I thought I'd point out
that if you're going to

place a month schedule in
March two more people.

Well I just want to point
out that next time.

Two weeks from today.

We'll also be doing it in sort of
the introduction part of the seminars

giving thank you to all the volunteers we
have the last couple of weeks with the K.

through twelve activities that hosting
the high school in middle school kids from

the last couple of weeks so worked out of
that will happen in two weeks from today.

So our speakers today is
curse America person with his

undergraduate work here at Georgia Tech
getting this sort of science in or

better of engineering temp and
nearing you done with us.

The University of Texas to do his
doctorate work getting his Ph D.

in chemical engineering and
then did a two year stint as a post doc

at nest in their Palmer's
division before coming back

to Georgia Tech first as an assistant
professor and I was an associate professor

in the Andes school of chemical and
biological engineering.

I work he is an associate professor
in the Simmons faculty fellow.

That's right and he directs the advanced
polymer then film and coordinates your

wheels Research Group and he's going to
assume tell us something about that today.

Yes OK This is a pointer.

OK so

it's always a little bit more daunting
to to speak to people that you know or

who are in your neighborhood and
then it has to go on.

To Chicago or somewhere and
speak to people he may never see again.

So feel free.

If you if you have questions
later to contact me at

the school of conclusionary
it's just Carson dot Meredith.

C H B E.

Well you know the rest.

What I want to talk today about is first

of all what I think is unique about our
work relative to nanotechnology and

that is how we actually do
things how we do experiments and

how we analyze the data and
many of our projects are involved

nano top nano materials nano
scale phenomena nano scale.

Composites for example and
many of our projects don't.

So we kind of have a mix of different
link scales in our projects and so

the topics that we actually that I'll
discuss may depart somewhat from

nanotechnology in fact you'll notice that
I don't have the word nano in my title.

So maybe I'm an outlier in that sense but
rest assured that the things that we're

working on are equally applicable to the
nano scale materials of my crime scale.

What we need or are fairly well known for
my group is developing and

applying what I'll call high
corporate screening methodology.

To advanced polymeric materials.

And in any talk.

At your own institution one thing
that's that's usually on my

mind is the opportunity for
building collaboration's with people.

And so
that's one reason why I wanted to focus.

Maybe a little bit more on
the methods that we use for research.

So people get an idea of
what we're capable of doing.

I want to talk about two
application through goods.

One is in mixing and
looking at fairly complex formulations

and I'll explain what I mean
by that in a few minutes but

the application there is developing
new proton exchange membranes for

fuel cells and the second major
application I want to talk about is

Michael pattern by a biomaterials OK So
two quite different applications.

So in terms of the word.

High throughput another word that people
oftentimes use it's common authorial and

it's a word that you will certainly
see in the pharmaceutical industry and

that you're seeing more and
more in materials materials synthesis and

as well as characterization and
one of the goals

of our research has been to develop
new proton exchange membrane switch or

a critical component of a fuel cell
to take only him use fuel cells.

And essentially this is an electrolyte.

Membrane that is usually a polymer.

Usually an organic problem or that is
responsible for transporting protons.

It's also responsible for preventing.

The actual break through a fuel or or
air to the to either the end of it or

Catto so energy is one of
humanity's greatest challenges.

I think that the motivation is there.

It's fairly obvious.

So some of the challenges in PM chemistry.

That I would like to point out
this is one of those slides that.

Comes in one line at a time.

So why do we want to try to
do this in a common toil or

high throughput method why not
do it in a conventional way.

One of the reasons is that these were to.

Those have very complex design goals
in that there are numerous properties

that we're trying to
optimize simultaneously and

all of and in many cases these properties
are working against each other so.

In other words optimizing
conduct to Vittie

meaning I don't conduct of
any proton connectivity.

Generally if you optimize that
without regard to the other ones

you end up with a material
that's not very durable.

So one of the approaches that's been

probably the mainstay of most of
the research in this area has been

the synthesis of novel materials
from the monomer level.

