►
From YouTube: DevoWorm (2023, Meeting #24): SAM and MedSAM, making monkey embryoids in culture, shear jamming
Description
GSoC Updates: SAM and MedSAM, integrating SAM into 2-D and 3-D models of the embryo. More developments in biological embryo models (embryoids in Monkeys). Shear jamming and nonlinear phase transitions in soft biological materials (embryos and cell collectives). Attendees: Sushmanth Reddy Mereddy, Bradly Alicea, Susan Crawford-Young, Morgan Hough, and Jesse Parent.
B
A
B
B
B
A
A
C
Two
weeks
on
the
same
problem,
why
couldn't
it
solve?
But
today
morning
I
was
able
to
clear
it
out
with
the
Sam
model.
Somehow
yeah
Mike
told
me
to
drop
the
same
idea,
but
I
was
still
continuing
on
it.
I'm
seeing
pretty
good
changes
in
the
model
and
my
model
will
be
trained.
I
started
working
on
the
devonet
model.
Also,
since
last
week,
I
was
able
to
give
most
of
my
time
to
gsoc,
and
this
can
I
share
my
screen
and
show
it
to
you.
Yeah.
B
C
C
In
a
classical
way,
I
couldn't
Implement,
because
the
model
architecture
was
completely
different.
Then
I
found
some
tutorial
on
the
Internet
or
Internet
met
Sam
this
repo.
They
are
trying
to
segment
a
3D
fine
tune,
a
3D
data
set
using
some
model
and
the
major
problems
using
this
code.
Repo
is
I
need
to
change
a
lot
of
things,
because
here
they
are
extracting
only
single
bounding
box,
not
more
than
one
but
right
now
we
have
the
images
of
this
type.
B
C
We
need
to
create
bounding
box
for
each
cell
and
then
we
need
to
give
as
a
prompt
to
the
Sam
model.
But
here,
whatever
they
code
wrote,
they
wrote
it
is
just
creating
only
one
bounding
box,
so
I
need
to
change
to
take
multiple
bounding
boxes,
then
I
need
to
start
training.
The
model
and
I
was
stuck
whole
week
on
the
training
of
Sam
model,
because
here
there
are
so
many
unknown
things
they
haven't
mentioned.
C
C
Here
we
are
giving
bounding
box
and
whatever
image
encoder
gives
us
the
output
that
will
go
to
this
mask
decoder
and
in
these
days
I
was
thinking
just
giving
an
image
I'm.
Giving
directly
image
to
must
be
Kodak
I'm,
giving
masks
this
image
directly
to
this
folder
and
it
is
not
working.
When
I
was
going
through
the
mid-sam
code,
then
I
understood
first,
they
are
giving
to
the
image
encoder.
Then
they
are
expecting
some
value
see
here
image.
C
C
Yeah,
this
is
the
thing
I
have
been
working
on
right
now.
We
have
the
images
2D
3D
images
that
we
need
to
convert
into
a
Nia
file
form
so
I,
converted
them.
I
wrote
this
code
and
I
tried
to
convert
them
and
they
are
converting
perfectly
after
that.
I
was
resizing
the
image
into
this
form
of
size,
1
0
to
4,
comma,
one
zero,
two,
four,
only
that
size
of
image
can
only
be
taken
by
Sam
model,
not
other
than
that
it
couldn't
take
any
other
mode.
So
we
are
here.
C
C
C
I
started
writing
the
code.
First
I
wrote
about
data
that
this
part
is
about
completely
of
taking
a
tip
file
because
whatever
we
have
the
data
set
from
cell
tracking
calendar
okay,
we
have
three
things
here.
First,
one
here,
sorry,
these
are
the
original
images
of
cells,
electric
Point
form-
and
these
are
the
ground
roots
of
the
cell.
C
Here
we
have
the
some
dots
kind
of
structure
where
we
can
give
to
the
model.
So
right
now,
I
have
wrote
the
data
sets
file
here.
So
first
we
are
taking
a
tip
file
as
input.
Then
we
have
the
3D
data
set.
Whatever
data
set
we
are
using,
it
is
a
3D
data,
so
I
created
a
function
about
it.
Taking
an
input,
a
3D
data
set
and
I
was
pre-processing
them
rotating
and
etc,
etc
and
I'm
taking
get
get
image.
C
A
function,
number
and
label
function,
get
label
label
means
yeah
gone
through
profit
and
trying
to
pre-process
them.
This
took
me
a
while,
because
I
need
to
understand
what
is
the
resolution
they
took
in
the
microscopy.
