Detecting Blood on Fabrics: Infrared Diffuse Reflectance Versus Attenuated Total Reflectance FT-IR

Aug 14, 2017
By Spectroscopy Editors

In forensic science, the detection of blood on fabric is a very useful tool. Therefore, it is important that the methods used for detecting blood be as accurate as possible. Michael L. Myrick and Stephen L. Morgan, both professors in the Department of Chemistry and Biochemistry at the University of South Carolina, have been investigating the use of infrared (IR) spectroscopy for this purpose, including comparing the effectiveness of infrared diffuse reflectance versus attenuated total reflectance Fourier-transform IR (ATR FT-IR). They recently spoke to Spectroscopy about their recent studies and the critical questions they have been addressing in how IR spectroscopy is used in forensic science.

Michael L. Myrick Stephen L. Morgan

 

 


You have recently published work on the use of ATR FT-IR spectroscopy and infrared diffuse reflectance spectroscopy for the detection of blood on fabrics for forensic purposes (1,2). First, why did you undertake this work?

Myrick: Our labs have been working together for a while on spectroscopic measurements of blood on fabrics, since well before 2010. Steve’s lab has worked for many years more on forensic analysis and chemometrics, while my work focused more on vibrational spectroscopy (Raman initially, then much more on IR and near-infrared (NIR) spectroscopy for the past 20 years). Our families are friends with one another, and Steve is “Uncle Steve” to my son Alex. Steve and I got started on this work together as a result of talks we had while walking Alex as a toddler around in a push car in my neighborhood after family dinners on weekends. In essence, we said, “Your lab does chocolate and my lab does peanut butter, so why don’t we make peanut butter cups”?

From that simple beginning we started doing spectroscopy for forensic applications, and the first thing we worked on was blood on fabrics—an organic analyte on an organic substrate. We chose that problem because it involved both of our interests: chemometrics and vibrational spectroscopy. Fortunately, our interests turned out to be new enough that the National Institute of Justice was willing to fund the work for several years.

Doing the actual work, though, we began to realize that there was a lot of missing analytical science in forensic measurements. There were already methods for measuring blood in the forensic field, ranging from made-for-TV luminol tests, to short-wave NIR reflectance measured with silicon-based video cameras. However, very little at the time had been done with long-wave NIR or mid-IR spectroscopy, and an important piece of information that we and everyone we talked to wanted to know was: How does what you’re doing compare to what we already know? Although our own interests were quite different from conventional spectrometer-based spectroscopy, we were inexorably drawn into answering questions about conventional methods.

These were difficult questions to answer for several reasons.

First and most surprisingly, no one had done a systematic study of detection limits using even the most well-known methods. Detection limits for the luminol method had appeared in a number of papers, but they differed by five orders of magnitude from one another, and they were in units of convenience rather than units of relevance.

Second, the unit of convenience used in the forensic literature of blood detection is “dilution factor,” which is very confusing when you realize that the measurements are made on a residue of blood solids after drying. Units like that are easy to use, but the meaning was ambiguous. We needed to figure out how to describe detection limits using quantifiable units. Eventually, we decided that different detection methods are best described and compared with different units. For example, NIR diffuse reflection has a fairly deep sampling depth, on the order of tenths of a millimeter, in fabrics. Blood that soaks into a fabric can be sampled with NIR. Units like % w/w or micrograms per unit area are well suited to describing detection limits in something approaching bulk type measurements. ATR, on the other hand, samples only the top surface of a fabric. Thus, transferred blood stains—where something with wet blood contacts the surface of a fabric, and a film of blood might remain only at the surface—could be sampled effectively with ATR, but not with NIR. For ATR and other surface-sensitive methods, a measure of film thickness seemed more appropriate for describing detection using ATR and other surface-sensitive methods. We decided to produce detection limits in all these units.

