Portable Raman Spectroscopy Systems for Field Analysis by Eckenrode, Bartick, Harvey, Vucelick, Wright, and Huff (Forensic Science Communications, October 2001)

October 2001 - Volume 4 - Number 4

Research and Technology

Portable Raman Spectroscopy Systems for Field Analysis

Brian A. Eckenrode
Research Chemist
Edward G. Bartick
Research Chemist
Forensic Science Research Unit
Federal Bureau of Investigation
Quantico, Virginia

Scott D. Harvey
Senior Research Scientist
Mark E. Vucelick
Student Intern
Bob W. Wright
Senior Staff Scientist
Radiological and Chemical Sciences Group
National Security Division
Pacific Northwest National Laboratory
Richland, Washington

Rebecca A. Huff
Chemist
Explosives Unit
Federal Bureau of Investigation
Washington, DC

Introduction | Raman Technique |
Raman Instrumentation With a Portability Emphasis
Laser Excitation Source and Wavelength | Fiber-Optic Probe
Optical Design and Collection Optics | Transducer
Data Processing and System Control | Commercial-Portable Raman Systems Applications and Practical Considerations | Conclusions
Appendix | References

Introduction

A major challenge confronting hazardous materials response teams involves the accurate and rapid identification of organic or inorganic chemicals outside the typical laboratory environment and under potentially dangerous conditions. Extreme care is required when analyzing unknown chemicals because of the possible instability of samples. Many chemicals can be sensitive to shock, heat, or light and could react violently including deflagration and/or explosion. In many cases, determining the nature of the unknown materials, whether biological (microorganisms) or chemical, is important and will dictate the handling, cleanup, disposal, and resolution of civilian issues. Addressing the need for rapid and accurate analysis of samples that are suspected to be dangerous has been difficult for most current commercial analytical instruments available to field agents, especially given the fact that analysis may prove lethal if handled incorrectly.

Raman spectroscopy is a rapid, nonintrusive, and nondestructive technique (in most cases) that can be used for the analysis of many classes of hazardous and potentially explosive compounds. With developments in fiber optics (Carrabba and Rauh 1992), filters (McCreery 1996), diode lasers (Imasaka and Ishibashi 1990), electromagnetic wave detectors (Sweedler et al. 1988), and data analysis software, Raman spectroscopy (Mulvaney and Keating 2000) has been able to move from the laboratory environment to the field (Smith 2000). Raman spectroscopy provides the ability to analyze bulk materials such as milligram (or kilogram) quantities of drugs present in bags or vials, as well as liquids consisting of solvent mixtures present in a variety of containers (Chase 1994). Raman spectroscopy is sensitive to slight differences in chemical structure and has been used to detect contaminants present within a medicinal tablet or gel cap. Raman spectroscopy can also easily and rapidly distinguish between the nitramine high explosives RDX (cyclo-1,3,5-trimethlyene-2,4,6-trinitramine) and HMX (cyclotetramethylene tetranitramine). Both laboratory and field-based Raman systems can provide reliable data with a minimal amount of sample preparation or manipulation and, as such, can be a powerful screening tool for the field analyst. New field-portable, Raman-based instrumentation will find important applicability in helping to ensure the safety of field responders in the FBI Laboratory’s Hazardous Materials Response Unit.

The first part of this article is intended to introduce the reader to Raman spectroscopy with particular emphasis on the hardware developments that have allowed this technique to be used more effectively in the field. This initial discussion is also intended to present the hardware and software issues that should be considered when evaluating the latest commercial-portable Raman systems. The latter part of this article addresses strengths and limitations that current commercial-portable Raman instruments have for field use by nonspectroscopists.

Raman Technique

Figure 1. Raman spectrum of acetominophen acquired from a Tylenol®-brand tablet placed on the sampling stage of a Chromex Sentinel portable Raman system. Click here to view an enlarged image.

The Raman spectroscopic technique involves directing a monochromatic light source (such as a laser) onto a sample and detecting the scattered light (Ferraro and Nakamoto 1994; Pelletier and Davis 1996). Most of the light is scattered elastically (Rayleigh scatter), but a small fraction is scattered inelastically (Raman scatter). Elastic scatter involves simultaneous emission of photons of the same energy as the incident photons. Rayleigh scattering is very intense relative to Raman scattering and corresponds to the frequency of the incident electromagnetic radiation (n^_ex). In contrast, Raman scatter is very weak and involves interaction of the incident electromagnetic radiation with the vibrational frequency of the molecule (n^_m) to produce a shift from the frequency of the excitation wavelength (n^_ex ± n ^_m). Specifically, Raman bands are generated by an induced oscillating dipole that is caused by interaction of the laser light with the electron cloud around the molecule. For a vibrational mode to be Raman-active, there must be a change in polarizability during the vibration. The vibrations involving symmetric stretches, multiple bonds, and vibrations of heavier atoms typically give rise to strong bands in the Raman spectrum, and therefore, Raman spectroscopy can be a powerful tool to characterize the structure and identity of molecules. An example of a typical Raman spectrum of acetominophen, an ASTM-recommended calibration standard, is shown in Figure 1.

Figure 2. A set of energy-level diagrams illustrating Raman scattering—Stokes (top) and Anti-Stokes (middle)—and the resulting Raman spectrum (bottom). If the chemical system is in thermal equilibrium, the equilibrium populations of the ground and excited states follow a Boltzmann distribution. Because the ground-state population is greater than that of the excited state, the Stokes lines are more intense than the Anti-Stokes lines (Ingle and Crouch 1988). Click for enlarged images of Stokes scattering, Anti-Stokes scattering, and the resulting Raman spectrum.