I
need
to
go
through
the
whole
data
set
documentation
to
understand
what
is
what
data
set?
Sorry
microscopy
resolution,
because,
according
to
that,
I
need
to
change
the
parameter.
C
C
If
you
see
yeah,
no
I,
think
I
don't
have
the
file,
but
they
have
wrote
this
code
in
another
framework.
Called
chainer.
Chainer
is
also
in
another
framework
like
pytots,
but
I
am
trying
to
write
it
in
pytos
and
implement
it.
So
the
data
sets
file
is
over
by
next
week.
I
am
trying
to
implement
the
model
architecture
mechanism,
nuclear
segmentation,
Network
nucleus
detection,
Network
and
a
trainer
file
for
that
yeah.
A
C
B
C
C
B
C
Yeah
yeah,
my
final
goal
would
be
I
will
be
comparing
the
owner
model
and
Sam
model
and
which
model,
because
Sam
is
trained
on.
Billions
of
images
and
download
is
just
crying
on
the
ceilings
I'll
see,
which
is
performing
better
over
the
cell
system,
mainly
in
ceiling
events,
and
my
own
motive
would
be
like
making.
C
B
C
A
C
A
C
Rgb
three
Channel
images
and
whatever
the
cell
images
we
have,
it
is
a
grayscale
images,
so
we
need
to
convert
them
from
grayscale
to
RPD.
Then
convert
things
after
converting
to
RTD.
We
are
giving
converting
it
into
some
image.
So
this
is
our
picture.
Someone
someone
wants
to
implement
some
model.
They
need
to
get
clear
idea
on
the
image
encoder
from
quota.
They
don't
get
idea
on
this
thing.
They
could
never
implement.
C
A
C
C
C
Just
segment
the
objects
we
can
take
out,
whatever
object,
we
want
and
yeah
generating
images
it
can
be
used
in
our
in
our
lineage
for
sorry
embryo
embryo
Gan
model.
Is
there
right
in
right
England
there
also
we
can
use,
it
has
different
type
of
implementations,
but
the
model
was
not
completely
released.
C
C
C
Model
is
also
working
fine
by
next
week.
I
will
try
to
implement
all
model
architecture.
According
to
the
image
resolutions
we
have,
after
that,
I
will
try
to
implement
a
loss,
function
for
and
I
need
to
talk
with
mayuk
once
for
this
loss
functions,
functions,
I,
don't
know
which
loss
function
can
be
used
for
microscopy
images
to
take
it.
B
C
I
mean
these
images
have
different
type
of
noise
in
them
actually
to
create
to
remove
those
noises
to
know
about
the
which
type
of
microscopy
they
used
and
which
type
of
lens
they
use
to
remove
the
noise
and
add
some
what
we
call
gaussian
noise
to
it
and
remove
the
noises
so
yeah.
Actually,
whatever
the
data
set,
we
have,
it
has
pretty
clear
documentation
about
the
microscope.
C
C
A
C
Bradley,
these
are
the
microscope
images
original
microscope
image.
This
is
the
segmentation
Maps,
and
these,
could
you
see
these
dots?
These
represent
the
centroid
position
of
this
cell.
If
this
cell
is
there,
for
example,
the
centroid,
if
dot
represent
centroid
of
that
cells
in
the
exact
place
by
this,
we
can
accurately
extract
the
centroid
of
a
different
time
lapse
and
we
can
segment
it.
We
can
extract
as
it
is
a
3D
data
set.
C
We
can
extract
the
volume
of
excel
also
and
future
work
around
this
model
would
be
implementing
the
lineage
population
model
for
C
elegans.
On
3D
data
set
using
this
data
set,
the
main
Feature
work
would
be
around
lineage
population
model.
Only
first,
we
will
try
to
implement
this
models
in
instant
segmentation.
After
that,
we
will
try
to
implement
lineage
population
tracking
the
cells,
which
is
a
parental,
which
is
the
daughter
cell.
According
to
the
colors
here,.
C
My
updates
for
next
week,
complete
week,
I,
will
work
on
the
demonet
and
some
parallelly
and
I
have
took
permission
from
my
college.
They
actually
provided
me
with
I'll,
give
my
full
time
on
g-soft,
Excel
and
yeah.
That's
it.
My
updates
are,
if
anything
is
there,
I
will
approach
you
or
my
I
will
I
will
think.
I
will
have
a
meeting
with
him
within
two
days
for
that
loss,
functions
and
I
will
let
you
know
if
any
improvements
are
there
or
if
I
found
out
any
problems,
yeah.
B
That
sounds
great.
