Non-Conventional Raman Spectroscopy
Sep 5, 2023
Matthew V. Schulmerich, Department of Defense Postdoctoral Fellow, Department of Bioengineering, UIUC
"Non-conventional Raman Spectroscopy applications towards the compositional analysis of soy bean seed"
is one good example of that.
So even though it says it's about soybean seeds,
you can think about biomedical applications
and cancer pathology and stuff like that as well.
So Matthew is currently a Department of Defense fellow
here at the University of Illinois
and working in Professor Rohit Bhargava's group.
He has a BS in chemistry and a BA in biology
from St. John Fisher College in Rochester.
And he got his PhD in chemistry from Michigan
where he worked in Raman spectroscopy.
So we'll like you since you came here from Michigan.
Don't go back.
So Matthew's won several great awards,
including the University of Michigan
Distinguished Dissertation Award,
the Thomas Hirschfeld Award in Applied Spectroscopy,
and the Casimir Feijan Award in Chemistry.
So welcome, Matthew, and look forward to your talk today.
Thank you.
So non-conventional Raman spectroscopy applications
towards the compositional analysis of soybean seeds.
So soybeans are actually a pretty big commodity
here in the United States and worldwide.
In the US, 76 million acres of land
are devoted to growing soybeans.
And this translates into about a $40 billion industry,
23 billion of which are exports.
Now in any industry, the goal is really
to gain the most value from your product that you can.
And in soybean seeds, not all soybeans are created equal.
Premiums are paid for nutritional content.
So a lot of work has gone into trying to develop approaches
to quantify various nutritional components of soybean seed.
So the things that people are interested in,
buyers are interested in, protein and oil content,
as well as various amino acids,
various fatty acids, and various sugars.
And the current approach to analyzing soybean
in a non-destructive way is near-infrared spectroscopy.
So near-infrared spectroscopy is a type of spectroscopy
that's been around since the 70s.
It's underwent quite a bit of development.
And right now it's the current industry standard.
Now I'm a big believer in Raman spectroscopy,
and I think Raman spectroscopy holds a lot of potential
for quantifying these specific components.
Currently, protein and oil is the only thing
that near-IR can quantify to the industry needs,
whereas Raman spectroscopy holds a lot of potential
for amino acids, fatty acids,
and specific nutrients like that.
So in this talk today, I'm first gonna go over
an introduction to spectroscopy.
Then I'll talk about how spectroscopy
can give you chemical contrast.
I'll follow that up with an instrument we built
that demonstrates you could use Raman spectroscopy
to quantify protein and oil content in single soybean seeds.
And then I'll follow that up with an instrument we built
for bulk soybean analysis.
So what is spectroscopy?
Spectroscopy is basically taking light,
breaking it into its component colors,
and trying to get information from those colors.
So an example of how you get information from color,
here's a picture of Waldo.
Now here, Waldo you know is red,
has a red shirt on with a red hat.
But in this picture,
you can't actually discern which one is Waldo.
By adding color to this, you get information.
Color gives you information.
So here you can see Waldo's right there.
So spectroscopy is converting light into color,
or light into information.
This can be qualitative
by observing what colors are present,
or it can be quantitative
by not only observing what colors are present,
but also the intensity of those colors.
Now the human eye can actually only see three colors.
It can see red, green, and blue.
All the colors on the screen here
are combinations of red pixels,
green pixels, and blue pixels.
Now with scientific instrumentation,
you can actually see a whole range of colors,
a continuum of the electromagnetic spectrum
where the visible region is actually a very small portion.
So for the work that,
for the spectroscopy that we do on soybeans,
the near-infrared spectrum is right in this range,
you can actually gain additional information
by looking maybe out in the mid-infrared
or using other parts of the electromagnetic spectrum.
So the typical spectroscopy experiment
would look like this.
You would take a range of colors with known intensities,
you would introduce it to a sample,
and you would collect the light
that comes out of the sample,
and you would then ratio the light that went in
to the light that went out,
and that would give you an idea
of how the sample modifies the light,
or how the light interacts with the sample.
So a general definition of spectroscopy
is the study of the interaction
between radiation and an analyte
as a function of wavelength.
Now if we were to do a spectroscopy experiment here
on water, you would take a whole range of wavelengths,
so the ultraviolet all the way out to the far-infrared,
you could put that through water,
and then you could ratio the light that went in
to the light that went out,
and you would get a spectrum that looks like this.