A more quantum mechanical viewpoint is illustrated in Figure 2. Incident radiation of energy hn_ex raises the vibrational state of the molecule from either the ground state (n = 0) or the first excited vibrational state (n = 1) to a virtual state. Emitted photons are either of the same frequency as the excitation wavelength (Rayleigh scatter) or are shifted by the vibrational frequency of the molecule (n_m). When the incident photon loses part of its energy to the molecule resulting in transition to a higher vibrational energy (Figure 2, top left diagram), the Raman-scattered radiation has a lower frequency than the incident radiation, and this is called Stokes scatter. The shift to higher frequencies is called Anti-Stokes scatter. Stokes-shifted Raman bands originate from molecules in their ground vibrational state, whereas Anti-Stokes bands originate from molecules in excited vibrational states. At equilibrium a Boltzmann distribution is assumed, therefore the greater population of vibrational ground state molecules results in an increased intensity of Stokes scatter compared to Anti-Stokes scatter. For this reason, Raman spectra are often plotted as a function of intensity versus the Stokes-shifted frequencies in wave numbers (cm^-1). Frequency shifts between the incident radiation, and the Raman-scattered radiation correspond to the vibrational energy levels of the molecule and can be used for chemical species identification. Although the positions of Raman bands are not dependent on the frequency of the incident radiation, their relative intensities are dependent because the intensities of Raman bands are proportional to the fourth power of the frequency of the incident light. In selecting incident wavelengths, the analyst must consider that longer wavelengths yield less intense signals, but with sensitive detectors, reduction in signal intensity may be offset by the minimization of fluorescence from either the sample, the containers, or from contaminants.

Because of different mechanisms, Raman spectroscopy is complementary to infrared (IR) spectroscopy, which depends on the change in dipole moment during the vibration, and thus offers many advantages. The following are among the analytical advantages of Raman spectroscopy:

• Little or no sample preparation is required.
  
• Raman is relatively unaffected by strong IR absorbers like water, CO₂, and glass (silica).
  
• No special accessories are needed for aqueous solutions because water is a weak scatterer.
  
• Fiber optics of varying lengths can be used to transmit the excitation laser light and collect the back-scattered light for remote analyses.
  
• The short wavelength excitation source can penetrate transparent and translucent container materials, and thus Raman measurements can be acquired through glass vials, envelopes, plastic bags, and several other (but not all) packaging materials.
  
• Properties of the laser sources make it relatively easy to probe micro-samples, surfaces, films, powders, solutions, gases, and many other sample types.
  
• When examining materials having inhomogeneous compositions, spatial resolutions greater than those obtained using infrared microscopy may be achieved because of the shorter wavelengths of light that are used for excitation.
  
• Raman systems are amenable to miniaturization for field applications.

Raman spectroscopy is a well-established analytical technique, and new portable systems are exploiting the many advantages for field applications.

Raman Instrumentation With a Portability Emphasis

Figure 3. Portable Raman systems typically align the laser to be coaxial with the collection axis. In this 180-degree back-scattering configuration, the laser and collection alignment can be maintained regardless of sample position or intervening optics (such as a window or cell wall). Click here to view an enlarged image.

There are several key instrument component issues that need to be adequately compared and contrasted when considering the various commercial-portable Raman analytical instruments. A block diagram of the typical portable Raman instrument hardware is shown in Figure 3. Each of the major functional components will be independently addressed in this communication for Raman systems in general. Although several of the portable Raman commercial systems share similar features, it is important to distinguish their capabilities to evaluate system strengths and limitations.

The major instrument design and device selection issues or components include:

• Laser excitation source and wavelength;
  
• Fiber-optic probe including working distance, filtering, and remote-shutter control;
  
• Optical design including the design of the collection optics and monochromator;
  
• Detection system;
  
• Size, power, weight, cost, and service or warranty;
  
• Signal-processing and data system including spectral database type or size; and
  
• Accessories for field work.

Laser Excitation Source and Wavelength

There are several factors that may influence the choice of the excitation source including power and wavelength. Because of these two important considerations and other sample-related issues (e.g., Raman scatter cross sections, concentration of the analyte, fluorescence, and the like), the excitation source must be carefully chosen. The best sources produce high-Raman intensities without causing photodecomposition, fluorescence, and/or absorption. In modern Raman instruments, lasers are always used as the preferred excitation source because of their high irradiance and monochromatic light. Most commercial-portable systems use small solid-state diode lasers as the excitation source. However, many have used 532-nm excitation (doubled Nd:YAG). Although shorter excitation wavelengths produce more intense Raman signals, short-wavelength excitation is also more likely to cause photodecomposition or fluorescence in some samples. Colored and especially black samples can absorb the incident radiation and/or the scattered radiation resulting in significant photodegradation of the sample and in some instances, the sample can be ignited. Most portable Raman systems use near-IR excitation lasers operating at 785 nm. Increasing the excitation wavelength towards the IR reduces the probability of sample fluorescence, but higher laser power is often needed to compensate for the decrease in Raman signal intensity. Additionally, as the excitation range moves farther into the red, the spectral range of charge-coupled devices (CCDs) as detectors is limited because of inability of these silicon-based devices to detect the bands shifted to longer wavelengths than ~1100 nm. It is best to use a laser wavelength long enough to reduce possible interference from fluorescence while providing the necessary power to maximize signal intensities.

Another major factor for portable Raman spectroscopy concerns laser stability in both power and frequency. Lasers used in portable Raman systems typically range in power from 50 to 500 mW. As laser power increases, the number of excitation photons increases for a given sample area, and the possibility of generating Raman photons is increased as well. Accurate and reproducible generation of a Raman signal requires the incident radiation to be stable. Wavelength instability of the excitation laser can change the Raman shift calibration. Additionally, diode lasers may “mode hop,” which contributes to the laser instability and could cause significant shifts of Raman spectra. In either case, field measurements may be compromised with a resulting loss in calibration validity, sensitivity, resolution, and identification ability.

