Tip:
Highlight text to annotate it
X
[ Music ]
>> Next up is Kristine Olsson, and Kristine has been
with the Johnson County Sheriff's Office Crime Lab
for nearly five years.
She's been in forensic science for 14 years, and is happy
to say that doing away with trace evidence section,
the Johnson County Sheriff's Office Lab enjoys the
distinction of reestablishing this discipline,
and she is going to be talking to us today about
"A Proposed Standard Test Method For Forensic Analysis
of Glass Using Capillary Micro X-ray Fluorescent Spectrometry".
Kristine?
>> Thank you.
I originally submitted this as a poster presentation,
so I alternate between thanking the organizers and cursing them
for putting me up here, so.
But no, I'm very pleased to be here.
[ Pause ]
My work today is presented on behalf
of the Elemental Analysis Working Group.
There's a number of posters and presentations
by this group throughout the week,
and basically this presentation concerns a proposed standard
method for XRF analysis of glass samples.
I'll be speaking with you today
about basically the history driving the proposed
standardized method.
Why did we feel that there was one that needed to be put out?
Also, we'll discuss some of the aspects
of the analytical recommendations that we make
in the method; some of the instrument parameters,
and some of the recommended QC measures that are in the method,
as well as a sample analysis scheme; an evaluation
of the results, which includes basically how you look
at your spectra or how do you evaluate your spectra;
and interpretation of results, which is inclusive
of recommended match criteria.
We'll finish up with a timeline for completion of the method
and submission, review and submission to ASTM
for acceptance, hopefully.
A little bit of background
on the Elemental Analysis Working Group.
It was established
by the International Forensic Research Institute
at Florida International University
under a grant from NIJ.
The aims of the working group were two-pronged
to improve the forensic analysis of glass
through validation standardization efforts
and examining the way the different instruments performed
from LIBS to XRF to the ICP-MS systems.
And also the other prong was
to develop standardized methods incorporating recommended
match criteria.
So, you've already heard how some
of the match criteria performed for Emily's work,
and there is a presentation coming up also
which discusses the way match criteria performed among all the
different instruments.
So why a standardized method for XRF?
In the 2010 CTS reports, approximately 69 percent
of the labs are using some sort of elemental analysis
to render their conclusions for glass.
Standard methods do currently exist.
There's an ASTM standard for solution-based ICP-MS, however,
micro-XRF is the preferred method for most of the labs
in the U.S. Approximately 35 to 40
of the labs are currently using micro-XRF, and right now,
there is no standard published method for addressing the use
of XRF for analysis and discrimination.
When we brought on glass in our laboratory, we tried to find
out what other people were doing as far as their match criterias,
and pretty much the answer was all the same is that you need
to talk to Scott Ryland, just go talk to Scott Ryland.
So, Scott Ryland was, we're trying to standardize Scott,
basically, and get him on the books before he retires
for good and leaves us.
The method that we've put together addresses only source
discrimination, so K versus Q comparisons.
And it uses, it discusses the use of mono-
and poly-capillary optics
and energy dispersive X-ray detectors.
The method does not address the end-use classification of glass,
which we did talk about, but didn't end up going that route,
and you can use XRF to determine whether a glass is from a window
or if it's from a container, but not in this particular method.
For those -- how many of you are XRF users out here?
Okay, quite a few.
For those of you who aren't, just a brief background on XRF.
It is a completely nondestructive method,
and it's based on the emission
of characteristic X-rays following excitation
of the sample by the X-ray source.
So you have your source here, it strikes the sample
and emits characteristic X-rays that are then detected
and separated by energy in the detector.
It works mostly for elements that are atomic number eleven
or greater, so basically,
for all practical purposes, sodium and up.
The method that we put together is all based on data
that was collected through the Elemental Analysis Working
Group, and it's based off the four round robin studies
that were done.
Each of those four round robins had a K/Q comparison aspect
to them, and they range from December 2008
to August of last year.
And they range from being very simple, so one K compared
to one Q, up to very challenging tests of multiple Ks,
multiple Qs from glasses that were all produced
in the same plants at very close time intervals.
One of the recommendations that did come
out of the Working Group and the round robins
that we performed is recommending the use
of the reference sample, something like 1831 in your --
included in every run that you do with your glass.