Thank
you.
Yeah
great,
this
looks
like
such
month
is
making
progress
on
the
Sam
model
and
getting
that
sort
of
integrated
into
the
d-volar
Divo
learn
Pantheon
I
guess
they'll
be
nice
to
have
like
this
whole
sort
of
you
know
having
the
three
sort
of
tool,
boxes
or
models,
and
then
you
know
having
a
bunch
of
stuff
that
people
can
use
around
that.
C
C
On
that
by
next
week,
I
have
a
thing
in
my
mind:
okay,
first,
because
right
now
for
me,
it
took
around
one
and
a
half
week
to
get
data
set.
We
want
to
use
I
need
to
understand
it,
but
if
we
have
something
kind
of
stuff,
if
see,
if
Susan
was
taking
some
microscope
images
and
sending
it
to
you,
we
don't
know
how
to
make
the
labels
of
it.
C
If
something
is
there
on,
the
internet
make
labels
of
each
cell
this?
This
is
the
daughter
cell,
enabling
it
by
a
b
or
sometime,
it
can
be
used
as
data
set
in
future
purposes.
I
can't
explain
clearly
in
like
this,
but
I
will
make
a
PPT
next
week
and
I
will
show
it
to
what
was
there
in
my
and
I
found
out
some
resources.
It
just
takes
around
three
days
to
implement
it
over
internet
and
host
it
in
like
a
web
thing
by
next
week.
I
will
show
you
to
a
small
demo.
B
B
C
We
have
something
something
that
weights
right
now:
Sam
model,
if
you
use
that
model
segment
it
after
that,
typically
after
segmenting
this
we
can
enable
it
right.
This
is
a
parental.
The
daughter
cell
I
couldn't
explain
this
way,
but
next
week
I'll
show
you
a
PPT.
You
will
get
a
clear
idea.
What
was
then.
B
So
thank
you
for
that
update
this
month.
Look
forward
to
next
week.
Himanshu
messaged
means
that
he
couldn't
make
it
today,
so
no
updates
from
him
right
now,
although
he'll
be
updating
Us
in
the
slack
I
think
everything's,
okay
with
him
and
I,
don't
know
if
I
don't
think
geohang
will
be
here
this
week,
but
he
last
week
he
gave
a
nice
discussion
about
some
of
his
ideas
for
sort
of
how
to
integrate
the
graph
neural
network
approach
with
some
of
the
network.
B
B
A
B
Morgan,
in
a
put
in
he
brought
my
this
to
my
attention
at
one
point
and
we
last
week
we
talked
about
creating
human
embryoids.
So
last
week
we
had
like
I
did
a
presentation
of
two
papers
that
have
been
written
on
human
embryoids.
This
is
where
they're
taking
human
embryos
in
there
or
they're
taking
stem
cells
from
humans
and
they're,
creating
these
human
embryoids.
B
If
we
just
did
an
assay
of
your
muscle,
we
would
find
these
same
genes,
and
so
you
know
it's
said
to
be
like
us,
a
muscle
cell
or
sometimes
they'll
use
the
term
muscle
like
cell,
because
you
know
we
don't
really
know
you
can
do
histology
on
the
on
the
emerging
cells
or
some
of
the
tissues,
and
you
can
see
commonalities
in
the
cell
morphology.
You
can
do
these
type
of
molecular
analyzes
and
then,
of
course,
we've
mentioned.
We
talked
about
the
role
of
implantation
in
primate
embryos
or
actually
mammalian
embryos.
B
These
embryoids
aren't
I,
don't
know
if
they're
implantable,
I
think
they're
just
growing
in
a
in
a
culture
and
they're
they're,
differentiating
and
they're
forming
these
tissues,
so
they
call
it
an
embryoid,
and
this
is
something
that
goes
beyond
some
of
the
organoid
work
that
we've
talked
about.
Where
you
have
these
three-dimensional
cell
cultures
and
you
have
different
cell
types
that
are
derived
from
stem
cells
and
you
get
you
know,
maybe
working
in
concert,
so
you
maybe
form
like
a
system
of
different
tissues.
B
Sometimes
it's
just
a
an
assortment
of
different
tissues
in
the
same
culture
and
when
I
say
tissue
I
mean
like
tissues
that
are
made
all
up
of
these
mature,
like
cells,
so
like
muscle-like
cells
or
fibroblasts,
like
cells
or
whatever
epithelial
like
cells.
So
we
don't
know
exactly
if
they're
exactly
the
same
type
of
cells
you'd
see
in
in
the
biological
context
like
in
the
organismal
context.