Commercial-portable Raman system manufacturers have usually improved their laser excitation wavelength stability by one of two ways. One approach is to install an expensive laser that comes equipped with an external cavity feedback system designed for single-frequency stabilization with thermoelectric cooling. A portable system using this type of laser stability can produce laser linewidths of less than 1 cm^-1 with a 0.2 cm^-1 shift during one week. Emission lines from a computer screen can be used for calibration of this type of system. Alternatively, by providing a neon line internally and continuously for auto-calibration, shifts from a lower cost laser can be minimized. The raw data, however, requires more manipulation. This latter approach results in similar wave number shift stabilization during long periods of time. An additional advantage of this latter approach is that the overall cost of ownership is reduced because less expensive lasers can be used. Either approach has proven to be successful in terms of library matching and overall system performance.

Fiber-Optic Probe

Figure 4. A fiber-optic probe with a parallel fiber design for Raman spectroscopy (U.S. Patent 5,112,127). The length of this head assembly is only a few inches with a diameter of 0.5 inches. Click here to view an enlarged image.

Use of a rugged fiber-optic probe in portable Raman systems is very important, especially for field measurements of potentially shock-sensitive compounds. The probe allows scene responders to distance themselves from a potentially hazardous environment as they impinge the laser energy on and into the unknown chemicals. Fortunately, excitation wavelengths used in Raman spectroscopy fall in the visible and near-IR, and therefore, silica fibers are able to efficiently transmit laser light and collect scattered light from the sample. The downside to this involves the large amount of Raman scattering caused by the silica in the fiber. This interference slowed development of fiber-optic probes until the design of optical systems within the probe assembly itself. An example of a probe design is shown in Figure 4. This optical design optimizes light throughput while also removing the fiber background. This is a parallel fiber arrangement employing two fibers—one fiber carries the excitation energy, and the other collects back-scattered radiation from the sample.

Light-filtering materials in various fiber-optic probes are key to the efficient transfer of unpolarized light and efficient exclusion of both Rayleigh light and fiber emission lines. A notch filter within the probe is designed to eliminate the Raman spectrum of the probe materials and spurious lines emitted by the laser. The dichroic filter within the probe performs two important functions. It first allows transmission of the laser frequency, and secondly, it reflects Raman frequencies generated by the sample and directs back-scattered light through a long-pass filter that transmits only Stokes-scattered light. The entire probe head assembly described here is sized appropriately for hand-held use and can be customized for special applications. The length of the entire fiber-optic apparatus including the probe head assembly can be greater than 100 m if necessary. A typical length, however, is 5 meters.

An important parameter to consider when working with explosive compounds is the system irradiance or laser power per unit area delivered to the sample. Ideally, a probe system should produce the largest spot size possible that maintains the light collection efficiency. Some compounds, especially black and brown-colored samples, experience localized heating if the power density is too high. This can generally be observed in Raman spectra when the baseline rises at higher shifts because of blackbody radiation. Although several systems have fixed dimension spot sizes, when analyzing a wide range of sample types, it is important to purchase a system capable of using different spot sizes. Typical spot sizes range from 100 to 500 micrometers in diameter. When using a fixed spot size, it is advantageous to allow continuous control of the laser power in the event that a sample may have the potential for localized heating or ignition. For example, in the event that a sample is dark in color and/or the Raman spectrum is rising at high shifts, it may be advantageous to begin acquisition at a lower power and increase the time to collect the data (due to the lower scattered light levels). When using a probe designed to deliver a fixed power, it may be advantageous to use a larger spot size. The ability to control both the spot size and the laser power allows greater flexibility for sample analysis.

Figure 5. New tripod arrangement for field positioning of the fiber-optic probe assembly. Click here to view an enlarged image.

Other important capabilities of a fiber-optic probe assembly include the ability to provide a flexible positioning apparatus and to control the shutter from a remote location. A convenient positioning device for Raman measurements using the fiber-optic probe is shown in Figure 5. In many cases, it is not desirable to shine a powerful laser on an unknown sample for fear of detonation or ignition, and a remotely operated portable system becomes especially advantageous for personnel safety. New portable Raman systems are equipped with a data system controlled shutter that allows the laser light to impinge on a sample only when the user is ready to acquire spectra and the proper safety precautions are in place. Use of a fiber-optic probe also could allow use of Raman spectroscopy on remote-controlled robotic systems. The size of current portable Raman systems makes this type of remote analysis feasible.

Optical Design and Collection Optics

In many portable Raman systems a bundle of optical fibers is used to create a “slit image” for the dispersion element, which then reflects light onto the detector. Alternatively, a single-collection fiber (typically about 200 micrometers) can be used. Regardless, small spectrographs can be characterized as “slitless” systems with the collimation of the beam at the grating being determined by the internal diameter of the collection fiber or bundle rather than by the entrance slit of the monochromator. A high-quality wavelength selection device such as a diffraction grating is required in Raman spectroscopy to ensure that relatively weak Raman lines are separated from backscattered Rayleigh radiation transmitted by the probe, and it also must separate Raman transitions of similar energy to produce adequate detail in the spectral fingerprint. The main considerations in specifying a diffraction grating are (1) the wavelength region where maximun efficiency is required, (2) the number of grooves per millimeter, and (3) the size of the grating (which will determine the resolving power). Most portable Raman systems use grating spectrographs as the wavelength dispersion element. A diffraction grating is a planar or curved optical surface covered with straight, parallel, equally spaced grooves. The improvement of ruled gratings and the development of holographic gratings have enhanced performance of spectrographs. Light incident on the grooved face of a reflection grating is diffracted by the grooves; angular intensity distribution of the diffracted light depends on the wavelength, the angle of incidence of the incident light, and the spacing of the grooves. Concave gratings are frequently used in spectrographs because they combine functions of both a dispersing element and focusing optics, thereby reducing the number of reflective surfaces in the instrument.

