
thing
stops
before
then,
but
eventually,
if
we
followed
it
long
enough,
instead
of
absorption
features,
we'd
start
seeing
emission
features
so
peaks
as
the
ejecta
starts
to
cool
down.
B
So
I
study
mostly
or
at
least
the
work
that
was
cited
for
this
award-
was
mostly
a
specific
type
of
supernova
called
a
type
1a
supernova
and
they're
characterized
in
these
ways
that
I
just
showed
you
with
a
light
curve
that
is
powered
by
the
radioactive
decay
of
nickel
56,
and
that
determines
the
peak
of
this
light
curve,
eventually
that
radioactive
nickel
56
decays
into
cobalt
56,
which
is
still
radioactive.
It
just
has
a
longer
half-life
and
reliefs
photons
more
slowly,
and
so
we
get
this
initial
nickel-powered
peak,
followed
by
a
cobalt
powered
tail.
B
The
spectra
at
if
we
take
them
right
here
around
this
peak,
show
absorption
features
from
intermediate
mass
elements
like
calcium
and
silicon
and
sulfur
telling
us
that
surrounding
this
radioactive
core
in
the
supernova
ejecta,
we
have
these
sorts
of
what
we
call
intermediate
mass
elements
in
a
shell
around
that
core
absorbing
photons
as
they
try
to
escape
the
really
cool
thing
about
type
1a.
Supernovae
is
traditionally
these
light.
B
B
Is
it
a
60
watt,
light
bulb
or
a
30
watt
light
bulb,
and
for
that
we
depend
on
really
really
good
data
of
supernovae,
that
we
can
measure
their
distance
some
other
way.
So
this
is
2011
fe.
It
is,
I
think,
still
our
most
well
observed
type
1a
supernova
and
it
was
very
close
and
we
have
a
lot
of
other
measures
about
how
far
away
the
galaxy
that
this
supernova
occurred
in.
So
we
can
use
things
like
this
and
we
can
use
cepheids
as
our
kind
of
calibrations
to
say,
okay.
B
So,
therefore,
the
universe
had
to
the
expansion
of
the
universe
must
have
accelerated
between
the
time
that
these
really
distant
supernovae
went
off
and
the
time
that
their
light
reached
us
today.
However,
this
was
really
phenomenal.
It's
noble
prize
worthy,
but
we're
still
arguing
about
the
exact
number
to
attribute
to
the
relative
rate
of
that
expansion.
The
hubble
constant
and
it
turns
out
over
time.
The
measurements
in
here
on
blue
from
the
supernovae
are
getting
pretty.
B
B
And
then
this
green
point
here
is
the
new
method
we'll
have
in
the
near
future
of
standard
sirens
as
we
get
more
gravitational
wave
objects
that
also
have
observed
electromagnetic
counterparts
right
now.
We
only
have
one
of
those,
but
so
why?
Why
is
this
happening?
It
can
be
a
couple
things.
B
It
can
be
that
the
hubble
constant
actually
was
a
different
value
at
the
stage
of
the
universe
that
the
cmb
is
measuring
and
the
supernovae
are
measuring,
or
it
can
be
that
we
don't
understand
some
of
the
errors
between
these
two
numbers,
and
this
is
not
my
field.
This
is
not
my
concentration.
This
is
more
of
why
my
field
is
important,
but,
to
start
with,
I
would
like
to
point
out
that
I
just
lied
to
you
turns
out
type
1a.
B
Supernovae
are
not
actually
all
exactly
the
same,
so
they're
close
when
we
measure
their
peak
luminosity
mb
here
and
plot
that,
against
the
amount
of
the
amount
they
had
dim
over
the
next
15
days,
delta
m15.
B
So
that's
kind
of
a
measure
of
the
width
of
this
light
curve,
as
well
as
the
peak
they
occupy,
a
very
they
originally
anyways
occupied
a
very
tight
region,
known
as
the
philips
relation
here
where
there's
there
is
a
variation
in
the
magnitude
and
hence
the
the
amount
of
nickel
power
in
these
and
in
the
width,
but
over
time,
as
technology
has
advanced
as
we've
developed,
better
better
telescopes
and
we've
moved
from
a
couple.
