ORIGINAL PAPER

Non-invasive detection of superimposed latent fingerprintsand inter-ridge trace evidence by infraredspectroscopic imaging

Rohit Bhargava & Rebecca Schwartz Perlman &

Daniel C. Fernandez & Ira W. Levin & Edward G. Bartick

Received: 29 January 2009 /Revised: 15 April 2009 /Accepted: 21 April 2009 /Published online: 5 May 2009# Springer-Verlag 2009

Abstract Current latent print and trace evidence collectingtechnologies are usually invasive and can be destructive tothe original deposits. We describe a non-invasive vibra-tional spectroscopic approach that yields latent fingerprints

that are overlaid on top of one another or that may containtrace evidence that needs to be distinguished from the print.Because of the variation in the chemical compositiondistribution within the fingerprint, we demonstrate thatlinear unmixing applied to the spectral content of the datacan be used to provide images that reveal superimposedfingerprints. In addition, we demonstrate that the chemicalcomposition of the trace evidence located in the region ofthe print can potentially be identified by its infraredspectrum. Thus, trace evidence found at a crime scene thatpreviously could not be directly related to an individual,now has the potential to be directly related by its presencein the individual-identifying fingerprints.

Keywords Latent fingerprints . Trace evidence . Infraredspectroscopy . IR . FT-IR . Spectroscopic imaging .

Chemical imaging

Introduction

Fingerprints and trace evidence are critical and integralconstituents of forensic investigations. As opposed tovisible patent fingerprints left clearly by blood or othersubstances, latent prints primarily contain residual materialfrom the person making the print and are difficult orimpossible to visually detect without treatment. Prints ofthis type typically require invasive techniques for theirdevelopment using powders, chemical reagents, and lightsources [1]. Two critical aspects of trace evidence charac-terization involve the identification of target areas toexamine and the subsequent analyses to accurately deter-mine the identity and the source of the material. Many ofthe approaches involved in collecting trace evidence areinvasive and destructive. For example, swabbing objectsdestroys fingerprint deposits within the area. A thorough

Anal Bioanal Chem (2009) 394:2069–2075DOI 10.1007/s00216-009-2817-6

A portion of this work was presented at the 16th Meeting of theInternational Association of Forensic Sciences, Montpellier, FranceSeptember 2–7, 2002.

R. Bhargava :D. C. Fernandez : I. W. LevinLaboratory of Chemical Physics, NIDDK,National Institutes of Health,Bethesda, MD 20892-0520, USA

R. Schwartz Perlman : E. G. BartickCounterterrorism and Forensic Science Research Unit,FBI Laboratory, FBI Academy,Quantico, VA 22135, USA

Present Address:R. BhargavaDepartment of Bioengineering and Beckman Institutefor Advanced Science and Technology,University of Illinois at Urbana-Champaign,405 N. Mathews Avenue,Urbana, IL 61801, USA

Present Address:R. Schwartz PerlmanIdeal Innovations, Inc,950 N. Glebe Road, Suite 800,Arlington, VA 22203, USA

Present Address:D. C. FernandezMount Sinai School of Medicine, One Gustave L. Levy Place,New York, NY 10029, USA

Present Address:E. G. Bartick (*)Department of Chemistry and Biochemistry, Suffolk University,41 Temple Street,Boston, MA 02114, USAe-mail: ebartick@suffolk.edu

examination for trace evidence is recommended beforeprocessing for prints and should be conducted by experi-enced and competent investigators to avoid disturbing areaslikely containing latent prints. Analysis of trace evidencespecifically within a print is not currently targeted. On aresearch basis, Raman spectroscopy has been used for theinvestigation of drugs of abuse detection in fingerprints[2]. However, spectra of the drugs were obtained withoutan image to show the spatial relationship within thefingerprints.

Optical techniques are employed to locate fingerprints asa preliminary step and may involve monitoring intrinsiclatent print luminescence or UV reflection [1]. However,fingerprint fluorescence is often not present or intenseenough to be detected with current instrumentation.Advances are being developed in time resolved methods,but these are still complex and typically geared toward usewith the europium-based treatment strategy which is aninvasive method [3]. Further, the use of high-poweredlasers or prolonged UV radiation may damage traceevidence contained in or found in the vicinity of finger-prints. Infrared (IR) spectroscopic techniques are attractivebecause they can be used to identify the intrinsic chemicalcomposition of materials from spectra produced by theradiation interaction with the substances. IR microspectro-scopy has been used to identify the chemical compositionof fingerprints by examining multiple, small sections offingerprints [4]. These techniques did not employ imagingcapabilities and consequently have not been used in thedetection or characterization of entire fingerprints. Our ownwork of developing this concept began in 2000 and hasbeen followed by recent other reports that further thisconcept of obtaining images of fingerprints. Fourier-transform IR (FT-IR) spectroscopic imaging technologyhas now matured to the level that allows the imaging andanalysis of many materials to be accomplished [5, 6]. Whilethe technology allows facile imaging, it is not portable and,in contrast to handheld laser systems, cannot yet betransported to scenes of forensic interest.

