Tip:
Highlight text to annotate it
X
>> Okay. Our final speaker this morning is Diana Wright.
Her graduate work concentrated on the comparison
of neutron activation analysis
and inductively coupled plasma mass spectrometry.
These techniques were used
to elementally profile ceramic architectural artifacts.
She earned her PhD.
Degree from University of Maryland
and where she joined the FBI laboratory.
In 2001, she joined the chemistry unit in the paints
and polymer subunit where she currently does casework
as a forensic examiner in specialties of tapes,
paint and polymeric analysis.
Doctor Wright is also a member of the American Academy
of Forensic Sciences, as well as a member of the E-30 committee
of the forensic science section of ASTM.
Her presentation today is entitled Discrimination
of Architectural Paints via Physical
and Chemical Methods of Analysis.
>> Diana Wright: Good morning.
Well, there will be no Raman in this talk today,
fortunately or unfortunately.
And depending upon how you feel about that,
my husband's Australian so I may be applying
for a job in New South Wales.
[ Laughter ]
Maybe. We embarked on this work because we find so often
in the literature a lot to use as a reference
for drawing conclusions on our automotive paints,
but we haven't found so much research on architectural paints
and we don't see as many architectural paints
in our laboratory and casework, but when it does come up,
you have a single layer of white paint transfer,
what can you say about it?
What's the significance of that?
So we embarked on this project
to see how well our SOP is working for discrimination
of paints and what sort of conclusions we can draw
from the analyses that we are performing.
Previous work was done 40 years ago now,
by Tippett.
Now in human years we all know that 40 is not old at all,
but in terms of work you want to bring into court,
you justify the basis of what you're doing,
we thought that we could probably update that a bit.
Tippett's work did however did study 2000 architectural paint
samples so it's quite a comprehensive study.
Using microscopic and microchemical techniques,
Tippett found that in these 2000 samples,
he could provide a discrimination where chance
of random association was one in a quarter of a million
so quite good discrimination.
And then when he combined instrumental techniques,
in his case, emission spectrography and PyGC,
that chance of random association jumped
to one in a million.
So, the purpose of this study,
then was to update Tippett's research
to assess more current paint formulations.
Surely, the paint industry has updated itself in 40 years.
To determine
if the discriminating power could be improved
with advanced analytical capabilities,
to attempt to address the significance of associations
and also to translate those significant assessments
into language that might provide clearer,
more stand alone reports as is recommended
in the recent NAS study.
Samples were collected by FBI field and lab personnel as well
as colleagues at other forensic laboratories in North America.
Over 950 samples were submitted.
They were collected from a range of structures, interior
and exteriors of residences, businesses, parks, restaurants,
and this was our sample collection form that we sent
out requesting samples.
The information we asked for was also quite similar
to what Tippett had looked for 40 years ago.
We wanted to know the sample color, the geographic location
of the sample, what type of building it was on,
whether it was a house or a business, whether it was
on a window sill or a wall as its substrate,
any kind of environmental conditions that we would want
to know about, whether it was an interior or exterior finish,
whether it was in direct sunlight, what have you,
the manufacturer if it was known, the approximate age
of the structure, the date
of its most recent paint application and the number
of coats that were applied.
We used our current analytical scheme
for comparative paint samples where, of course,
you start with macroscopic and microscopic examinations.
We recorded the top coat color, the number of layers,
the sequence of the layers.
We did record what the substrate was on the samples.
We then went onto FTIR where again,
for paint examiners this is going to be quite familiar,
we analyzed, looking for the organic binder
and also any inorganic pigments
and fillers in the sample.
We used a microscope accessory on the FTIR,
we used the diamond cell, diamond compression cell
for our analysis and our detector cuts off at 650.
We then went on SEM/EDS both back scatter
and EDS analysis, again to image the samples which proved
to be quite helpful with some of the trickier layer structures
and also to assess inorganic pigments and fillers
and then finally, on to Py-GC/MS where again we were confirming
and in some cases differentiating
the inorganic binders.
Samples were initially divided into groups
by their top coat color.
We had a blue group, a green group, a red group, etcetera.
Doing this, we had about 200 samples
that were classified as white.