B
Supernovae
per
you
know,
month
to
several
per
day
turns
out
the
diversity
of
type
1a
supernovae
is
pretty
large.
We
have
these
normal
standardizable
candles
on
the
philips
relation
here,
but
we
also
have
overbright
type
1a
supernovaes
that
still
look
pretty
normal,
these
91
ts
and
then
really
overwrite
super
c's.
That
don't
look
that
normal.
B
We
have
subluminous
ones.
We
have
guys
that
the
delta
m15
is
too
small
and
then
there's
a
whole
bunch
of
other
sort
of
also
thermonuclear
transients.
That
may
or
may
not
be
related
to
the
type
1as
we
know,
but
the
question
becomes:
how
do
we
actually
standardize
all
of
these
and
if
not
all
of
them
are
standardizable?
A
B
B
If
you
might
have
guessed
by
me
winning
a
nurse
award,
I
do
use
super
computers
and
I
model
supernovae,
and
this
process
generally
involves
three
kind
of
major
steps.
There
is
stellar
evolution.
We
need
to
know
what
we're
exploding.
What
does
it
look
like
at
the
time
of
explosion?
What
kind
of
companion
does
it
have?
How
does
this
affect
the
age
and
distribution
and
density
of
the
star
that
explodes?
B
Then
we
need
to
actually
explode
the
thing
we
need
to
perform.
Hydrodynamic
simulations
with
full
nucleosynthesis
involved
in
that
hydrodynamic
steps
concurrently,
because
the
energy
released
by
nuclear
reactions
actually
can
change
the
hydrodynamics,
and
this
determines
the
ejecta
composition
of
the
star
and
we
could
stop
there.
We
learn
things
from
that,
but
if
we
want
to
be
able
to
apply
this
to
the
real
problems
in
astronomy
right
now,
we
need
to
take
it
a
step
further.
We
don't
need
to
just
ask
what
does
this
explosion
look
like
from
a
hydrodynamic
sense?
B
We
need
to
say
what
would
that
supernova
ejecta
look
like
if
I
viewed
it
through
a
telescope
if
or
a
real
explosion.
You
know
at
a
red
shift
of
whatever
that,
I'm
viewing
through
my
telescope.
What
would
the
light
curves
inspector
look
like
and
that's
called
radiative
transport
calculations
to
determine
all
the
physics
that
goes
on
with
the
photons
as
they
release
from
that
ejecta
and
travel
to
us?
I
do
two
out
of
three
of
these
things.
B
I
do
the
hydrodynamic
simulations
with
nuclear
synthesis
and
the
radiative
transport,
and
that's
what
I
will
talk
to
you
about
today.
Oh,
yes,
I
threw
in
the
the
lovely
picture
of
corey
because
all
of
my
thesis
work
was
was
performed
here
at
nurse,
so
anyways
going
back
to
type
1a
supernovae,
let's,
since
I'm
skipping
the
stellar
evolution
part,
because
I
know
pretty
much
what
I'm
exploding
from
the
light
curve
and
the
spectra
of
observations.
B
I
know
again
my
injectable
needs
to
look
like
a
center
core
of
radioactive
nickel
56,
surrounded
by
a
shell
of
intermediate
mass
elements.
That
tells
me
what
is
happening
here
is
I
am
exploding
a
carbon
oxygen
white
dwarf.
That's
what
gives
us
this
kind
of
distribution
as
that
carbon
oxygen
burns
up
the
alpha
chain
into
heavier
and
heavier
elements.
So
at
the
densest
regions
of
the
core.
We
get
these
radioactive
elements
as
the
burning
mechanism
moves
out
through
the
star.
B
We
reach
less
and
less
dense
regions
and
we
burn
less
and
less
completely
until
we
even
can
have
a
layer
outside
here
of
unburnt
material
of
carbon
oxygen.
B
However,
there
is
more
than
one
ways
to
blow
up
a
white
dwarf,
so
the
original
idea
here
was
pretty
ingenious
of
if
all
type
1a
supernovae
look
the
same.