The spectral resolution required for qualitative and quantitative analyses by means of IR or Raman is under debate. It is still believed that high-resolution data is required to analyze gaseous components. Several studies, however, have indicated otherwise (Jaakkola et al. 1997; Qin and Cadet 1997). For pattern recognition studies, researchers (Bangalore et al. 1999) have concluded that a nominal resolution of 16 cm^-1 provides sufficient selectivity. It is also suggested that if the instrument resolution is similar to the natural linewidth of the vibrational bands for a given analyte, then a further improvement in resolving power is unwarranted and may prove detrimental in terms of data collection time and data storage space. Another report (Kawai and Janni 2000) estimates that if resolution is better than 8 cm^-1, precise identifications can be expected for compounds in their library. Natural linewidths of most of the compounds in their database range from 3 to 100 cm^-1 with typical linewidths of less than 8 cm^-1 from crystalline powders.

The most accurate results for identifying unknowns by spectral searching and/or visual examination occur when the resolution of the Raman instrument matches that of the spectral library being used. However, test comparisons of simulated 6 cm^-1 spectra and 4 cm^-1 spectra generated on a bench-top Raman system yielded statistically identical results (Harvey and Wright 2000). When the library has been acquired at a higher resolution (lower wave number) than the spectrometer being used for analysis, the spectra can be deresolved to match that of the instrument. However, the searching capabilities and issues involved when using lower resolution instrumentation, specifically in light of new portable Raman instrumentation for field analysis, are being explored.

Collection optics for a given set of lenses can also be characterized by a parameter known to photographers and optical scientists as the f-number (f/#). The f/# of an optical element indicates the solid angle of light collected if the object is at the focal point. A trade-off occurs between the amount of light signal that is able to be collected and the depth of field within the sample. Optics with a small f/# in most field analysis applications collect light from a larger solid angle. The depth of field is smaller, however, and focusing becomes more important. In contrast, optics with a larger f/# collect light from a smaller solid angle with a resulting larger depth of field, and thus there is less reliance on an accurate focus. In practice, an intermediate f/# is selected, such as f/2, on most portable systems, so that measurements are not unduly affected by small distance variations of sample from the probe face due to different probe positions. The focal length relates the distance from the probe lens to the region within the sample defined by the depth of field. A focal length of 5–10 mm is common for most laser excitation probes. Many containers may have thick walls, and it is advantageous for the probe to have a longer focal length, such as 10 mm, to minimize the Raman signal originating from the container material.

Spectral range for portable Raman systems is typically between 200 and 2000 cm^-1 and is limited by issues such as detector-response range and spectrograph performance. In some systems, manufacturers provide the capability to extend the range. However, a physical change or movement of the grating spectrograph is required, which is less desirable for portable first-responder applications. Fortunately, many of the compounds in the databases generated thus far for hazardous materials have a majority of their Raman transitions between 300 and 1800 cm^-1. It is in this spectral region where the vast majority of fundamental transitions of organic and inorganic compounds occur. Compounds that have strong Raman transitions beyond the spectral range of the portable system will require careful characterization. For example, cyanides, which have medium-to-strong Raman transitions around 2220-2255 cm^-1, may be difficult to distinguish using a truncated range spectrometer. Although these compounds are not totally precluded from identification with a portable Raman system, their identification may be especially difficult in mixtures.

Another important consideration for portable systems is the number of moving parts. As the number of moving parts increases, the probability that calibration and alignment of optical elements will vary with shipping, vibration, or handling also increases. Ideally, the portable Raman system will have no moving parts and thereby simplify ruggedization and stabilize performance.

Transducer

Figure 6. The basic operation of an integrating CCD is animated at the Ocean Optics Website. Click here to view an enlarged image.

The transducer or detector of the light dispersed by the spectrograph in portable Raman systems is most commonly a CCD primarily because they are several orders of magnitude more sensitive than photodiode arrays alone. The CCD evolved as part of a unique signal-processing scheme, using silicon technology presently used for analog memory, digital memory, and other functions. CCDs contain devices that respond to radiation with the generation of electron-hole pairs in the depletion region of these devices. The basic CCD concept is to detect the charge generated in these photodiodes. By applying suitable step potentials to the CCD elements, it is possible to store or move a charge that has been optically introduced into the CCD device by means of a shift register system (Barbe 1980; Sweedler et al. 1988). CCD arrays can now be fabricated with as many as 10⁶ pixels, or picture elements, per square centimeter so that a CCD can also serve as an optical imaging device of high sensitivity, speed, and resolution. Many CCDs are fabricated from a monolithic silicon chip, and miniature spectrometers typically use a two-dimensional CCD array. The silicon-based CCDs have a wavelength dependence that will determine the excitation range of the lasers suitable for use as a source for dispersive Raman spectroscopy. The usable wavelength range is from 400 nm to about 1000 nm with a maximum response between 500 and 700 nm. Portable systems typically use 785-, 810-, or 852-nm excitation, which is not optimal for a CCD. However, there is a benefit from reduction of the fluorescence typically generated by using shorter wavelength excitation lasers. Another advantage of the CCD is that it is a full-spectrum detector because it simultaneously measures all wavelengths that impinge on it. An illustration of how a CCD functions is shown in Figure 6. The imaging capability offers speed in data acquisition because scanning is avoided. It is this dual capability of signal processing and imaging that has made the CCD such an important primary sensor, with applications in astronomy, infrared technology, and Raman spectroscopy.

