Nanosensors for trace

explosive detection
Selective and sensitive detection of explosives is very important in
countering terrorist threats. Detecting trace explosives has become
a very complex and expensive endeavor because of a number of
factors, such as the wide variety of materials that can be used as
explosives, the lack of easily detectable signatures, the vast number
of avenues by which these weapons can be deployed, and the lack of
inexpensive sensors with high sensitivity and selectivity. High sensitivity
and selectivity, combined with the ability to lower the deployment cost
of sensors using mass production, is essential in winning the war on
explosives-based terrorism. Nanosensors have the potential to satisfy
all the requirements for an effective platform for the trace detection of
explosives.
Larry Senesac and Thomas G. Thundat*
Oak Ridge National Laboratory, Oak Ridge, TN 37831-6123, USA
*E-mail: thundattg@[Link]

We discuss the potential of nanosensors as a platform for trace and concealment and weapon delivery schemes. Trace detection of
explosive detection with an emphasis on nanomechanical sensors explosives typically involves collecting vapor or particulate samples
for achieving high sensitivity and selectivity. Nanomechanical and analyzing them with a sensitive sensor system. Currently, many
sensors, because of their versatility, can be used for receptor- different techniques exist for detecting trace explosives4–9. The most
based as well as receptor-free detection, and integrate well into a common include ion mobility spectrometry (IMS), mass spectrometry
multiplexed multimodal sensor system. (MS), and gas chromatography (GC) followed by detection using a
sensitive sensor. Most of these devices are, however, rather bulky,
The need for explosive trace detectors expensive, and require time-consuming procedures. Because of these
In recent years, explosive-based terrorism has grown enormously limitations, such systems are sparsely deployed at strategic locations
because explosive-based weapons are simple, easy to deploy, and can such as airports and government buildings. Further complications arise
cause enormous damage1–3. Detecting explosives is a challenging task when one considers not only airports, where there is a reasonably
because of a number of issues, such as the low vapor pressures of most controllable environment for sensing and detection, but also the
explosives, frequent introduction of novel explosive compositions, virtually uncontrollable entry points to public places, transportation

28 MARCH 2008 | VOLUME 11 | NUMBER 3 ISSN:1369 7021 © Elsevier Ltd 2008

 Nanosensors for trace explosive detection REVIEW

networks, infrastructures, and road networks with unpredictable positives. To facilitate continuous operation, the sensor should be
vehicular and pedestrian traffic. Protecting against explosive-based readily reversible at room temperature. These sensors should also have
terrorism can thus only be accomplished by mass deployment fast detection and regeneration times for efficient operation. Finally,
of miniature sensors that are sufficiently sensitive and selective, the sensors should have the capability for mass deployment because of
inexpensive, and amenable for mass production. the breadth of terrorist threats involving explosives. Currently available
For optimum sensitivity in explosive trace detection in air, the sensor platforms are unable to meet these stringent requirements.
proven strategy is: Sensors based on nanoscience, however, offer a clear path to the
1. Collect air and particulate samples from around or on the suspected development of trace explosive sensors that satisfy these criteria.
object;
2. Preconcentrate the explosive molecules on a sorbant material; Properties of explosives
3. Adsorb the released molecules from the preconcentrator on a signal Before we discuss trace explosive detection, it is important to
transduction sensor element for detection; and understand the physical and chemical properties of explosives.
4. Analyze, characterize, and report the data. Explosives are chemical compounds that can be initiated to undergo
Central to detection is chemical recognition using a selective self-propagating decomposition resulting in the sudden release of
agent and signal transduction. In an integrated system, the signal heat and pressure. Explosives are classified as low or high explosives
is then processed for reporting. The most important performance based on their burn rates. Low explosives that burn at low rates
characteristics of trace explosive sensors include high sensitivity, (centimeters per second) include propellants, black powder, etc. High
selectivity, reversibility, and real-time operation. Table 1 shows explosives, which detonate at velocities of kilometers per second, are
commonly used sensor performance characteristics. Trace explosive further divided into primary and secondary explosives based on their
detection requires extremely high sensitivity and low limits of stability. Primary explosives, such as lead azide, are extremely sensitive
detection (LODs) because of the relatively small number of molecules to external stimuli such as friction and thermal or electrical sparks for
that can be collected as a result of their very low vapor pressures. initiating explosion. Secondary explosives, such as 2,4,6-trinitrotoluene
High selectivity is essential in order to have an acceptable rate of false (TNT) and hexogen (RDX), are very stable and require primary
explosives to initiate explosion.
Table 1 Definition of performance parameters for sensors Many commonly used explosives are organic compounds and can be
Performance parameter Definition classified into six broad classes based on their chemistry10–13:
1. Aliphatic nitro compounds, such as nitromethane, hydrazine nitrate;
Sensitivity The slope of a calibration curve or change
in unit sensor response with change in unit 2. Nitroaromatic compounds, such as TNT, dinitrobenzene (DNB),
analyte concentration. hexanitrostilbene, picric acid;
Limit of detection (LOD) Lowest analyte concentration value that can 3. Nitramines or nitrosamines, such as octogen (HMX) or RDX;
be detected. 4. Nitrate esters, such as pentrite (PETN), ethylene glycol dinitrate
(EDGN), nitroglycerine, and nitroguanidine (NQ);
Resolution Smallest concentration variation that can
be detected when the concentration is 5. Acid salts, such as ammonium nitrate; and
continuously changed. 6. Organic peroxides, such as triacetone triperoxide (TATP) and
Dynamic range The analyte concentration from LOD to hexamethylene triperoxide diamine (HMTD).
maximum concentration that can be reliably Explosives from the last group, known as home-made explosives
detected.
(HMEs), have very high vapor pressures because of the presence
Selectivity The ability to detect a specific analyte in the of volatile organic compounds (VOCs) such as acetone. Recently,
presence of other interfering molecules. HME-based terrorism has grown rapidly because of the ease with which
Reversibility The ability of the sensor to return back to its they can be manufactured. However, HMEs are extremely unstable and
original value when the analyte is removed. require special handling to avoid detonation.
Response time The time required to respond from zero Most common explosives have extremely low vapor pressures at
analyte concentration to a step in the ambient temperature. Table 2 shows the vapor pressures of common
concentration. explosives8. The very low vapor pressures indicate that these molecules
Linearity The range where the sensor response is in are extremely sticky and tend to adsorb to surfaces very easily. It
direct proportion to the analyte concentration.
should be noted that the vapor pressures of explosives increase rapidly
Hysteresis The difference in sensor characteristics with temperature. Explosive vapors created by heating a sample
for increasing and decreasing analyte condense rapidly on cooler surfaces. Adsorption will be higher on
concentrations.
surfaces with a high surface energy such as metals, metal oxides, etc.

