Philippe Lagueux

The need for protection against chemical attacks has been a military priority for decades. The same need has been added more recently on the list of authorities in charge of preventing and monitoring terrorist activities. The requirements for future remote chemical detectors are truly defining the state of the art in field instrumentation. In this article, we analyze these needs and show that passive imaging spectroradiometers are well suited to meet the requirements for remote chemical agent detectors.

Military personnel face a number of threats on the battlefield. Chemical and biological attacks are some of the most insidious threats since they are silent and hard to detect and can have devastating effects. Chemical and biological weapons have been referred to as "the poor man's atomic bomb" because of the low cost and relative ease of fabrication using materials and supplies that are used extensively for civilian purposes and thus easily accessible. Despite the Convention on the Prohibition of the Development, Production, Stockpiling and Use of Chemical Weapons and on their Destruction (1) signed by many nations in 1993, there is evidence that some countries are fabricating and stockpiling such weapons.  

Because of their ease of fabrication and use, chemical weapons might also be used by terrorist cells. The tools to detect chemical attacks are thus very relevant for homeland security. In fact, the threat of chemical attacks is more likely to come from nonmilitary-related toxic industrial chemicals (TIC) because of their availability and low cost. As a consequence, the requirements for modern chemical agent detectors now also include a long list of TICs.  

The ideal chemical agent sensor must detect, identify, and ultimately quantify several chemical species (chemical warfare agents [CWA] or TICs) in the areas interrogated. Because of the lethality of the target species, the ideal chemical sensor must operate remotely, up to 5 km from the threat and in near real time, to provide the alert in about 1 min. Chemical agent analysis requires identification capability, to select the course of action (for example, to identify what type of suit to don) and spatial localization capability, to determine where the threat is present. With the localization and meteorological information, one can predict when the threat may reach populated areas, what actions to take, and where not to send troops (contamination avoidance).  

Given this great need, various technologies are already in existence for the detection of CWA and TICs. The most developed and deployed solutions are point sensors (2) that operate using several techniques, including ion mobility spectrometry, conventional infrared absorption spectroscopy, colorimetric chemistry, and flame ionization. These sensors can be characterized as sniffers and operate on the principle of sampling the surrounding air to analyze its constituents, searching for chemical agents. While most of these techniques can be very sensitive and good at differentiating various chemical species, they lack the capability to detect chemicals remotely. This leads to deployment problems when large areas need to be protected and to difficulties of decontamination after an event has occurred. One can thus understand why the development of remote sensors has been supported for decades to complement point sensors.  

The most mature operational chemical remote sensor is the U.S. Army Joint Service Lightweight Standoff Chemical Agent Detector (JSLSCAD) (3). According to General Dynamics, "The JSLSCAD is a small, fully automatic, passive standoff chemical, nerve, blister, and blood agent vapor detector. The sensor is a portable sensor system capable of detecting chemical clouds from a fixed site (tripod) as well as 'on-the-move' mounted on a vehicle."  

The JSLSCAD sensor is a spectroradiometer operating in the thermal infrared region of the electromagnetic spectrum. This passive method of chemical analysis relies on the unbalance of emission and absorption of chemical species in the field of view of the instrument. If the gas is warmer than the temperature of the background, the spectral bands of the molecules in the gas appear in emission, while the same bands are seen in absorption when the gas is colder (4). This technique is analogous to the conventional laboratory infrared absorption spectroscopy except that in the case of passive spectroradiometry, the source of radiation is the background or the gas itself. The JSLSCAD sensor features a single field of view that is scanned 360° in the surrounding environment using an optical–mechanical scanner. When a gas is detected, the sensor applies verification algorithms based upon spectral information to identify the chemical species and confirm the presence of the threat.  

Although current chemical remote sensors have proven their value, there is still room for improvement given the critical need for detection. There exists a consensus that more or better spectral–spatial–temporal information would lead to better detection. In particular, it is believed that the gathering of several fields of view (that is, imaging) will improve the quality of the detection. Imaging is now included in the roadmaps of future generations of chemical remote sensors (5). In theory, a single-pixel sensor can be configured to do imaging, but because of the overhead of the dead time between spectra, the use of single-pixel sensors is quite inefficient at collecting high-resolution images. Imaging spectrometers with large focal plane arrays operated in the staring mode are approximately 10–100 times more efficient at this task.  