A CCD is the transducer in many of the portable Raman systems available commercially, and their pixel area and temperature stabilization requirement can characterize the “detectors” for these portable systems. A typical CCD found in portable Raman systems has a pixel array size of 1024 x 128 with 2 micrometer pixels. To achieve a compact size with no moving parts and to maximize the spectral range, a greater Raman shift range is required on each pixel. In bench-top Raman systems, when physical space or moving parts are not an issue, the entire multidimensional CCD array can be more effectively used. For example, many bench-top Raman instruments have a data spacing or spectral element spacing that is less than 0.5 cm^-1. This implies that the spectral range is limited to ~500 cm^-1. However, because more of the CCD can be used, the spectral range can be wider because of the multidimensional advantage of the array. With portable instruments, the spectral element spacing is compressed to approximately 1.7 cm^-1/pixel. The effective spectral range, using only one dimension of the multidimensional CCD array, is therefore limited to ~1600 cm^-1 (200 cm^-1 to 1800 cm^-1). It was found that although portable systems have a higher pixel spacing relative to commercial bench-top Raman spectrometers, this was not a resolution-limiting factor and thus was not detrimental to the library search algorithms employed for identification purposes.

Additionally, most CCDs in portable Raman systems are thermoelectrically cooled and temperature stabilized at –45 degrees Celsius; however, it is not the value of the temperature, but the dark counts that should be considered when comparing instruments. This cooling lowers the dark signal and its associated shot noise. Typically, a detector that can cool to achieve a dark count of fewer than 3 electrons/pixel/sec or less is adequate for most portable Raman systems. The temperature stabilization to the low dark counts should occur within 5–10 minutes for rapid field deployment situations. Low-noise level measurements are important in the low-light conditions of Raman spectroscopy. When the capability for rapid signal averaging is combined with low-noise characteristics, the CCD is the transducer of choice for maximizing the signal-to-noise ratio (S/N) achievable by Raman spectroscopy.

Data Processing and System Control

Many commercial-portable Raman systems rely on a computer system (typically a laptop) to perform signal manipulation and to provide an interface for the user. The computer performs many functions such as digital filtering, smoothing, line-shape analysis, spectral matching by means of an integrated library, wavelength calibration, and instrument response correction. One primary function of the computer system is for data acquisition and signal averaging. Control of the signal integration time is important to balance acquisition speed with S/N enhancement. The S/N for Raman involves a measurement of the Raman-scattered signal (generally weak in nature) over and above other radiation signals. Several approaches to maximizing S/N ratio include selecting an alternate excitation source and enhancing the collection efficiency of the scattered signal. Additionally, by cooling the CCD detector, a reduction in the background or dark current is realized, and by using spectrographs with better stray-light rejection and filters to remove Rayleigh scatter, the S/N can be improved.

A major obstacle to enhanced S/N for portable Raman systems includes reduction in background signals from the environment and sample fluorescence. Several vendors of portable Raman systems have developed field accessories that cover the sample or enclose the container to be analyzed, which effectively limits the amount of background light from the environment and improves the S/N ratio. Unfortunately, reducing sample fluorescence is often one of the more difficult tasks in Raman spectroscopy. One method to minimize sample fluorescence involves selecting an alternate laser wavelength. For example, a laser source operating at 1064 nm will produce significantly less fluorescence (in many samples) than a laser operating at 514 nm. Although the fluorescence signal may be reduced or eliminated at longer wavelengths, a reduction in signal intensity is a result of the lower frequency used (because the Raman signal intensity is directly proportional to the fourth power of the frequency), thereby necessitating a higher source irradiance for increased Raman signal.

It is well known that silicon-based CCD detector response drops as the wavelength increases. Therefore, to compensate for this reduction, a white light spectrum is generated, and all of the data collected are divided by the spectrum of this white light source. This produces a normalized spectrum having relative intensities of Raman peaks that are not dependent on the instrument response function. This white light correction is performed on some systems automatically, whereas other systems allow for a manual correction. Janni (Janni and Kawai 2000) used a National Institute of Standards and Technology (NIST)-traceable source and compared it with their mathematically generated white light correction function. Their white light curve is used to ratio acquired sample Raman spectra and to produce standard data sets. To check their white light correction function, they analyzed a sample of cyclohexane and compared the sample spectrum (frequency and intensity ratios) to values previously published (Frost and McCreery 1998). The frequency and intensity values for cyclohexane are shown in Table 1.

Table 1. Cyclohexane Reference Standard Data Used for
Raman Instrument Response Correction

One important issue for field-portable Raman systems is the user interface. Often in a field situation, the user is not able to perform numerous and intricate software functions easily because of personal protective equipment (PPE) worn that limits vision and dexterity. An interface that provides big, soft buttons with transparent data analysis and library searching would be of great value. For example, a screen with selections such as CALIBRATE, DETECT, and CHECK LIBRARY would present an access or user level that allows a first responder to simply collect data. Although some vendors have developed their own user interface (Moleclue, EIC Laboratories, Norwood, Massachusetts; Sure-Cal, Chromex, Albuquerque, New Mexico), many rely on a software package called GRAMS (Galactic Industries, Salem, New Hampshire, www.galactic.com), which contains several data manipulation features described earlier. Although the software is powerful, it can involve many steps to perform a desired analysis. This can prove daunting to personnel not familiar with the software and requires more extensive training for primary responders. A modification in the user interface to make the software more simple and direct is needed. Furthermore, software control of library searches is also an area where further research and improvement is required. How the search algorithms handle spectral resolution, fluorescence, or other elevated baselines is critical for proper identification of an unknown substance in the field. If spectra are collected at a resolution of 4 cm^-1, will the search algorithm be able to accurately identify a compound in the library database that was collected at a significantly poorer resolution? Alternatively, if the library was generated at a resolution of 4 cm^-1, will the search algorithm be able to accurately identify a compound when the Raman spectra are collected at 18 cm^-1? It was determined that the best search algorithm provided by the GRAMS software involves using a single-sided peak matching (enabling shoulder detection), nonbaseline corrected, 8-bit data construction, 1600 spectral range data points, and the first-derivative least squares algorithm. Kawai and Janni (2000) made a similar determination.