Now this is actually four different types of spectroscopy.
This is UV vis spectroscopy, visible spectroscopy,
near IR spectroscopy, and IR spectroscopy.
The method's the same.
You ratio the light that went in
to the light that went out,
and ultimately the only thing that changed
is the range of wavelengths
that you put through your sample.
So the range of wavelengths that you choose
actually give you different chemical information.
For example, UV visible spectroscopy
is electronic spectroscopy,
and what this does is light absorption
is a result of electronic transition,
so electrons moving from a ground state
to a higher energy state.
Infrared and Raman spectroscopy
are vibrational spectroscopies.
So what these are looking at
are these are looking at groups of molecules
and specific vibrations that are moving
as a result of light incident on them.
Now, Raman spectroscopy is different
than the other types of spectroscopy I'm showing you here.
With Raman spectroscopy,
you use a single wavelength of light,
and you put that single wavelength of light
through your sample.
So it's common to use something like a laser
to get many photons at one wavelength
and then put that through your sample.
So for an illustrative example,
I'm gonna use 785 nanometers
as our Raman excitation frequency.
So if you think in terms of photons,
when a photon interacts with a sample,
a couple things can happen.
Number one, it could be absorbed.
So if it's IR light, the light would be absorbed,
and that would promote some vibrational states.
If it's UV or visible light,
that can promote electronic transition states
and the light's absorbed.
If the light's not absorbed,
one of three things can happen.
The light will be scattered through the sample,
and how it scatters depends on
the molecules that are present.
So the first type of scattering is Rayleigh scattering.
With Rayleigh scattering,
what happens is a photon comes in,
it interacts with the sample,
it promotes molecules to some kind of virtual energy state,
and then the same photon comes off.
So a 785 nanometer photon comes in,
and then a 785 nanometer photon leaves the sample.
Second type of scattering is Stokes-Raman scattering.
And with Stokes-Raman scattering,
what happens is a 785 nanometer photon might come in,
and then it gains,
or it leaves some of its energy with the sample,
and that results in a lower energy photon coming out.
So this might be a 800 nanometer photon.
Alternatively, you could have anti-Stokes scattering,
and this is where a 785 nanometer photon will come in.
The molecules are already vibrating
because they're at room temperature,
so the photon can actually gain energy from the molecule,
and as a result,
you would have a higher energy photon coming out.
So if you put a 785 nanometer photon in,
something like a 700 nanometer photon could come out.
So what this looks like in terms of instrumentation
is you would have a photon come in, 785 nanometers,
interacts with the molecules,
and then in the Raman spectrometer,
if 785 comes out, this is Rayleigh scatter light,
that gets rejected by the Raman spectrometer.
Alternatively, a 785 nanometer photon could come in,
interacts with the molecules.
This time, you have anti-Stokes-Raman scattering,
so it gains energy from the sample,
and you have a 782 nanometer photon come out.
This makes it through the Raman spectrometer,
and then gets sorted by wavelength.
Alternatively, you have a 785 nanometer photon coming in,
interacts with the molecules.
This time, it loses some of its energy to the molecules
and as a result, you have a lower energy photon
that makes it through the Raman spectrometer,
and it gets sorted by wavelength.
Now in reality, you have a laser,
so you have billions and billions of photons
interacting with your sample.
The vast majority of those photons
are actually Rayleigh scattered,
so you don't actually look at those.
Those get rejected by the Raman spectrometer.
The ones that are Raman scattered make it through,
and then get sorted by wavelength.
So a Raman spectrum looks like this.
It's basically a histogram of photon counts
that's collected over time.
So if you take this, put it on its edge,
you end up with a Raman spectrum that looks like this.
So on this axis, you have Raman intensity,
which is photon counts,
and on this axis, you have a shift
from an excitation frequency.
So if we excite at 785 nanometers,
you have an 810 nanometers to 915 nanometers,
and this is generally considered
the Raman fingerprint region.
And it's usually reported in terms of wave number shift
in order to compare different laser wavelengths
if you excite with a different laser other than 785.
So that is an introduction to spectroscopy
and how Raman spectroscopy is different
than the other types of absorption spectroscopy
that are out there.
Now, spectroscopy gives you contrast.
So to compare the different types of contrast,
I got a couple slides
that just compare near IR to IR to Raman.
So with near IR spectroscopy,
the light tends to scatter more than it's absorbed,
so you have a greater penetration depth.