IR spectroscopy has, historically, been widely used tocharacterize a large variety of trace evidence such as singlefibers, paint chips, drugs, and explosives [7–9]. A broadknowledge base exists of comparing and relating observedspectra of unknown materials of forensic interest to knownmaterials. This process often involves the removal of traceevidence from the crime scene and is, thus, invasive anddoes not preclude potential damage to latent print evidence.The recent introduction of imaging capabilities to infraredspectroscopy allows simultaneous and rapid acquisition ofspatially localized spectral data over millimeter to centime-ter sized fields of view [10]. The chemical specificity ofinfrared spectroscopy is retained in the imaging modalitywhile high throughput acquisition capabilities and imaging

visualizations are added. FT-IR spectroscopic imaging hasbeen conducted successfully on latent fingerprints that havebeen fumed with cyanoacrylate [11, 12]. While this methodhas been successful for enhancing the development of latentprints, it is invasive and contaminates trace evidencelocalized the in the fingerprint area rendering this materialunidentifiable. Attenuated total reflectance-infrared (ATR-IR) imaging has been used to identify evidence after a tapelift has been conducted [13]. However, this is also aninvasive approach and disturbs the deposited print forfurther analysis. ATR-IR imaging has also been used toconduct studies on latent fingerprints with time andtemperature studies [14] but have limited universal utilityas the print has to be in close contact with the crystal. Non-invasive FT-IR spectroscopy using the reflection absorption(R-A) mode for spectroscopic imaging of latent fingerprintswithout simultaneously acquiring data on trace materialshas been demonstrated in our laboratories on a wide rangeof substrates [15, 16].

By combining this mode of data recording withemerging FT-IR spectroscopic imaging instrumentation,we present a non-invasive, non-destructive approach todetect superimposed fingerprints and fingerprints withassociated trace evidence. This method has the potentialto obtain the molecular structure of constituent materialsand to provide results in an integrated form that is useful forphysical evidence identification. Some critical issues to beaddressed in the assessment of this new methodologyinclude the evaluation of the validity of the approach andthe derivable equivalence between the new approach andtraditional methods. The benefits imparted by the newmethod will likely justify its acceptance as a standard toolonce the required challenges are met. Latent fingerprintsfound at crime scenes consist of natural secretions plusextraneous deposits that may or may not be related to thecrime in any manner. This makes simultaneous processingof prints and recovery of trace evidence a challenging task.In this study, we present a spectroscopic imaging approachto non-invasively detect and develop latent prints whileobtaining information on the presence and identity of traceevidence contained in the vicinity of the print. We presenttwo examples, namely overlapped print resolution andextraction of multiple trace evidence within prints, wherethe benefits of the new approach are enumerated anddemonstrate how current forensic techniques can beenhanced by the introduction of spectroscopic imaging.

Experimental

Spectroscopic imaging data were acquired using a com-mercial FT-IR spectroscopic imaging system (Spotlight 300from PerkinElmer, Shelton, CT). This system allows

2070 R. Bhargava et al.

acquisition of spectral data from large areas at two differentprojected pixel sizes, namely, 6.25 µm×6.25 µm and25 µm×25 µm. The instrument is equipped with a lineararray of 16 Mercury–Cadmium-Telluride (MCT) detectorsand a precision stage that allows the sample to besequentially moved to allow data acquisition from largeareas. Spectra were acquired in the rapid scan mode at amirror velocity of 1 cm/s and at a spectral resolution of16 cm−1 resulting in a data acquisition time of approxi-mately 0.5 s for 16 simultaneously acquired spectra. Thetime required for imaging a larger area is simply scaled bythe numbers of pixels. Reflective, aluminum-coated glasssubstrates were used to image all samples. While reflectiveglass is not typically found in a forensic scenario, itsimulates a condition of a reflective substrate such as adoorknob, knife blade, or handle. The IR microscope wasoperated in the reflection mode, thus employing a R-Asampling method. Custom-written software implemented inIDL/ENVI software (ITT visualization systems, Boulder,CO) was used to process and display images from the fullspectral data set.

