This.

Is that and.

This is four years for this.

Thank you.

Thanks everyone for showing up today.

I hope you have a jacket on.

It is a little chilly in here and thank
you David for organizing all of this.

Today I'm going to talk about
some of my group's work.

Silicon on insulator that
nano ribbon biosensors and

this work actually was done I joined
Georgia Tech about two years ago.

So all of this was actually done by
students who were at U.T. Dallas.

Just me.

First I thought I'd introduce myself my
group does a variety of nano electronic

materials and
devices work to be materials graphene

molybdenum by soft fied we do work
on resistive change memories and

other types of memory technologies we also
have had a lot of effort in three five

in silicon seam OS devices specifically
looking at interfaces with high K.

dielectrics And of course I want to talk
today about our work on bio sensors and

I think one of the things independent of.

Which technical area one of the theme
that pervades all of these is that their

electronic properties of the devices
are extremely sensitive to the material

properties and really that's the theme
that pervades all of the research in my

group and you're going to see
that today with bio sensors and.

So before I get going.

First of all life's some acknowledgements
first Dr Tarasoff and a new Ph D.

student mending side will start some
work that I'll describe at the end of

my talk today on biosensors but

the work that you're going to see was done
by Nicky Fernandez who was you to Dallas.

She is now research engineer
at Texas Instruments.

Curtis Kant Lee who's an assistant
professor now at Boise State University

mood and two senior research
scientists in my group Dick Chapman and

Harvey Stigler and we've had a lot of
collaboration with Professor should balls

group specifically on self assemble model
layers and I'll be talking about that this

work is funded now by an N.S.F. award
it was previously funded by an S.R.C.

center called the Texas analog
center of excellence and

direct funding from Texas Instruments.

So I'll give a brief background and
motivation for these types of biosensors

will talk about the sensor fabrication
steps that are functional is ation of

the silicon and then issues related
to mobile ions and sensor biasing and

modeling a little bit of work on
circuits followed by a brief summary.

So first of all for
those of you who don't know what these.

Bio fets are and
give a description of how these work.

Hopefully most of you know what a mosque
that is you have a metal gate separated

from safety type silicon by
a gate oxide a dope source and

drain region and as you know if you put a
voltage on the metal gate you get a charge

that accumulates in the metal gate you
get near a charge that cumulate so

inverts in the channel of the silicon and

that's what allows current to flow
from the source to the drain.

These bio fets are actually quite similar

except we're taking off this
top metal gate electrode.

And we replace this with surface chemistry
and so we're going to talk about some of

that surface chemistry it's
first we have selfless.

Model layers which just act as linking
groups and then we have specific receptor

sites that are designed by
biochemists I'm not a biochemist but

they designed these receptors.

So that they will only attach to
the molecules that you're trying to sense.

So that's how we get the selectively
out of these devices that we design

these receptors siller only
bind to what we want to sense

when the protein or whatever it is
you're trying to sense comes in.

It is bound to the surface
of these receptor.

All of these proteins will have
a certain charge associated with them.

And so now we have a charge that is close
to the silicon channel separated by

the gate oxide and
these this surface chemistry and

just like the case where you're
adding charge to the gate.

We're adding a charge here and what we're
measuring then is the conductance or

the current that flows through
the transistor and so we can get very high

sensitivity because Ma SPETZ
are very good at amplifying charge

that's on the gate through the conductance
associated with the transistor.

There's also been a lot of work recently
at least in the early two thousand folks

looking at things like nano wires
instead of conventional fats and

that the idea is similar.

You have a nano wire that is backdated so

you could apply a voltage to
change the charge on the nano wire

you functionalize the Nano wire and
then use that inductance change.

What we'll talk about actually what we
found is that the sensitivity of nano

wires actually for most conditions is the
same as the sensitivity of a nano ribbon.

So the fact that it is
a wire is not too important.

But if you have very thin silicon you're
going to get about the same sensitivity as

you have for and then a wire.

Ideally you don't need to
modify the species so.

This is one of the pluses
of this type of technology.

You are only modifying your sensor and
you don't have to modify your species so

this is called label free sensing.

So just to give you a little bit more
specific sense of what's going on.