Other areas of software improvement with the goal of making a user-friendly interface include developing:

• Libraries for Raman functional group analysis;
  
• An advanced library search algorithm especially for spectra that contain high levels of noise from fluorescence or thermal instability of the CCD;
• An improved differentiation of mixtures by means of ranking of preprogrammed spectral features or deconvolution methods; and
• An expansion of the current database for hazardous materials, chemical warfare agents/precursors, explosives, solvent mixtures, frequently encountered matrices, and the like.

There are several Raman databases or spectral libraries in the open literature. However, the FBI Hazardous Materials Response Unit requires a customized database. Organic and inorganic explosives including precursors, chemical warfare agents including precursors, pesticides, narcotics, solvents, and naturally occurring toxins are some of the types of compounds that need to be in a Raman database for the FBI Hazardous Materials Response Unit’s portable instrument. Two libraries have been independently assembled for the FBI Hazardous Materials Response Unit’s Raman projects and thus far total more than 450 compounds. One major issue that often arises pertains to the spectral resolution of the instrument that acquired the spectra(um) found in the library. Data collected at 6 cm^-1 resolution can be searched with similar results on an instrument having 4 cm^-1 resolution and vice versa. However, accuracy of search results will degrade for Raman instruments that can only acquire data at greater than 8 cm^-1 resolution (for example, 12 cm^-1 or 16 cm^-1) as mentioned previously. In this instance, a separate database should be collected or the questioned spectrum can be searched against a library based on mathematically deresolved spectra. Provided that the instrument line shape function (ILS) is known, high-resolution reference spectra can be convoluted with the ILS to yield the exact spectrum at lower resolution. However, testing is required to determine how low (poor) the resolution can be to produce accurate search results, specifically for hazardous chemicals, using both deresolved spectra and spectra acquired at the same resolution as the searched library.

Commercial-Portable Raman Systems

A spreadsheet showing the different commercially available Raman systems with an overview of the major standard features of each system is shown in Table 2. Included in this spreadsheet are portable Raman systems that have been commercially available for at least one year and weigh less than 30 pounds.

Figure 7. Person-portable Raman systems: InPhotote™, www.inphotonics.com; Sentinel (top right), Chromex, Inc., Albuquerque, New Mexico, www.chromexinc.com; Raman Systems R-2001 (bottom left), Ocean Optics, Inc. and Raman Systems, Inc., Dunedin, Florida, www.oceanoptics.com; Solution 633 (bottom right), Detection Limit, Inc., Laramie, Wyoming, www.dlimit.com (See Appendix for additional vendors.)
Figure 8. Example of a drug-data set collected on a portable Raman system. Click here to view an enlarged image.

Jordan (1998) has presented a brief overview of current commercial-portable Raman spectrometers. However, there have been significant recent improvements in instrument capabilities as discussed previously. A picture of each of the particular systems highlighted in Table 2 is shown in Figure 7A–D. Some Raman system suppliers have opted to target a given market and user such as law enforcement for drug analyses in the field. Representative data from a portable Raman instrument is shown in Figure 8. The Ocean Optics system is the least expensive and smallest implementation with their targeted users being university teaching and training facilities. The R-2000 can be used for a limited range of qualitative applications. However, quantitative performance suffers from laser fluctuations and spectral artifacts (Mann and Vickers 2000). The new R-2001 system has an improved approach to the cooling of their CCD, and recent results are encouraging (Battiste et al. in press). Ocean Optics systems have also been deployed for studying undersea spectroscopy problems. The Detection Limit system has a wide spectral range (although stepper motor controlled) and has been used primarily in process applications. It is clear that all of these portable Raman systems have overlapping capabilities, and each can be used in a variety of ways.

The Chromex, Inc. Sentinel system and the InPhotonics, Inc. InPhotote™ system have been compared by the Edgewood Chemical Biological Center for testing with Chemical Agent Identification Sets (CAIS) standards and actual recovered CAIS material (Christesen et al. 1999).

Applications and Practical Considerations

The FBI Hazardous Materials Response Unit’s field chemical analysis requirements are quite expansive. However, their primary concern at any accident or crime scene is for the safety of their response and support personnel. The nonintrusive nature of the Raman measurement provides a clear safety advantage when interrogating potentially hazardous chemicals. Raman spectroscopy can be used effectively to identify unknown container contents and maintain the integrity of the evidence without exposing personnel to the sample. In addition, this technique, for the most part, can provide a nondestructive analysis. Because Raman spectroscopy uses visible light or near-IR light and many containers efficiently transmit these wavelengths, a good rule of thumb is if you can see a container’s contents, Raman should be considered for the analysis.

Although Lewis et al. (1995) have addressed Raman analysis through different container materials, the issue was reexamined using Raman systems specifically designed for portable applications. Several different containers were examined to see if reliable Raman spectra could be generated through them. The best results were obtained by taking a background spectrum of the container at a position above the sample and subtracting this signal from the spectrum of the analyte of interest taken through the container wall. In these experiments, uncolored and amber glass vials showed no adverse effect on the Raman signal intensity or the measured bands. Polypropylene and polyethylene plastic containers (such as resealable plastic bags) were examined. It was found that only an increase in acquisition time was required because of attenuation of the laser and backscattered radiation. A typical acquisition time of 20 seconds was increased to 50 seconds for the plastic containers. It was also found that fluorescence and/or scattering from the containers was not detrimental to the acquisition of reproducible spectra. Raman spectra generated from solid surfaces were generally straightforward except when the background surface was dark-colored or porous as with concrete or soil. In these instances, the heating, fluorescence, or absorption of the sample interfered with the analysis by means of Raman spectroscopy.