So it's run, every time you run your glass analysis you include
1831 as a sample and run it under the same parameters
that you run your samples under.
And this can be used for various things.
It could be used to normalize your elemental ratios,
which we'll discuss in a minute.
It can be used to normalize for inter-laboratory comparisons.
So you can compare your data
with data developed from a different lab.
And it normalizes basically instrument response
between different laboratories.
It can also be used
for inter-laboratory data collected from different days.
So if you run your samples on different days,
you can still compare them back
to each other using 1831 to normalize.
And finally, it's a QC check to monitor performance
of your analytical method.
This is a spectrum of sheet glass, basically.
This is 1831, and when we talk about elemental ratios,
for those of who don't do XRF currently, it's referring
to making a ratio of the relative intensity
of your calcium peak against other elements in your sample.
So, for example, you could ratio the intensity of the calcium
to the intensity of iron and use
that as a semiquantitative comparison for your sample.
Although you can do quantitative data with XRF,
for glass analysis, because of some of the limitations,
we do comparison on a semiquantitative basis.
This is my data, so right here these are my instrument
parameters for my particular lab.
Every lab doing the round robin studies came to the table
with their own parameters, and then over time,
we kind of optimized parameters that worked best
from what we were seeing in the studies.
And this is some of those instrument parameters included
in a sample analysis scheme.
These are some of the recommendations that we make,
a spot size range is needed for glass analysis
between 10 micrometers to 2 millimeters;
an excitation voltage for your instrument
of at least 35 kilovolts to excite your higher end,
higher energy elements; a beam current not
to exceed 50 percent dead time; a midrange time constant
which gives you the best compromise
between counting statistics and resolution;
a life time providing reasonable counting statistics.
For example, in a mono-capillary system
with 100 micrometer spot size, 1200 live seconds
with give you good counting statistics
for a good spectrum for comparison.
And then these two aspects here, presentation of a flat surface
of your glass sample to the X-ray beam
which will minimize your takeoff angle effects.
On full-thickness glass fragments that's really not a
problem since generally you have two flat surfaces
and they're nice full-thickness.
On small fragments, that gets to be a little bit more difficult.
And when evaluating K and Q against each other,
it's very critical to use fragments of similar size
and thickness to minimize your critical depth effects.
And basically when we did our round robins, we were requested
to run exactly the same number of replicates on all
of our samples, so these parameters were standardized
and they performed very well among all the
different laboratories.
So the recommendation is that on your question fragment,
a minimum of three replicate measurements should be taken.
And when you're characterizing your known standard,
a minimum of three different, three replicate measurements
on three different fragments from your known should be run
for a total of nine replicates.
When evaluating your results, or evaluating the spectra
that you've obtained, these are the recommendations
that we make.
The first is the overall assessment of your spectra,
so assessing your spectra to determine if--
what elements you have present
and discriminating your artifact peaks
from your two element peaks.
The recommendation is that in order
to determine whether an element is present,
there should be a signal-to-noise ratio of three
or greater and that element may be identified.
Troy Ernst, who is also in the Working Group,
is going to be speaking after the break about
"when is a peak a peak?"
So when do you actually call a peak in your XRF spectra?
The next step is to do a spectral overlay
when you're doing your K/Q in comparisons,
so you overlay your spectra one on top of the other.
For some laboratories, including some that participated
in the Working Group, this is their only basis for comparison.
They don't do any other statistical calculations;
it's simply a spectral overlay.
And when you overlay your spectra, you want to ensure
and verify that the same elements are present
in the same relative intensities of your spectra.
And the last evaluation is the semiquantitative evaluation
where you calculate the ratios and determine statistically
if they are different
or significantly different or the same.
You want to, when you're calculating your ratios,
you want to use elements, and this is recommended
to use elements that are close
in atomic number or close in energies.
Those will suffer the same takeoff angles similar
to each other, and for, in order to use an element
in a quantitative ratio or semiquantitative ratio,
it needs to have a signal-to-noise ratio
of 10 or above.
And again, Troy will be addressing the applications
of these two recommendations specifically
to XRF data in his talk.