First of all you can use these as
ph sensors quite simply even if you

have the silicon only terminated by
oxide and has associated hydroxyl groups

at low PH You can see these hydroxyl
groups are positively charged and

at high PH You can see these groups
are negatively charged so you can

use these sensors quite easily almost any
sensor actually that has hydroxyl groups.

You can use as a PH sensor.

For the case of say protein
most proteins contain things

like means that are negatively charged
groups that are negatively charged.

And the charge of these
proteins depends on the P.

H. and
something called the P I of of the protein

that you're trying to sense it's really
the combination of these charges and the.

And the PH That gives you
the charge that's on the silicon.

So these biomolecules they attach in
these biomolecules to receptor is

form what's called a semi permeable
membrane or more talk about that.

Keep in mind that one of the things and

we're not going to talk about too much is
it does take a certain amount of time for

approach the proteins in your
solution to reach the sensor.

So the time it takes.

Once the protein is attached for
your fet to sense that is quite quick but

it does take a good bit of time for these
proteins to diffuse to your sensor and

it depends on the device geometry and
reaction right constants.

So this hopefully gives you an idea of
how these sensors are supposed to work.

And if you look even in more detail.

I'm plotting here the potential as a fun.

Action of position.

So here is our semiconductor oxide.

First we have something called
the Helmholtz layer the Helmholtz layer is

just the charges associated with the
hydroxyl groups on top of the silicon or

the silicon oxide.

We also have the charge associated with
salt ions of course if you're sensing

something like blood you're going to
have sodium chloride in that blood and

you get almost like at the pollution
layer near the surface and

in this the pollution layer these salt
ions are not hydrated with water and so

they're that's part of the charge that
will be associated with this device.

And when we get out to
the outer Helmholtz plane.

You can think of this is moving into the
bulk of a conventional semiconductor where

three electrons would screen say
your dope in atoms for the case

of an electrolyte it's these water
ions that screen your electrolyte X..

Excuse me extreme your salt ions and
so the charge of

the salt ions the effect of charge is
smaller than it otherwise would be.

Now when we add a protein and many times
you'll hear us talk about the membrane so

we're going to add a protein that attaches
to a receptor molecules in this membrane

you get a change in the charge and
you get a change in the potential drop.

So this is actually showing the solid line
is the potential drop that you would have

a what expect if you only have
the receptor molecule there.

If you then attach a protein to it and
have that charge in the membrane

you're going to get a change in
the distribution of the potential and

when you have a change in the distribution
of the potential you can see this decrease

in this case of the potential that will
show up again as charge in mirror charge

in the silicon and we measuring
the conductance then of that silicon so

what I want to point out that will
see this later the screening and

the amount of change that you have for a
given molecule depends on both the PH of.

The solution because of the charges
on the top surface of the oxide or

of your Sam as well as
the salt concentration.

So both of these things and electrolyte
will affect what the fat finally reads.

So that gives you
the background of these and

there's been a lot of work on
demonstrating applications so

I'm just showing a few of the applications
that have been demonstrated so

Labor's group for example has shown that
you can detect things like cancer markers

one of the cancer markers is telling
a race and Winnie flows tell him a race.

Containing solution over his sensor he
sees a change in conduct the conductance

of the Nano wire in this case single
viruses have been detected with these

types of sensors prostate specific
antigen which is a marker for

prostate cancer has been detected.

So if you can fabricate a stable
biosensor they could be used for

a whole wide range of applications simply
by changing these receptor molecules.

Now the issue however is that a lot
of the issues with the manufacturing.

Processing and
reliability of these sensors

has really been neglected in much of this
application oriented work and they show

these pretty results they show one result
with the change in the conductance but

you have really no idea how
statistically meaningful that is and

how reliable the sensor is and
so since our funding originally

was from Texas Instruments they really
want to know is this something that we

can envision manufacturing something
that we could put on a chip.

So a lot of the work that
I'll be talking about

is not actually I won't be talking
about applications you can go through.

I have a list of just a few of
the references there's been lots of

application work but I'm going to be
talking about is more of a detailed

understanding of issues related
to reliability integration

the interaction between the sensor and
the electrolyte and the.

Models that you would need if
you ever wanted to really see

these become a true technology.

So first a little bit more about
I gave you a description of how

the sensor works I'll talk to you
about how we fabricate our sensors so

this is actually the design of our chip.

What we have here is on the sides of
our chip we actually have room for

test devices as well.