The rest of the samples had some sort of a hue to them
and the largest group of hued samples was off white,
off course and there were about 300 samples in that category.
So this is a pie chart break
down of the samples that we received.
As you can see, almost 80 percent were hued,
about 20 percent were white and within those hued categories,
was off white, black and gray were combined,
yellow and peach were combined,
brown and tan were combined, etcetera.
And here's a depiction of some
of the larger samples we received.
We did also receive casework sized samples, but this is just
to give you an idea of how the categories were breaking out.
The first step in the examination process
of each individual sample was, of course,
to determine was it even paint and 15 samples turned
out to not even be paint.
So they were just eliminated from the study.
That left 960 samples to be intercompared.
That was over 460 thousand pairwise comparisons
that needed to be performed.
Is this a good time to say there were three
of us working on this study?
If the sample was paint, we then assessed the layer structure.
What was the sequence of the layers, the color,
relative thickness of the layers, features of the layers
such as were there air voids where the paint's delaminating
in certain interfaces, etcetera?
We did record what the substrate was,
but we didn't factor it into comparative assessments.
So, in other words, if there was a paint that was on wood
and a paint that was on drywall,
the purpose of the study was could we discriminate
between the paints, not could we discriminate
between the entire system to include the substrate.
So, for example, if this is one of the samples that we received,
you might call it red, you might call it red brown,
you might call it brick,
you might call it something else depending
on how good my monitor is here today, but because we didn't want
to miss any intracomparisons, this sample would be compared
to both the red group and the brown group.
So, where I said there were 460 plus pairwise,
460 thousand plus pairwise comparisons, this sample here
in the middle was compared to two different groups.
So, that number was artificially low in some sense.
Here's an example of two samples that we compared.
You can probably tell that they both have more than one layer
to them, but if you just looked at the top coat,
those samples were indistinguishable.
If you then looked that the underlying layers,
the one on the left was characterized
as having a taupe colored top coat.
We also got quite good at our naming, stone, tan, etcetera.
The second layer had air voids in it and was classified
as a medium gray, then there was a thick, white powdery layer,
etcetera, a five layer paint system.
That sample being compared to the one on the right,
where the top coats were comparable,
showed an underlying layer chemistry of white, turquoise,
white, turquoise, white and then brown, seven layer system,
quite different from its pair and quite different
from anything else in the study as it happens.
I also want to note here,
the sample on the right was originally characterized
as a gray, so it was compared to all the gray, black group
and the one on the left was in the brown group.
But, because the analyst who looked at the sample
on the right thought there were brown undertones to it,
that sample was brought over into the browns and that's how
that intercomparison took place.
So, again, even though the sample started out in one group,
if the color was questionable, it might be compared
into another group to not miss any pairwise comparisons
that would, should be performed.
So, in this example, these samples were eliminated,
or discriminated from everything else and from each other.
Here's another example.
These are a yellow top coat that finishes similar
and so they were considered to be indiscriminated,
undiscriminated up to this point.
If you look at the underlying layers,
these were both two layer paint systems, a yellow over a white.
FTIR analysis was then performed
on the top coat, the yellow layer.
The samples were still considered
to be undifferentiated, but then FTIR
of the underlying white layer, clearly shows one is kaolin
and one is talc based and so again those samples were able
to be discriminated based on FTIR of the underlying layer.
So that's kind of the approach that we were using
for the samples and now we'll talk
about the actual discrimination power we found
within the hued group, so 80 percent of the population
that we're looking at.
That consisted of 760 samples that were compared.
That was over 290 thousand intercomparisons.
From that, 109 pairs remained, so from 290 thousand pairs,
just macroscopic and microscopic exams,
brought us down to 109 pairs left undifferentiated.
One hundred and eight of those pairs,
which constituted 79 samples, was brought forward to FTIR.
One pair was brought forward to SEM
because the layer structure was too complicated to assess
up to this point and I'll discuss
that sample a little bit more later.
Then from that FTIR group and that one pair
in SEM, 32 pairs remained.
So, we've gone from 290 thousand pairs to 109 to 32 just based
on visual, microscopic and FTIR.
Now let's talk about the white samples,
20 percent of the population.