MARCH 2008 | VOLUME 11 | NUMBER 3 29

REVIEW Nanosensors for trace explosive detection

Table 2 Vapor pressures and molecular weights of some the sorbent material is designed to have a high partition coefficient
explosives for explosives, all the explosive molecules collected over a period of
time can be desorbed in 40 ms for detection by a sensing unit. The
Explosive Molecular Vapor pressure
weight at 20oC (Torr) NRL preconcentrator is based on a suspended array of micro hot plates
(g/mol) designed using complementary metal-oxide-semiconductor (CMOS)
Ethylene glycol dinitrate (EGDN) 152.1 5.2 x 10–2 technology and coated with a thin polymer layer that selectively

2,4,6-trinitrotoluene (TNT) 227.1 4.8 x 10–6

adsorbs the explosive molecules.

Pentaerythritol tetranitrate (PETN) 316.1 6.2 x 10–8 Signal transduction using nanosensors
2,4,6-trinitrophenol (picric acid) 229.1 3.1 x 10–8 Nanoscale effects provide unprecedented opportunities for developing
ideal trace explosive sensors that satisfy all the requirements outlined
Tetranitro-triazacyclohexane (RDX) 222.3 8.3 x 10–10
above. Here we define ‘nano’ in a more general sense, where reduction
Tetranitro-N-methylamine (Tetryl) 287.1 3.7 x 10–10 in size results in effects that cannot be observed with macroscale
materials or provides the ability to achieve disproportionately
compared with low surface energy materials such as polymers, plastics, higher sensitivity18–22. Nanoscale dimensions can, in certain cases,
etc. The sticky nature of explosive molecules on surfaces at room enhance physical and chemical properties, and this enhancement
temperature also causes condensation of the molecules in delivery can originate from various sources such as increased surface area,
lines into sensor systems. Therefore, trace sampling of these families confinement effects, etc. Recent advances in top-down fabrication
of explosives with very low vapor pressures is a challenging problem techniques have spurred both micro- and nanotechnologies. Micro and
because of the small number of molecules in a sampling volume. nanoelectromechanical systems (MEMS and NEMS) are fabricated using
similar top-down processes. As fabrication leads to smaller and smaller
Sample collection and preconcentration dimensions, better physical performance can be expected for certain
Sample collection is the front end of any integrated sensor system, types of signal transduction. For example, an increase in sensitivity
and for trace explosive detection it is the most challenging task. can be achieved by making the sensor dimension smaller. However,
Since vapor pressures of most explosives are extremely low, a large increasing the detection sensitivity alone cannot satisfy all the ideal
amount of air must be sampled in order to obtain enough molecules sensor requirements. The sensor also needs to have high chemical
of a particulate for detection by the sensor element. The higher the recognition capability. Despite their high sensitivity, many nanosensor
sensitivity of the sensor, the smaller the number of molecules that platforms suffer from poor selectivity. However, the development of
need to be collected for analysis. Often the sensitivity of sensors is materials with nanoscale effects that can provide chemical speciation
insufficient to detect trace explosive vapors in realtime and, therefore, is still in its infancy. An example is the use of polymer particles and
preconcentrators are essential for detecting explosives. Typically, a nanoparticles that, when bound to an explosive molecule, change one
large volume of air, which is a mixture of many molecules including of their measurable properties providing high specificity detection
very low concentrations of explosives, is collected using a pump, and (see section on receptor-free detection below). Sensor platforms
explosives and particulates are trapped using special materials. The that use nanoscale effects to enhance physical properties such as
trapping material is then heated to desorb the trapped molecules. The electrical, magnetic, mechanical, and optical properties, however, are
mechanism of trapping is based on the adsorption phenomena on abundant. By combining the high sensitivity offered by nanosensors
large-area surfaces. with the chemical selectivity provided by some receptors, it is possible
Conventional preconcentrators are bulky, require large amounts to develop miniature sensors that are highly selective, sensitive, and
of power, and have slow response times. Recent advances in reversible.
microfabrication have allowed the development of miniature In the nano arena, various bottom-up approaches have been
preconcentrators that can be heated rapidly because of their low developed using individual atoms and molecules to build useful
thermal mass. The microfabricated preconcentrators are coated with structures, such as nanoparticles and nanowires, and functional groups
high affinity materials that can withstand temperature cycling and for molecular recognition and self-assembly. These functional groups
can have potentially higher efficiency than their bulky predecessors are then combined with micro- and nanodevices to achieve sensors
in collecting explosive molecules14–17. Recently, researchers at the US with better performance characteristics, as well as signal transduction
Naval Research Laboratory (NRL) have developed a sorbent-coated to the real world. The combination of micro/nanoscale top-down
microfabricated preconcentrator device where the temperature of the and nanoscale bottom-up processes is paving a clear path to the
device can increase to 180°C in 40 ms17. The adsorbed molecules, development of nanosensors by combing a wide variety of old and new
including explosive, desorb very fast during this heat cycling. Since concepts.