OK So developing polymers at home and
polymers that possess all of these

properties generally by combining
monomers that would contribute

one of more of these properties and then
searching for the right way the best way

to synthesize them the types of catalysts
that are used the compositions.

Another approach is to look at this is
a formulation or a blending problem and

that's the approach we're
actually taking and here.

The idea is to use distinct precursors.

Meaning polymers that
are already synthesized.

They may be polymers some of
polymer they may even be short.

MacRumors.

Or look them are but they're already
synthesised beyond the monomer level and

the idea is to combine properties that
you want in a blending approach and

then do something like cross-linking for
example to stabilize the membrane.

OK So
an example that I'll talk a lot about

is probably been the lead for
I had as an inert phase.

Many of the phase that doesn't.

Transmit protons but

is fairly stable towards hydrolysis
which is one of the major.

Mechanisms of degradation.

Combined with polystyrene sulphur made
which is a very academically model

material to look at in
terms of a poly electrolyte

it's probably not a material
that would ultimately

find its way into an application but
it's a good one for us to work with.

And they demonstrate.

The instrumentation we have plus
I'm allowed to talk about it.

So most of this work was developed

in a collaboration with a large
chemical company our come up.

Who make that actually the largest
supplier of poly vanilla bean fluoride.

And this is actually it's
kind of like tough one.

Except it only has two Florian's.

On each monomer and instead of four so.

It's a lot cheaper and

Arkham is goal is to find a way to
use poly vanilla The fluoride in.

P.P.M. membranes.

So intimately seldom.

So idea was we want to build
a response function and

other words we have these controllable
parameters things we can go in the lab and

and actually turn the dials and
things like chemistry.

The chemistry is that we
select as a building blocks

the composition the ratios
we actually put those and.

The way that we actually perform
the mixing of those various

components the order that they go with.

And the time training affects
the temperature of course

that we carry out the fabrication had and
the dimensions the thickness.

For example of the membrane all of these
things are things we can directly control

and ultimately influence
the approximately five properties.

Well property classes that were
interested in optimizing OK So

the way that we go about doing
this in terms of a common a way is

we don't make a few samples at a time.

OK so
in a in the traditional view you would

use some type of design algorithm and that
may be something as simple as guessing.

To looking at the literature and
seeing what other people have done or

if you have a good model it may
actually be running a model.

It's very rare that you actually
have a good model that you can

design from on that and that level.

So you could actually what people would
do is pick what they think are the best

options.

OK and make some samples
meaning a few tens of samples

at most maybe and then spend time
characterizing those in great detail.

And in that case the emphasis
is on careful selection of a few

compositions and very detailed very

precise accurate measurements of
the physical and chemical properties.

We're taking a different viewpoint.

Another way of looking at it and
that is still selecting carefully.

The the regions that we want to look
at here because this is a huge space

of variables.

With billions of possible
combinations even for.

Even if you look at two or three
the building blocks like the poly but

nobody in Florida probably started Cell
Phone eight they're still on the order of

ten to the nine combinations of these
five things that would be relevant.

So we still locked in
there are never that down

graduate students tend to get nervous
when you tell them they have to do.

Tens of the nine experiments.

So there's still
a considerable narrowing down.

All of this based upon previous
knowledge even when we do that.

Our goal is to make several thousand
samples several thousand combinations

of these and to map out what
the properties look like for

those combinations and
our goal is to do that quickly.

We need to do it in a few weeks or less.

And the way that we need complex
that is by building libraries.

So samples of distinct temperatures for
example and distinct compositions.

Are combined on a run one
sample with a small footprint.

So maybe a few inches square.

And we can do this using gradients and
I'll talk to you about how we do that and

that and then the Grady instead of metal
and compositional are combined on one

one substrate and
you get a two dimensional library so

to speak of different
conditions then the next

step of course is to screen this and
to screen it fairly rapidly.

OK So the basic idea is to go

point to point with some type
of measurement technology.

To measure the properties
you're interested in and

there's three properties we focused
on one is proton connectivity and

others water transport and
another is mechanical properties.