And this is the reason that near IR
is commonly used in quality control applications
for the agricultural industry and soybeans as well.
So because the light scatters more than it's absorbed,
it can go through very thick path lengths,
and as a result, you can get the light
through an entire soybean or through an entire apple,
something along those lines.
This spectral information arises
from broad spectral overtones as a result,
and as a result,
it makes it difficult to assign spectral features
to specific chemical components.
So the chemical information you're achieving here
is basically groups of molecules moving together,
and as a result,
you don't know anything about the specific vibrations
that are present,
so they're spectral overtones.
Light scattering tends to be a problem,
so corrections are needed.
So spectroscopy tends to follow Beer's law
in order to do quantitative things,
and in Beer's law,
you need to know the sample path length.
So because the light scatters,
it doesn't go directly through the sample,
so you don't really know the path length.
There's a lot of very good corrections
for scattering regimes
that actually work pretty well in the near IR,
but this is something to consider
if you're doing near IR spectroscopy.
IR spectroscopy,
light tends to absorb more than it's scattered,
and as a result,
you have very little penetration depth,
and for this reason,
in order to do IR spectroscopy,
you have to work on very, very thin samples.
Samples tend to be on the order
of maybe five to 10 microns thick.
The spectral information arises
from specific vibrations
arising from molecular functional groups.
As a result,
you can assign spectral features
to specific chemical components.
So the chemical specificity in IR spectroscopy
is superior to that of near IR spectroscopy.
Sample thickness and preparation
tends to be the biggest difficulty with IR spectroscopy.
So if you want to take IR spectrum
of something like tissue,
you have to microtome it to fairly thin sections,
and you also have to make sure
there's not a lot of water present
because water has a very large absorption
in the infrared regions.
Now, Raman spectroscopy
is actually the best of both worlds.
With Raman,
you can use any wavelength you want.
So you can choose a 785 nanometer laser,
a near IR laser,
that achieves the penetration depth
that you gain with near IR spectroscopy,
but at the same time,
it's also a vibrational type of spectroscopy
that gives you similar chemical information
to what you achieve in mid IR spectroscopy.
So you have very high chemical specificity
in a Raman spectrum.
Now, the quantum efficiency of Raman is low,
and that's why it hasn't really been developed
for things like agricultural applications
where you're analyzing whole soybeans.
Quantum efficiency being low
means it's a low light event.
You need very, very sensitive detectors.
You need very high powered lasers
in order to get a nice Raman spectra.
And the reason that it's actually applicable
in the research I'm showing you here
is because detectors, lasers, filters
are becoming better and better each year.
And as a result,
we're able to do things
that we couldn't have done 20 years ago.
So all spectroscopy follows Beer's Law.
So I'm gonna show you some calibration curves
in future slides here.
And these calibration curves are based on this principle
where you have a concentration.
So you can get the concentration
using wet chemistry approaches.
And you have basically a known amount of a sample
that you're looking at.
And then what you do is you take a spectrum of that,
and that spectrum gives you a response.
And this response is gonna be linear.
So then what you could do
is use this linear relationship.
So you get an equation for this line,
and you can take a spectrum,
follow that over to the line,
and then figure out what the concentration is.
So that's how you quantify using both near IR,
mid IR, and Raman spectroscopy.
Now, conventional Raman spectroscopy,
as well as mid IR, and not so much near IR,
you can do spectral imaging.
So spectral imaging is basically
getting your spatial resolution using point mapping,
and then getting a spectral resolution
by taking a spectrum at each one of these points.
So you get your X, your Y,
and then your Z-axis is gonna be a spectrum, okay?
So in conventional Raman spectroscopy,
what happens is the light would come in,
the light goes through a lens,
it focuses to a point,
the Raman spectra is generated,
the Raman signal scatters in all directions,
that gets collected by the lens,
and then goes back up to a Raman spectrograph.
Now, by taking multiple points over time,
you get a spectrum at each one of these points here,
and you can put together a map of Raman spectra
that looks like a Raman image that you see here.
And in order to visualize the distribution of chemistry
in this image, you could pick a couple spectral bands,
and I chose red, green, and blue
because those are the colors your eyes are sensitive to.
And this gives you an idea of the distribution of chemistry
within the soybean sample.
So if you compare the three different types of spectroscopy,
you have near IR here,
you can kind of see the distribution of chemistry
shown by the near IR spectrum.