For match criteria, it's recommended or it's --
you may use either of a range overlap or the calculation
of three standard deviations of your calculated elemental ratios
as your statistical measure.
So what we have basically is, again, this is a spectra
of sheet glass, it's 1831.
And just looking at your, that first step of the evaluation
of your spectra and determining whether you have artifact peaks,
like the copper, this isn't really copper in your sample,
but it is an artifact peak, and this is a pile-up peak of some
of the other things that are present in your sample.
These are common elements that are seen in glasses.
There certainly can be other elements included
in there and not just these.
And you also want to check your spectra for artifact peaks
such as escape peaks or sum peaks.
Once that's been established, then you move
on to a spectral overlay of your K
and your Q directly on top of each other.
Just looking at this, we'll be looking at the,
asking you to focus on the iron and the manganese in here.
This is from round robin 4 that we did in the Working Group,
and this is a K/Q comparison, your K is blue, your Q is red.
And for the XRF users, and even for the non-users,
if you were looking at this and assessing it, would you say
that those glasses, by a show of hands, are different?
I see some people putting hands up.
How about now when you can see the peaks a little bit better.
There is a difference here in the iron peak
and the intensities that you see here.
So the difference is a little bit notable.
These are different glasses.
They were produced in the same plant approximately one month
from each other.
At this point, if you were doing a spectral overlay
and assessing this, at this point you could discriminate
those samples and stop there and say that they are different.
Or you could go on and calculate your statistical calculations
to determine if they are different or not,
or basically prove that they're different.
Looking down here at the manganese peak, down here,
it's kind of down here in the weeds,
and it looks a little bit messy.
The question comes into play is that a peak or not?
And these are the kind of samples
that actually inspired a little bit Troy's talk and his work,
so you don't want to miss "When is a Peak, a Peak?"
But this down here, if you apply what he's going to be talking
about to that, you can do a calculation
that determines that, yes, in your Q, you can call
that present, as manganese as being present,
but not in your K. If you'll notice the calculation is not 10
or above, so you can't use it
in your semiquantitative comparison,
but it is distinctive that it's there
in one and not in the other.
That was the first Q. This is the same K compared
to a second Q. And looking at this, by a show of hands,
would you say this is different, these glasses are different?
Nobody? Gosh, you guys are so good.
This is, these glasses are indeed the same;
they are from the same source.
But, if you do notice here, there is a little bit
of a difference in the intensity of your peak, and at this point,
this is the kind of situation where you could go on
and do your statistical calculations to show
that they are significantly different.
These ratios here are recommended ratios
in the method, and again, if you have something else
in your sample, you're not beholden to just use these.
If there's another element that is applicable in your sample,
you could certainly incorporate those calculations
for your comparison.
And also by that same token, if there's something
in these ratios that is missing from your sample,
then you're not beholden to include that ratio.
For the interpretation of results or the match criteria,
again, the recommendation is
that you can do one of two things.
The elemental ratio range overlap.
So you would characterize your known and get the complete range
of ratio calculations in your known.
Do the same thing for your Q
and determine whether those ratios overlap.
In some of our labs that participated,
this is their preferred method of statistical evaluation,
and it has two benefits in that it's nonparametric,
it's very easy to explain to a jury.
The other thing is, you could do three standard deviations.
So from your K you determine the mean,
determine your three standard deviations,
and that becomes your range to compare to.
Then you determine the average of the calculated ratios
in your Q and see if that average falls
into the range created by your known.
In doing the round robins,
these two statistical applications perform very,
very well, comparable to each other.
And this is applying it to the two samples
that we've already looked at, the Ks and Qs
that we've already looked at.
So in blue you have K1, the calculated calcium
to iron ratio, the mean,
the standard deviation, the created range.
The Q1, the calculated calcium to iron ratio is 2.98,
which does fall in, so they are the same source, K and Q,
and there's a nice overlap in your spectra as far
as range overlap goes.
This is range overlap calculated for the K that we looked at,
the Q1, which did not match and does fall,
when you do your calculated range overlap,
the ranges do not overlap,
so this statistically is found to be different.
And then the K2, which is from the same source,
all the ranges do overlap with your K.
For the three standard deviations,
it performs the same way.
