Noninvasive Detection of Concealed Liquid Explosives Using Raman

Anal. Chem. 2007, 79, 8185-8189

Noninvasive Detection of Concealed Liquid Explosives Using Raman Spectroscopy C. Eliasson, N. A. Macleod, and P. Matousek*

Central Laser Facility, Science and Technology Facilities Council, Rutherford Appleton Laboratory, Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0QX, UK

We present a Raman spectroscopic method for the noninvasive detection of liquid explosives within bottles, and other packaging, of substantially higher sensitivity and wider applicability than that currently available via conventional Raman spectroscopy. The approach uses a modification of the spatially offset Raman spectroscopy (SORS) concept, which permits the interrogation of a wide range of containers, including transparent, colored, and diffusely scattering plastic and glass beverage, medicine, and cosmetic bottles, with no change in experimental geometry. The enhanced sensitivity is achieved by the technique’s inherent ability to effectively suppress fluorescence and Raman contributions originating from the wall of the container. The application is demonstrated on the noninvasive detection of hydrogen peroxide solution, a critical component of a number of liquid explosives. In contrast to conventional Raman spectroscopy, the modified SORS concept enables the detection of concealed hydrogen peroxide solution in all the studied cases. The noninvasive detection of liquid explosives through containers or packaging represents a considerable analytical challenge, which has attracted particular attention in recent years due to the heightened terrorist threat.1 For example, liquid explosives were used in the failed bombing of the London transport system in July 2005 and in the foiled bombing of transatlantic flights in 2006. The difficulty in detecting these substances derives from the wide variety of packaging in which they can be readily concealed. The problem is compounded by the general insensitivity of the existing noninvasive security screening methods to liquid compounds. For example, extreme measures had to be adopted at UK airports during a recent heightened alert state; including, for example, requesting passengers to drink baby milk from their bottles in front of a security officer as a means of verifying the authenticity of the milk. Clearly an urgent need exists for an effective highthroughput screening method. Raman spectroscopy2 holds particularly great promise in this area due to its high chemical specificity, its experimental simplicity, and its ability to be deployed as a portable battery-powered device. Presently, the key problem * Corresponding author. Tel: +44-(0)1235-445377. Fax: +44-(0)1235-445693. E-mail: [email protected]. (1) Poulli, K. I.; Mousdis, G. A.; Georgiou, C. A. Anal. Bioanal. Chem. 2006, 386, 1571-1575. (2) Pelletier, M. J. Analytical Applications of Raman Spectroscopy; Blackwell Science: Oxford, 1999. 10.1021/ac071383n CCC: $37.00 Published on Web 09/20/2007

© 2007 American Chemical Society

preventing its use in this area stems from interfering fluorescence and Raman signals originating from the packaging overwhelming the subsurface Raman signals one aims to recover and analyze.3 Here we demonstrate an approach, based on an adaptation of the spatially offset Raman spectroscopy (SORS)4 concept, that facilitates a dramatic reduction of these interfering signals for a wide variety of packaging types, permitting a much wider applicability than possible before. The applicability of this method stretches well beyond this area and includes, for example, the authentication and quality control of products in the pharmaceutical and chemical industries. The SORS concept utilizes the diffuse component of light in analogy with NIR absorption tomography5 or fluorescence spectroscopy6 and is based on earlier investigations into the Raman photon migration process.7,8 The approach relies on the collection of Raman spectra from spatial regions offset from the point of illumination on the sample surface (see Figure 1). The laterally offset Raman spectra contain different relative contributions from sample layers located at different depths due to the wider lateral diffusion of photons emerging from greater depths.5,7,8 Consequently, for larger spatial offsets, a greater suppression of the surface layer signal (relative to the subsurface layers) is achieved. For a two-layer system such as beverage containers containing liquid, the residual surface signal contribution can potentially remain within the offset (subsurface) spectrum. This can be numerically removed using a scaled subtraction of the Raman spectrum obtained at the zero offset to produce the Raman signature of the liquid within the container. Importantly, the SORS approach is also capable of effectively suppressing interfering fluorescence if it originates from the surface layer of the probed medium. The basic SORS concept has been demonstrated in a number of applications including the Raman tomography of turbid media,9 in the noninvasive probing of bones,10,11 and in the (3) Matousek, P.; Parker, A. W. Appl. Spectrosc. 2006, 60, 1353-1357. (4) Matousek, P.; Clark, I. P.; Draper, E. R. C.; Morris, M. D.; Goodship, A. E.; Everall, N.; Towrie, M.; Finney, W. F.; Parker, A. W. Appl. Spectrosc. 2005, 59, 393-400. (5) Das, B. B.; Liu, F.; Alfano, R. R. Rep. Prog. Phys. 1997, 60, 227-292. (6) Pfefer, T. J.; Schomacker, K. T.; Ediger, M. N.; Nishioka, N. S. Appl. Opt. 2002, 41, 4712-4721. (7) Everall, N.; Hahn, T.; Matousek, P.; Parker, A. W.; Towrie, M. Appl. Spectrosc. 2001, 55, 1701-1708. (8) Everall, N.; Hahn, T.; Matousek, P.; Parker, A. W.; Towrie, M. Appl. Spectrosc. 2004, 58, 591-597. (9) Schulmerich, M. V.; Finney, W. F.; Fredricks, R. A.; Morris, M. D. Appl. Spectrosc. 2006, 60, 109-114. (10) Schulmerich, M. V.; Dooley, K. A.; Morris, M. D.; Vanasse, T. M.; Goldstein, S. A. J. Biomed. Opt. 2006, 11, 060502.