And finally the desire is to build
some kind of response function so

to take the known compositions.

And when I say composition.

Usually I mean the composition
of these parameters.

So a set of parameters
several thousand for

example and then to relate those
in some way some rational way to

the the measurements we get and
that's data analysis.

It's.

So.

So how do we actually do the make
these compositions variants and

what I was going to point out is that
actually these are nano scale materials if

you look at the domains.

They're from warmed in
the most optimal materials.

You have this inert matrix that
provides mechanical stability and

resistance to fuel crossover and
inside of that matrix

you have face separated domains
of the ionic conductor and

in the best materials that we found so
far those those face separated domains of

electrolyte are only on the scale of
a few tens of nanometers at most.

So this is one case where we
are doing nanotechnology.

So one way that we make composition
gradients to actually make these high

density libraries is by doing
a match up that mixing process.

In which we actually miss say two
components although we can do.

Many of them if we want in a small
vial and what we would do is say

start out with pure B. So we have
a solution with material B. and it.

And pump it into a file and then we start
pumping in material a maybe the other

component in the blend makes
this vigorously and the mixture.

OK back into this file.

And so what happens is we be
going from here to be pure a.

As time goes on.

And as this composition is changing from
B. to A We sample it into a small syringe.

OK.

And then this is when you're done
with the sampling process you have

a gradient meaning a column of the liquid
that has a variation in the composition

this that gets deposited on
the substrate as a as a thin.

Stripe.

And then we have a knife edge
that comes down and coats it kind

of in a Dr blade type method and
unbelievably it actually works.

And so to my amazement it actually works.

Simply because you have so
many things working against you here

in terms of diffusion or
merely diffusion mass transfer

which ultimately would would wipe out
the gradient if you wait long enough.

As well as flow of this right.

Pressure driven flows do
the height differences.

But amazingly it actually
works quite well.

So just to give you an example.

This is a sample it's a blend
of two polyesters which isn't

which is for a project I'm going to
talk about second and biomaterials but

you can see along this axis is
the composition of one of the polyester

and that that polyester is Crystal and
so you can see

the different the color of the haze not
color but the appearance of haze light

scattering due to the crystalline
of the varying if you go across

and then here's the temperature that
we have the older that in this range.

For some reason instead of
flipping the image over.

The student who prepared this
decided to make the act the axes go.

In the opposite direction that
you normally would put them but.

So one of the things that we discovered
about this method although it

works really well for for some materials.

It's very sensitive to the Scotty So
if it's those two materials

that you're mixing have the Scots cities
that are very different from each other.

The two solutions.

Then you get into trouble because you have

you're trying to cut a gradient
where you have a high.

A lot of this gossipy.

And what happens is.

Instead of getting nice straight lines
like in this case you begin to get.

Instability some of the diff
whether it's a disk.

If the viscosity gradient is too
steep the flow becomes unstable.

OK.

So it turns out that this
doesn't work very well for.

The types of materials we're
looking at in fuel cells

fuel cell membranes because
generally we're looking at blending

not ideal materials the per floor
Palmers that we're working with

tend to have very non ideal
behavior in terms of viscosity and

then we're generally blending this
with a low viscosity amount of mine or

a low molecular weight.

Compound and then Alton times we were also
looking at composites nano composites

adding an inorganic face that's dispersed.

So you have three different very
very unique types of reality

three different discussing these
sometimes different by a thousand or

so and it doesn't work well
here's another example

of one that doesn't work well
this is a library temperature and

composition of those same two polyesters
the image I showed you before.

This is probably a couple locked
on that's the white face and

the holy left eye coke like
the light is the dark face.

So if you go in with a microscope and
look at the detail areas

even though it's a continuous gradient
we can go in and look at areas.

You can see that there is their face.

Separated microstructure and then they go
from if you go in with and they have film.

You can show that these go from.

Tens of nanometers to tens of microns.

OK.

So.

But in cases where we're looking at.