So here's the range for your K glass, the Q1,
which did not match, the calculation, the mean of all
of your replicate measurements on your Q. The calculation
for that is 1.98, which falls outside of this range.
So they are found to be significant,
there's a significant difference between that Q and that K.
For K2, which was from the same source,
the calculated average does fall into the K1 range.
And again, these glasses were, K1 and K2,
K1 and Q2 were produced in the same plant at the same time --
oh, I'm sorry, those are from the same source.
K1 and Q1 were produced
in the same plant a month apart from each other.
What the method does not address is a significance
of the elemental association.
The only thing we do as far as go to match criteria
and interpretation that way,
there's no addressing of significance.
There's also no discussion of a consensus
for language describing that significance,
and all that is still under discussion in our Working Group
and I know in other circles.
And again, this method does not address the use of XRF
to determine end-use classification
for your question particles.
The timeline from here on out is
that hopefully the draft will be distributed
to SWGMAT following this Symposium.
SWGMAT will distribute the draft to all their members
for comments, and hopefully within one month,
following review, we'll get those comments back
and incorporate those into the draft.
Then it will be submitted to ASTM in October
for the fourth quarter 2011 vote.
And if you are interested in getting copies of the method
to review and comment on, Dr. Almirall is available
to provide those copies for you,
and he'll be speaking later on too, as well.
These are all the XRF participants that participated.
There were nine labs total
that participated in the round robins.
You can see right there, Scott Ryland participated,
so I'd like to thank him because he drove a lot
of work done on this.
And thank you to the solution ICP-MS users,
the laser ablation ICP-MS and LIBS users
that participated, as well.
Thank you.
[ Applause ]
>> So we will entertain questions for any
of our three speakers now if anybody has a question.
Yes, sir.
>> I have a question here for Emily.
It's not clear to me why you're assuming two unique glass panes,
which is [inaudible], why you're assuming they should be the
same [inaudible].
>> We're assuming that it would be produced
in the same manufacturing plant around the same time
to make the windshield pane, we would assume
that they would come off the sheet around the same time.
>> And I agree that in most cases,
that's a correct assumption, but there could be an instance
where the one pane could come off the pallet,
that pallet's now empty, they bring in another pallet
and the next pane [inaudible], and they could have come
from conceivably weeks apart [inaudible].
I agree that your assumption would be valid
for most cases, but --
>> Yes, and that might be true
for why not all 11 pairs were associated
for this particular set.
>> Anyone else?
Yes, sir.
>> I have a question for Stefan.
Do you use the same energy that's [inaudible].
Have you tried to anneal the glass fragment [inaudible].
>> Actially, we haven't.
This is one of the things we were still do,
because in the course of thinking
about what does a red curve really mean.
It's not really, you know, printed in the user manual
of Foster and Freeman GRIM 3, what is a red curve
and what does it mean?
And in a way, so I digged a little deeper
and after some time I got pretty good communication to Foster
and Freeman, and they said, okay, a red curve means
that you really have a problem, not you, but the instrument,
of course, with contrast and the variation of the contrast
and red cursors -- not cursors, ah, curves, [inaudible] also
when the glass has been close to radiation also,
radioactive radiation, and this can be overcome by annealing.
So certainly we would go into this point and select a couple
of glasses and do the annealing, certainly not for all glasses,
but for some of them, yes.
Of course, then we have to live
that we can produce fresh etches, we just have
to get the embedded material, we have to you know, clean it,
do the annealing and the fragments are really tiny now,
so it'll be a pain, but I'm certain Carolyn will give it
a try.
>> Yes, sir.
>> Kristine, this may be a silly question, but for the labs
that do spectral overlay, is it done visually?
>> Yes.
>> Not numerically?
>> No.
>> Ma'am?
>> A question for Emily.
You said that you use silicon as your internal standard.
Were you concerned about silicon coming off say, the torch,
and producing extra air into your method?
>> No, because it's present in such a large concentration
in glass that any sort of other variation wouldn't contribute
so much to the peak intensity, most likely.
>> Any other questions?
Going once, going twice, okay everybody,
you're a few minutes early for your break.
So if you'd like to go ahead and take your break now,
we'll come back here and we'll start again at 4:30.
[ Applause ]