Of circuits.

So we have conventional mosse transistors
that we can fabricate on these

sides of the chip.

We have a microfluidic channel that you'll
see in a second that goes down the center

so that's this dotted
line down the center and

that microfluidic channel is the channel
is over the Nano ribbons and then

the Nano ribbons of course are connected
to the source and drain electrode so

we can put this on a probe station
probe our source and drain electrode.

We can apply a gate bias to the back.

So you'll hear us talk
about a back gate voltage.

So by applying a bias to
the back we can change

the initial charge state of
our then silicon nano ribbon.

We also have pads that we
can use that are inside and

actually it's right here
we have pads that are say

gold pads that we can also change
the bias of our electrolyte solution.

So we can apply a voltage to our
electrolyte as well as a pliable the to

the back gate and
this is about a six mask level process.

So what we do is we take
a silicon on insulator wafer

typically we've been buying thicker.

S.L.Y. maybe one hundred nanometers.

And we actually before we do
the mesa isolation we will use

a very high quality dry oxidation
to thin our space so why the thin

the silicon later that layer down to about
ten nanometers or so thirty nanometers.

We then use the fogger feet to isolate
the silicon we deposit a Phillips side.

We X. hole.

As for our source train.

We either diffuse or
implant our source drain regions.

We then actually oxide
from the channel region.

And this is actually a very important step
but I'll talk about we do a high quality

seam OS dry gate oxide so
this might be a couple of

nanometers thick two nanometers or
so two or three nanometers thick.

We deposit our source
drain metal DEP step and

then we have my trial
which covers our pads.

So this just gives you a sense
of how we fabricate this device.

And when it's all done we also have to
add of course the fluidic flow system.

Now we're not experts in microfluidics So

we just use a fairly simple
microfluidic process but

we're all done we use these machine
will polycarbonate slides and P D M S.

We have tiger on tubing and
we simply screw our down our.

Our microfluidics to the and
align it to our chip so

this is what it looks like
when we are all done.

And so one of the things I just
thought I would mention is that

I have quickly gone through
the process flow one

we have seen an even using sed process
steps that have been in the literature.

There are lots of issues associated with
the processing of this device that really

never come out in the literature and
I'm not going to talk about all of them.

I won't only going to talk about one but
it took us about three

years actually to get to the point that
we had a reliable sensor process and

there were all kinds of things that we had
to worry about we had leakage originally

through the buried oxide associated
with how we were our silicon layer.

We had to be very careful when we have
very thin S.L.Y. layers even things

like an R.C.A. clean which you don't think
affect the silicon actually an R.C.A.

clean will remove a couple of nanometers
of your silicon and if you're doing.

Three or four R.C.A. cleans when
you're done you have no silicon left.

Talk about cleaning the surfaces we had
some issues with serious resistance

oxide quality all kinds of issues that we
had to solve in order to make this final

biosensor I'll give you one example
though if you look at the literature.

Most people actually use plasma
cleaning to clean the photo

resist residue from the gate oxide one
of the things that is different between

a conventional seam OS
filled effect transistor and

these biosensors is that at the end we
need to leave this gate oxide open and

you have photo resist and
those of you who work

on these types of things know it can be
very difficult to remove photo resist.

And it's not very controllable.

And we use the plasma claim that we
had seen many times in the literature.

The problem is with specially within
gate oxides this is the drain current vs

the back gate voltage of our
transistor our bio sensor and

you can see before we do this plasma step
we have a nice looking ID characteristic.

However after this plasma step that
everyone uses we get a lot of stretch out

which means that you have defects
that are forming between the box and

the silicon layer and
those defects are going

to cause history Sisson all kinds of
problems associated with the biosensor So

one of the things that we had to do
was change for example our resist.

And instead use H M D S and
remove that with an S C one solution.

And we did a lot of experiments we
actually published some of the results

in the journal materials chemistry.

You can see that now we get
a much better looking ID

characteristic after the S C one clean.

It's a little bit shifted that's probably
because the surface of the oxide

obviously different after an S C one for
example but

it certainly doesn't have
this large stretch out.

So this gives you a sense of the types
of issues that we're dealing with.

So the next step now that
we've built our sensor.

We have to design these linkers
these self assemble model layers and

they link the hydroxyl groups on the gate
oxide to the receptor molecules that

we're going to use to sense.