The white samples were a beast.
Walking into the lab every day, 197 pill boxes staring at us.
How do we do this?
We decided that the best thing to do was having gone through
and assessed all these samples as having a white top coat layer, we went back
to the paperwork that we as analysts had used
to characterized the underlying layer structure
of these samples.
Anything that had five or more layers,
was pulled from the population and assessed just based
on that visual microscopic assessment initially.
So, anything that had five or more layers, there were a couple
that were less than five layers,
but had very unusual layer structures, a white,
a forest green, a white or a white, a bright red, a white.
All of those, which constitutes 77 samples, were compared just
by intercomparison of that visual microscopic assessment
on the paperwork.
Those 77 samples were distinguishable from each other
and of course, from the rest of the population.
That brought us down to a smaller subset,
but we then tried to continue along this path and say,
okay these samples are single layer,
they'll all be inner compared.
These samples have two layers, they'll be inter compared.
These samples have three or more layers,
they'll be inter compared.
That got very difficult, very quickly
and I'll just flip forward and then come back to this again.
If this is an example of one of the samples we were looking at,
you might be able to tell that there are three layers
in this example and the top layer is white.
If we then tried to compare that visually to sample next to it,
we weren't sure if that sample was a very thick, single layer
or if that was multiple applications
of the same paint, etcetera.
But the tops were white
and to this point were visually indistinguishable.
Then you have a sample on the right, where the white,
the top layer's also white and the second un,
first underlying layer might be comparable
to the one next to it.
It was getting very confusing as to how to assess these samples.
So what we did was, we said, okay we're just going
to take the FTIR analysis of thin peel of the top layer
of all these samples and analyze them by FTIR.
In doing that, we were then able to form groups very readily
where some contained kaolin,
some contained calcium carbonate, some contained both
and some contained neither
and that provided much smaller populations to deal with.
We then performed FTIR analysis, as I said, and even groups
where the samples might look comparable visually,
we were able to discriminate based on FTIR of the top coat.
There were some where we couldn't discriminate by FTIR,
but then doing a visual comparison of the samples
of that subset color did discriminate between samples.
So we dealt with the hued samples,
we've dealt with the white samples, but we still felt
that we needed to go back and assess any samples
within what we called the off white group that might need
to be compared to the whites, some that were very,
very light in color and there were 54 samples we identified
from what we originally termed as hued that we thought needed
to go back and be assessed against the whites.
Many of those samples were visually consistent
with the whites, and so again FTIR analysis was performed
on the top coats and based on that assessment,
12 pairs required further analysis.
Performing, then microscopic analysis on those 12 pairs, discriminated 6 more.
Performing FTIR analysis on additional layers,
discriminated an additional pair.
So of those 54 samples that had to go back and be compared
to the whites, 5 pairs remained.
So, kind of bringing it all back
around to our original analytical scheme,
960 samples were originally part of our population,
200 top coats were analyzed by FTIR.
I'm just talking about top coats here to try
to provide some context of the number of samples
that had to go forward.
I'm not taking in to account where we needed
to do underlying layers by FTIR.
So, from those say 200 samples,
43 had to go forward to SEM at this point.
This is the sample that I originally talked
about that we didn't want to bring forward to FTIR
because we had absolutely no idea where one layer started
and where one ended just based
on visual microscopic assessment.
So, we embedded the sample and by back scatter, we could see
within this pair, these layers structures were comparable.
You can see there were 10 layers in the samples
and you can see going across the screen,
if your eye tracks across,
you can see the layers are looking pretty consistent.
Both these samples were off white visually
and they were from interior walls.
So back scatter did delineate the layers for us,
but EDS could not differentiate
between samples in any of the layers.
We then went back and performed FTIR on the top layer
and the bottom layer within this pair
and again there was no discrimination.
So, at this point, we considered this pair undifferentiated
and we did not bring it forward to pyrolysis.
In addition to that pair, 31 pairs went forward to SEM.
That constituted 27 samples.
That's from the hued group only,
the 80 percent of the population.
Back scatter and EDS together discriminated 24 pairs
from those 31.
From the white and off white group, 10 pairs went forward
which constituted 14 samples and from
that 7 pairs were discriminated.