Figure 9. Spectrum of Bis-(2-chloroethyl) sulfide (distilled mustard) in a clear vial. (The spectrum acquired through an amber vial is identical.) Click here to view an enlarged image.

Figure 10. Analysis of DIMP present in a resealable plastic bag. DIMP is a Sarin simulant and is an appropriate example because Sarin was deployed from plastic bags in the Aum Shinrikyo cult Tokyo subway terrorist attack in March 1995. These data were collected on a bench-top “transportable” Raman spectrometer. Click here to view an enlarged image.

Figure 11. Raman spectrometry can easily differentiate TNT (top) and C4 (bottom) in closed containers. Click here to view enlarged images of TNT and C4 graphs.

Figure 12. Raman spectrum of urea pellets contained in an unopened standard white business envelope (top) and in an unopened security envelope containing a patterned blue obscurant dye (bottom). Although only two prominent bands matched under these circumstances, the use of a custom database illustrates the capability of a Raman instrument to collect valuable chemical data without tampering with evidence. Clear here to view an enlarged image of the top and bottom spectra.

Examples of selected spectra obtained through different containers are shown next. A spectrum of distilled mustard (a blistering agent) in a clear glass container obtained with a bench-top “transportable” Raman spectrometer is shown in Figure 9. The spectrum of diisopropylmethyl phosphonate (DIMP) collected through a resealable plastic bag is shown in Figure 10.

Several experiments were performed to examine the different problems that hazardous materials responders may encounter in the field. The following list provides some practical findings as they relate to laboratory and field Raman system detection and identification:

• Background lights and sunlight proved problematic to the spectra collection, and a shroud enclosed within an additional cardboard box was necessary.
  
• Samples that are blue or dark-colored need to be approached with caution because they will tend to absorb the laser energy and convert it to heat that could lead to the ignition of, for example, gunpowder or flammable solvents on an asphalt surface.
  
• Spectra of solids in an open environment such as 2-(dimethylamino)ethane thiolhydrochloride, atrazine, and C4 (a mixture of RDX + polyisobutylene +
  di(2-ethylhexyl)sebacate + fuel oil) (see Figure 11), were easily obtained.
  
• Spectra from small quantities of the explosives HMX, RDX, and TNT were successfully obtained.
  
• Fiber-optic probe distance and orientation was not critical for liquids.
  
• Fiber-optic probe distance and orientation was critical for solids with the optimal placement of the focus close to the interior container wall.
  
• Spectra of ricin, cocaine hydrochloride and free base, and urea were obtained.
  
• A library match to the spectrum of urea was obtained for urea placed in two different types of envelopes (see Figure 12).
  
• Multicomponent solvents or formulations such as turpentine and Dursban (contains 6.7% chlorpyrifos in an inert filler) can be successfully differentiated and identified.
  
• Spectra from samples with fluorescent impurities such as anthracene and CEES (2-chloroethyl ethyl sulfide, a distilled mustard simulant) were also obtained successfully.

Frequency regions are used to identify functional groups just as they are in IR, and for hazardous materials the chemicals can be classified and frequency matched for identification purposes as illustrated in Table 3. More specific bonding types can be classified, with Raman being particularly useful for chemical substructures such as -C – S-, -S – S-, -C – C-, -N=N-, and —C=C-. In many cases Raman spectra are sharp and simple and thus provide advantages over IR in terms of frequency discrimination and compound identification.

Conclusions

Raman spectroscopy has been applied to organic systems since the 1930s and is often used in conjunction with IR spectroscopy to give qualitative and structural information. In many cases, Raman spectra are sharp and simple and thus provide advantages over IR in terms of frequency discrimination. Although there has been a great deal of progress in bringing Raman spectroscopy to the field, there remain several areas where improvements and/or further testing are warranted. The data system (including the software and library-searching capabilities) will need to improve both in terms of the user interface and the library-search algorithms. A group frequency searching and screening capability would be highly desirable. This capability would provide the responder with an assessment screen that can be used to make educated decisions in the event that the chemical or mixture in question is not in the library. Although not intended to be exhaustive, Table 3 lists a few selected compound classes and their characteristic frequency ranges that would be important. A major goal of our continuing research efforts is to build chemical libraries for the types of chemicals that the FBI Hazardous Materials Response Unit personnel encounter in the field (Harvey et al. in press). An expansion of the library to include frequently encountered mixtures will be a necessity for FBI Hazardous Materials Response Unit applications and requirements. New instrumentation for remote positioning of the fiber-optic probe with feedback optimization also will be valuable, as is the capability for remote-shutter operation. Control of the laser power by means of pulsing, beam chopping, or beam spreading may prove necessary as potentially ignitable samples are interrogated. A portable Raman system is needed that can discriminate between signal and background light so that shielding will not be necessary. Finally, although there have been different approaches by commercial vendors and national laboratories (Thinnes 2000) to make the portable Raman systems rugged enough for field use, issues such as shock, vibration, dirt and dust susceptibility, decontamination, and other environmental effects will have to be measured.