Then
they
all
should
kind
of
be
blowing
up
the
same
right
and
we
can
take
advantage
of
the
fact
that
we
know
carbon
oxygen.
White
dwarfs
are
supported
by
electron
degeneracy
pressure
and
that
electron
degeneracy
pressure
has
an
intrinsic
mass
limit,
the
chandrasekhar
mass,
where
above
this
mass
limit
it
can
no
longer
support
itself
through
electron
degeneracy
pressure.
B
So
if
anything
happens
to
this
white
dwarf
that
pushes
it
above
that
limit
something
needs
to
happen.
Something
needs
to
heat
up.
Something
can
then
explode
it
just
it
can't
be
a
white
dwarf
anymore,
and
so
our
original
thought
was
just
take
a
carbon
oxygen
white
dwarf
the
chandrasekhar
mass
limit
and,
let's
detonate
it
at
the
center,
turns
out.
When
you
do
that,
you
get
too
much
nickel,
you
get
too
bright
of
an
event.
B
Okay,
so
maybe,
instead
of
a
detonation,
we
should
definitely
grate
the
whole
star.
We
should
burn
it
subsonically,
and
then
we
get
the
opposite
problem.
We
don't
get
enough
nickel.
We
get
too
dim
of
an
event
to
to
explain
kind
of
either
side
of
that
philips
relation.
So
why
don't
we
do
both
right?
Why
don't
we
definigrate?
B
For
a
little
bit
to
puff
things
up
and
make
them
less,
dense
and
hot
enough
that
by
the
time
we
then
transition
to
a
detonation,
we
can
burn
some
nickel
56,
but
we
can
still
leave
behind
some
of
these
intermediate
mass
elements
and
that's
pretty
successful.
We
can
vary
the
amount
of
nickel
we
create
doing
that
a
little
bit,
but
we
don't
actually
understand
how
that
happens
physically.
B
B
So
if
we
can't
explain
the
entire
diversity
of
type
1a
supernovae
by
varying
how
much
deflagration
we
do
before
detonating,
what
else
do
we
have?
Well,
we
can
change
the
mass
of
the
white
dwarf
exploding.
Maybe,
but
how
do
we
do
that?
If
the
whole
concept
was
they
explode
because
they
exceed
the
chandrasekhar
mass
limit?
B
B
The
computational
power
at
the
time
only
had
resolution
capabilities
to
be
able
to
do
this
with
a
significant
amount
of
helium
on
the
surface
of
the
white
dwarf
and
as
I'll
show
you
later,
those
don't
look
like
type
1a
supernovae,
there's
too
much
helium
on
the
surface
that
we
don't
expect
in
those
explosions,
so
they're
kind
of
tossed
out
for
a
long
time
until
around
2010,
lars
bilson
and
ken
chen,
amongst
others,
with
updated
nuclear
networks
and
updated
resolution
capabilities,
show
that
aha,
we
can
actually
use
this
mechanism
to
explode
white
doors
with
very
thin
helium
shells,
and
so
thick
helium
shells
in
nature
would
come
from
a
non-degenerate
companion
like
a
helium
sdb
star,
but
thin
helium
shells
would
come
from
a
degenerate
companion
like
a
white
dwarf
white
dwarf
binary.
A
B
B
So
here's
a
movie
of
what
I
was
just
talking
about
getting
into
my
work.
We
have
on
the
left
here,
temperature
and
on
the
right
density
of
a
white
dwarf
with
a
helium
shell
sitting
on
the
surface,
we're
about
to
ignite
the
helium.
B
On
the
surface
burning
around
that
helium
shell,
you
see
a
shock
front
traveling
to
the
center
of
the
white
dwarf,
and
I
apologize
because
I'm
not
good
at
the
zoom
jumps
here,
but
the
shock
wave
will
converge,
we'll
get
thermonuclear
runaway
and
hopefully
a
1a
like
explosion,
but
again
in
order
for
it
to
be
1a.
We
want
that
center
radioactive
material
and
intermediate
mass
elements
and
then
unburnt
material.