Third, no one had developed good methods for reproducible sample preparation. In most cases, samples were prepared by dripping blood solutions onto a substrate from a pipette or something similar. This was done manually, producing very heterogeneous and irreproducible samples unsuited to calibration work. The logic behind this sample preparation method was that you don’t get reproducible stains in the field. But as chemists and spectroscopists and scientists, we understand that finding calibrations and detection limits requires standards. For that we need control of the process of making standards so they are as homogeneous and well-understood as we can get them. We ultimately developed two methods for making blood-on-fabric standards. The first was via dip-coating fabrics in diluted blood and developing a calibration between dilution, draw speed, and mass of blood solids deposited for each fabric. This method provides very reproducible and homogeneous samples. The second method specifically developed for luminol testing used hydrophobic polymer rings to confine a diluted blood sample to a specific area on a fabric, in this case cotton. Blood dilutions were made in units of liquid blood mass percent on a Mettler-Toledo AG204 analytical balance, using the absolute gravity of water at 25.5 °C (0.9969 g/mL) and blood (1.0595 g/mL). Aliquots of these known solutions were transferred to the rings and dried to produce standards used mostly for studies of the common luminol method.

Fourth, there was confusion about what infrared means in the forensic field. The infrared region spans a large range of wavelengths and energies, and previous work had mostly covered the shortest wavelengths detectable with conventional charged-coupled device (CCD) cameras. Detection in this wavelength range is relatively poor, but a forensic practitioner wouldn’t necessarily realize that chemical spectroscopists consider the shortest wavelength NIR as very different from longer wave NIR, which is again quite different from mid-IR. We felt it was important not to categorize infrared as a single thing, but to show how different bands of infrared wavelengths might behave differently in blood detection.

Fifth, we knew that substrates matter. Fabrics are different chemically from one another, and therefore are different spectroscopically. This difference between substrates matters a great deal in vibrational spectroscopy. It also matters in other types of testing, because the fiber sizes, hydrophobicity, and response to moisture and temperature can vary greatly between different fabrics. We made an effort to characterize our fabrics as well as possible.

We began our work with basic diffuse reflection spectroscopy. We spent a few months studying all the commercial FT-IR instruments we could locate at our institution to figure out which showed the lowest noise for different wavelength regions before settling on the instrument in our laboratory to use for our study as a good compromise. Because that was the first paper in this direction of our work, we included an extensive table of literature results for comparison of methods (3). We defined detection limits in relevant units and introduced our first method for making reproducible samples. We also broke the infrared region into three distinct bands: mid-IR, long-wave NIR, and short-wave NIR. We also looked at multiple fabric types. We followed that experimentation with another basic technique: ATR FT-IR.

Can you please briefly describe the differences between the two approaches?

Morgan: Diffuse reflection methods involve directing a beam of light onto a diffusely reflecting sample like a powder—or a fabric. Diffuse reflection is generated by the light penetrating into the material via multiple reflections, scattering around and sometimes penetrating through particles (or fibers) in the sample. Light that penetrates the sample top surface can be absorbed by the sample, or can emerge after multiple reflections from the opposite side (diffuse transmission) or can emerge from the side on which it entered (diffuse reflection). Diffuse reflection is a function of particle size and shape, so although there are theoretical studies that describe the process (the simplest being the Kubelka-Munk model, which we recently reviewed [4]), the actual process is messy and not readily predicted from theory. Diffuse reflection is always mixed with a little specular, or front-surface, reflection. Generally, diffuse reflection has a penetration depth on the order of millimeters, very dependent on the strength of absorption. For the same reason, it is also quite nonlinear because strong absorption reduces the effective pathlength and penetration depth.