30 MARCH 2008 | VOLUME 11 | NUMBER 3

 Nanosensors for trace explosive detection REVIEW

A large number of sensing elements/platforms exist that can There are also many polymers with high partition coefficients for
be considered nanosensor platforms. This includes micro- and explosives28. Molecularly imprinted polymers (MIPs) are another way
nanofabricated structures, quantum dots, nanowires, nanotubes, of obtaining chemical speciation. MIPs for explosives are specially
nanobelts, etc. These sensors are based on surfaces that undergo generated via the interaction of functional monomers, explosive
physical changes as a result of molecular adsorption. For example, molecule templates, and a cross-linking agent. The templates are
micro- and nanofabricated resonant devices for mass detection then removed prior to sensing. Only the molecules that match the
undergo a variation in their resonance frequency after the adsorption of template shape and properties can occupy the cavity, providing high
molecules. Nanosensors, however, suffer from some serious drawbacks. selectivity. MIPs that are very specific for TNT have been developed
First, receptor-based sensors often have poor selectivity23,24. Second, and demonstrated29,30.
since the sensors are extremely small physically, they do not have
a large enough cross section to collect molecules on a reasonable Electronic nose approach
time scale25. Sensors with large surface area are better for collecting Since most reversible receptors for explosives are not very selective,
explosive molecules than miniature ones. The enhanced sensitivity is it is possible to use them in an array format for creating a response
thus offset by a lack of sufficient molecules for rapid detection. Also, pattern similar to an artificial olfaction system31–33. A wide variety
nanosensors based on adsorption generally do not have a very large of highly sensitive physical sensor platforms are available for
dynamic range because of the limited number of adsorption sites in the developing such an electronic nose concept. The responses from the
sensor. array are analyzed with pattern recognition algorithms. However,
each receptor group in the array must provide a unique signal for
Receptor-based trace explosive detection the pattern recognition to work. At present, an array format with
We can categorize nanosensors as receptor-based or receptor-free. a pattern recognition algorithm can identify a single chemical with
Receptor-free nanosensors are based on detecting physical properties certainty. However, the confidence level of the pattern recognition
of explosives, e.g. thermodynamic, chemical, or optical properties. goes down with binary mixtures, and totally fails for mixtures of three
Receptor-based sensing achieves selectivity through the specific or more components. Simply increasing the size of the array will not
interaction between the receptor molecule and the explosive analyte. increase the selectivity. This is mostly because of the limited number
A change in a measurable property, such as mass, conductivity, or of interactions that can serve as the basis for a given interaction. It
absorbance, is then used to quantify the interaction. However, since is, therefore, doubtful that any method based on sensor arrays and
there are no reversible receptors that are specific for explosives yet, pattern recognition will achieve selectivity when relying on simple
an array of different reversible receptors is used in an electronic nose interactions for speciation of small molecules, especially explosives.
format for speciation. Concentrated efforts are presently underway to develop receptors with
Chemical selectivity in explosive detection based on receptors high selectivity.
originates from the chemical interaction between the explosive As described above, coatings that can identify explosive molecules
molecules and the receptor molecules. For reversible detection, with no interference do not exist at the present time. This would
explosive molecules must bind to the receptors with weak chemical require the use of partially selective coatings on arrays able to provide
bonds that can be broken at room temperature, for example, van der responses that are orthogonal or close to orthogonal (see section on
Waals interactions, hydrogen bonding, etc. Chemical selectivity based nanomechanical sensor platforms). For each analyte vapor (explosives
on weak interactions is very unselective and, therefore, should be used as well as interfering molecules), an ideal detector array would produce
in an array format such as an electronic nose. There are a number of a unique chemical response signature34,35. A catalog of these signatures
coatings that can be used for imparting partial selectivity, such as self- may be produced for a large number of analytes of interest and may