We demonstrate three cases commonly encountered inforensic investigations. In the first, we examine latent printsformed using unwashed hands on a reflective glass slide.The second case is one of overlapped prints, where eachwas made under different hand washing conditions. In thethird case, a forensic scientist handled RDX powder andsubsequently touched a reflective slide, thereby making alatent print and transferring particles on hand to the slide.Undeveloped prints were visualized with side lighting anddigitally photographed. The prints were processed using the“Microburst” cyanoacrylate fuming technique used by theFBI [1]. Cyanoacrylate was heated to 300°C in a fumingchamber prior to placing the glass substrate inside. Oncethe cyanoacrylate was actively fuming, the substrate wasplaced inside the chamber for 30 s to reach optimumdevelopment. In the same manner as the undevelopedprints, the developed prints were digitally photographed.

Results and discussion

Figure 1 shows an image generated using the absorbance ofa vibrational mode corresponding largely to the distributionof oil-rich material based, for example, based on2,920 cm−1 (A) and protein-rich matter based on1,568 cm−1 (B). The typical spectra obtained in ourexperimental setup are shown for both types of matter inFig. 1c. It should be noted that the contrast in the infraredimages was generated by the spectral absorption resultingfrom the intrinsic chemical composition of the print and didnot involve the use of any chemicals that would interferewith subsequent analyses. As demonstrated by the respec-

tive vibrational mode absorbance distributions, the chemi-cal specificity allows us to determine that the oil-richfeatures typically form continuous patterns along thefriction ridges, while the protein-rich material is depositedas flakes along the ridge producing a speckled pattern. Theutilization of particular wavelengths, with the knowledge ofthe characteristic frequencies of the finger deposit compo-nents, permits the enhancement of friction ridge patterns. Itshould be noted that the recorded data are a convolution ofthe bulk absorption and microscopy induced opticalcoupling in the data. Hence, a simple correlation betweentransmission mode and the R-A mode is unlikely

Frequently, multiple sets of prints are encountered inclose proximity. For example, Fig. 2a shows a photograph(optical image) of a print overlaid by another print that wasdeveloped using cyanoacrylate fuming to enhance thecontrast of the image. Prior to fuming with cyanoacrylate,the prints are also visualized using infrared spectroscopicimaging by monitoring the absorbance of the CH stretchingvibrational mode near 2,940 cm−1 shown in Fig. 2b. Thefirst major aspect to note is that infrared imaging is capableof providing the same level of morphological informationon prints that is currently available with chemical develop-ment, followed by optical imaging. There is furtheradditional information that is not available in simple opticalimages. Differences observed in the absorbance of the CHstretching mode and other vibrational modes in the spectraindicated that the two prints have different chemicalcomposition. In particular, the print on the left likelycontains an excess of an oily substance (esters). Thiscondition results from rubbing skin on the forehead, next tothe nose or other oily portions of the body. The oil isproduced from sebaceous glands in the skin. Fingerprintdeposits coming strictly from eccrine excretions solely fromthe fingers and palms have a higher salt than oil content.The print to the right did not demonstrate any unique peakscompared to the print on the left showing only a variationin peak intensities. Hence, the carbonyl stretching modefrom the esters was used to isolate one print, as shown inFig. 2c, while completely suppressing the other, resulting ina print usable for identification purposes. By using a ratioof the carbonyl near 1,730 cm−1 to amide I stretchingvibrational mode absorbance near 1,640 cm−1 from thedeposit of dead skin particles, the second print can beisolated as in Fig. 2d, showing that it is partially obliteratedby the other print. A comparison of the binarized prints,shown in Fig. 2c and d, reveals that the prints are from thesame finger, but the spectral profiles indicate differences inratios of the chemical composition. It is noteworthy thatpart of the print on the right can be recovered in IR imagingbut not in optical imaging methods. The ability tomathematically extract prints using the linear spectralunmixing approach by selecting specific features is unique

Superimposed latent fingerprints and inter-ridge trace evidence 2071

to a chemical imaging method and requires both spectraldata and its spatial distribution (imaging).

Latent prints containing trace evidence also can beexpected to contain oil and protein-rich regions as,observed in previous case. However, there are also likelyto be unique chemical signatures associated with the traceevidence spectrum. Just as the prints could be resolved,spectral subtraction can be used to eliminate the effects oflatent material on trace evidence. Unique spectral features

of the trace evidence can then be used to provide images ofthe distribution of trace evidence. The full spectrum fromtrace evidence can be obtained from pixels that aredominated by the material and compared to databases toidentify the material. We demonstrate the power of thisapproach in examining trace evidence of explosives withinlatent prints. A clear delineation of the three (twocomponents of fingerprints and the trace evidence) isshown in Fig. 3a using a composite red–green–blue