Here you have a mid IR spectrum,
and you can see there's a lot more information here,
there's a lot, there's a different distribution
than what you're seeing in the near IR.
And then here's a Raman image of the same soybean section.
And you can see there's pockets of,
I guess, teal here and orange here.
Those correspond to sugars,
the purple here and the purple here correspond to oils,
and the red might correspond to protein
in various different chemistries.
So understanding the relationship
between what spectral bands you observe
in the Raman spectrum in the image,
you have to kind of look at known standards
in order to figure out
what you're actually visualizing in this distribution.
So using the mid IR imaging,
we can visualize the distribution of protein and oil
in soybean sections.
And we see that the distribution
is actually fairly heterogeneous.
It's not a uniform distribution of protein.
And these are four different soybean sections
that we took infrared images on.
You can see this soybean section,
which has a protein content of about 50%,
has these splotches of protein here,
whereas this one that has a protein content of 51%
has actually large pockets of protein,
whereas the other portions of the seed
are low in concentration.
So what this means is if you use
conventional Raman spectroscopy
to develop a calibration model,
if you took just a point spectra here,
you would get a very different calibration
than if you took a point spectra here.
So you need to take representative sampling of the soybean
in order to actually develop a proper calibration model.
So in order to do that,
we turn to actually transmission Raman spectroscopy.
So conventional Raman spectroscopy,
if you take a point sample on an entire soybean,
what you would get is you would get a signal
that's actually surface weighted.
So we have a pharmaceutical capsule here
that has a gel cap,
and then inside is a powder
that has pharmaceutical ingredients.
And if you take a conventional Raman spectra,
what you end up seeing is basically just the gel cap.
This is a fluorescent signal that arises from this gel cap.
If you take a spatially offset Raman measurement,
so this is different than a conventional Raman approach,
what you would do is you would illuminate in one position
and then collect spatially offset from that.
And the idea is that because this sample is light scattering,
the photons would diffuse through the sample,
and then if you collect offset from that,
you would get some kind of representative sampling
from the path that the photons took in the diffusion process.
And through this approach,
you end up getting a spectra that looks like this.
So you actually do get
some of the active pharmaceutical ingredient
in this Raman spectrum.
Now, if you take a transmission approach,
so you basically take the illumination
and bring it down to the bottom,
you end up getting a much better representative sampling
of this pharmaceutical.
So that spectra looks like this.
So taking a similar approach with soybeans,
the idea was we could take a representative sampling
of the soybean and then get a representative spectrum
that we could develop calibration models on.
So this is actually a general problem.
We apply it to soybeans,
but it could be applied to process analytical technology
where you might be looking at fluids mixing through tubing.
You could use it for biomedical applications
where we've actually used it
to recover bone spectra through the skin.
You could use it in pharmaceuticals
where you wanna look at the distribution
of active pharmaceutical ingredient,
or we're using it in soybeans here
where we're trying to quantify economically
and nutritionally valuable components of soybeans.
So light scattering, it looks like this.
If you have no light scattering,
what I have here is a beaker full of water
and I have just the Illinois emblem inside the beaker.
And if you increase the scattering,
it's pretty clear that you can no longer
actually see this eye emblem.
And the reason for that is because the light
as it travels through the beaker
ends up spreading out, diffusing,
and you can't actually image anything inside of that
without doing some kind of math on it.
And then this is what it looks like in a cuvette.
You have a cuvette full of water.
And then if you increase the scattering,
you can see the laser beam profile spreads out,
diffuses as the scattering levels increase.
Now, the reason this is important for ramen
is you're only gonna get a ramen signal
where the excitation photons actually propagate to.
So in highly scattering systems,
you can see the laser actually spreads out
and propagates pretty significantly,
which means if this was a soybean,
which is highly light scattering,
you would get laser light kind of spread out
throughout the entire soybean.
And then there is some Monte Carlo modeling
that basically demonstrates that you get
representative sampling in a transmission approach.
So here we have a light scattering sample,
and what we're showing is a histogram
of where the ramen photons actually come from.
Okay, so the single soybean instrument.
Okay, so on that last slide,
I showed a Monte Carlo simulation.
We actually did this for soybeans.
So here we took a CT scan of the soybean
in order to figure out what the internal structure
of the soybean looks like.
We then generated a computer mesh
so we can model photons propagating through the soybean.
We labeled this with the proper absorption
and scattering coefficients.