assembled monolayers (SAMs), polymers, metal oxides, single-stranded then be used for the identification of unknown vapors. For single
DNA, etc. The choice of coating has to take into account the response component analytes, this task is relatively straightforward, but for
time and the recovery time. mixtures of several different analytes a sophisticated computational
An example of a partially selective coating for explosives is a SAM algorithm is required.
of 4-mercaptobenzoic acid (4-MBA; also known as thiosalicylic acid)26.
SAMs of 4-MBA provide carboxyl end groups for acid-base reactions Nanowire platform
with explosive molecules. A 4-MBA SAM immobilized on Au-coated Nanostructures such as nanowires and nanocrystals offer a highly
nanosensor surfaces has a pKa in the range of 5–7 and binds strongly sensitive platform for the detection of molecular adsorption on their
with basic groups such as nitro-substituted molecules of explosive surfaces36–40. Nanowires such as carbon nanotubes have diameters
vapors. Recently, Zuo et al.27 reported the use of 6-mercaptonicotinic in the nanometer range. Inorganic nanowires and nanocrystals have
acid (6-MNA) monolayers to increase the sensitivity of TNT detection. unique electrical and optical properties that can be exploited for highly

MARCH 2008 | VOLUME 11 | NUMBER 3 31

REVIEW Nanosensors for trace explosive detection

Fig. 1 Diagram showing molecular-adsorption-induced bending of a cantilever beam arising from differential adsorption. Differential adsorption is obtained by
immobilizing a selective layer on one side of the cantilever. The adsorption decreases the free energy, which changes the surface stress.

sensitive molecular adsorption detection. The optical properties of Cantilevers are usually microfabricated from Si using conventional
nanocrystals can be modulated by molecular adsorption. Electrical photomasking and etching techniques. Typical dimensions of
conduction in a nanowire changes drastically through field effects a cantilever can vary from 200 µm to a few microns in length,
induced by molecular adsorption. Snow et al.41 have demonstrated an depending on the mode of operation. For miniature sensor systems,
interdigitated electrode capacitor modified with single-walled carbon detecting cantilever response using variation in piezoresistance is very
nanotubes (SWNTs) that can detect classes of chemical vapors with attractive57. In piezoresistive cantilevers, any change in deflection of
high sensitivity. When an electrical potential is applied, the huge the cantilever results in a variation of the cantilever resistance. The Si
electric field radiating from the SWNTs changes the polarization of piezoresistive cantilever fabrication technique developed by Boisen’s
the adsorbed molecules, which affects the measured capacitance. group58 produces cantilevers with high sensitivity, low electronic noise,
The change in capacitance depends on the class of molecule and can and very small drift. Fig. 2 shows a scanning electron micrograph of
provide partial selectivity. By coating the SWNTs with chemically a piezoresistive cantilever array. The piezoresistive readout technique
selective layers, however, it is possible to achieve even higher is miniature and compatible with electronic packaging. Fig. 3 shows
selectivity. Nanowires can be grown from many different materials the differential bending response of a piezoresistive cantilever beam
such as Si, ZnO, etc., thus providing different surface properties for coated with 4-MBA with respect to an uncoated cantilever when
chemical sensing and modification with functional groups42–45. exposed to TNT vapors. A reference cantilever (uncoated) makes it
possible to collect the response signal in common mode rejection.
Nanomechanical sensor platform
Molecular-adsorption-induced nanomechanical effects offer
unprecedented opportunities for trace explosive detection.
Nanomechanical sensors such as cantilever beams have many modes
of operation46–55. For example, the resonance frequency of a cantilever
beam varies with mass adsorption. The sensitivity of detection depends
on the resonance frequency of the cantilever. Cantilevers with very
high resonance frequencies – nanocantilevers – show very high
sensitivity to mass adsorption56. If molecular adsorption is restricted
to one side of the cantilever, the beam undergoes bending because
of adsorption forces. Cantilevers with very low spring constants
(low resonance frequency) show high sensitivity for bending mode
operation. Fig. 1 shows the bending mechanism of a cantilever sensor.
The force causing the cantilever to bend is independent of the adsorbed
Fig. 2 Scanning electron micrograph of a piezoresistive microcantilever array.
mass and depends solely on the binding energy of the adsorbate- The electrical resistance of the cantilever changes with bending. Cantilever
surface interaction. bending from submonolayer coverage of adsorbed molecules can be detected.