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007 8185

Figure 1. Schematic illustration of (A) conventional backscattering Raman geometry, (B) the standard SORS approach and (C) adapted SORS with transparent and (D) turbid samples. (E) Illustration of the overall experimental setup.

pharmaceutical field.12 Recently, a more sensitive variant, inverse SORS, was also developed10,13,14 and demonstrated in applications such as the security screening of powders in envelopes.13 A review of these techniques is presented in ref.15 In its standard form, SORS does not permit, automatically, the simultaneous interrogation of diffusely scattering and transparent samples. The proposed geometry satisfies this requirement by utilizing the insensitivity of the SORS scheme to the laser beam incidence angle at the sample. This insensitivity stems from the fact that the photon directionality is quickly “forgotten” as the photons undergo a massive number of scattering events within the diffusely scattering medium upon entering the turbid sample.5 This enables the angle of incidence to be set such that the laser beam, when presented with a transparent medium, passes through the container into the sample to the object plane of the Raman imaging system, which can be placed several millimeters below the sample surface within the probed liquid. Such an arrangement is shown to be effective with both transparent and turbid media. Another key distinction from conventional backscattering Raman spectroscopy is the ability to displace the laser beam incidence point on the sample surface to beyond the direct line of sight of the Raman collection system. Consequently, fluorescence or Raman signals emanating from transparent bottle materials, which can often overwhelm the weaker Raman signal of the contained liquid, can be strongly suppressed without compromis(11) Matousek, P.; Draper, E. R. C.; Goodship, A. E.; Clark, I. P.; Ronayne, K. L.; Parker, A. W. Appl. Spectrosc. 2006, 60, 758-763. (12) Eliasson, C.; Matousek, P. Anal. Chem. 2007, 79, 1696-1701. (13) Matousek, P. Appl. Spectrosc. 2006, 60, 1341-1347. (14) Schulmerich, M. V.; Morris, M. D.; Vanasse, T. M.; Goldstein, S. A. In Advanced Biomedical and Clinical Diagnostic Systems V; Vo-Dinh, T., Grundfest, W. S., Benaron, D. A., Cohn, G. E., Raghavachari, R., Eds.; Proceedings of SPIE 6430, SPIE: Bellingham, WA, 2007; p. 643009. (15) Matousek, P. Chem. Soc. Rev. 2007, 36, 1292-1304.