We then illuminated a soybean on one side,
and we just watched the photons
randomly walk through the soybean
and had a detector that had a certain numerical aperture,
and we figured out where the ramen photons originated
that we actually detected.
And if you look at a histogram of this,
so this cross-section through the sample here,
you end up seeing something like this.
So you end up getting photons largely concentrated in here,
but you also get some from the surrounding area too.
And this is center-weighted
as opposed to being surface-weighted.
So if you look at a soybean,
a soybean actually has this outer skin
on the outside that isn't useful.
So you would wanna get all the information
from the interior of the soybean
and not the exterior of the soybean,
which is what conventional ramen spectroscopy
would give you.
Then I got an image here of a soybean
that we illuminated with a 785 nanometer light.
This is 200 milliwatts,
and you can just see it's kind of glowing red,
and this just kind of supports the idea
that the photons just diffuse and bounce around
throughout the entire soybean.
Okay, so we wanted to see,
can we actually collect a ramen signal from the soybean?
So in our first set of experiments,
we had a 785 nanometer laser
that was focused onto the back surface of a soybean,
and then we just collected all the light
that came out of it.
And we ended up seeing spectra that looked like this.
Now, each one of these spectra took an hour to collect,
and that's largely because of the inefficiency
of our setup, and this was only a 40-milliwatt laser.
But we get a definite ramen spectrum
from four different types of soybeans.
So you get a phenylalanine band here,
which corresponds to protein
as well as the amino acid phenylalanine.
You get CH2 bands, and you get ramen spectra.
This is a ramen spectra that is recognizable.
But the acquisition time was quite long.
So based on that data and some of the simulations
that we did, we were able to apply for funding,
and we got funded to put together a instrument
that could measure ramen spectra
from single soybeans efficiently.
So this is the instrument we ended up building.
And what you have here is you have a sampling stage,
and the sampling stage consists of a series of holes
that you would put the soybeans into.
And then there's a hole that passes through this plate.
The laser light comes from the bottom,
illuminates the bottom of the soybean.
This is a collection lens that is automated.
It moves up and down and focuses onto the soybean.
Then that light gets collimated from the sample,
goes to fiber optics here,
which brings it down to the ramen spectrometer.
So this thing's automated.
It's controlled by LabVIEW.
We're able to load it up with soybeans,
set whatever parameters, and then come back
maybe the next morning,
and all our data's stored in the computer.
So with this, the spectra that we acquire
with this instrument looks like this.
This is a hyperspectral system,
which means that we collect about 256 spectra at one time.
And the CCD frame would look like this,
where this is the spectral axis shown in pixels,
and this is the slit height,
which basically shows the different spectra we collect.
If you take this and put it on its side,
you get your 256 ramen spectra look like this.
Now, to develop a calibration model for one soybean,
we only need one spectra.
So basically, in order to improve signal-to-noise,
we just bend the spectra and get a single spectra.
So acquisition time was one thing we wanted to look at.
We wanted to take good data,
but at the same time, do it as quick as possible.
And for this study, what we did
is we changed our acquisition time
from two minutes up to about 10 minutes.
And you can see the different spectra
corresponding with those measurements here.
So this is the two-minute measurement.
This is the 10-minute measurement.
And we did a signal-to-noise comparison.
And we find that after about five minutes,
there actually is no improvement in signal-to-noise.
And this is largely because we're gaining noise
from the readout of the CCD.
But if you look at this,
the two-minute measurement actually looks pretty good.
So just to compare those different measurement times,
we developed a leave-one-out
cross-validation calibration model.
And what I'm showing you here is a correlation plot.
So I just wanted to explain what a correlation plot is.
A correlation plot, you take wet chemistry.
So these are reference values,
where you take a soybean, you grind it up,
and you're able to get ground truths
for protein and oil content.
And then this is the spectra response
predicted by the calibration model.
So in order to figure out how well you're doing,
what you wanna do is you wanna
take one of these points,
see what the wet chemistry said,
see what the calibration model said.
And if it matches up perfectly,
you're gonna be along this X equals Y line right here.
So for our instrument, the different acquisition times,
we took a two-minute measurement,
up to 10-minute measurement.
You can see that the root mean standard error,
so this is basically the actual minus predicted,
is about the same, statistically the same,
for all 10 measurements.
So this means that we basically took
our calibration and validation set
for the next set of samples
at a two-minute acquisition time.