PM membrane formulations we
had to develop a new way of

doing this gradient blending and
so we actually developed a more

dynamic approach where we we
actually built a dynamic.

Mixing apparatus where we
could achieve high shear.

And look something like that and
we essentially code in pump and

that's the solutions we desire to look
at in our library and these a pumpkin

at differing rates so this one might be
speeding up while this was slowing down.

They go into the mixer and then they're
coated into a knife edge blade.

But we've edged into the sides of
the knife edge these channels.

So it distributes the flow out evenly.

OK And so what's coming out now
is in time the solution coming

into these channels and getting coated
is changing in composition with time.

OK And so while it's being coated and
mixed the substrate is moving under it.

At and the rate that it's moving at
determines the thickness of the film.

So as you might imagine this is
a well this is a continuous mixing

process as opposed to the batch one.

And you might imagine you have to be a lot
more careful about what the flow rates

are or else you want to get the nice
linear gradient we'd like to look at in

fact we had to develop a model to look at
this which is come a chemical engineering

bread and butter mixing models and
we found that for any given combination.

There's only one.

You know any given type of material
viscosity so there's only one solution.

There's only one set of those two
pumping rates where one speeding up and

the other slowing down.

So only one solution that results
in a linear compositions.

At the end.

OK all the other solutions this either
goes up kind of exponentially or

or it goes over exponentially and
approaches and I say asymptote So

every time you do a different library you
have to design how you're going to do it.

Fortunately we have a program
that does this for us.

Computer program.

So once we make the library.

Fortunately most of the materials we look
at have distinct infrared fingerprints so

we can use them in the metrics go in and
find the distinct peaks.

For example this is these are blends of

kind are twenty one which
is a product number.

It's simply a type of probably
even a wooden fluoride and

it's blended with a poly electrolyte
which is probably starting cell phone.

Eight and we look at these are different
ratios of the blend going from zero

to one hundred percent.

And this was actually calibration data.

So you can you can calibrate
this quite nicely and

then as soon as you made
the library you can do I.R.

measurements and
get the compositions in your library and.

For example this is one
that was made where and

this is an acceptable grading and

you can tell it's not absolutely perfect
but for the purposes of screening.

It works really well and
it takes about three minutes to make this.

So you take computer stock solutions.

Drop them into the syringe problems and

three minutes later you
have this gradient.

So how do we actually so
that means that a student can for

a pair easily tens of these
gradients in a day and depending

on how large the measurement size is
of say the conduct of any measurement.

You can easily generate hundreds.

Distinct measurements in a day.

So we're getting we're approaching that
limit where we could look at a large

parameter space in a short amount of time.

So how do we actually generate the data.

The proton connectivity is the one
I'll talk about the most essentially

we're doing and kind in spectroscopy
using a a four point probe.

So the probe is placed down
it's something that I'm sure

most of you are familiar with from
work in microelectronics perhaps.

But essentially it's a pick and place
measurement where we have this this head.

That measures that has the four
probes there tungsten carbide.

Probes and

the head is designed to always put down
a pre-determined and constant load.

So the needles are actually loaded with
a constant force on the sample and

the membrane is immersed
in water that's free.

That's essentially free.

And in other words we've measured
the of the resistivity of the water and

then we're looking at
the intrinsic proton connectivity

of the proton center in the membrane.

On the self unaided polystyrene.

Has been acid treated so that.

We get rid of the salts
that might be present.

Anyway we're measuring
the impedance in this membrane and

using impedance to back out the real
part of the connectivity and

this is an example of what it looks like
we centrally modify we had a commercial

instrument which you can barely
tell what where that is any more.

Right.
There's a commercial Here's the base and

that's the commercial had and

then we have modified it with
the robotics here to essentially allow.

That very precise placement of the head.

This is actually a second
head that measures thickness.

So that the contact to Vittie.

Requires that you measure
the resist to video in the film and

you measure the thickness and
then divide those so

this one measures the thickness to
within plus or minus a half Micron It's

a simple gauging probe and
then the connectivity had goes and next.