B
So
let's
take
a
look
at
the
same
simulation
but
broken
down
into
what
is
created
so
again.
Here
we
have
that
same
density
movie.
B
Here
will
be
a
sum
of
the
mass
fraction
of
the
sum
of
all
radioactive
elements
and
here's
all
the
intermediate
mass
elements,
so
you
can
see
immediately
that
we
might
have
a
problem
that,
with
this
case,
which
is
the
case
of
a
very
large
helium
shell,
we
create
all
of
this
radioactive
material
on
the
outside
of
the
ejecta,
and
this
is
what
I
was
talking
about
with
stan
woozley's
models
in
the
1980s.
This
doesn't
look
like
a
type
1a
supernova.
B
But
this
is
a
hard
problem
to
do.
This
is
a
hard
problem
to
do
computationally.
This
was
a
2d
simulation.
Sorry,
I'm
seeing
the
the
the
questions
in
the
thing
now
feel
free
to
interrupt
me.
These
are
these
are
what
I'm
working
on
now.
The
published
work
is
all
wendy,
but
I'm
showing
the
preview
of
the
2d
stuff.
B
So
this
is
a
really
hard
simulation
to
perform,
because
we
need
to
resolve
a
huge
number
of
orders
of
magnitudes
of
scales,
nuclear
burning
fronts
in
order
to
actually
resolve
them,
which
we
typically
don't
we
typically
play
other
games
would
require
a
resolution
of
about
one
kilometer,
but
we
also
need
to
expand
it
to
the
point
that
all
of
the
burning
is
done.
All
of
the
hydrodynamics
is
done,
we're
in
a
steady
state
and
that's
about
10
to
the
fifth
kilometers
for
homologous
expansion.
B
B
So
what
I
do
is
I
use
castro
a
massively
parallel,
compressible
hydrodynamics
code
developed
by
a
group
here
at
well
there
at
lbl,
with
ann
algren
and
don
wilcox,
and
all
of
those
guys,
as
well
as
in
santa
barbara,
not
santa
barbara
stony
brook
with
mike
sengali
they're
all
amazing
and
have
been
so
helpful
to
me
through
my
phd.
B
But
I
use
castro
because
it's
capable
of
adaptive
mesh
refinement
also
it
has
the
radioactive
net,
not
the
radioactive,
the
nuclear
networks
involved
in
it.
So
I
forgot
to
mention
a
slide
ago.
B
The
just
the
hydro
resolution
is
10
to
the
five,
but
you
do
need
to
double
the
time
steps
in
order
to
do
nuclear
burning,
because
you
need
two
burning
time:
steps
for
every
one,
hydro,
time
step
so
back
to
amr
and
I'll
go
ahead
and
show
the
pretty
2d
picture
of
amr
amr
allows
me
to
create
regions
of
selective
high
resolution
refinement
so
where
I'm
performing
nuclear
burning
or
where
I
need
to
really
really
pay
attention
to
my
shot
fronts.
B
I
can
look
at
regions
that
are
a
higher
resolution
than
these
areas
that
I
still
need,
but
can
be
static
for
now-
and
this
is
a
2d
example
of
this
courtesy
of
ken
chen,
but
you
can
see
we
can
have
very
large
regions
for
the
circumstellar
material,
but
like
really
really
high
resolutions.
In
order
to
look
at
this.
As
a
case,
I
think
this
is
a
set
off
not
as
anyways.
B
B
So
I'm
going
to
pause
now
and
say
very
very
clearly.
I
just
showed
you
awesome,
2d
movies,
as
the
question
just
asked,
but
now
I'm
going
to
show
you
sedona
results
from
our
1d
published
model
set
of
hydro
models,
so
all
of
the
the
results
now
are
going
to
be
1d,
I'm
working
on
multi
id
now
and
we'll
get
back
to
that
at
the
end.
Hopefully,
but
I
just
always
need
to
pause
and
make
that
very
clear,
because
a
1d
movie
communicates
nothing,
but
these
are
1d
results.
B
So
my
1d
results
as
I've
alluded
to
this
whole
time.