Now, with the current instrument,
and even with the instrument in this configuration,
we probably could get down
to maybe a 30-second acquisition time.
But this is all reasonable
for what the people that we're working with
wanna do with this instrument.
So the 30-second acquisition time's actually pretty good.
So we developed a calibration model
where we did 236 soybeans for oil
and 237 soybeans for protein.
And this is what our calibration model looked like
compared to the wet chemistry.
What you see in the green, the yellow, and the red
is the one standard deviation.
The green is one standard deviation in the wet chemistry,
so that's basically the error of the reference method.
And our calibration model's within those error bounds.
And this is the validation set.
So we validated on a slightly larger set.
So we had about 300 soybeans for oil
and 300 soybeans for protein in the validation.
And you can see we're doing pretty well.
So if given a soybean,
we could predict the protein or oil content
within about a 1% standard error.
So that's not bad.
And this is basically about the same
as what NIR-IR could do.
Now, this is actually the first demonstration
of using Raman spectroscopy
to quantify non-destructively the protein
and oil content of soybean seeds.
So that's fairly significant.
Now, NIR-IR, again, is the industry standard,
and that actually is used to quantify
protein and oil content.
Now, what NIR-IR has trouble with
is amino acids, fatty acids,
looking at specific components.
So in order to tackle that,
we moved to a bulk soybean instrument.
The reason we went to a bulk soybean instrument
is because if you look at the relative contribution
from the various amino acids,
these things add up to make your protein content,
and the fatty acids add up to make your oil content.
So the concentrations of the amino acids
are much smaller than protein or oil.
So as a result,
you have to actually do the reference method,
the wet chemistry on actual, on bulk samples.
So in order to compare,
we had to take bulk soybean measurements.
So we went from the single soybean to the bulk.
So we designed an instrument, an AutoCAD,
and also used some ZMAX to do some optical modeling.
And the instrument looks like this.
It has a hopper on the top where we can,
a hopper on the top where we can
just basically pour a bunch of soybeans into.
It has a little turn screw
that feeds the soybeans from the hopper
into a sample chamber here.
So there's some soybeans in the sample chamber.
The optical path goes like this.
After the measurement's collected,
there's a trap door here that opens up
and drops the soybeans into this box here.
And then we can just repeat the measurement
and do multiple samples with this,
with a specific variety of soybean
that fill the hopper there.
So the optical path looks something like this.
We have a laser beam come in.
We collimate the laser beam to about a two inch waist.
That collimated light goes through the samples.
We collect the samples with a lenslet array.
And that lenslet array then focuses the light
onto collection fibers.
And this is a hyperspectral measurement as well.
This is actually the same spectrometer
and set up from this side over
that was used for the single soybean instrument.
So in our labs, we basically have both of these on a table
and we just flip a couple mirrors
and we're able to use the same spectrometer
for both instruments.
Now the lenslet array is actually kind of a neat way
for solving a problem.
We wanna collect a light over a large surface area
and at the same time have high efficiency,
high numerical aperture.
So the way to do this is to have multiple focuses
that collect a lot of light, so higher NA lenses.
And then we take all that light that's collimated
from the lenses and then focus those into the fiber.
So this is actually fairly efficient.
Now the one problem with this was that
each one of these spots is fixed.
So in order to, and you can imagine if the sample loaded
into the, the soybeans loaded into the sampling chamber
differently each time.
And one of these focuses was like in between two soybeans
that would cause a problem.
To solve that, we basically made this rotating
lenslet array.
So now over the acquisition time,
which could be anywhere from 30 seconds
to maybe two minutes,
the signal basically gets averaged.
And as a result, you get signal from two concentric rings.
And just to kind of show you what that looks like,
I just put laser through the other side
and you can see how the collection spots move
as a function of time.
So the Raman spectra we're able to obtain
with this instrument looks like this.
It's pretty good.
This is about 12.5 millimeter path length.
And in the experiment, in order to test out this instrument
and see how good we can collect specific amino acids
and fatty acids, we had 52 soybean varieties
that are representative of the different soybeans
that grow around Illinois.
We had 300 milliwatts of laser power.
We did five runs and each run was 10 minutes.
So for a given variety, we had five replicates basically.
And each replicate took 10 minutes to acquire.
And then we did the wet chemistry for a couple
of specific amino acids and a couple fatty acids
that people are interested in.
And these are basically the results for that.