And so you end up with for
example this is a screen of the thickness

of a typical membrane with
the height here and micro meters.

This is an absolute height.

So the variation is approximately about
fifteen microbe years and the reason that

there's that variation there is there's
some tilt in the instrument and what not.

But we know what that plane is and

we subtract it out later
to get the thickness.

So here's actual conduct Tiffany's and
Milli semen for

centimeter of some of the formulation.

So this is five different kinds of current

kind are which is the poly dependability
in fluoride So we shouldn't trade names.

OK but these are different molecular
weights some of them are copal mothers.

With two different fractions
of the nobody in Florida and

then we're looking at the content
of one of our Probably electrolytes

going from twenty to sixty percent.

And so each row here is one
coming from one library

and what we did is on one library we
can make these measurements and then.

Along the composition direction and

then along the other direction we just
have repeatability repeated samples and

you can see there's a wide range of
conduct Tiffany's explored here and

this is actually for example max's
out at about one hundred twenty.

No.
Seaman's percent of the year so we have

some that are performing much more poorly
and some that are performing better.

So another thing that we need
that's important is understanding

the conduct of it.

He is a function of relative humidity so
we take to take a relative

humidity enclosure and
sit it down over this whole apparatus.

And then do controlled humidity and
look at how the kind of committee

changes as it goes from a dry state all
the way to the liquid water environment.

So one other thing I'll say a few words
about mechanical characterization.

Simply because this is a technique we
developed several years ago that seems to

attract a lot of interest
collaborators and it's ideal for

looking at for example nano composites
where you want to explore compositional or

chemical dependency is so this is actually
an apparatus where we can do rapid.

Screening properties in
the way that it works.

And we see what you see this grid here.

This is a grid.

That's about two inches square and

we have one hundred holes
in that in that grid.

Ten by ten array and there's another
plate just like that on the bottom and

they align well and in between
those two plates we put a film and

so that film can either be a uniform film
or it could be one of the gradients on it.

And down here although it's hard
to tell there's actually a little

needle here invented and a strain gauge.

And essentially what happens is this
is positioned under each hole and

then this is driven down
of the normal rate and

we get stress versus strain
curves then it moves to the next.

All and does the whole thing over and
over again.

So this is a side view this this
time the needle is going down.

And the samples on the bottom.

This is an enclosure to
control the humidity.

Or the environment.

All right so just to

let you know if you look at one hole in
this grid and you took a cross-section.

It would look something like
this where you have a frame

sandwich the plates
holding the samples and

then you have under defamation
the film looks something like this.

Where it's being in reality there
are some curvature to the film.

Generally it's a small enough amount of
curvature that we can treat it as a linear

film and we can calculate the stress and
the strain.

That would actually correspond
to a unique axial defamation.

So where we're doing a biopsy
go test because that's nice and

convenient for doing rapid assays but
the stress and

strain we calculate is comparable
to the the instruction or

the S.T.M. test that better standard.

OK.

Let's see.
So for example.

This is a nylon stress versus strain where
you can see an initial elastic response.

This is actually building behavior and
then we have some strain crystallization

or toughening before
the material finally fails and

this was just a film where we've
measured by for more creative holes.

You can see that it's
pretty nice repeatability.

So even though this is a high to put test.

We can do one hundred a grid of one
hundred and about fifteen minutes.

It actually is fairly detail and also
you can look at the whole sort of low.

Behind after you puncture them and

you can get an idea of what the crack
here was like brittle or ductile.

And we actually get pretty good agreement
with the conventional assays which

are generally in the axial tests where
someone makes a doggone type sample and

measures the tensor
properties along the axis and

we compare that to our measurement
we get a pretty nice correlation.

This is a modulus For example you
know a set of polyurethanes and

we're very in the amount of curative in
the polling your thing and by doing so

we vary the modulus and this ratio.

You know the correlation of R.S.A.

to the conventional one is is
a number that's predicted by theory.

So.

So I'd like to put a few
acknowledgements of that work.