B
We
don't
you
know,
we
really
don't
know
if
they
work
together
in
the
same
way
as
you
would
see
in
an
organizable
context,
but
it's
promising
work,
and
so
this
week
this
is
all
a
preface
to
what
I'm
going
to
talk
about
now,
which
is
that
we
talked
about
humans
last
week
now
they're
doing
they
also
have
work
in
monkeys.
So
we're
going
to
talk
about
that.
B
So
this
is
an
article
in
nature,
and
this
is
of
course,
was
shared
by
Morgan
lab
grown
monkey.
Embryos
reveal
in
3D
how
organs
begin
and
at
25
days
old
specimens
could
be
the
oldest
primate
embryos
ever
grown
outside
the
womb,
and
so
this
is
the
25
dealed
monkey
embryo
stained
with
a
conventional
dye,
which
is
the
pink
image
and
all
right.
So
this
is
the
conventional
die,
and
then
this
is
the
fluorescent
dye
coming
up.
So
you
can
see
that
there's
a
there
are
multiple
colors
here.
B
The
blue
patches
in
the
fluorescent
dinos
denotes
cell
nuclei.
So
you
have
a
lot
of
nuclei
there.
The
done
Central
patch
of
green
and
red
mark
trophoblasts.
So
you
can
see
the
cells
as
blue
and
then
these
trophoblasts
is
green
and
red,
which
let
me
do
it
again:
okay,
so
green
and
red
are
in
here
and
those
are
the
trophoblast
cells.
So
this
is
where
this
is.
These
are
cells
in
the
embryos
outer
layer,
so
this
is
a
25
year
old
embryo.
B
This
is
what
they
look
like
in
utero
at
25
days,
so
there's
a
structure
of
different
types
of
cells.
Already
you
have
an
inner
cell,
mass
and
outer
layer
and
so
forth.
We
have
this
sort
of
structure
coming
up
and
so
okay,
so
scientists
have
cultivated
monkey
embryos
in
the
laboratory,
long
enough
to
watch
the
beginning
of
organ
formation
and
the
development
of
the
nervous
system.
B
The
embryos
reached
the
age
of
25
days,
making
them
what
might
be
the
oldest
primate
embryos
to
be
grown
outside
the
womb.
So
this
again
is
where
their
actual
growing
primate
embryos,
and
so
this
is
actually
based
on
two
independent
papers
from
I
think
this
is
from
2023.
This
is
from
this
year.
This
is
11th
of
May
in
2023,
so
there
are
two
papers
here:
I
think
this
gong
paper
and
the
joy
paper,
if
I'm
not
mistaken,
yeah
that
those
are
one
and
two
so
the
gong
paper.
If
we
look
at
that.
B
B
B
Early
neuralation
features
are
demonstrated
in
cultured
embryos
and
epigenetic
features
of
three
germ
layers
in
primate
are
unveiled
in
cultured
embryos,
so
they're
actually
able
to
get
the
three
germ
layer,
sort
of
epigenetic
features
of
that
and
I,
don't
know
what
they
mean
by
epigenetic
features,
but
the
graphical
abstract
is
here
where
they
have
this
prolonged
embryo
culture.
In
this
special
system.
B
Here
you
get
it's
pretty
small.
You
can't
really
see
it
but
you're
starting
to
get
an
ergulation
here,
you're
starting
to
get
a
neural
plate
border
and
differentiation
of
the
neural
crest.
Then
you
get
the
neural
tube
forming
a
little
bit
later,
and
then
you
get
the
amnion
and
the
mesoderm
and
the
endoderm.
So
you
get
all
these
different
things
happening
in
the
in
this
embryo
model.
B
So
that's
those
are
those
two
papers
that
were
published.
Let
me
go
back
up
to
the
top
here
or
where
we
were
before.
So
it's
very
impressive.
This
is
actually
one
of
the
authors
of
the
human
papers.
We
talked
about
a
developmental
biologist
at
Caltech.
It's
going
to
bring
a
lot
of
new
insights,
so
basically
they're
going
3D
they're
building
this
three-dimensional
cell
culture.
B
So
previously
both
teams
had
managed
culture,
monkey
blastocysts
in
Petri
dishes
for
up
to
20
days,
a
blastocyst
is
a
ball
of
dividing
cells.
So
it's
like
this
massive
cells
and
a
sphere.
That's
what
you
have
a
blastocyst,
so
that's
kind
of
what
they
had
been
able
to
manage
previously
and
those
last
two
papers.