Whether
or
not
a
double
detonation
is
going
to
look
like
a
type
1a
supernova
is
really
dependent
on
the
mass
of
the
helium
shell.
So
I
broken
these
light
curves
these
synthetic
light
curves
from
our
models
into
two
categories.
On
the
left.
Here
I
have
a
number
of
light
curves
of
varying
masses
of
white
dwarf
and
these
all
have
a
very
thin
helium
shells,
one
hundredth
of
a
solar
mass
of
helium
on
their
surface
on
the
right.
A
B
B
Now
these
guys
have
an
extra
bit
that
we
don't
see
in
type
1a
supernovae,
this
initial
early
flux
access,
this
initial
peak
that
comes
before
the
standard
nickel
peak,
and
this
is
due
to
that
radioactive
material
burnt
in
the
heavy
helium
shells
that
I
showed
you
in
the
2d
movie
turns
out
that
radioactive
material
is
mostly
not
nickel,
it's
mostly
iron
and
chromium
which
have
shorter
half-lifes
than
nickel,
so
it
decays
very
quickly.
B
It
also
sits
at
the
outside
of
the
ejecta,
so
it
escapes
the
diffusion
time
is
very
small.
It
escapes
very
quickly
and
that's
why
we
get
a
very,
very
early
flux,
excess,
but
like
curve
wise,
if
you'd
hide
that
early
part,
which
I
can't
do
with
my
hand
right
now,
you'd
get
you
know.
It
follows
by
a
more
standard
light
curve,
a
nickel
powered
peak
then
and
then
falling
into
a
tail.
But
this
early
double
peak
feature
is
not
something
we
see
in
type
1a
supernovae.
B
The
story
holds
true
when
we
look
at
the
spectra
again.
These
are
synthetic
spectras
split
into
thin
helium
shells
on
the
left
here
and
thick
helium
shells
on
the
right,
where
the
different
colors
are
different
masses
of
underlying
white
dwarfs
from
heaviest
at
the
top
to
lightest
at
the
bottom
again,
the
thin
helium
shell
spectra
look
like
type
1a
supernovae.
We
see
these
absorption
features
from
these
intermediate
mass
elements.
B
I
pointed
out
to
you
earlier,
like
silicone
and
calcium
and
as
we
get
to
the
lower
mass
funds
a
little
bit
of
titanium,
but
the
thick
helium
shells
again
don't
look
like
type
1a
supernovae.
B
The
helium
shell
ashes
are
so
optically
thick
that
they
actually
block
photons
from
escaping
in
this
blue
region,
and
that's
this
flat
line.
This
line
blanketing
we
see-
and
these
are
spectra
taken
at
peak
brightness.
So
this
this
line
blanketing
continues
through
a
lot
of
the
supernova
evolution
of
just
photons,
these
blue
photons
not
being
able
to
escape
through
the
helium
shell
ashes,
and
so
they
look
like
almost
similar
to
a
supernova
type,
1a
supernova,
but
just
blocking
a
huge
chunk
of
the
spectrum.
B
So
I'm
going
to
focus
now
on
the
type
1a
supernovae
because
or
at
least
the
secondary
supernovae
candidates,
because,
as
I
said
earlier,
those
are
the
ones
we
care
about
for
cosmology.
Those
are
the
ones
we
see
in
the
real
world.
Those
are
the
ones
we're
hoping
to
explain
at
least
some
of
with
these
sub
chandra
models.
B
So
when
we
plotted
our
spectra
sequence
like
this
again
heaviest
or
most
most
luminous
event
at
the
top,
because
again
the
heavier
the
mass
of
the
white
dwarf,
the
more
nickel
56,
it
creates
the
more
luminous
the
event
we
plotted
that
in
order-
and
we
see
a
lot
of
the
trends
we
hope
to
see
in
type
1a
supernovae,
like
these
intermediate
mass
features,
growing
less
pronounced
for
the
brighter
ones
but
kind
of
more.
Interestingly,
we
noticed
this
trend
that,
as
we
get,
we
increase
the
mass
of
the
white
dwarf.