32 MARCH 2008 | VOLUME 11 | NUMBER 3

 Nanosensors for trace explosive detection REVIEW

for speciation. Fig. 4 shows the response from an array of cantilevers

coated with different partially selective coatings. The exposure of the
array to all analytes is conducted in the same manner as described
above for TNT. The weak bonding between the analytes and the
cantilever coatings makes it possible to desorb the analytes with a flow
of nitrogen, thus regenerating the detector. The array response patterns
can then be analyzed with pattern recognition algorithms to provide
speciation of the analytes.

Receptor-free trace explosive detection

A cantilever platform offers the possibility of detection methods based
on thermal processes because of its low thermal mass59–61. When a
microcantilever with adsorbed explosives is heated rapidly at a rate of
105 °C/s, the explosive undergoes thermally induced decomposition/
deflagration. Heating a cantilever at such a high rate is accomplished
Fig. 3 Differential bending response of a piezoresistive cantilever coated with
a 4-MBA functional layer with respect to an identical but uncoated cantilever by passing a current pulse through the cantilever (10 V, 10 ms).
when exposed to TNT vapors. The cantilever is exposed to the TNT vapor for During heating, the temperature of the cantilever increases from room
20 s, followed by a 60 s desorption period where it is flushed with dry nitrogen temperature to close to 500°C. Differential cantilever bending with and
only. As the TNT desorbs from the cantilever surface, the bending returns to the
original baseline value and the device is ready for another exposure. without adsorbed explosives shows a deflagration response of adsorbed
TNT (Fig. 5). A sensitivity of 40 pg is demonstrated for deflagration-
This makes it possible to reject environmental interference, such as based cantilever sensors. In this technique, the cantilever cleans itself
temperature fluctuations, that would otherwise show up in the signal. at the end of each heat cycle. The technique has the potential of being
The device is exposed to TNT for a short period followed by a flow of miniaturized.
dry nitrogen. This produces the response peak shown in Fig. 3 followed Another approach using the very high thermal sensitivity of
by a desorption period during which the adsorbed material is removed a cantilever beam is photothermal deflection spectroscopy62–65.
by the nitrogen, leaving the device ready to detect a new analyte. Bimaterial cantilevers have been demonstrated to have very high
Since the cantilever coatings are only partially selective, an array where sensitivity to changes in temperature. A bimaterial cantilever
each cantilever is coated with a different selective coating is needed undergoes bending because of the changes in temperature resulting

Fig. 4 Response patterns of an array of six cantilevers each coated with a different SAM (columns A-F) when exposed to vapors of TNT, ethanol, acetone, and
water. Each column shows the 80 s bending response (see Fig. 3) of one cantilever/coating (A-F) when exposed to each of the four analytes. The rows in the figure
are produced by placing all six responses to one analyte side by side to produce a response pattern that is unique to the analyte. These response patterns are then
analyzed with a pattern recognition algorithm for speciation.

MARCH 2008 | VOLUME 11 | NUMBER 3 33

REVIEW Nanosensors for trace explosive detection

Fig. 5 Deflagration response. As the cantilever exposed to an explosive is heated, its temperature first lags behind the temperature of the reference cantilever
(no explosives) because of the increased thermal load. Then as the explosive reaches the deflagration temperature, the exothermic process causes the temperature
of the loaded cantilever to increase and exceed that of the reference cantilever. As the explosive leaves the cantilever, the temperature returns to that of the
reference cantilever.