Table 3. A Selected List of Chemical Classes and Substructures Used to Assist in Identifying Unknown Chemicals in the Field

Raman spectroscopy has proven itself as a valuable technique for analyses in and out of the laboratory. Many of the factors that affect the performance of bench-top Raman spectrometers (Bowie et al. 2000a and 2000b) also will affect the performance of portable systems. The unique requirements of field analysis, however, places different requirements on the instrumentation. Several manufacturers of portable Raman systems are continuously improving their instrumentation from the laser and fiber-optic probe through to the transducer. The nonintrusive nature and remote-sensing capabilities of Raman spectroscopy are real advantages for the FBI Hazardous Materials Response Unit teams working in and around potentially hazardous chemicals (including shock-sensitive compounds), and a Raman system will be a valuable front-line screening tool for FBI field investigations.

Appendix

Additional suppliers of portable Raman systems or components.

• Renishaw plc, Schaumburg, Illinois, www.renishaw.com
  
• Control Development, Inc., South Bend, Indiana, www.controldevelopment.com
  
• Raman Systems, Inc., Boston, Massachusetts, www.ramansystems.com

References

Bangalore, A. S., Demirgian, J. C., Boparai, A. S., and Small, G. W. Effect of spectral resolution on pattern recognition analysis using passive Fourier transform infrared sensor data, Applied Spectroscopy (1999) 53:1382–1391.

Barbe, D. F., ed. Charge-Coupled Devices. Springer-Verlag, New York, 1980.

Battiste, D. R., Knudsen, R. D., Kreischer, B. E., McDaniel, M. P., Martin, J. L., Benham, E. A., and Womble, M. E. Online Raman analysis of ethylene and hexene in the Chevron Phillips 1-hexene and polyethylene processes, Applied Spectroscopy (in press).

Bowie, B. T., Chase, D. B., and Griffiths, P. R. Factors affecting the performance of bench-top Raman spectrometers: Part 1: Instrumental effects, Applied Spectroscopy (2000a) 54:164A–173A.

Bowie, B. T., Chase, D. B., and Griffiths, P. R. Factors affecting the performance of bench-top Raman spectrometers: Part 2: Effect of sample, Applied Spectroscopy (2000b) 54: 200A–207A.

Carrabba, M. M. and Rauh, R. D. Apparatus for measuring Raman spectra over optical fibers, U.S. Patent 5,112,127, 1992.

Chase, B. A new generation of Raman instrumentation, Applied Spectroscopy (1994) 48:14A–19A.

Christesen, S., MacIver, B., Procell, L., Sorrick, D., Carrabba, M., and Bello, J. Nonintrusive analysis of chemical agent identification sets using a portable fiber-optic Raman spectrometer, Applied Spectroscopy (1999) 53:850–855.

Ferraro, J. R. and Nakamoto, K. Introductory Raman Spectroscopy. Academic Press, New York, 1994.

Frost, K. J. and McCreery, R. L. Calibration of Raman spectrometer instrument response function with luminescence standards: An update, Applied Spectroscopy (1998) 52:1614–1618.

Harvey, S. D. and Wright B. W. Final report: Validation report for field-portable fiber-optic Raman spectrometer for detection and analysis of hazardous material in closed containers, Contract Number DE-AC06-76RLO 1830, February 2000.

Harvey, S. D., Vucelick, M. E., Lee, R. N., and Wright, B. W. Blind field test evaluation of Raman spectroscopy as a forensic tool, Forensic Science International (in press).

Imasaka, T. and Ishibashi, N. Diode lasers and practical trace analysis, Analytical Chemistry (1990) 62: 363A–364A, 367A, 369A–371A.

Ingle, J. D. and Crouch, S. R. Spectrochemical Analysis. Prentice-Hall, New Jersey, 1988.

Jaakkola, P., Tate, J. D., Paakkunainen, M., Kauppinen, J., and Saarinen, P. Instrumental resolution considerations for Fourier transform infrared gas-phase spectroscopy, Applied Spectroscopy (1997) 51:1159–1169.

Janni, J. A. and Kawai, N. T. Final report: Generation and expansion of a forensic Raman spectral database for laboratory and field applications, Contract Number JFBI98082, February 2000.

Jordan, J. R. Raman spectrometers: A field guide to instrumentation, Inside Laboratory Management (December 1998) 24–26.

Kawai, N. T. and Janni, J. A. Chemical identification with a portable Raman analyzer and forensic spectral database, Spectroscopy (2000) 15:32, 34, 36–41.

Lewis, I. R., Daniel, N. W., Chaffin, N. C., Griffiths, P. R., and Tungol, M. W. Raman spectroscopic studies of explosive materials: Towards a fieldable explosives detector, Spectrochimica Acta A (1995) 51A:1985–2000.

Mann, C. K. and Vickers, T. J. Chemical analysis with a low-resolution Raman spectrometer, Applied Spectroscopy (2000) 54:742–746.

McCreery, R. L. Analytical Raman spectroscopy: An emerging technology for practical applications, American Laboratory (1996) 28:34X, 34BB–34DD, 34FF, 34HH, 34JJ.

Mulvaney, S. P. and Keating, C. D. Raman spectroscopy, Analytical Chemistry (2000) 72:145–157.

Pelletier, M. and Davis, K. Raman spectroscopy: The next generation, American Laboratory (February 1996) 34C–34N.

Qin, D. and Cadet, G. Quantitative analysis of process streams by online FT-IR spectrometry, Analytical Chemistry (1997) 69:1942–1945.

Smith, J. P. Spectrometers get small: Miniature spectrometers rival benchtop instruments, Analytical Chemistry (2000) 72:653A.

Sweedler, J. V., Bilhorn, R. B., Epperson, P. M., Sims, G. R., and Denton, M. B. High-performance charge transfer device detectors, Analytical Chemistry (1988) 60:282A–284A, 286A, 288A, 290–291A.

Thinnes, G. Idaho National Engineering and Environmental Laboratory (INEEL), (2000) glt@inel.gov.