So, again, 43 samples have gone forward to SEM
and now 19 samples need to go forward
to pyrolysis to try
further differentiation.
Seven of those pairs or 14 samples,
were part of the hued group,
four pairs or five samples were part of the white
and the off white group.
From the hued group, one was discriminated leaving 6
indistinguishable pairs.
So at this point of the analytical scheme,
this is where we stopped and we say there are 6 pairs
that we can't differentiate in that hued group.
In the white and off white group,
one pair of two layer samples, which was a white
over cream layer structure, both layers were indistinguishable
by pyrolysis.
There were three two layer samples, again white over cream
and both layers in each of those samples were
also indistinguishable.
Here's just a pyrogram depiction of pairs
that were discriminated by pyrolysis.
These happen to be one of the hued pairs.
You can see the styrene discriminates these
samples quite readily and here's a pair
that was not differentiated by pyrolysis.
So where does that leave us?
Over 950 samples were submitted and evaluated,
one 10 layer plus pair was indistinguishable through SEM,
10 pairs were indistinguishable through pyrolysis.
So, going back to Tippett's work and determining
which techniques provided the most discrimination,
in his study, he looked at microchemical testing
and microscopic examinations.
If we lumped together, our macroscopic
and microscopic exams with FTIR, we found that we had 42 pairs
that were indistinguishable after that suite of exams.
That already provided 99.991 percent discrimination.
If you added further instrumental techniques,
in that 4 percent of the original 960 went forward to SEM
and less than 2 percent went forward to pyrolysis.
You have an overall discrimination
where only 11 pairs remained indistinguishable
which is a 99.998 percent overall discrimination.
What were those indistinguishable pairs?
Well, there were two that were single layer pairs,
one was a dark blue.
One pair, one sample was originally assessed as having a
brown top coat and one was assessed
as having a green top coat and I know
that sounds a little bit funny when you think of brown
and green, but the colors were quite light
and when you put them within the brown group it was
indiscriminated, within the green group it was
indiscriminated and then you put them together
and the colors were quite close.
Most of the samples were two layer systems, there was one
that was 10 layer system and you'll look in the pair numbers
and you'll see there's three pairs
that are listed seven, eight, nine.
Those were three samples that formed three pairs so that's,
of the 11 indistinguishable pairs, that's three
within that listing of off white two layer samples.
So, conclusions.
Tippett found that two pairs of samples
from different sources were comparable and I should note
that in his discrimination power of one in a quarter million
and one in a million, depending on what techniques he used,
he excluded from that discrimination's assessment any
pairs that he found originated from the same source.
So, having anything that originated from the same source,
he did not include in his discrimination assessment
and he still found in his study, two pairs of samples
from different sources that were comparable.
For each indistinguishable pair in this study,
the samples were collected from the same building or structure.
Therefore, within this study there were no random pairs
observed to be indistinguishable.
Everything that we paired up,
came from the same building, structure, etcetera.
So, in conclusion, macroscopic and microscopic exams
in combination with FTIR remain the most
powerful discriminators.
I don't think anyone's really surprised to hear that.
SEM and Py-GC/MS though can provide additional
discrimination and should be utilized.
It saved us in the case of the 10 layer system.
We definitely needed the back scatter
to elucidate those layers.
And single layered or neutral colored samples can contain
enough characteristics to allow for a strong association
in a comparative architectural paint exam.
We'd like to acknowledge everyone
who submitted samples to the study.
It was quite a process, but Jeff Bryant and the Centre
for Forensic Sciences and our ERT squads definitely submitted
a lot of samples and I'd also like to thank Andrea
for getting us going on this project.
It was her initiative that got us going
and I really think it will be a benefit to the community.
If there are any questions, I'll take them now.
>> What, in your screening of the cross sections,
I didn't hear you mention it, apparently you didn't do it,
fluorescence microscopy, did you try or consider
that as a potential way
of making further discrimination at that level?
>> Dr. Wright: We did not, we did not try it.
I'd be open to talking about it more, but we did not try it.
>> Is there a reason why you didn't try it
or you just don't normally?
>> Dr. Wright: We don't normally because it isn't something
that we've done historically, but we're open to ideas.