Three
and
four
were
examples
of
that
work
past
that
point
all
of
the
embryos
had
collapsed,
making
it
impossible
to
see
more
advanced
stages
of
their
development,
such
as
early
signs
of
the
nervous
system
and
organ
formation.
B
So
if,
in
previous
work,
they've
been
able
to
get
these
sorts
of
balls
of
cells
and
they
were
dividing,
but
they
couldn't
get,
you
know
any
sort
of
structure,
they
couldn't
get
a
lot
of
differentiating
cells
and
and
keep
them
in
this
anatomical
context.
So
they
say
the
embryos
had
collapsed.
That
just
means
that
they
didn't
survive
to
that
stage.
So
that
was
the
problem
that
they
had.
B
They
couldn't
get
past
that
stage
of
development,
but
in
the
new
studies
the
researchers
grew
monkey
embryos
and
small
vials
of
culture
medium,
which
allowed
the
embryos
to
grow
in
three
dimensions
as
they
would
inside
the
womb.
So
they
had
these
three-dimensional
cultures
that
we
saw
an
example
of
in
the
second
paper.
B
Both
teams
coaxed
their
embryosa
survive
25
days
after
fertilization,
and
then
this
is
where
they
have
some
content
here.
If
you're
interested
about
some
of
this-
and
so
these
are
so,
they
obtained
egg
cells
and
fertilize
them
in
the
lab,
with
sperm
collected
from
their
male
counterparts
a
week
later,
they
placed
the
resulting
blastocysts
into
a
gel-like
substance
in
some
most
wonderful
containers
and
watched
them
grow
for
25
days.
B
So
about
two
weeks
after
fertilization,
more
than
half
of
the
embryos
had
an
embryonic
disc
or
a
flat
mass
of
cells.
These
discs
eventually
form
the
three
main
layers
of
the
cell
body.
These
are
the
three
germ
layers.
The
lab
grown
embryos
also
showed
genetic
features
similar
to
those
seen
in
natural
monkey
embryos
within
the
same
time,
frame
by
day
20,
the
embryosa
developed
a
neural
plate.
One
of
the
early
Hallmarks
of
the
nervous
system
is
a
natural
embryos,
this
plate,
thickened
and
bent
into
a
tube
that
forms
the
basis
of
the
Bremen
spine.
B
That's
that's
all
observable
on
this
model,
and
you
can
also
see
the
three
germ
layers
and
then
you,
you
know
you
have
earlier
stages
as
well.
They
also
pinpointed
cells
would
eventually
become
motor
neurons.
The
insights
gleaned
from
the
labyrin
embryos
will
help
researchers
develop
a
better
understanding
of
early
embryo
development
and
primates,
so
this
is
another
example
in
primates,
a
monkeys
not
in
humans,
but
you
have
this
these
two
competing.
So,
basically,
primates
were
beginning
to
be
able
to
do
this
in
different
species.
So
you
know
every
species
is
different.
B
We
have
model
organisms
like
Mouse,
and
human
and
fruit
fly
and
things
like
that,
and
they
all
have
differences
in
terms
of
how
they're
reared
and
some
of
their
developmental
features
and,
like
you
know
it's
always,
you
know
a
struggle
to
apply
techniques
across
these
model
organisms
and,
in
this
case,
they're
actually
able
to
kind
of
do
this
across
model
organisms.
In
a
sense,
they're
able
to
you,
know
kind
of
apply
the
same
type
of
techniques
and
observe
some
of
these
things,
so
they
also
were
able
to
get.
B
They
were
able
to
do
some
interesting
thing
with
the
origins
of
blood
so
so
actually
to
add
stronger
mechanical
support
for
the
embryos.
So
one
of
the
problems,
of
course,
is
the
mechanical
Integrity
of
the
embryo,
these
two
different
types
of
cell
culture,
so
they
use
that
cell
culture
and
they
added
glucose
to
provide
them
with
energy
as
they
grew.
So
they
were
able
to
do
certain
things
in
the
cell
culture
to
make
this
to
provide
stronger
mechanical
support,
as
the
embryos
grew.
B
When
tan
and
colleagues
took
a
closer
look
at
the
embryos
mesoderm
cells,
they
found
that
some
had
differentiated
into
heart
muscle
cells
and,
as
it
had
matured
into
cells
found
in
the
lining
of
blood
in
lymphatic
vessels,
so
they're
starting
to
get
a
circulatory
system
they're
starting
to
get
some
heart
muscle,
starting
to
get
some
vasculature.
That's
growing
and
they're,
getting
perhaps
some
blood
and
other
types
of
cells
associated
with
that.