Fig. 2 a Cyanoacrylatedeveloped latent prints that aresuperimposed. b Infrared absor-bance distribution using the CHstretching mode at near2,940 cm−1. c The carbonylband at near 1,730 cm−1 can beused to distinguish the imagefrom the underlying substrateand provide a usable fingerprintfor identification. d The ratio ofthe carbonyl band at near1,730 cm−1 and the amide bandnear 1,640 cm−1 are used topredominantly visualize theprint on the right

Fig. 1 Images of a latentfingerprint developed by usingdifferent vibrational modes tohighlight different aspects of thechemical composition of thedeposited material. a Printimage developed by the absor-bance magnitude at 2,920 cm−1.and b print image developed byabsorbance magnitude at1,568 cm−1. c Example spectrafrom the oil-rich region (top,dark line) and flake rich regionare shown

2072 R. Bhargava et al.

(RGB) image constructed from characteristic vibrationalmodes in the three species. The oil-rich material is indicatedby blue being obtained at 2,920 cm−1, a C–H stretchingpeak. This mode is included in the other components, but ispredominantly from the oil. The skin flakes were imaged asgreen was obtained at 3,240 cm−1, the N–H stretching ofthe protein. The particulate matter of RDX was depicted inred and was obtained at 3072 cm−1. While the spatial imageis dominated by the oil-rich regions of the print, the greenand red regions are visibly distributed over the print. Sincethe image is digitally recorded, expanding a small spatialregion for enhanced clarity allows observations of thedistribution of red particles as shown in Fig. 3b. It is

noteworthy that the particulate matter lies entirely betweenthe ridges of the print. This behavior is found generallywith particulate matter on hands and we expect this giventhe topology of the skin surface. Only inter-ridge presence,however, establishes a high probability that the personforming the prints handled the material. Thus, ambiguity inaccidental co-localization of trace evidence or tamperingcan be ruled out. Characteristic spectra of the threesubstances are obtained and shown in 3C by the samecolors as the RGB image. The spectrum of the particulatematter was found to clearly show the peak near 3,072 cm−1

from the explosive, cyclotrimethylenetrinitramine, com-monly known as RDX (Research Department (composition)

Fig. 3 a RGB image of afingerprint obtained using oil,protein and particulate matter-specific absorbance modes. bExpanded view of the box in A.c Spectra of the componentsshowing the bands indicative ofthe specific components. Asingle fiber can be seen in thecenter of the print (green–bluecolor)

Fig. 4 A reference spectrum ofRDX explosive pointing to thespecific 3,072 cm−1 peak

Superimposed latent fingerprints and inter-ridge trace evidence 2073

X). A reference spectrum of RDX obtained at 4 cm−1

resolution is shown in Fig. 4. Note that in Fig. 3c there areadditional spectral contributions from the oils of the latentfingerprint. However, as mentioned above the 3072 cm−1

peak is a clear identifier of RDX. For more complicatedmaterials or for materials that have less distinct spectra, adatabase-matching approach may be utilized. Anotherinteresting result of the image is the visibility of a singlefiber in the middle of the print. Just as the RDX waschemically analyzed, the fiber may be spectroscopicallyanalyzed or extracted and subject to conventional forensicanalyses.

The sensitivity of conventional infrared spectroscopy inidentifying trace evidence depends on the differentiation ofthe spectrum of the material from the spectral response ofthe matrix and fingerprint as well as on the signal-to-noiseratio of its spectrum. Reflective substrates, similar to thosereported here, and inter-ridge evidence localization provideeasy and rapid elimination of fingerprint and matrixcontributions to spectra of trace evidence and the sensitivitydepends only on the quality of acquired data. Typicalsignal-to-noise ratio levels of 200:1 may be expected inreflective imaging configurations. Under these assumptions,a single pixel measuring 6.25 µm×6.25 µm must containapproximately 1 ng of material to be accurately identified.Higher sensitivities have been claimed with newer instru-mentation and the limits may be further decreased byproper sampling and large co-adding processes [17]. Whilethis sensitivity is highly comparable to the most sensitiveinfrared spectroscopic methods, the minimum concentrationlevels are greater than those required by some otheranalytical techniques. For example, gas chromatography/mass spectroscopy (GC/MS) is among the most sensitiveand widely used techniques to characterize trace evidence.The approach described in this paper offers severaladvantages to aid and enhance MS analyses. First, theinfrared spectrum of the evidentiary material providesadditional information about its identity and can beemployed to deduce probable molecular structures of thematerial. Second, the spatial localization information can beemployed to extract evidence selectively without the needto sweep the entire area, resulting in the extracted materialbeing less contaminated by other matter in the vicinity ofthe trace evidence. Hence, spectroscopic imaging has thepotential to lead to a more careful and thorough forensicexamination of trace evidence.