B
We
also
noticed
a
blue
shift
in
this
in
in
a
lot
of
the
lines
you
can
see
it
most
clearly
in
this
silicon
2
line
that
we're
increasing
the
velocity
of
this
line
as
we
get
into
brighter
and
brighter
events.
So
the
regions
that
are
carrying
the
silicon
are
going
faster
in
these
events,
and
so
this
was
our
first
thing
that
we
said
aha.
B
So,
in
order
to
do
this,
we
chose
this
population
of
type
1a
supernovae
from
this
paper,
saying
it
all
2018,
and
these
were
all
chosen
from
popular
surveys,
but
chosen
because
they
had
fully
resolved
rise
times
of
the
light
curve.
So
these
are
the
ones
we
caught
early
on
enough
to
have
a
full
light
curve,
but
other
than
that
no
cut-offs
for
whether
or
not
they're
a
little
weird
or
a
little
dim
or
a
little
bright.
B
So
we
plotted
these
using
mb.
Here
is
the
absolute,
the
intrinsic
brightness,
so
the
absolute
brightness
of
the
peak
of
the
light
curve,
not
the
apparent
magnitude
but
the
absolute
brightness
of
the
the
peak
of
the
light
curve.
So
the
brightest
they
get
the
most
luminous
and
we
plotted
that
against
the
velocity
of
the
trough.
The
minimum
of
the
trough
of
that
silicon
line
and
then.
B
Models
on
top
of
here
and
we
saw
something
cool.
Hopefully
we
saw
that
there
were
kind
of
two
populations
of
observed
supernovae.
B
And
so
we
looked
at
this
and
said:
okay.
Well,
maybe
this
could
be
it.
This
could
be
our
two
populations
of
supernovae,
those
that
originate
from
a
subchandra
progenitor
via
a
double
detonation
mechanism,
and
maybe
these
are
something
else
and
if
you
do
the
back
of
the
envelope,
calculations
of
you
know
calculating
binding
energy
and
figuring
out
the
expected
mass
from
binding
energy
and
velocity
here.
The
the
mass
of
these
would
sit
right
around
the
chandrasekhar
mass.
B
So
maybe
these
are
the
chandrasekhar
mass
explosions
and
these
are
the
subchandras,
and
that
would
be
really
really
cool
and
it
looks
like
maybe,
but
you
know
that's
only
to
the
machine
learning
tools
of
our
eyeballs
there's,
not
enough
data
here
to
really
tell
a
compelling
story
just
on
this,
so
we
started.
Looking
at
other
axes
and
so
far,
we
have
seen
a
number
of
ways
that
these
two
populations
of
type
1a
supernovae
separate
themselves.
B
So
when
we
look
at
their
color,
so
their
b
band
their
maximum
b
band
magnitude
and
their
minus
their
maximum
v
band
magnitude,
the
ones
that
sit
in
this
cluster
are
significantly
bluer
than
the
guys
that
lie
along
this
relationship
and
they're
bluer
than
any
of
our
models.
I'll
come
back
to
that
point.
In
a
minute
I
had
a
student
this
summer
working
on
looking
at
carbon
absorption
features
in
the
supernovae,
because,
due
to
the
detonation
nature
of
these
subchandra
explosions,
they're
very
efficient
at
burning
carbon
and
leave
very
little
behind.
B
So
you
would
not
expect
to
see
a
carbon
absorption
feature
from
a
subchandra
explosion,
but
you
would,
from
you,
can
anyways
from
china's
cycle
mass
explosions,
not
all
of
them,
but
some
of
them.
So
my
summer
student
hayden,
working
with
a
carnegie
summer
program,
went
through
all
this
data
and
showed
that
the
ones
in
red
here
are
the
carbon
absorption
feature
ones
and
they
all
kind
of
sit
in
this
cluster.
B
So
that's
pretty
copacetic,
then
the
last
one
is
looking
at
nebular
emission
features.
We
also
noticed
that
the
ones
that
lie
in
this
distribution
show
stronger,
calcium
to
emission,
just
like
our
models
than
the
ones
that
sit
in
this
cluster.