>> Okay, thank you.
>> About 30 percent of the work we do is architectural,
but a lot of the architectural work we do is historic samples,
so we look at architectural paint from 1600 to 2009,
and like Skip said, fluorescence microscopy can be
incredibly helpful.
Generally speaking, synthetic binders will exhibit a very dim
fluorescence where natural binders, oils,
some alkyds, nitrocellulose will tend
to fluoresce.
So if you don't have access to SC Medias,
which we do fortunately,
fluorescence microscopy is an add on of maybe 5,
8 thousand dollars to a polarized light microscope.
It's a very quick discriminator.
Further, some pigments exhibit characteristic fluorescence
which can be helpful.
When that technique doesn't work,
another tool we found useful is to cut
or microtome a thin section and go back
to the first technique I learned
which was polarized light microscopy where you can start
to compare layers based on the particle composition,
particle size, distribution, etcetera
and I should have prefaced this by thanking you for this study
because I think it's fabulous.
Very good work.
>> Dr. Wright: Thank you.
>> Tom: Go ahead, Ed.
>> Ed: Thanks Tom.
I also want to thank you.
I think that was an excellent study.
Something else that I thought I might mention is
that we've looked at quantitating titanium dioxide
in fibers, as fillers and this might be considered
as an additional study if one might what to do anymore
than you've done, but if you look at the concentration
of titanium dioxide which is readily done using Raman
spectroscopy, that can also be useful discriminating method.
>> Dr. Wright: Thanks Ed.
>> Tom: Again, a nice piece of work, a lot of work.
To add onto Skip's comment,
a technique that's highly interused
in forensic laboratories
is cathode luminescence.
I did a paper in 1996 at the Trace symposium
and Chris just did a chapter in a book and Joanne, I think,
will back me up on this, especially white paints,
off white paints cathode luminescence every time
distinguished between them just by putting them
under the scope and looking at them.
>> Dr. Wright: Thanks Tom.
>> Hello, I'm Sal Bianco [assumed spelling]
from the Maryland State Police crime lab.
Just a quick comment, I enjoyed your paper.
On the last proficiency test that came around,
it was on white paint and I was sort of pressed for time,
so I started to think, what could I do quickly.
So I grabbed a crime scene scope and tried a variety
of wave lengths and different goggles
and I could distinguish all the white paints,
they all fluoresce differently.
>> Dr. Wright: Which technique?
I'm sorry, I couldn't here you.
>> Sal Bianco: It's called a crime scene scope.
It's an alternate light source.
It's like a 500 watt light bulb and down
and dirty quick screening, I could discriminate them.
>> Dr. Wright: That would have come
in handy with 197 pill boxes.
[ Laughter ]
We just walked in the lab every day
[ sigh ]
>> Ed: By the way, in regard to the comments I made
that the work that was done with that study
of the titanium dioxide was done in Steve Morgan's group
in South Carolina and there's, it's been published
in Applied Spectroscopy.
>> Dr. Wright: Thanks Ed.
>> One more just very... one more very,
sorry, can you hear me?
One more, just very brief comment.
White paints, which can be incredibly difficult
to differentiate, suffer from the problem
that the natural binders, like alkyds and oils, those binders
with discolor in the absence of light.
So, if it's covered by a top coat, you,
look in your kitchen cabinets, they'll be yellow,
outside of the cabinet is bright.
So, if you find evidence that it's protected from sunlight
and white paint that has been exposed to sunlight over time,
a simple macroscopic examination may cause those two colors
to appear very different.
So one thing when we do in trying
to determine an historic paint color,
is often to photo bleach the sample
to eliminate the discoloration.
That, it's a process that Rubens used back in 1600s
after rolling paintings, putting them on ships and sending them
across vast distances.
He would say, unroll the painting and expose it
to sunlight for three days
and the whites will be white instead of yellow.
So if your visual discrimination shows two different hues
or two different off whites, it might be helpful
to photo bleach that, you can do
that with a fluorescence microscope
or a long wave UV lamp over a period of hours.
>> Dr. Wright: Thank you.
>> Any other questions?
Thank you.
[ Applause ]