And if I have a few more minutes

I was going to say a few things
about another project area.

And looks like I do.

So don't worry I'm not
going to show every slide.

Another area that we are working in is in
biomaterials problem if I am materials and

what our goal has been in this
area is to learn to understand and

to utilize the way that cells living
cells interact with material surfaces

and this is relevant to the right
range of applications and

really the key questions in
this in this technology are how

do you control cell responses
in a predictable manner.

When they attach to and grow on a surface
and how do you sense the cell or sponsors.

And the way that we are addressing these
questions is by using pattern surfaces.

So so prisoners that have
a scale to Micron scale.

Relief features on them.

Where the chemistry may be homogeneous
where it might be heterogeneous.

So we're looking at chemical and

physical acts as well as
a combination of those two.

And.

Essentially.

Let's see it just so

essentially we're looking at
combinations of multiple link scales.

So perhaps a little islands
in the nanometer regime

combined with islands in a micro meter
regime and then in a two dimensional

pattern and trying to find out what are
the spacings and what are the heights and

the lateral link scales of those patterns
that give you the optimal behavior.

So for example.

What's the what's the spacing.

What's the pattern that give you
the fastest growth rate for bone cells for

example.

In doing this we use the library
technique I just talked about to make.

Surfaces that have gradients
in composition and

then we can deal them on
a temperature gradient.

And the only reason for doing that
is that due to the reactions that

might be there or the face separation
crystalline of the it might be there.

We can force the materials face
separate and achieve structures.

You know from you know
tens of animators up to

this as large as we
really want to make them.

And then we have a way to
generate on a microscope slide

size sample hundreds of different
surface features combinations.

Surface features.

And if you look at some established
bioinformatics methods.

So how can we understand.

The interaction of these many site scales

with the physical chemistry
the chemistry chemistry of the surface.

How can the correlate that to
the biological work spots it's a very

complicated problem.

And if you look at bioinformatics
methods that are designed to take

the data from high
throughput experiments and

things like the genome project
protein sequence and so on.

There are stablished tools
that people in the community.

These communities use.

To analyze state of but
if you look at C.S.I. stands for

cell surface interactions I was trying
to be cute with the acronyms and

never seem to catch on though.

So in looking at these are all
chemical type problems

where the chemistry generally involves
a small discrete set of combinations like

amino acids or nucleotides in
materials though we're looking at

structures that are generally
non-deterministic and

they're not discrete.

Meaning that we have gradients or we can
adjust composition over a large range and

we can adjust it.

Smoothly.

We can pick from.

An almost infinite variety
of material chemistries OK.

So we have a huge amount of information
and possible combinations to look at and

what we really want to understand is how
to relate that information together.

So these are lives.

You may have seen before.

So from a simple experiment we make
a library and then we can hear her eyes

the features that are on the library for
example looking at the diameter of done.

Mains as well as the height
of the roughness.

And you know we have variations
from a few tens of nanometers

to hundreds of animators in height and for

less than a micron to close to one
hundred microns in diameter and

that's on one library and these are little
crystalline islands that are popping out

of the surface which is a part
biodegradable polyester.

Here's another library where
we did a coke limitation and

the basic idea is to generate
a diverse variation in structures so

the bright areas are crystalline
in the darker of one of us here.

So then what we do is we take cells.

Usually with a collaborator in this case.

These are bone fell.

And we call for
them on this library all at once.

So that all of the cells are exposed to
the surface is the same surface features.

The library of surface features is it's
close to the same cell population and

we look on the surface of this library.

At various types of cell assays to find
out what was the biological response.

Of cells that were attached to different
positions on the library and for example.

This is circularity which is
an indication of cell shape

that it's been shown to to correlate
with certain kinds of cell behaviors.

And you can see that if you go to zero.

It becomes in the lips.

If you go to one it
becomes a perfect circle.

It turns out that we see a strong
correlation between cell shape and

position in the library.

This could be here is
the phase transition for

the two problem are sort
of blended together.

And one thing that we notice over and over
again is so in here the polymers are to.