B
The
team
also
pinpointed
cells
that
develop
into
connective
tissues
and
one
that
form
the
foundation
of
the
digestive
system,
so
they're
actually
getting
these
precursors
to
certain
parts
of
the
different
different
organs
that
are
going
to
form.
So
not
only
are
they
getting
like
these
precursors
of
the
nervous
system
but
they're
getting
other
precursors
as
well,
and
these
are
less
further
along
than
the
neural
crest
but
they're
able
to
they're
able
to
identify
them.
B
A
B
B
B
When
we
talk
about
the
embryo,
there
are
a
lot
of
correspondence
between
what's
going
on
the
embryo
and
what
we
see
more
generally
with
these
soft
materials,
and
so
one
of
the
things
that
we've
talked
about
in
the
meetings
and
that
come
up
in
some
of
the
you
know,
migratory
patterns
of
the
embryo
cell
is
shear
jamming,
and
so
this
is
a
nice
review.
This
is
on
the
archive
one
in
the
condensed
matters
off
materials.
B
Section
looks
like
it's
really
complete.
It
has
a
lot
of
well,
they
gave
a
whole
outline
I.
Guess
that's
only
16
pages,
but
I
mean
it's
good.
So
this
is
the
abstract
here.
Jamming
is
a
ubiquitous
phenomenon
that
appears
in
many
soft
matter
systems,
including
granular
materials,
Foams,
colloidal,
suspensions,
emulsions,
polymers
and
cells.
So
everything
and
then
plus,
sells
when
jamming
occurs.
The
system
undergoes
a
transition
from
flow
like
to
solid-like
States.
B
A
B
Water
and
you
know
the
the
if
it's
flowing
through
a
pipe
it
thickens
as
it
gets
colder
and
then
it
freezes
solid.
So
this
is
a
common
phase
transition.
These
kind
of
systems,
conventionally
the
jamming
transition,
occurs
when
the
system
reaches
a
threshold,
jamming
density
under
isotropic
compression.
B
B
Actually,
with
respect
to
some
Cellular
Systems,
where
you
just
get
these
local
jamming
thing,
you
know
local
jamming,
where
you
might
get
a
cluster
of
things
that
become
a
solid
and
then
the
rest
become
kind
of
fluid
like
so
that's
that's
kind
of
an
interesting
State
as
well
or
a
macro
State
I
guess,
but
recent
studies
reveal
that
jamming
can
also
be
induced
by
shear.
So
Chris
we've
talked
about
shear
and
the
role
of
share
in
in
cellular
mechanics.
B
Sheer
jamming
has
attracted
much
interest
in
the
context
of
non-equilibrium
phase
transitions,
of
which
jamming
phase
transitions
are
one
mechanics
in
reality
of
amorphous
amorphous
materials.
Here
we
review
the
phenomenology
of
share
jamming
and
its
related
physics.
We
first
describe
basic
observations
obtained
in
experiments
and
simulations
and
results
from
theories
share.
Jamming
is
demonstrated
as
a
bridge
that
connects
rheology
of
a
thermal
soft
spheres
and
thermal
hard
spheres,
so
these
are
I
guess
with
they're
doing
thermal
characterization
with
in
terms
of
temperature
dependence
based
on
a
generalized,
generalized
jamming
phase
diagram.
B
A
universal
description
is
provided
for
sheer
jamming
in
a
frictionless
and
frictional
system,
so
they
do
this.
You
know
they
have
friction
and
frictionless
systems.
We
further
review
the
Ice,
ISO,
isostasticity
and
criticality
of
the
sheer
jamming
transition
and
the
elasticity
of
shared
Shear
Jam
solids.
B
So
this
is
where
we're
looking
at
the
sure,
jamming
transition
and
some
of
its
properties
and
then
the
elasticity
of
some
of
these
jam
solids.
So
even
though
it's
a
solid,
it's
still,
you
know,
has
some
soft
matter
properties
that
are
interesting.
They
have
a
greater
greater
elasticity
say
than
something
that
isn't
that
you
know.
B
B
This
is
the
jamming
density
here
on
this
axis
and
then
they're
simulating
this
in
frictionless,
poly
dispersed
thermal
hard
spheres
and
a
their
the
GM
states
that
value
J
or
compress
compression
quenched
from
equilibriums
liquid
States
at
EQ
that
are
prepared
using
the
swap
algorithm.
The
density
separates,
Shear,
yielding
and
Shear
jamming
behaviors
in
thermal
hard
sphere
glasses,
then
B
is
the
dependence
of
this
parameter
on
the
intraparticle.