8186

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

ing the collection efficiency of the Raman system. This permits the sensitive interrogation of challenging samples such as green glass bottles, which can pose an insurmountable obstacle to conventional backscattering Raman spectroscopy.16 The past development of SORS focused solely on its use with highly scattering media. The study presented here demonstrates that the SORS concept is also applicable with mixed, stratified, samples such as those consisting of a diffusely scattering container with a transparent liquid. In this case, the mechanism of photon transport to the Raman detection region is different from that of standard SORS where photon diffusion dominates.15 Here, the container acts as a source of diffuse radiation illuminating a large area of liquid. Subsequently generated Raman photons emanate into all directions. The two mechanisms facilitate the effective illumination of the container wall (and the liquid content) directly under the Raman collection area with impacting Raman photons, which then pass through the wall via diffusion to the surface where they can be detected. The spatial offset ensures that the Raman and fluorescence contributions from the container wall do not overwhelm the weak Raman signal of the liquid. The effectiveness of this mechanism has not been demonstrated before. The concept also naturally works with the reverse situation where the container is transparent and the liquid diffusely scattering. In this case, the photon propagation mechanism is similar to that utilized in standard SORS, i.e., photon diffusion within the liquid medium. These characteristics make the proposed concept particularly effective and applicable to a wide variety of packaging and samples. The experiments presented here were performed using an 830nm laser with 250-mW average power. The Raman spectra were collected using a bundle of optical fibers and dispersed through a spectrograph onto a thermoelectrically cooled CCD camera. The full details of the experimental setup are given in Supporting Information. The acquisition times were 1 s in all measurements. The Raman signal was collected at two spatial offsets through two subsequent measurements, at zero offset and at 10 mm. The zero spatial offset Raman spectrum is used to remove the residual Raman spectrum of the surface (container wall), which can potentially contaminate the 10-mm spatially offset spectrum using a scaled subtraction approach to cancel this residual surface contribution. To demonstrate the potential of the deployment of this technique in the field by a nonspecialist, the scaled subtraction of the two offset spectra was performed “blind” in an automated way with no human intervention. The pure Raman spectrum of the liquid obtained can then be compared with a library data set containing known explosive constituents using standard procedures.17 The concept has been demonstrated by performing measurements on a range of common containers typically carried by air travel passengers in their carry-on luggage including transparent, colored, and opaque plastic containers with the original content intact and with the content substituted with a 30% aqueous solution of hydrogen peroxide (H2O2 (aq)). Hydrogen peroxide is a critical constituent of a number of liquid explosive mixtures and is readily available as a hair coloring developer to oxidize hair. It can be used, for example, to form the extremely unstable explosive (16) Nordon, A.; Mills, A.; Burn, R. T.; Cusick, F. M.; Littlejohn, D. Anal. Chim. Acta 2005, 548, 148-158. (17) Smith, E.; Dent, G. Modern Raman Spectroscopy. A Practical Approach; Wiley: Chichester, 2005.

Figure 2. Conventional (CR) and SORS Raman spectra of a white plastic jar with a thickness of 1.2 mm filled with H2O2 (aq). The bottom spectral trace is the Raman spectrum of the empty jar itself and is essentially identical to the conventional Raman spectrum of the jar containing H2O2 with no obvious trace of the Raman signature of H2O2 (indicated by the dotted line). The SORS spectrum, on the other hand, shows a close resemblance to the reference spectrum of H2O2 (aq) (top trace) with some residual spectral features originating from the jar. All spectra have been automatically background corrected.

material hexamethylene triperoxide diamine by combining 30% hydrogen peroxide with other chemical constituents.18 Hydrogen peroxide-based explosives were used in the failed terrorist bombing of London on 21 July 2005, when detonators failed to initiate the explosions. In this case, a mixture of hydrogen peroxide, chapatti flour, acetone, and acid was used.19 Hydrogen peroxide could be smuggled to the terrorist scene, e.g., on board an aircraft, either premixed with other chemical constituents or as a separate component to be mixed on board. In either case, for the explosive to be effective, a relatively concentrated solution of hydrogen peroxide is required18 such as that used in this study. RESULTS AND DISCUSSION The result of probing a small white plastic jar containing H2O2 solution noninvasively is shown in Figure 2. Such plastic jars are readily available in drug stores and are commonly used by travelers to transfer a smaller amount of moisturizing cream or other products for the purposes of travel to save space in carryon luggage or used as a pill bottle to store medication required during the flight. Figure 2 compares the performance of the conventional backscattering Raman method with the proposed SORS approach. The Raman spectra of individual components are (18) Becker, W. S.; Dale, W. M. Forensic Science Communications; http:// www.fbi.gov/hq/lab/fsc/backissu/oct2003/index.htm, October 2003. (19) http://news.bbc.co.uk/1/hi/uk/6261899.stm.