from, for example, adsorption of infrared (IR) energy by the cantilever absorption of IR energy, and is directly related to the vibrational
or adsorbed molecules. Resonant absorption of IR radiation by a modes of the adsorbates and to the heat capacity and thermal
monolayer of adsorbed explosive molecules results in bimaterial conductivity of the cantilever beam. Fig. 6 shows the mechanical
cantilever bending caused by the nonradiative de-excitation of the photothermal deflection spectrum of TNT adsorbed on a cantilever.
molecules. A mechanical IR spectrum of the adsorbed explosives can The observed photothermal spectrum corresponds to the excitation of
be obtained by sequentially exposing the cantilever to IR radiation various vibration modes of the explosive. The thermal energy is then
from a monochromator. The bending of the cantilever as a function transferred to the bimaterial substrate causing cantilever bending.
of illuminating wavelength reveals the IR absorption spectrum of The extent of bending is proportional to the released thermal energy.
the molecules adsorbed on the cantilever surface. This photothermal Almost all of the peaks observed in the photothermal spectra match
deflection signal depends on the temperature changes arising from very well with the conventional IR absorption spectra. For example, the

Fig. 6 Photothermal deflection spectroscopy of adsorbed TNT on a cantilever. When a bimaterial cantilever that is exposed to TNT is sequentially illuminated with
infrared (IR) radiation, the cantilever bends as the adsorbed TNT molecules absorb the energy. The plot of the bending as a function of the IR wavelength creates a
mechanical IR absorption spectrum of TNT.

34 MARCH 2008 | VOLUME 11 | NUMBER 3

 Nanosensors for trace explosive detection REVIEW

symmetric stretching vibrations of the C–NO2 bonds cause the peak energies (for example, Lewis donor-acceptor, Bronsted acid-base, and
at around 7.5 µm, while the 6.5 µm peak is caused by the asymmetric charge transfer interactions that have enthalpies (H) in the range
stretching vibration of the same bonds. However, the relative 100–400 kJ/mol). Although these binding mechanisms are irreversible
intensities of the peaks are slightly different for the photothermal at room temperature, sensor regeneration can be accomplished by
spectrum compared with conventional IR spectra because of surface heating the cantilever by passing an electrical current through it. The
effects. The extent of bending is directly proportional to the adsorbed temperature at which these analytes desorb also carries selectivity
explosives, the impinging power of the IR radiation, the absorption information. Selectivity and sensitivity can also be accomplished by
mode, and the thermal sensitivity of the cantilever. The disadvantage, combining different platforms into a single sensor unit for measuring
however, is the need to incorporate a tunable IR source. different chemical and physical properties. For example, the electrical
Another highly successful platform for trace explosive detection is polarization effects of adsorbed explosives can be measured as changes
based on amplifying fluorescence polymers (AFP)66–69. These polymer in capacitance of a SWNT sensor with mass and stress measurements
fluorescence characteristics change when TNT molecules are adsorbed made using cantilever arrays. It is also possible to combine some of
on the polymer. The AFP fluoresces when continuously excited by these ultrasensitive platforms with separation methods, such as gas
an ultraviolet light source. However, when a single molecule of TNT chromatography, to obtain chemical selectivity. However, separation
adsorbs on the polymer, the fluorescence is quenched. A chain of such methods can increase the detection time. Fast micro-GCs are under
conjugated polymer quenches is observed when TNT molecules adsorb development that, in combination with ultrasensitive sensors, can
anywhere on the molecular wire. The fluorescence can be detected detect explosives vapors.
with simple optical components. The AFP signal transduction is highly
selective and very sensitive for TNT vapor detection. A commercial Challenges in moving toward practical
version of this sensor is marketed by Noamdics, which uses thin films application
of AFP. Many variations of the fluorescence platform exist70–72. For Creating a sensor that is both sensitive and selective is only the first
example, nanocomposite films, such as silica doped with nitrogen- part of the challenge. The second part involves bringing explosives
containing macrocyclic molecules (porphyrins), show fluorescence from the environment to the sensor element. The practical application
quenching in response to explosive molecule adsorption73. of nanosensors for trace explosive detection will require integrating
Receptor-free detection of nitroaromatic explosives can be nanosensor elements into a system that includes vapor collection and
carried out using electrochemical techniques that make use of redox preconcentration.
properties74–76. However, these methods are useful only if the The integration requires many challenges to be overcome.
explosives can be first dissolved in a solution. Most explosives are Extremely low vapor pressures limit the number of explosive molecules
not very soluble in aqueous media. Pulsed voltammetry has been that can be collected in a reasonable detection time. Also, the
demonstrated for the detection of TNT in a marine environment. In this concentration of the explosives decreases exponentially as a function
method, nitroaromatic groups are first reduced into hydroxylamines of distance from the source. Very low concentrations of explosive
followed by conversion of the hydroxylamines into amines. In molecules can increase the false positive rate. Effective techniques for
electrochemical methods, the reduction potential with respect to a sample collection and preconcentration are, therefore, very important
reference electrode is unique and thus provides selectivity in detection. in developing practical sensor systems.
Variations of electrochemical techniques can be miniaturized. The The preconcentration step will increase the detection time.
technique is very simple and easy to use. Electrochemical detection Since explosive molecules stick to surfaces, delivery lines and other
has limitations, however, for example the explosives must be in an components need to be maintained at a higher temperature. Operating
electrolyte for detection. Electrochemical techniques also have limited at higher temperatures decreases sensitivity. However, a decrease
sensitivity. in sensitivity is preferable to sacrificing selectivity, especially when
using receptor-based signal transduction that depends on low-energy
Combined platforms molecular interactions. For example, water molecules that occur at
It is possible to achieve true chemical selectivity for explosive molecule orders of magnitude higher concentrations than explosives can severely
detection by combining different sensor platforms, or different restrict the selectivity when certain receptors are used. Techniques
operational modes of the same sensor platform, to obtain orthogonal based on arrays can provide partial selectivity. However, a multimodal
signals. For example, the mass, stress, and thermal signals of explosives array approach, where a combination of receptor-based and receptor-
measured with cantilevers are truly independent of one another and free techniques is used for detection, can lead to high selectivity
can provide orthogonal signals for pattern recognition. In addition, and sensitivity in practical explosive detection. Nanosensors offer
the ability to heat cantilevers to well above room temperature the freedom of multimodal detection, and therefore make an ideal
allows the use of coatings that bind explosive molecules with higher platform for trace explosive detection.