Friction
coefficient
mu,
Sub,
Zero
and
simulated
monologus
per
say
thermal
soft
spheres.
B
So
this
is
where
you
have
this:
this
is
this
parameter
here.
Let's
say
I
think
this
is
the
isotropic
isotropic
jamming
density
on
this
axis.
This
is
one
over
P
hat.
This
is
I,
guess
this
symbol,
which
is
jamming
density,
and
they
showed
us
for
three
different
values:
EQ
J
and
C,
and
then
this
is
where
you
turn
this
into
an
inter-particle
friction
coefficient
or
you
look
at
the
dependence
of
this
on
the
interparticle.
Friction
coefficient.
B
B
B
And
that's
then,
the
experiments
they
get
into
direct
granular
matter
and
dense
suspensions,
so
this
is
sure
jamming
and
granular
experiments.
This
is
a
coetz
system
used
in
the
experimental
photoelastic
disks
where
the
state
is
unjammed,
then
Force
network
of
a
sheer
Jam
State
visualized
by
an
image
processing
technique.
So
this
is
a
force
Network
here
we
talked
about
networks
last
time
with
respect
to
embryo
networks,
but
we
also
have
these
Force
Networks
that
can
form
well.
B
B
And
then
this
is
the
pressure
string
curve
here.
This
is
sheer
jamming
accompanied
with
an
Abrupt
onset
of
finite
pressure
can
be
identified
at
a
value
of
0.2.
So
it's
like
right
here
where
this
share
jamming
occurs
down
here
at
this
end
of
the
function.
So
you
have
this
phase
transition.
Where
you
have
fluctuation
around,
you
know
this
pressure
and
then
it
falls
off
and
that's
your
phase
transition
and
so
there's
something
that
goes
on
in
this
network.
That
changes
when
you
get
this
share
jamming
transition,
and
so
it's
kind
of
interesting.
A
A
B
About
earlier
yeah
yeah,
so
this
is
a
unified
screen
density
phase,
diagram
of
sure
jamming,
yielding
and
what
they
call
Gardener
transitions
in
a
thermal
soft
spheres
and
thermal
heart
spheres.
So
the
gesturing
curves,
you
saw
in
the
last
figure
that
I
showed
kind
of
translate
into
this
phase
diagram.
So
for
these
different
values
you
have,
if
you
compare
them
across
the
range
of
values,
you
get
these
different
regions
that
have
like
different
properties.
So,
for
example,
you
have
so,
let's
see
open
cyan
triangles
represent
the
sheer
jamming
line.
B
So
you
have
these
triangles.
This
is
the
sheer
jamming
line
obtained
in
a
thermal
soft
spheres.
Filled
blue
triangles
represent
the
sheer
jamming
line
obtained
in
thermal
hard
spheres.
So
that's
here
these
blue
triangles,
let's
see
if
they
have
any
other
symbols
or
okay,
but
these
show
the
different
state,
the
different
states
that
you
get
so
you
get
in
Tropic
flow.
B
Here
you
get
stable
glass
in
this
region.
You
get
a
jam
of
solid
here.
You
get
mechanical
flow
here
and
then
there
are
these
intermediary
States
that
they
don't.
Let's
see,
mg
is
marginally
stable,
glass
and
then
SJ
is
a
sheer
jammed.
So
these
are
different
types
of
like
this
is
not
a
jam
solid.
This.
A
B
B
So
there
are
these
different
types
of
philanthropic
flow,
mechanical
flow,
and
so
those
are
also
there's
an
entropic
yielding
line
with
the
brown
diamonds.
That's
entropic
flow
phase
and
yeah.
So
that's
that
just
basically
characterizes
these
different.
It's
basically
telling
us
that
these
phase
transitions
move
us
from
one
type
of
regime
to
another
and
that
there's
some
transitional
regimes
in
there.
This
is
share
jamming
and
two-step
yielding
so
this
is
kind
of
where
they
kind
of
get
into
the
details
of
sheer
jamming,
and
they
talk
about
that.
B
There's
this
jamming
phase
diagram
at
zero
temperature.
So
this
shows
like
the
jammed
and
unjammed
components,
there's
isotropic
jamming,
which
is
IJ
over
here
in
this
graph
a
and
then
there's
sheer
jamming
on
this
side.
So
sheer
jamming
is
a
different
thing
than
isotopic
jamming
and
then
there's
the
unjammed
part
of
the
graph
here,
and
then
you
see
this
process
of
sheer
jamming,
dilatency
and
hardening.