While fingerprints remain a primary and widelyemployed identification tool in forensic investigations, theiracquisition and correlation with trace evidence hasremained a two-step process. Employing the spectroscopicimaging approach proposed here provides an opportunity topreserve and obtain the classical identifier, the fingerprint,as well as any trace evidence. Subsequently, the identified

individual is linked directly with physical evidence. Thecombined evidence provides strength to a suspect’s in-volvement with a crime scene or materials identified astransferred from a victim to the suspect during a violentcrime. In addition to explosives, i.e., RDX, fibers, anddrugs have been identified from between the ridge lines oflatent fingerprints and developed by infrared spectralimaging [18, 19] in other studies.

Conclusion

While initial results are promising, further work on non-reflective substrates and field conditions is required toestablish this technique as a useful tool for routine use byforensic examiners. While conventional methods requiredevelopment time and disturbing the locale where printsexist, infrared imaging is non-destructive, non-invasive, andinherently incorporates digital recording and image en-hancement capabilities. The quality of evidence gatheringwith this approach produces fingerprints that are high-levelindividual characteristics. This quality can potentially leadto highly probable connections between the suspect andtrace evidence. For the first time, this study also linksindividual fingerprints with trace evidence, i.e., fibers orparticulate matter in an automated, non-invasive, andunambiguous form. These characteristics of the proposedapproach will facilitate remote, automated, and real-timecomparisons of latent fingerprints and trace evidenceagainst databases for identification, providing an enhancedlevel of support to law enforcement and anti-terrorismactivities.

References

1. Hollars ML, Schwartz RL, Trozzi TA (2001) Processing Guide forDeveloping Latent Prints. Federal Bureau of Investigation,Washington DC

2. Day JS, Edwards HGM, Dobrowski SA, Voice AM (2004)Spectrochim Acta 60A:563–568

3. Menzel ER (2006) In: Greene JR (ed) Encyclopedia of PoliceScience, 3rd edn. CRC, Boca Raton, FL, pp 526–530

4. Williams DK, Schwartz RL, Bartick EG (2004) Appl Spectrosc58:313–316

5. Levin IW, Bhargava R (2005) Annu Rev Phys Chem 56:429–4746. Bhargava R, Levin IW (eds) (2005) Spectrochemical Analyses

Using MultichannelInfrared Detectors, Blackwell Publishing, UK7. Bartick EG (2003) In: Chalmers JM, Griffiths PR (eds) Handbook

of Vibrational Spectroscopy, Vol. IV. Wiley, New York8. Suzuki EM (2009) In: Saferstein R (ed) Forensic Science

Handbook, Vol III, 2nd edn. Pearson/Prentice Hall, Upper SaddleRiver, NJ, pp 75–251

9. Bartick EG (2009) In: Saferstein R (ed) Forensic ScienceHandbook, Vol III, 2nd edn. Pearson/Prentice Hall, Upper SaddleRiver, NJ, pp 252–306

2074 R. Bhargava et al.

10. Lewis EN, Treado PJ, Reeder RC, Story GM, Dowrey AE,Marcott C, Levin IW (1995) Anal Chem 67:3377–3381

11. Tahtouh M, Kalman J, Roux C, Lennard C, Reedy BJ (2005) JForensic Sci 50:64–72

12. Tahtouh M, Despland P, Shimmon R, Kalman J, Reedy BJ (2007)J Forensic Sci 52:1089–96

13. Ricci C, ChanKLA, Kazarian SG (2007) Appl Spectrosc 60:1013–102114. Ricci C, Phiriyavityopas P, Curum N, Chan KLA, Kazarian SG

(2007) Appl Spectrosc 61:514–52215. Bartick EG, Schwartz RL, Bhargava R, Schaeberle M, Fernandez

DC Levin IW (2002) Baccino E (ed) Proceedings of the 16th

Meeting of the International Association of Forensic Sciences,Montpellier, France, September 2-7, 2002, Monduzzi Editore, IPDBologna, Italy pp 61-64

16. Crane NJ, Bartick EG, Perlman RS, Huffman S (2007) J ForensicSci 52:48–53

17. Park K, Lee J, Bhargava R, King WP (2008) Anal Chem 80(9):3221–3228

18. Chan KLA, Kazarian SG (2006) Analyst 131:12619. Bartick EG (2004) Presented At: The Scientific Working Group

for Trace Materials (SWGMAT), Quantico, VA, March 29-April 1,2004

Superimposed latent fingerprints and inter-ridge trace evidence 2075