MARCH 2008 | VOLUME 11 | NUMBER 3 35

REVIEW Nanosensors for trace explosive detection

Conclusions as highly sensitive and very selective signal transduction platforms for
Trace explosive detection with high sensitivity and selectivity is a an integrated explosive sensor system.
formidable challenge because of a number of operational factors, such
as the acute dearth of explosive molecules that can be collected in a Acknowledgments
reasonable time, lack of selectivity because of interference from other
We would like to thank the Bureau of Alcohol, Tobacco, Firearms and
molecules, and the breadth of chemical compounds and combinations Explosives (ATF), the Federal Aviation Administration (FAA), the Transportation
that can be used as explosives. However, nanoscale effects can Security Administration (TSA), and the US Department of Homeland
Security (DHS) for financial support. We are indebted to Richard Lareau
be exploited to offer the possibility of sensors that satisfy all the and Eric Houser for educating us on trace explosive detection techniques.
requirements for trace explosive detection. Nanosensors with increased Experiments presented here were carried out with Dechang Yi, Charles
sensitivity and selectivity and the ability to operate in a multimodal VanNeste, Adam Krause, and Eric Finot. SEM images are provided by Zhiyu
Hu and Arnab Choudhury. Oak Ridge National Laboratory is managed
platform offer a potential paradigm for deploying a large number of by UT-Battelle, LLC for the US Department of Energy under contract
sensors for detection. We have shown that nanosensors have potential No. DE-AC05-00OR22725.