B
So
we
have
now
three
different
axes
by
which
these
supernovae
distinguish
themselves,
which
is
a
start
to
saying.
Well,
maybe
we're
looking
at
two
different
classes
of
supernovae,
which
is
really
important
for
cosmology,
because
if
these
guys
are
different,
we're
not
going
to
want
to
standardize
them
the
same
way
that
we
standardize
the
ones
that
sit
in
this
cluster
or
the
normal
ones,
and
what's
even
cooler
about
this
plot
b
here,
the
color.
B
It
turns
out
that
a
lot
of
the
time,
these
fast
red
events
up
here,
have
been
historically
problematic
for
our
hubble,
constant
measurements,
because
they're
too
red
in
order
to
include
them
in
our
hubble
residuals.
B
We
need
to
invoke
maybe
they're
too
red
because
there's
extra
dust
in
their
environment
and
they're
extra
reddened,
but
what
I'm
saying
is.
Perhaps
they
are
more
intrinsically
red
than
these
events,
and
so
in
order
to
include
them
in
our
hubble
measurements,
we
need
to
come
up
with
a
new
kind
of
color
correction
magnitude
correction,
to
standardize
these
guys
than
these
guys
and
maybe
someday
we'll
be
able
to
do
that.
B
Now.
There's
a
really
really
important
question.
You
all
should
be
asking
me
at
this
point,
and
that
is
if
I
want
you
to
believe
me
that
there
is
a
significant
portion
of
type
1a,
supernovae
and
most
rates
at
this
point
say
maybe
up
to
30
that
occur
from
this
double
detonation
progenitor
system.
B
B
And
this
is
a
problem-
or
at
least
it
was
a
problem
for
a
while,
because
we
know
these
these
systems
exist.
We
have
observed
binaries
with
a
white
dwarf
and
a
helium
sdb
star
with
separations
that
would
turn
into
the
stable
mass
of
creation.
We
need
to
build
up
a
helium
shell
and
explode
them
like
this
and
at
the
time
that
I
started
this
at
the
time
that
I
published
his
first
paper
on
these
1d
models.
B
B
I
think
it
was
summer
2018
and
I
stopped
here
and
a
graduate
student
from
cal
tech
keisha
day
who's
now
starting
a
postdoc
at
mit.
He
just
got
a
hubble
fellowship,
I
believe
actually
came
up
to
me
and
he
said
I
have
an
event
you
have
to
see,
and
he
showed
me
this
and
there's
a
little
bit
of
spoilers
in
the
title
of
the
paper
there.
But
this
is
supernova,
18,
b-y-g
or
or
sorry
19
your
faces
are
covering
the
no
it's
18.
B
your
faces
are
covering
my
titles
of
my
slides,
but
so
this
was
a
thermonuclear,
odd
type,
1
supernova.
We
didn't
want
to
call
it.
I
mean
technically
it's
a
1a
because
it
doesn't
have
hydrogen
or
helium,
but
it's
not
a
normal
one
right.
So
it's
a
peculiar
type
1..
It
exploded
significantly
offset
from
its
host
and
keisha
showed
it
to
me
because
of
two
things,
because
this
light
curve
here
this
r
band
light
curve.
B
It's
too
broad
for
a
type
1a
or
as
I
would
like
to
say,
it
has
an
early
flux
excess
in
the
early
times.
And
if
you
look
at
the
spectra,
it
is
significantly
line
blanketed
in
the
blue
for
much
of
its
evolution.
B
B
B
So
I
did
I
made
this
guy,
which
I,
for
some
reason
have
left
the
masses
off
of,
but
it's
actually
the
the
2d
movie.
I
showed
you
is
the
same
mass
of
this
there's
a
0.76
solar
mass
white
dwarf
with
a
0.15
solar,
mass
helium
shell,
which
is
a
huge
amount
of
helium.
It
is
more
helium
than
I
included
in
my
original
1d
paper
in
any
of
those
models,
because
we
thought
that
it
was
right
out.
B
B
This
was
the
fit
for
the
spectra
at
the
time
of
peak
and
it
is,
I
think,
to
date,
still
the
best
model
fit.