And again there's not been a lot of work
in the literature about issues with this.

What is supposed to be fairly simple
molecular functional is ation.

So in collaboration with
Professor Hall we looked at

a lot of these types of sands and
we specifically in

the literature will use what are called
amino silences the linker and

I'll talk about that and we did work
to optimize some of the conditions.

So we looked at four
different amino silence and

here are the chemical names and
their skeletal diagrams one is

called apt has all of these actually are
amino terminated by amino that just means

you can see you have an an H two or
H two and kept to these molecules.

They're all silent based which means
that we're taking the oxide and

the bottom is silicon so
we're going to be attaching silicon

to the hydroxyl groups or
the oxide that's coming from R S.

IOW two and then they have different
properties in between we have

one case that has purple chains that
are the case that has metal chains and

the length of these are different but a
lot of these are used for bio sensing and

so we wanted to do a very critical
study of these Sams and I'm not

going to again be able to go through all
of this I can share the publication but

we looked at optimizing things optimizing
functional is ation by characterizing

with things like F.B.I.
our lips Samba tree contact angle.

We also did some basic bio sensing
detection of strip tab and using.

Ten and tried to look at home
a genie how much and they.

Reproducibility instability of the sand
something that actually even though

there's been all of this work
published on these biosensors

these things have really not been
well published in the literature.

So I'll give you one example.

So if you look at the literature
this apt has this shorter chain

is actually something that is used over
and over again for these biosensors.

And so we do our functional is ation and
we do it as a function of time.

And we see it first.

What we expect to see
first of all this is I

should point out this is
something called F T I are so

you're basically measuring
the vibrational modes of these molecules.

And that's as a function
of the wave number so

the wave number tells you what
vibrational mode that you are looking at.

And the absorbance tells you whether or

not you have that certain
vibrational mode.

And initially we can see for

example an increase in the silicon oxygen
carbon stretch which is associated with

this initial functional is ation
of the surface and we also see.

C H two stretches associated with
the formation of these Sam So we see

an increase in the peak intensities with
time which is what we have would expect.

The problem is what we found is that after
about thirty minutes we expect this Sam

to be nicely uniform and I'll explain why
in a second and the expected thickness of

this app has to supposed to be about
seven or eight angstroms or so.

What we find is after about thirty minutes
the thickness continues to increase and

what we believe now what's happening
is instead of forming a very nice two

dimensional model layer of Sam we instead
get something that's called polymerization

So these chains are now interacting with
one another and set of having this nice

leaf form Sam we have this three D.
structure that is formed.

The problem with this three D.
structure and I'll show this

intellectual measurements in a second
is things like water can easily.

Get through this three D. structure and
interact with their surface or

other types of molecules can come in and
cause stability problems and so this app

has looks like it's going to be a problem
and I'll show that again in a second.

On the other hand we also looked
at a longer chain amino Cylon a U.

tests which is longer chain and

what we find is in this one is not
quite as studied in the literature.

That's similar to apt as we see
the structures that we expect to see.

But unlike ab tests
the thickness doesn't change.

And so we can put this in for
fifteen minutes and

almost immediately the same forms but we
do not see an increase in the thickness.

If we say overnight so
it's a much more controllable process and

also suggests that this a you
test is not polarizing.

So now we have a very densely packed Sam.

Where we're really stopping any molecules
from because we're bonding is closely as

we can we're stopping any molecules
from penetrating this Sam So

we have no polymerization and
a very fast reaction.

So we did a lot of these types of
studies for many of these Sams and

again I'm not going to go
into all of that detail.

So what's the problem the problem that
we found from a sensor standpoint

we're showing here in this case.

These are out of hives these are apt
tests and a test will come back those out

of hives or just another type
of self assemble monolayer.

C four out to hide is much
shorter than c eleven how

to hide the shorter chain
Sam's tend to polymer arise.

And the problem is as we sort of expected.

If we put these in buffer solution.

The buffer solution is just the biological
solution that we're going to do our

sensing in.

After about two hours and
buffer all of these Sams have left.

So we use F.T. I.R. to characterize and
we get almost all of the C four out

to hide have left because
it's being attacked by the.

Groups in the water in the solution on
the other hand the longer chain out of

hides after two hours there's only a small
drop in the longer chain aldehyde So

it means that from the standpoint of using
these sensors these long chain out of

hides are much more stable.