also shown for comparison. The container is highly diffusely scattering and presents an insurmountable challenge to conventional Raman spectroscopy; the signature Raman band of H2O2 at 876 cm-1 is completely swamped by Raman signals originating from the container wall. A conventional Raman instrument would therefore be unable to detect the presence of this compound in this container. In contrast, SORS, after the “blind” automated spectra processing using two spectra obtained at the zero and 10-mm spatial offsets (∆s in Figure 1c), dramatically suppresses the surface Raman signatures and clearly, and unequivocally, identifies the H2O2 marker band. Another example is illustrated in Figure 3, where a plastic sun lotion bottle (L’Oreal) is probed with its original content (Figure 3a) and with its content replaced with the hydrogen peroxide solution (Figure 3b). In both cases, the conventional Raman method is ineffective due to intense Raman bands originating from the container wall swamping the weaker Raman signatures of the content. In contrast, SORS effectively removes this overwhelming signal and permits the unhindered interrogation of the internal content. In both cases, SORS provides Raman spectra of the interior with sufficient clarity to permit unambiguous identification of the chemical composition of the internal content. Figure 4 shows the results of similar measurements on a, semitransparent, facial moisturizing cream jar (Olay). SORS was again able to conclusively identify the presence of hydrogen peroxide within the jar. Although the H2O2 marker band is visible within the conventional Raman spectrum, it would be difficult to unequivocally conclude whether this band originates from a compound, either genuine or illegitimate, within the container material or whether it derives from the wall of the container since only a single spectrum is available for analysis. The ability of the SORS approach to recognize and reject the surface Raman signature provides a powerful tool for noninvasive screening. From these measurements, it is estimated that the Raman signal could still be detected if the H2O2 solution was diluted by 1 order of magnitude, placing the detection limit on H2O2 concentration in aqueous solution at around a few percent. The Raman technique (in its basic form) is clearly not capable of detecting trace quantities of explosive materials; however, since concentrated solutions of H2O2 are essential for the effectiveness and viability of liquid explosives, this does not present a major obstacle for the deployment of the technique. Supporting Information shows the results of probing a number of other typical containers including transparent and nontransparent white or colored plastic bottles, as well as clear and colored glass bottles. These experiments further demonstrate the inherent ability of SORS to suppress the overwhelming fluorescence originating from some surface materials such as green glass bottles, which can otherwise swamp the much weaker Raman signal of the content. In all the studied cases, SORS either outperformed or equaled the conventional backscattering geometry in permitting the conclusive identification of hydrogen peroxide solutions in these containers. The Supporting Information data also emphasize one crucial point stemming from the proximity of the key hydrogen peroxide Raman band at 876 cm-1 to the dominant Raman band of ethanol at 882 cm-1, a constituent of many beverages brought on board aircraft. The overlapping Raman bands, which are separated only Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8187

Figure 3. Conventional (CR) and SORS Raman spectra of an L’Oreal sun lotion bottle, filled with (A) original content and (B) 30% H2O2 (aq), respectively. The bottle was made of white, opaque semisoft plastic material with a thickness of 1.1 mm with colored print (black, yellow, and blue). The dotted line indicates the signature band of H2O2 (aq) at 876 cm-1.

by a few wavenumbers, render any approach reliant solely on monitoring the position of the most intense Raman bands or the intensity of a selected spectral region prone to errors, particularly if a mixture of the two compounds is present and the H2O2 marker band may appear only as a poorly resolved shoulder on the Raman band of ethanol. In this case, it is preferable to obtain the Raman spectrum of the internal content across the entire spectral region to provide a definitive identification of the chemical constituency of the liquid (see Supporting Information Figures 7S and 8S). Such 8188 Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

Figure 4. Conventional (CR) and SORS Raman spectra of Olay facial moisturizer cream, filled with (A) its original content and (B) 30% H2O2 (aq), respectively. The bottle was made of semitransparent hard plastic material with a thickness of 6.5 mm. The dotted line indicates the signature band of H2O2 (aq) at 876 cm-1.

a spectrum is typically not available from conventional Raman spectroscopy but can be attained using the described method. CONCLUSIONS We have demonstrated an effective Raman spectroscopic approach for the detection of liquid explosives concealed in a wide range of plastic and glass containers. The technique, which has a substantially higher sensitivity than that available from conventional Raman spectroscopy, allows the effective suppression of both the fluorescence and Raman light emanating from the container wall. In contrast with conventional Raman spectroscopy, the SORS method was capable of identifying the presence of

concealed hydrogen peroxide in both transparent and nontransparent samples. Other uses include quality control and the authentication of food and other chemical products. ACKNOWLEDGMENT The authors thank Dr. Darren Andrews, Professor Anthony Parker, Dr. Tim Bestwick, and Professor Mike Dunne of the Science and Technology Facilities Council for their support of this work and the financial contribution of CLIK Knowledge Transfer, EPSRC (grant EP/D037662/1), NESTA, and Rainbow Seed Fund.

The Home Office Science Branch is thanked for useful discussions. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review June 30, 2007. Accepted August 7, 2007. AC071383N

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8189