Face and I hear one face.

So are you.
There's no microstructure to speak of.

But inside you get a range
of micro structural sizes.

As well and the structures and
the cells almost always prefer

to if we see optimal performance.

It's generally in here in
the micro structured regime.

So.

Here's another ass
a looking at proliferation.

So the rate at which cells divide.

We can take the library
after the culture is over

as they sell proliferation and do simple.

Fluorescent staining and.

Look at the areas where you
have highest proliferation and

compared to areas where you have
no proliferation no cell growth.

OK.

So the whole question is how do
you at the end of the day relate

the experimental details of the library

meaning to temperature in
composition to the cell response.

And then when importantly how do we
relate the polymer that's under there.

It has certain size ranges.

It has certain chemistry.

Sitting underneath the cells.

How do we relate the properties
of that polymer.

To the behavior of the cells in culture.

Because we have and

on top of that considering that we have
literally as many as many distinct bots

as you want to measure which is generally
a couple hundred on a from the library.

We have two hundred spots.

Different compositions temperatures.

But each spot has a different
set of micro structures on it.

And a different ratio of the chemistry's.

So how do we take all
of that information and

understand which of those
factors is the oil.

The ones that are controlling
cell behavior.

So the way that we do it is putting
everything in a database an Oracle

database and carrying out some
statistical tests to figure out

which are the important factors and what
are the trivial or not important ones so.

The traditional method that
we would use as engineers and

scientists is a global approach where we
would take an average of each property

on the library and from the cells
as well as from the surface and

we try to make a plot of those two and
see if there is a relationship.

OK.

And.

When we do that we get
something that looks like this

if we if we plot the island size of
the microstructure versus the cell.

Proliferation rate meaning the fraction
of cells that are dividing

we get this nice shot down of blood.

Which is very useful as you can tell.

So we know that something's
happening with island size but

we can't really say what it is.

So this is another one where we looked at

proliferation ratio as versus cell
density as the cells credit each other

they're supposed to stop start
shutting down their proliferation.

And as you can see we've got
another shotgun plot here.

So what we've developed is
a new informatics concept for

the biological materials interaction
it's what we call local metrics and

it's based on the idea that the cells.

This is an image where we have cells
these green yellow dots are the nuclei.

Of cells that have divided and
you can probably barely see the red.

That's the cell body.

The other large purplish domains

are probably copper locked on islands that
are that are raised up above the surface.

So we're looking here at how
the cells are related to.

The material they're sitting on
spatially and we have some information

the bright colored nuclei have divided and
the dark colored ones.

Have not divided.

OK so.

How can we correlate that well based
on the concept that the cells don't

they don't respond globally to the
microstructure state of respond locally.

So the hypothesis is that the cell.

Here is going to be most sensitive to
the structures in its local neighborhood.

Not so much to structures that might
be over here one hundred microns away.

So what we've devised is a methodology for

picking a cell in an image measuring
all the distances to other cells and

classifying those distances
as either proliferated or

nonproliferation and then doing it for
all of the cells in the image and

then doing this for
the cell to the microstructure.

In that image and
then we know from our database we have

the positions of all of these
micro structures and cell nuclei.

In this image and we have a couple
of hundred other images and

we know where those images were
were positioned on the library.

And then for each object in
the image we know its size and

various parameters from image analysis.

So from all of this we put it
into a database and we can build

probability functions as a function of
local distance between the objects.

OK.

So instead of taking a microscope
image over the library and

averaging the data in it.

And in a kind of conventional
regression analysis.

We're actually taking that image and we're
only looking at the local interactions

those who are within about
one hundred microns or less.

And so we're in a way we're
doing a huge histogram normal.

Ation all the data and
looking at it and in a different way.

Surprisingly this this
hadn't been done before.

As far as we can tell.

And the analysis of of cells interacting
with us with features on a surface.

What we notice is that these shapes
have very distinct features in them.

So this is the.

This is the probability that a cell at
a certain position from another cell.

Would be proliferating or would divide.