This comes back then to this app has an A.

You Tez remember that app has a short
chain and what you can see is that we get

this precipitous drop here and that's
related I don't show the F.B.I. our But

we have shown that this is related to the
complete removal of this app has Sam So

you can imagine if you were
doing bio sensing and you saw.

You know you don't know when this protein
is going to common you saw this big this

change like this you might think
that you are sensing a protein but

all that's happened is you have
lost your Sam Now we do see

a slight decrease in the current for a
test but we don't see any of these sort of

very quick drops associated with the loss
of the molecules so again discovering some

evidence that using longer chain Sams is
likely something that you will want to do.

So our conclusions we
have a lot of things.

The one that is most commonly
used in the literature A.P.

tests we don't think is a good choice for

doing this type of bio sensing some
other Sams that we looked at A.P.

we think is a better alternative
as well as this A You tests.

Now the other interesting thing
is that if you have your Sam.

Attached to your receptor receptor
molecule like by a ten attached to your

Sam All four of these Sam's are stable.

So you might think well then it doesn't
really matter which one I use for

biosensors But as we're going to see at
the end a lot of times what we will have

is a reference sample that does not
have the receptor molecules and

then we'll have a reference sensor and
then we'll have the sensor of

interest that has the reference
sample the reference receptor.

And so if we do not have this biotin
attached to that reference and

you're using a P.T.S..

That's going to give you obviously a wrong
result in terms of sensing and so again

you can look through our papers to look
at all the details associated with this.

So the next thing I want to talk about
briefly are Mobile Iron affects I mean to

see how we're doing here.

How good.

So another thing is most of you probably
know that you don't want to do if you're

making a C. mosse chip.

Is throw a lot of salt on it right
that was actually a big problem

early in the semiconductor industry
was Mobile Iron contamination

of people using their bare hands to
make transistors to get mobile ions and

your device was drifting all over the
place but there actually had no been no

studies done on how this may affect
how these may affect these biosensors.

So this is an example of where we have

our biosensor in the first case
that I'm going to show you here.

We just have the S.R.O. two and
we do not have self assemble model layers.

And the idea is that we would expose
our biosensor to either sodium and

potassium based ph solutions
we would actually clean it.

So we would just take solvents
we would claim the solution

runs that with the eye water
drive a sample and measure it.

So will first look at sodium
in the case of two million

dollars of sodium what you can see is that
you get this large history rhesus loop so

a history says that a half a volt or
a volt or

so again indicating
that the bio sensor is.

Seeing the sodium in the sodium as we're
applying a voltage is drifting in and

out of the gate oxide.

So again these changes this history
rhesus could be picked up incorrectly for

example as a protein that's attaching.

Now as we decrease the sodium Concentra.

You can see that this history says
gets smaller but it's still there.

If instead that we use
potassium based solutions.

We see basically no history says sloop and
the reason is that potassium

has about a ten to the tenth
power lower diffusive A-T.

compared to sodium so potassium is
larger it diffuses a lot slower and

so we don't see these history
rhesus affects with potassium.

Now the question is what happens if
we take and put for example a U.

test so this is that longer chain Sam.

What we find is that we see no history
cysts with these Sam terminated

two samples so
the SAMs are actually quite good.

If you have a long chain Sam
that doesn't polymerized and

is not going to be released as
it's in biological solution.

These long chains Sam's are quite
good at stopping the fusion of.

Mobile ions into the oxide.

This is important.

Of course if you're designing your fats
and you have open areas of gate oxides

you might want to have some open areas of
oxide what this points to is you're going

to have to really worry then
about Mobile Iron contamination.

I won't go through the details we've
also did some comparative studies C.V.

measurements of a story been used to look
at mobile contamination and basically we

see the same results using C.V.S.
we saw for our mosfet measurements and so

both measurements are confirming that you
have to be careful with these mobile ions.

So we've given you a little
bit of a sense of some

of the materials related issues molecular
functional ization and mobile ion and

now we're going to start moving on to
more of the device related issues so

talk a little bit about sensor biasing and
then modeling of these biosensors.

So here is what again our
sensor looks like and

if you remember we have this back
eight voltage that we can apply.

We also have the possibility
of either measuring or

applying a voltage to the electrolyte.