I
have
ever
made
for
an
event.
We
show
the
line
blanketing
in
the
right
region.
We
show
the
right
velocities
for
these
absorption
features
and
the
right
width
for
this,
this
calcium
trough.
This
is
the
spectra
at
peak,
but
the
rest
of
the
spectral
evolution
was
also
great
up
through
the
point
that
I
can
actually
trust
my
rated
transport
lte
models.
B
So
we
were,
we
were
thrilled,
we
published
the
paper
and
to
date
I'm
pretty
confident
in
saying
this
is
the
most
direct
evidence
that
we
have
that
there
is
more
than
one
way
to
explode
a
white
dwarf
that
this
double
detonation
mechanism
can
explode
a
white
dwarf
because
there
is
yet
to
be
any
other
pending
theory
that
could
that
can
make
this
model.
B
There
could
be
some
day,
but
there's
not
right
now.
So
since
then,
we've
had
two
other
exciting
events
that
we
also
think
are
thick
shells
and
there
are
more
coming
out
both
through
archival
data.
We
found
a
couple,
and
hopefully
we'll
more
will
happen
in
nature
over
time,
so
keep
paying
attention
for
the
rate
of
these
guys,
I
think,
will
be
constantly
evolving.
B
But
what
about?
What's
next,
this
is
the
last
I
have
for
you,
these
slides
of
what
about
the
malt
id.
Why
haven't
I
published
them
all
tidy?
It's
been
a
couple
years.
Why
isn't
this
happening?
It
turns
out
it's
almost
prohibitively
expensive.
B
B
I
think
that
is
the
hope
of
these
kinds
of
simulations
in
the
future
and
I'm
hoping
that
the
other
modeling
groups
are
also
working
towards
making
their
simulations
gpu
capable
and
that's
how
that's
the
only
way
we'll
be
able
to
do
something
like
this
in
3d
in
anything
more
than
one
simulation,
that's
all
I
have
for
you.
So
if
you
guys
have
any
questions
for
me,
that
would
be
great,
but
thank
you
again.
A
A
So
I
guess
I'll
start
with
one,
and
maybe
you
said
this,
but
I
missed
it.
It
was
so
these
these
white
dwarfs
that
exploded
by
this
mechanism
are
dimmer
than
the
other
ones
is
that
is
that
right,
mostly.
B
Yes,
so
they
can
get
almost
as
bright
as
a
typical
as
like
the
what
we
classify
as
the
normal
1a
that
might
change
a
little
bit
in
multi
d.
You
never
know
and
there's
a
debate
of
exactly
what
the
peak
brightness
is
on
them,
but
I
would
say
I
would
more
favor
the
subluminous
end
of
the
1as
for
these
guys.
A
B
Yeah,
that
is
a
lovely
question
right
now,
we're
still
not
getting
enough
thick
shell
events
to
explain
the
number
of
binaries
we
see,
and
specifically
we
have
a
slightly
bigger
problem
of
the
ones
we
have
caught,
have
all
been
fairly
offset
from
their
galactic
host
and
we
need
to.
We
don't
actually
expect
them
all
to
be
in
older
environments.
B
Yes,
so
we
need
to
start
seeing
them
in
other
environments
as
well.
In
order
for
this
to
actually
work
out
or
we
need
to
start
seeing
the
binaries
in
older
environments,
you
know
one
or
the
other,
but
but
again,
I
think
we
also
had
a
problem
historically
that
we
didn't
know
what
these
were.
So,
if
you
saw
one
you
probably
ignored
it,.
B
I
submitted,
I
submitted
a
you
know,
a
request
for
time
on
pearlmutter,
when
that
is
a
thing
that
can
happen,
and
I've
also
requested
to
start
being
able
to
use
my
current
allocation
on
promoter.
But
I'm
still.
A
Oh
sure,
yes
yeah,
if
you're
not
on
there,
already
we'll
get
you
on
right
away.
Okay,.
B
A
No
well,
if
not
thank
you
again,
I
really
enjoyed
the
talk
and
thanks.