So
you
get
this
even
within
the
jammed
phase.
B
Something
that's
initiated
through
the
share
process,
we're
held
together
through
the
share
process,
and
then
the
unjam
state,
of
course,
is
what's
not
jammed.
It's
moving
around
so
nice
physical
characterization
here
and
so
it's
Isis
stasity,
is
as
an
isotropic
jamming.
The
share
jamming
transition
always
occurs
under
the
isostatic
condition.
B
So
this
is
what
they're
referring
to
when
they
refer
to
that
the
removal
of
a
single
Bond
of
an
isostatic
system
could
result
in
a
floppy
mode
that
extends
in
the
entire
system.
Basically,
it
can
introduce
an
instability,
thus
approaching
the
jamming
transition.
There
is
a
diverging
mechanical
length
which
resembles
the
diverging
correlation,
like
they're,
on
a
standalone
thermal
critical
point.
So
there
are
these
aspects
of
of
stability
that
are
important
here,
and
this
is
just
where,
if
you
remove
parts
of
the
system,
you
can
get
different
effects.
B
So,
basically
is
it
becomes
an
isotropic.
You
get
you
get
instabilities
that
either
result
in
a
phase
transition
or
not,
and
so
that's
that's
an
interesting
kind
of
point
because
you
usually
think
of
like
you
know,
we
don't
really
think
of
a
lot
of
the
sort
of
an
isotropic
aspects
of
this
where,
if
you
have
variation
in
the
system,
what
is
the
consequence
of
that?
Well,
especially
over
time,
because
these
systems
do
evolve
over
time
so
and
then
they
talk
about
approaching,
share
jamming
from
above.
B
So
they
talk
about
summarizing
the
counterparts
and
isotropic
jamming,
which
have
been
accurately
measured
in
many
studies.
They
give
a
lot
of
math
here.
It
basically
describes
this
according
to
equation
10,
which
is
this
the
sheer
Jam,
the
shear
modules
G
changes
this
continuously
at
Shear,
jamming
and
sharp
contrast
to
the
case
of
isotropic
jamming,
which
so
we
talked
about
sheer
jamming
and
isotropic,
jamming
being
different,
where
G
vanishes
continuously
at
jamming.
So
this
is
G
in
in
equation
10.,
which
actually
is
a
Shear
modulus.
B
The
macroscopic
friction
coefficient,
which
is
Mu,
is
also
equivalent
to
that,
and
then
it's
also
equivalent
to
P
so
they're
relating
all
that
then
they're
talking
about
how
G
changes,
discontinuously,
Shear
jamming
and
that's
different
from
isotropic
jamming
in
equation.
Six,
which
is
up
here
down
here
actually
G,
is
related
to
how
these
two
things
here:
the
share
modules
G
vanishes
continuously
at
the
jamming
transition.
So
they
talk
about
this
relate
set
of
relations.
B
G
vanishes
continuously
jamming.
This
difference
is
explained
by
the
elasticity
Theory.
The
last
scaling
in
equation.
11
is
a
consequence
of
the
anisotropy
of
bond
orientations
denoted
by
a
unit
Vector
at
distance
R.
So
they're
they're
laying
out
the
mathematical
argument
for
this
and
they're
pointing
to
this
difference
between
this
sort
of
share,
jamming
and
isotropic,
jamming
and
and
linking
that
to
elasticity,
Theory
I
think
that's,
that's
all
because
they
talk
about
related
rheological
phenomena
and
then
that's
the
end
of
the
paper,
and
they
have
a
lot
of
nice
references
here.
B
A
B
A
It's
kind
of
interesting
because
bone
is
opposite
of
blood,
but
yet
bone
is
what
produces
blood
right.
It's
bone
is
the
hardest
material
in
the
body
and,
basically
blood
is
the
softest
or
most
flowing,
of
course,
right
and
so
I
don't
know.
The
contrast
is
neat
and
then
bone
produces
blood.
So
I
don't
know
just
sort
of.
B
B
Well,
not
only
that,
but
you
start
out
apparently
with
something
that
is
like
a
set
of
stem
cells
that
are
differentiating
into
these
components
of
the
circulatory
system,
and
then
what
then,
your
bone
comes
into
play
where
it's,
the
marrow
is
actually
producing
the
blood
or
rejuvenating
the
blood,
and
that
you
know
that
has
to
be
kind
of
the
switch
over
between
just
kind
of
producing
this
stuff
through
stem
cells
that
are
freely
out
there
versus
being
in
a
place.
I
guess
it.
You
know
it's
like.