REFERENCES 38. Allen, B. L., et al., Adv. Mater. (2007) 19, 1439
1. Colton, R. J., and Russell, J. N., Science (2003) 299, 1324 39. Cui, Y., et al., In Nanowires and Nanobelts – Materials, Properties and Devices,
Wang, Z. L., (ed.), Kluwer Academic Publishers, Dordrecht, (2003), 3
2. Hallowell, S. F., Talanta (2001) 54, 447
40. Wang, Z. L., Materials Today (2004) 7 (6), 26
3. Fainberg, A., Science (1992) 299, 1531
41. Snow, E. S., et al., Science (2005) 307, 1942
4. Yinon, Y., (ed.), Counterterrorist Detection Techniques of Explosives, Elsevier,
Amsterdam, (2007) 42. Xie, C., et al., Anal. Chem. (2006) 78, 8339
5. Woodfin, R. L., (ed.), Trace Chemical Sensing of Explosives, John Wiley & Sons, 43. Forzani, E. S., et al., Small (2006) 2, 1283
Hoboken, New Jersey, (2007) 44. Kong, J., et al., Adv. Mater. (2001) 13, 1384
6. Steinfeld, J. I., and Wormhoudt, J., Annu. Rev. Phys. Chem. (1998) 49, 203 45. Star, A., et al., Adv. Mater. (2004) 16, 2049
7. Singh, S., J. Hazard. Mater. (2007) 144, 15 46. Thundat, T., et al., Appl. Phys. Lett. (1994) 64, 2894
8. Moore, D. S., Rev. Sci. Instrum. (2004) 75, 2499 47. Gimzewski, J. K., et al., Chem. Phys. Lett. (1994) 217, 589
9. Kolla, P., Angew. Chem. Int. Ed. (1997) 36, 800 48. Chen, G. Y., et al., J. Appl. Phys. (1995) 77, 3618
10. Furton, K. G., and Meyers, L. J., Talanta (2001) 54, 487 49. Pinnaduwage, L. A., et al., Rev. Sci. Instrum. (2004) 75, 4554
11. Federal Register (2002) 67 (81), 20864 50. Berger, R., et al., Science (1997) 276, 2021
12. Dionne, B. C., et al., J. Energetic Mater. (1986) 4, 447 51. Datskos, P. G., et al., Sens. Lett. (2003) 1, 25
13. Pella, P. A., J. Chem. Thermodyn. (1977) 9, 301 52. Lavrik, N. V., et al., Rev. Sci. Instrum. (2004) 75, 2229
14. Matzke, C. M., et al., Proc. SPIE (1998) 3511, 262 53. Feng, X. L., et al., Nano Lett. (2007) 7, 1953
15. Lu, C.-J., et al., Anal. Chem. (2003) 75, 1400 54. Ekinci, K. L., et al., Appl. Phys. Lett. (2004) 84, 4469
16. Potkay, J. A., et al., Proc. 16th IEEE MEMS (2003), 395 55. Husain, A., et al., Appl. Phys. Lett. (2003) 83, 1240
17. Voiculescu, I., et al., IEEE Sens. J. (2006) 6, 1094 56. Yang, Y. T., et al., Nano Lett. (2006) 6, 583
18. Pinnaduwage, L. A., et al., IEEE Sens. J. (2005) 5, 774 57. Boisen, A., et al., Ultramicroscopy (2000) 82, 11
19. Ekinci, K. L., et al., J. Appl. Phys. (2004) 95, 2682 58. Thaysen, J., et al., Sens. Actuators, A (2001) 92, 96
20. Ekinci, K. L., and Roukes, M. L., Rev. Sci. Instrum. (2005) 76, 061101 59. Nelson, I. C., et al., Proc. SPIE (2006) 6223, 62230
21. Ren, M. H., et al., Anal. Chem. (2005) 77, 2700 60. Pinnaduwage, L. A., et al., J. Appl. Phys. (2004) 95, 5871
22. Su, M., et al., J. Am. Chem. Soc. (2003) 125, 9930 61. Pinnaduwage, L. A., et al., Nature (2003) 425, 474
23. Walt, D. R., Anal. Chem. (2005) 77, 45A 62. Barnes, J. R., et al., Nature (1994) 372, 79
24. Hsieh, M. D., and Zellers, E. T., Anal. Chem. (2004) 76, 1885 63. Datskos, P. G., et al., J. Vac. Sci. Technol. B (2001) 19, 1173
25. Sheehan, P. E., and Whitman, L. J., Nano Lett. (2005) 5, 803 64. Datskos, P. G., et al., Sens. Actuators, B (2001) 76, 393
26. Pinnaduwage, L. A., et al., Appl. Phys. Lett. (2003) 83, 1471 65. Krause, A., et al., unpublished data
27. Zuo, G., et al., Nanotechnology (2007) 18, 255501 66. Zhou, Q., and Swager, T. M., J. Am. Chem. Soc. (1995) 117, 12593
28. Houser, E. J., et al., Talanta (2001) 54, 469 67. Yang, J. S., and Swager, T. M., J. Am. Chem. Soc. (1998) 21, 5321
29. McCluskey, A., et al., Org. Biomol. Chem. (2007) 5, 3233 68. Yang, J. S., and Swager, T. M., J. Am. Chem. Soc. (1998) 120, 11864
30. Walker, N. R., et al., Anal. Chim. Acta (2007) 593, 82 69. Toal, S. J., and Trogler, W. C., J. Mater. Chem. (2006) 16, 2871
31. Persaud, K., and Todd, G. H., Nature (1982) 299, 352 70. Toal, S. J., et al., J. Forensic Sci. (2007) 52, 79
32. Albert, K. J., et al., Chem. Rev. (2000) 100, 2595 71. Naddo, T., et al., J. Am. Chem. Soc. (2007) 129, 6978
33. James, D., et al., Mikrochim. Acta (2005) 149, 1 72. Tao, S., et al., J. Mater. Chem. (2007) 17, 2730
34. Hsieh, M. D., and Zellers, E. T., J. Occ. Environ. Hyg. (2004) 1, 149 73. Sohn, H., et al., J. Am. Chem. Soc. (2003) 125, 3821
35. Archibald, R., et al., Anal. Chim. Acta (2007) 584, 101 74. Lu, Q., et al., Anal. Chim. Acta. (2002) 469, 253
36. Patolsky, F., and Lieber, C. M., Materials Today (2005) 8 (4), 20 75. Wang, J., and Thongngamdee, S., Anal. Chim. Acta (2003) 485, 139
37. Grobert, N., Materials Today (2007) 10 (1–2), 28 76. Masunaga, K., et al., Sens. Actuators, B (2005) 198, 427

36 MARCH 2008 | VOLUME 11 | NUMBER 3