Requirements for Remote Chemical Agent Detectors

The requirements for chemical agent detectors usually focus on the list of agents, threshold concentrations, probability of detection, time for detection, and probability of false positives. See for example the specification for the "Commercial joint services lightweight standoff chemical agent detector" (6) issued by Joint Program Executive Office for Chemical and Biological Defense.  

System modeling is necessary to flow down these agent requirements to instrument requirements. This operation usually is performed by the provider of the solution and requires fairly complex radiative transfer models. The light reaching the sensor is governed by the emission and absorption of all gas layers in front of the target gas as well as the emission from the background behind the target (4).  

The emission of any material is given by its emissivity times the Planck function, while absorption in a gas is modeled by Beer's law (7). Both Planck's relation and Beer's law are nonlinear functions. It is possible to implement numerical algorithms (8,9) to solve this radiative system by coding the full nonlinear physical model and an iterative solving procedure.  

To simplify the calculations, it is also possible to linearize the problem. The brightness temperature difference between the spectrum measured in the presence of a gas and the spectrum measured in the absence of the gas can be expressed by equation 1, assuming the gas concentration is low, which is expected for detection close to threshold (10).  

Once this linear assumption is made it becomes possible to exploit the wealth of existing methods developed in the field of chemometrics and multivariate analysis. Several researchers have developed efficient methods (11,12) in order to retrieve simultaneously varying quantities from complex observables such as spectra data. The choice of a detection algorithm among the various methods is generally based on three key factors: the composition of the analyzed pixel, the type of model used to estimate the variability of the target and background spaces, and the model used to describe the mixed pixels. Several examples of retrieval can be found in the literature (10,13,14).  

The spectral resolution is an important instrumental parameter because it defines the capability of the instrument to resolve the various spectral features of the chemical species of interest. For gases around room temperature and pressure, the FWHM of individual gas lines is dominated by pressure broadening and is on the order of 0.1 cm

  (15). However, it is not necessary to resolve individual lines but rather to choose a spectral resolution in which most species can be distinguished from one another. Figure 1 shows the spectral signatures of nine gases at a spectral resolution of 4 cm

  (7.7–1.8 μm) spectral range. These gases are spectrally representative of CWA signatures. For example, it is clear that the TEP and DMMP signatures are quite similar around 1050 cm

. A certain spectral resolution must be utilized for the DMMP shoulder at 1075 cm

  to be resolved and usable to discriminate against the TEP measurements. Similarly, a certain spectral resolution must be utilized to correctly represent the sharp CH

  to distinguish this gas from the ozone, which otherwise displays a very similar spectrum.  

The motivation for utilizing a coarser spectral resolution is to improve the signal-to-noise ratio (S/N) of the measurement. Equation 2 describes how the noise equivalent spectral radiance (NESR) varies with spectral resolution Δσ. For an equal observation time 

, the NESR is inversely proportional to the spectral resolution Δσ. For example, the NESR at 4 cm

  should be four times higher than at 16 cm

. Consequently, choosing an unnecessarily fine spectral resolution adversely affects the sensitivity of the instrument.  

 where NEP(σ) is the noise equivalent power (W/ √Hz)

 AΩ is the instrument optical throughput (cm

The spectral resolution chosen for most passive FTS chemical agent detectors ranges from 4 cm

. The coarser spectral resolution is typically used for searching and detecting without definite identification. The finer spectral resolution is typically used for more detailed spectral analysis and identification of gases on interest.  

The spectral range of chemical agent detectors is chosen with three criteria in mind:  

To be a remote sensor and detect over long distances, the instrument must use a spectral range in which the atmosphere is transparent. The atmosphere absorbs in several regions of the electromagnetic spectrum, mainly because of water vapor and carbon dioxide molecules. The chemical agent detector must then operate in the so-called atmospheric windows where the absorption is low and the transmission is high. Figure 2 shows the conventional 3–5 μm MWIR window and the 8–12 μm LWIR window over 1-km and 5-km path lengths.