And most of the studies that you see in
the literature most of the recent studies

they just keep the electrolyte floating so

they don't apply a bias the electrolyte
they simply apply a bias to the back gate

to change the initial charge
state of the silicon channel.

What we found is that if you
sweep this back to voltage and

measure the potential in the electrolyte.

That this electrolyte voltage
even though it is isolated

from this packet voltage this electrolyte
voltage moves with this back voltage.

So we apply back at voltage and we measure
a voltage on the topside electrolyte and

this is something again that had not
been really shown in the literature and

it depends the exact amount
the electrolyte voltage how that couples

the pens a bit on how you functionalize
the surface of your sensor

and what we did is that we actually
developed a model and the idea.

The basic idea is that when you
have this electrolyte on top.

What you really have is a big capacitor
you have the electrolyte that's kind of

like a metal you're plying a voltage and
so

if I apply a good voltage on this side and
you're sweeping at a slow enough rate.

The potential for the charge in the
electrolytes going to change is just like

a metal insulator metal capacitor and
it's actually quite complex at the pens

a lot on exactly how you design the sensor
and what the various capacitance

is are what the box capacitance
is what if you have nitride for

example that is covering your source
drain the amount of and we did three D.

modeling to show this
the amount the selection.

Voltage changes will depend sensitively
on a lot of these conditions.

Why is this important.

So if you do a simulation.

So I'm showing here two simulations in
the first case some simulating the drain

current vs the should say back
gate voltage for this sensor.

In one case you have no
coupling that's these circles

the other case you have
eighty five percent couple.

So whatever voltage you apply to the back
is about eighty five percent coupled to

the front.

And this is our experimental
data of this these open squares.

You can see that the expected sensor
result because of a very thick back gate

oxide is very different than what
you actually see experimentally.

So all of the Ivy characteristics or
most of the Ivy characteristics you see in

the literature have this steep kind of Ivy
relationship which doesn't make sense and

the reason is that you have this
coupling that is going on and

what we found is that this
does affect the sensing So

this is showing actually ph sensing and

showing that if you leave the electrolyte
floating or you pin the electrolyte

voltage to a given voltage of about the
same as the back it voltage you actually

get about the same ph sensing results and
I just realized I forgot to put on here.

The case if the electrolyte is at some
different voltage you're going to get

different sensitivity for the PH sensor
and so it's very important that you either

pin this electrolyte voltage to
some value or at least know what

the electrolyte voltage is if you're going
to be able to predict the sensor results.

The other thing that we wanted to do
was start moving this towards our

towards ability to do circuit simulations
so you can imagine a company like.

Texas Instruments.

They're going to want to take the sensors
and rap amplifier circuits and

other type of circuitry and
functionality around these sensors.

But in order to do that.

Of course you need a spice model for
the bio sensor in there actually for these

types of nano ribbon biosensors there
had not previously been a spice model

developed so we went about developing
a spice model for these biosensors and

you can find all the details in
a publication that came out last year.

The basic idea is first of
all we take a conventional

Well a somewhat conventional fat.

So the model is actually
comes out of Berkeley.

It's a quite recent model.

It's independent multiple gate model for.

S.O.I. transistors so
basically it's a model

the FET is a model of a silicon
on insulator technology.

And then we coupled that.

To some affective potentials and sign not.

Which determine the potential
drop in the electrolyte solution.

So if you were back at
the beginning of the talk.

I mentioned that the voltage that you see
going through this electrolyte depends

on things like the salt concentration and
the PH and there are actually detailed

electrochemistry equations that you can
find that for a given ph a given salt

concentration determine the voltage
drops within the electrolyte and so

we put those equations inside of spice
we have one that's associated with

reference potential one that's associated
with this how Holt's layer and

the inputs to these are things
like the ph the salt concentration

the density of molecular charge.

That's attached to this S.O.I.
mosfet And so

this is actually showing how well
our results fit and we actually.

Don't have any fitting
parameters we picked for

example the ph obviously
is something that we know

we know we can measure or have a sense
of the amount of hydroxyl groups.

So there's really no input fitting
parameters the drain current

versus the back gate voltage in this
case we're actually tying the back and

front gate together.

So this is the ideal characteristic
that we model and the field

symbols are our data and so you can see
as we change the ph the threshold voltage

of the device changes quite a bit and that
our model fits this ph sensing quite well.