ATR is quite different. Light reflects from the high-index side of a high-refractive-index crystal surface, like a germanium crystal or diamond. The fact that the light strikes the interface from the high-index side (for example, the diamond side of a diamond-air interface) is what makes it “internal” reflection. This type of reflection, when the angle of incidence is greater than the critical angle, is the most efficient reflection of light known to man, far better than the best mirrors. It is so efficient we call it total reflection; this phenomenon is used to transmit light through optical fibers over kilometers of distance, and the loss from reflection is negligible inside the fibers—all the loss can be attributed to absorption in the medium or defects. However, it has been known for hundreds of years that when light strikes an internally reflecting surface, it doesn’t appear to bounce back from the surface. If one traces the incident ray and reflected ray to their point of intersection, they appear to meet just across the interface in the low-index medium. This point of intersection allows the light to sample the low-index medium. If it happens that the low-index medium can absorb the light, then the reflection is no longer total—it is attenuated, hence the term attenuated total reflection, even though that name seems like a self-contradiction. This method is useful for sampling thin films or other materials in very close proximity to the interface. As a rule of thumb, the depth of penetration of light in ATR is around one-tenth of the wavelength of the light. This means that long wavelengths produce longer pathlengths, and thus ATR is most commonly used in the mid-infrared spectral window where wavelengths are on the order of several micrometers and penetration depths can reach 1 µm or more.

Why is ATR FT-IR better than diffuse reflectance spectroscopy for blood stain detection on fabrics?

Morgan: We’ll give you the academic answer first: The two methods have different strengths and weaknesses. For instance, diffuse reflection is a noncontact method, while ATR requires contact. Diffuse reflection works in all spectral windows, while ATR is most sensitive in the mid-IR. But in terms of ultimate or best detection limits, ATR definitely outperforms diffuse reflection.

The practical answer is this: Blood solids on fabrics are heterogeneous samples at which ATR excels. When measuring one organic material coating or another, spectroscopic detection limits will be affected by the spectroscopy of the substrate unless there are unique spectroscopic bands present in the coating that clearly distinguishes it. When few blood solids are present, there is still a spectrum—it’s just the spectrum of the substrate fabric. Our best detection limits arise when we can minimize the spectrum of the substrate relative to the coating. ATR allows us to have strong surface sensitivity. Blood solids sit on the surface of the fibers, so surface sensitivity allows the blood solids to dominate the spectrum when they are present. If the coating of blood were on the order of micrometers thick and the coating were uniform, ATR might not be able to see the substrate fabric at all. It is this surface sensitivity that makes ATR better at detecting small quantities of blood.

Your methods also involved gap derivative processing and partial least squares (PLS) regression. What challenges, if any, did you face in developing the calibration model?

Myrick: Spectra of fabrics can be quite variable depending on many factors, but mostly the effects are in the form of baseline shifts and curvatures. A simple way to deal with those effects is through derivative processing, so we began looking into different derivative processes. And there are many different ways to process derivatives—too many to make a fully comprehensive study in any single paper.

Around the time the work was getting started, we attended the International Diffuse Reflectance Conference in Pennsylvania, and saw the well-known NIR spectroscopist Karl Norris talk about fourth-derivative “gap-segment” processing. It seemed pretty interesting and we decided to look into it in our work (5,6). Little did we realize at the time that it would present us with big mysteries.

The problems with calibration fell into two categories—first, brute-force calculation of all the myriad variations of four gaps in the spectrum, and then producing calibration maps for all of them; and second, understanding what the results meant. The calculation part was straightforward compared to the understanding of two mysterious observations. First, we found our detection limits were strongly dependent on the exact combination of wavelength gaps chosen. And although the analyte was always the same (blood solids), the optimum combination of wavelength gaps varied strongly with the fabric type. We didn’t understand those results.

Trying to understand drove us crazy for about 18 months until we finally began to see the answer: Derivative processing by the gap method in fourth-derivative (and probably better with higher derivatives as well, though we didn’t test it) allows the processing to be sensitive to patterns that differentiate the analyte from the substrate, a form of matched filtering first used many years ago in radar detection of aircraft signatures. It’s a technique for minimizing the signature of the nontarget materials (fabric substrate) compared to the target material (blood solids).

What detection limits were you able to achieve on the various fabric types you tested? Is the method effective on all fabrics you tested?