So here we have fatty acids.
So these are some specific fatty acids
that animal feed people are interested in.
We developed a calibration model using the open circles
and then the validation models on an independent set
of 26 soybean varieties are shown
in the closed circles here.
And then for amino acids, we're also doing very well here.
So again, this is actually the first example
of using Raman spectroscopy to predict fatty acid
and amino acid content.
And based on stuff we looked up in the literature,
this is actually quite a bit better
than what NIR-IR has been able to do
in the same application.
So one thing I will mention is these acquisition times
were 10 minutes per run,
but we're only using 300 milliwatts of laser power.
Since this study, we've actually upped our laser power
to about 3.5 watts.
And because we're not focusing the light down,
we're using a collimated beam,
we're not actually burning any sample at that power
and we get a much faster acquisition time.
So we're at about 40 seconds for the same signal of noise
that you're seeing here.
Instead of 10 minutes per run, it's about 40 seconds per run
which is pretty good.
Industrial folks would like to see it a little bit faster,
but I think we can actually,
we have room to even improve above and beyond 3.5 watts.
We can go up to five watts right now
with a commercially available laser.
Okay, so in summary,
I gave you an introduction to spectroscopy
and showed you how Raman spectroscopy is slightly different
than the absorption approaches to spectroscopy.
I compared the different chemical information
you get from the three methods.
I showed you our single soybean instrument
and our bulk soybean instrument.
So with that, thank you for your time.
I can take questions.
Yeah.
You showed a picture where you were actually imaging
within a beam, right?
Yeah, so this was, which-
You were early on, you showed pictures
that looked like beams and you were imaging
and I'm just wondering why,
is that an important piece of information
the distribution, say, was in
or is it just whole people information
So it depends who you talk to.
Plant breeders are very interested
in looking at the distribution of various amino acids.
And one thing that is of interest
is apparently as a soybean plant grows,
different environmental,
even if the soybean had the same genetics,
environmental influences affect
what amino acids actually get deposited
and how they get deposited and how they accumulate.
And apparently sometimes amino acids
that are there during one stage of seed development
disappear later on in the seed development.
So if this is an amino acid that you're interested in,
maybe it makes more sense
to harvest the plant sooner than later.
So that's one thing that's of interest.
And that's where the distribution comes into play, yeah.
Is this information, could this affect things like price?
Oh, absolutely, yeah.
So knowing, basically anytime you know what you sell,
you can get more money for it
if that's something that people are interested in.
So being able to quantify the amount of lysine in a soybean,
you can actually sell that
to people that formulate animal feed,
and then they know what they're putting
into their animal feed,
so then they can turn around and sell it
to their customers at a higher price.
Yep.
Yes, sir.
So I know you said that you're using,
you occasionally said how much power
you're using for each setup.
Is there a limit on how much you can do?
Because I know you said these methods are non-destructive
and that you're not cutting up the soybeans,
but how much light can you put into it
until something happens,
and have you reached that limit?
So on the single soybean instrument,
we actually focus down the light
in the first iteration of that.
And that's kind of unnecessary
in anything that's a light scattering approach.
So when you focus down the light,
you have higher fluence,
and fluence is basically the power per unit area.
And that could cause burning.
So even at near IR wavelengths,
there is some absorption.
And if you have more photons
coming into contact at the same spot,
the light that absorbs would basically cause
increases in temperature, which could cause burning.
So in applications where you're focusing the beam down,
the power actually matters quite a bit
because you don't want to destroy your sample.
In the bulk soybean instrument,
we have collimated light hitting our sample chamber.
And that collimated light,
I mean, you can put that up to pretty much as high
as you can buy a laser that would go up to that.
Right now, we got a laser that was 3.5 watts,
which is a pretty powerful laser.
The strongest one on the market that I know of is five watts.
And that's increasing every day,
but a five watt laser is pretty powerful.
Yeah, because I know with lasers,
especially you need to say, oh, it's only five watts.
It's like 1.6, well, not 1.6.
You know, it's a fraction of what comes into most appliances.
Or a laser, like it's a lot.
Because a laser is coherent and it's also monochromatic.
So like light bulbs, for example,
give a broad spectrum of light.
So you couldn't necessarily like take a filter
and get that much.
And not only that, with light bulbs,
you can't collimate the light
because it's not coming from a small point.
So collimating stuff with extended sources like light bulbs.
Okay, cool, sure.
Thanks.