We also looked at comparing this to
results in this case in the literature for

different bio sensing and

we chose some literature results
because of how well documented a lot of

the parameters that were used were
in some of these papers so this is

the case of trying to detect protein
called Abbott in using biotin receptor.

And they had results for

different protein concentrations as
well as different salt concentrations.

Now again we had no fitting parameters we
didn't try to change any of the parameters

in our model and so this is the threshold
voltage shift that they obtained for

these conditions and this is the threshold
voltage shift that we got from our model.

Now you can see that there
are some differences.

About a factor of two in this case
a little bit smaller here a little bit

smaller here but for
the most part we're able to model

the magnitude at least of this change in
the threshold voltage for the sensor.

If you wanted to use this specifically for
biotin ABBOTT Of course we could go in and

start fudging and doing some fitting
parameters to try to fit this better but

we wanted to show that we could
fit that without doing it.

Similarly folks have done D.N.A.
sensing with these types of sensors.

Where you have.

Immobilize D.N.A.
and then you hybridise that D.N.A.

and you get a shift in the threshold
voltage and again we have

very good fitting compared to
this experimental data and so

the point is that we now have a we think a
very nice spice model that you can use to

model bio sensing and
ph sensing using these biosensors and.

So the other nice thing with this

thing of another nice thing
that this model does that

allows us to look at how other parameters
are expected to affect our results.

So one of the other parameters inside
of the small model are the number of

groups or the number number of groups that
are attached to our silicon dioxide and.

So we can change the number
of hydroxyl groups and

that changes how the PH is
affecting that surface charge.

And then we can look at bad as
a function of the number of

molecules that we're attaching for
example.

And so we can see how much of a change.

Do we expect in one case if we have
no hydroxyl groups versus another

case we can also look at
the effect of salt concentration.

So this is for one.

Protein concentration five
times ten to the nine.

Per cubic centimeter and
then increasing the amount of protein.

That's attached.

Excuse me one protein concentration
changing the salt concentration.

What you can see is if your salt
concentration is quite low.

Then the shift you get.

Is quite large and

the idea is that salt is not going to be
able to screen that molecular charge.

However if you go up to much
higher salt concentration.

You're not going to be screening
that molecular charge and

you're going to get a much smaller
change in the threshold voltage

compared to the case of
no molecules attached so.

We can use this model to actually help
us predict what we would expect for

certain types of sensing conditions and.

What we're showing here is what types
of membrane charge densities that we

think we would be able
to see using our sensor.

So this is showing
the threshold voltage shift.

As a function of
the membrane charge density.

So you can see that on this scale.

We're on a logarithmic scale and on here
for this line we're on the linear scale.

You can see that for

a membrane charge density of about
three times ten of the eighteen for

sort of typical salt concentrations
typical ph the threshold

voltage shift is only a couple tenths of
a mellow bolt that you're expected to see.

So we can get some idea of if we
assume that you know we may be able

to detect something like a millivolt we
will get a sense of what the how well

the sensor would be able to detect
different concentrations of protein.

And if we know how if we
know this membrane charge

density is a function of
the concentration in solution so

this membrane charge density
is a number a volume.

Fractions of the number
per cent a meter cubed.

You can actually read late this and

there's papers that tell you how
to do this you can relate that to

in steady state the concentration
that you have in solution.

So for example if we assume that we can
sense something like three times ten

of the eighteen molecules percent
a metre cubed we can determine then

the concentration limit of the cup
of the protein in solution.

So for example for
by a construct tablet and

if we assume three times
ten of the eighteen.

We should be able to see something like
Pico molar concentration for the case.

Of D.N.A. It's actually not quite as
good that you would be able to only see

something like ten nano Moeller or
so detection limit for these types

of molecules so it allows us to predict
the detection limits of our sensors.

So the final thing I want to
mention is that all of these

models don't worry at all about noise.

Right.
You see just very smooth curves all

the way down to micro volts.

The problem is that especially if
you're in biological solution and

you're using this.

There are all kinds of
sources of noise and so

the last part of this project was
actually beginning to look at

some of these circuits issues and
I'm going to give you one brief example.

So what we did is we took our sensors and
to start within spice model and

we have a reference sensor where we
do not have the receptor molecule.