And so
you notice that as you get within fifty

microns they probability goes less
than one it will only goes to zero.

Because the cells were not divide if
there are overlapping there are chemical

pathways to shut that down but
locally very close together.

It's and hence in this small region.

And there's another reason why that
happens chemically So we've got this nice

this nice tool mile to detect
these cities phenomenon.

And we've used it for example to map out.

This is going over some of the statistics.

Let's go back for.

I don't want to see all this.

All right so we've used it to map out cell
interactions with the materials features.

So.

Here for example we're looking at the size
microscope girl size in micro meters or

little islands on our library and
we're looking at here.

The center to center distance between
each island and each cell and

the color tells us whether the cells were
resting or whether they were proliferating

Girling dividing and we can see that
there are these hot spots of activity.

This is actually a very
rugged landscape so

to speak in that there are a lot of peaks
and valleys local minima and Maxima.

But there are a couple of
areas that show up as being.

Areas with high activity
where cells really like

to grow and there are a number of
theories about why this might happen.

It's a it's a fairly new area of
research across biomaterials research.

So we don't exactly understand
completely why this happens but

the point is that trying to do a
conventional analysis where we might pick

a few combinations of these things and
then look at it we can easily miss

these these two combinations that
show up at literally unpredictable

positions in the in the face space but
by doing it on a library.

We're able to find things like
this that show up sporadically.

And that's all I want to say we
can just skip forward here think

I have a knowledge of slide and
that'll be the end of the talk.

So if there's any questions.

Maybe there's a little bit of
time left for those thanks.

This was very good.

So the variation on a library is
always going to be more than it is in

a uniform sample where you've
controlled everything really well but

it's still it's acceptable so if you hard
to quote dumb burst stuff in my head but

it's what the it's an acceptable standard
deviation maybe within say fifteen percent

on the diameter of the.

Yes.

So when I see that going
in terms of applications

one application is tissue engineering
where you would desire to

select out certain kinds of cell to
have years for certain applications so

you can use the features of the surface
not just the chemical features but

the physical ones as well to select
out preferred self behaviors.

Maybe causing cells not to grow
in some area of the device and

causing them to grow more on another area.

Those are things we can
learn how to control.

Another major application is in
diagnostics so designing surfaces of

say lab on a chip devices and
where those surface surface is pre-select.

Certain types of cell behavior in
order to do an ass a kind of thing.

Certainly.

Absolutely.

Yeah there are a lot about
the patients where a gradient will

artificially influence their
response that whole question

there is usually these are cases where
some kind of instability is created

by the presence of a gradient
kinetic instability and so

usually that instability doesn't
become active until the gradient

exceeds a certain magnitude and

all I can say is that we are several cases
like the face separating polymer case and

as well as polymers that segregate.

We've looked at how the gradient effects.

There was segregation across the seas and

so there's a large range of gradients
that have virtually no effect.

But then you hit a gradient like
a maximum gradient above which

it changes the phenomenon drastically and
usually it's when the size scale

when the gradients is changing some
feature like the thickness significantly

over the same size scale that
the phenomenon is occurring at

that same thing.

So the surface interactions have
to worry all about silt room.

But you know when we look
at that same thing now.

Exactly.

And in fact that's that's exactly
what we've done as a matter of fact

that's it's very important
when you're looking at

trying to correlate
the cell material behavior.

The fact is there are millions of
cells there and they're all in a sense

communicating with each other and you have
to accurately you have to appropriately.

D. convolute those two effects and

the truth is nobody really
knows how to do that properly

simply because there isn't there
is no general mathematical theory

that takes into account all the complex
cities of the soul so interaction but

we can do it at least in
the numerical level or

a phenomenological level because we
can have a distribution function for

the cell cell interactions for at least
certain properties like proliferation and

we have a distribution function for
the cell material interactions and.

By looking at also the pier
like uniform services for

example that feature
list surfaces compared to

the libraries where there are features
weekend and essence to convolute

roughly what's going on
think there's a bigger day.

Thanks.