The spectral range obviously must coincide with the location of spectral features of the chemical species of interest. In analytical chemistry, the region from about 850 to 1300 cm

  (7.7–11.8 μm) is called the fingerprint region. The bands in this region originate from vibrational modes, resulting in a rich but unique absorption pattern. It is thus possible to differentiate between most compounds by measuring this region. Figure 1 illustrates this statement well: each gas is quite distinguishable from the others. Other spectra regions, including the 3–5 μm atmospheric window, feature less usable bands and are used much less often. 

The spectral range also must be chosen to maximize the S/N. Because the driver in a passive chemical measurement is the temperature contrast between the gas cloud and the background, it is more appropriate to express the noise of the sensor using the noise equivalent temperature difference (NeDT) instead of the noise equivalent spectral radiance (NESR). Assuming typical values of NESR = 2 nW/cm

  for a sensor operating in the MWIR (3–5 μm) and NESR = 10 nW/cm

  for a sensor operating in the LWIR (8–12 μm), we calculate the corresponding NeDT for a reference emitter at room temperature (300 K) as shown in Figure 3. The MWIR range rapidly degrades toward the shorter wavelengths and requires a larger temperature contrast to yield the same S/N. The temperature contrast varies greatly according to the environment and the viewing geometry. On sunny days, the objects are heated while the sky (viewed directly or through diffuse reflection) is very cold, yielding good contrast. On cloudy days or at night, the gradients are smaller. In normal environments the temperature gradients vary from 0.2 to 2 K. For such environments, Figure 3 demonstrates that the MWIR S/N would be insufficient except for a very narrow spectral range. The situation is even worse for other atmospheric windows at shorter wavelength than the MWIR. Atmospheric windows at wavelengths longer than the LWIR also exist, but good detectors for these ranges are not practical to use in the field.  

The published requirements for detection response time range from a fraction of a minute to 2 minutes (6). These specified times must be shared among all of the sensor views performed to cover the field of regard.  

Figure 4 shows the temporal evolution of a field release of SF6 (purple cloud) at a rate of 1.25 kg/min and demonstrates that the temporal evolution of a gas cloud is relatively slow on a field of regard shown (6.4° horizontal). The time interval between frames is 6.2 s. In the horizontal direction, the cloud is moving with the wind at a speed of approximately 5 m/s. The diffusion of the cloud in the vertical direction is approximately 1.3 m/s. It is interesting to note the presence of gas swirls indicative of turbulence.  

Although the clouds move relatively slowly, the movement is fast enough and the shape of the plume changes so that any raster scanning must be extremely rapid to adequately build an accurate representation (imaging) of the plume.  

The sensitivity is the cardinal requirement for any passive remote chemical agent detector and is usually represented by the NESR. The NESR, described in equation 2, represents the minimum radiance detectable by the sensor (that is, the radiance for which the S/N is equal to 1). In typical passive chemical agent sensing scenarios, an NESR of 5–10 nW/cm

  meets the detection and identification requirements for most chemical agents. This is also visible in Figure 3, considering that the typical temperature gradients are above 0.2 K.  

The clouds generated during chemical attacks are relatively dense and produce fairly large absorption/emission features. The concentration levels specified are integrated column densities in mg/m

Equation 2 shows that the sensitivity variation is inversely proportional to the square root of the observation time. Sensitivity comparison between various instruments must be performed for an equal field of regard (spatial range), for an equal instantaneous field of view (spatial resolution), for an equal spectral resolution, and for an equal observation time.  

Field of Regard and Instantaneous Field of View Requirements

Articles

In this article, the authors discuss the need for protection against chemical attacks and the role of passive imaging spectroradiometers in the detection of remote chemical agents.

Philippe Lagueux

Articles

Chemical Warfare Agent Spectral Imaging for Real-Time Identification and Localization