And we have the sensor where
we have the receptor molecule.

And we went through a lot of detail but we
could feed these sort of realistic current

or voltage versus time characteristics and
so although you can't see it here this

sensor actually this small change in this
case is what we said you might see for

a certain concentration of the protein.

So there's tiny little I should have put
an average in here this tiny little change

whereas both sensors are seeing a lot of
noise and we want to start getting a sense

of how much of this noise will
affect our ability to sense.

And so what this is showing is
the change in the voltage that we expect

say the threshold voltage is a function
of the concentration of molecules

that are attached and again you have this
nice curve that goes down all the way.

So this would be the ideal
case with no noise.

And then in the simulation we started
adding different values of noise in this

case six hundred sixty three Microvolt.

Six point six Microvolt.

It's actually fairly small
are mess noise values.

And so if you want to be able
to detect Pekoe mole or for

example for Abbott and biotin reaction.

What you can see is that the idea is
that you need to be in this linear

part of the curve so once we get down
here basically our noise we can no longer

see any change with change in the molecule
concentration because our noise is just

overriding that change in potential
due to your molecules and

so this can tell us about
what we would expect so

we would need an R M S noise voltage less
than about six hundred sixty micro bolts.

In order to be able to detect this
three times ten of the eighteen.

Limit for a given reaction so we can
use these circuit models now to tell

us even more realistic
scenarios where you have noise.

What are the detection limits of our
sensors as well as the associated

circuitry.

So with that I'll just summarize.

So some of the key things I think we found
first of all sense of the sensors are very

sensitive to surface preparation and we've
done a lot of work trying to understand

how to do surface cleaning how
to prepare reliable Sam's.

The fluid obviously putting these in
a fluid dramatically alters the device

characteristics so we have mobile ion
issues you have to worry about as well as

these complex distributed R.C. circuits
that affect the biasing in the device.

We've developed the spice model that
agrees well with experiment and

we can predict what happens as
a function of salt concentration

site binding charge and
a lot of other parameters.

And find finally under
realistic conditions these bio

facts have a small signal
to noise ratio and

this low frequency one over after
noise is a particular problem and

increase the minimum detectable
concentration of species.

So as I said much of that has
already done all of this work.

We're now actually funded.

Through a new and S.F.

partnership where we're looking at
disposable sensors so similar technology.

Except on plastic substrates and
a variety of sensors so multiplex sensing.

And we're also beginning to explore other
types of materials beyond silicon such as

Millennium by sulfide two D. materials
on flexible substrates for sensors and

that's really where we're expecting
to go over the next couple of years.

So with facts.

I thank you for your great attention and
I'll be happy to answer any questions.

Thank you thank you.

Yes.

Right

so silicon for

these types of applications does not have
a bias stress instability problem right.

These are sort of you know
high quality transistors so

those problems of typically been
engineered one of the reasons actually.

We're looking at things.

If we want to move to
flexible substrate so

that's where the organic world plays T.F.
TS and organics on a flexible substrates

one of the reasons that we want to look
at things like most of them by sulfide.

Which you can process and actually this
is just showing this is actually graphene

that we have moved into a solution.

These are metal contacts on top of
just graphene sitting in solution.

And this is the kind of idea that we have
is these types of semiconductors unlike

organics are quite stable the basal plane
is almost completely unreactive And

so in terms of using this technology more
in the flexible realm we think that these

may have a better chance than things
like organics where reliability

is a big problem.

And.

It was

a very

small one day.

That ran this.

This is concentration.

Because by systems.

I mean.

That's very much.

How do you.

I think that that is a big issue and
that's you know when we come back

to these types of calculations where we're
looking at the number of the membrane

charge the number of molecules at
a given concentration and looking at

Device size that's one reason that these
nano ribbon sensors we think actually.

Are a better way to go
then the Nano wires.

Because you have a much larger
surface area so the concentration.

You know you can average that out.

You're still getting though
the electrostatics from the Nano wire

primarily comes from the electrostatics in
the vertical direction as you can imagine.

So if you're very thin the electrostatics
it's about as good as a nano wire that

happens to be three D.
can find it and show that here but

that's something that we have actually
shown using three D. modeling and so

we think that you actually do want to have
nano ribbons instead of nano wires but

I think that's definitely a big issue.

All right thank you thanks.