Morgan: Our 2015 paper (1) showed that measurements of blood detection limits in the mid- and near-IR using diffuse reflection spectroscopy can be based on characteristic strong protein absorptions due to the amide I band at 1650 cm-1 and the amide II band around 1550 cm-1, the amide A band near 3300 cm−1, and the amide B absorption band near 3100 cm−1. With appropriate selection of gap derivatives, partial least squares can yield detection limits for blood solids ranging from 14 µg/cm2 (930x diluted) on acrylic, 17 µg/cm2 (500x diluted) on cotton, and 6.2 µg/cm2 (610x diluted) on polyester. Although detection limits can be measured in other wavelength regions, characteristic background from fabrics can also impede or aid detection. For example, valid calibrations were not possible with diffuse reflection for blood solids on nylon because of the similarities between IR spectra of blood and nylon.

ATR FT-IR measurements coupled with gap derivative processing and partial least squares improved detection limits for several fabrics. Our recent 2017 paper (2) demonstrated that ATR FT-IR with partial least squares provided improved detection limits of 0.0026 µg/cm2 (2700x diluted) on cotton (arguably the most common forensic target), as well as the first measured infrared detection limit of 0.0077 µg/cm2 (250x diluted) on nylon. Detection limits for polyester and acrylic were estimated at 0.0066 µg/cm (280x diluted), and 0.011 µg/cm2 (270x diluted), respectively. The values in units of µg/cm2 are so much better because ATR only sees blood solids at the very surface.

What are your next steps in this work?

Myrick: We have another paper in press right now on detection limits with luminol that will complete our study of detection limits by conventional methods (3). When we began our work, we didn’t expect, plan, or even want to do the work described in these three papers; we really wanted to focus on infrared imaging. This work was driven by necessity because of the poor state of knowledge of the detection of blood by conventional means, without which we couldn’t compare the work we did expect, plan, and want to do.

So with these three papers completed we are hoping to be able to continue with our main focus of research in imaging. Still, one can’t help but notice that more questions have been raised by these three papers, some in forensic analysis and others for the general application of spectroscopy and mathematical preprocessing of spectra. And one suspects that there are other fields where very fundamental questions of chemical measurement have not yet been addressed in peer-reviewed literature. These are not problems caused by lack of technology, but rather lack of application of existing chemical and spectroscopic science to current technology. Forensic science is a field where there’s such a drive toward applications that basic understanding can be left behind. We often hear research in forensic science devolve into a question of “But when can you deliver an instrument?” without a lot of patience for basic science.

References

  1. S. A. DeJong, Z. Lu, B. M. Cassidy, W. L. O’Brien, S. L. Morgan, and M. L. Myrick, Anal. Chem. 87(17), 8740–8747 (2015). DOI: 10.1021/acs.analchem.5b01825
  2. Z. Lu, S. A. DeJong, B. M. Cassidy, R. G. Belliveau, M. L. Myrick, and S. L. Morgan, Appl. Spectrosc. 71(5), 839–846 (2017). DOI: 10.1177/0003702816654150
  3. B. M. Cassidy, Z. Lu, J. P. M., S. K. Tazik, K. A. Witherspoon, S. A. DeJong, R. G. Belliveau III, K., S. M. Ervin, M. L. Myrick, and S. L. Morgan, Forensic Sci. Int., in press, available online (2017). DOI: https://doi.org/10.1016/j.forsciint.2017.06.031
  4. M.L. Myrick, M. Simcock, M.R. Baranowski, H. Brooke, S.L. Morgan, and J.N. McCutcheon, Appl. Spectrosc. Rev. 46(2), 140-165 (2011). DOI: 10.1366/10-06137
  5. S.A. DeJong, W.L. O’Brien, Z. Lu, B. M. Cassidy, S.L. Morgan, and M.L. Myrick, Appl. Spectrosc. 69(6), 733–748 (2015). DOI: 10.1366/14-07693
  6. S.A. DeJong, Z. Lu, B.M. Cassidy, S.L. Morgan, and M.L. Myrick, Appl. Spectrosc. 70(6), 1044–1054. (2016). DOI: 10.1177/0003702816641273
native1_300x100
lorem ipsum