Wireless sensors and sensor networks for homeland security applications

Abstract

New sensor technologies for homeland security applications must meet the key requirements of sensitivity to detect agents below risk levels, selectivity to provide minimal false-alarm rates, and response speed to operate in high throughput environments, such as airports, sea ports, and other public places. Chemical detection using existing sensor systems is facing a major challenge of selectivity. In this review, we provide a brief summary of chemical threats of homeland security importance; focus in detail on modern concepts in chemical sensing; examine the origins of the most significant unmet needs in existing chemical sensors; and, analyze opportunities, specific requirements, and challenges for wireless chemical sensors and wireless sensor networks (WSNs). We further review a new approach for selective chemical sensing that involves the combination of a sensing material that has different response mechanisms to different species of interest, with a transducer that has a multi-variable signal-transduction ability. This new selective chemical-sensing approach was realized using an attractive ubiquitous platform of battery-free passive radio-frequency identification (RFID) tags adapted for chemical sensing. We illustrate the performance of RFID sensors developed in measurements of toxic industrial materials, humidity-independent detection of toxic vapors, and detection of chemical-agent simulants, explosives, and strong oxidizers.

1. Introduction

Detection technologies for homeland security applications are used in numerous scenarios ranging from airport screening to inspection of shipping containers, and to surveillance of public places. New detection technologies must meet the key requirements, which include sensitivity to detect target agents below risk or threat levels, selectivity to provide minimal false-alarm rates, and response speed to enable effective screening at high-throughput checkpoints [1]. Miniaturization of detection systems is also of critical importance, making possible measurements on-site in real time, rather than collecting samples for delayed laboratory analysis [2].

Chemical sensors are analytical devices where a sensing material is applied onto a suitable physical transducer to convert a change in a property of a sensing material into a readable form of energy [3]. The signal obtained from the transducer is further processed to provide useful information about the concentration of species in the sample. The energy-transduction principles that have been employed for chemical sensing involve radiant, electrical, mechanical and thermal types of energy [4]. Chemical sensors have their niche among modern microanalytical instruments because of several operational advantages, e.g.:

1. real-time determination of the concentrations of specific sample constituents;

2. little to no power consumption;

3. operation without consumables and frequent maintenance;

4. unobtrusive sensing; and,

5. deployment in multiple locations forming distributed sensor networks.

However, one drawback of chemical sensors is their relatively low response selectivity to analytes of interest in complex samples that contain high levels of interferences. Long-term stability of sensors is another general important unmet need in chemical sensors.

While the design of a sensor for a particular application will be dictated by its requirements, it is useful to set down the features for an ideal sensor for chemical species (e.g., low false-alarm rate, broad dynamic range, high sensitivity, high selectivity, high long-term stability, simplicity of maintenance, fast response, low initial and operation costs, response reversibility, small size, low power consumption, and self-calibration) [5].

In real-world applications, the qualities of an ideal sensor are weighted according to application. For homeland security applications, sensitivity, selectivity, and response speed are among the most important requirements [1,6,7]. Significant improvement in sensor sensitivity has been achieved using new transducer designs and sensing materials with large surface area. Sensing materials with large surface area also significantly increase the response speed [5,8]. However, high selectivity and long-term stability remain the most significant challenges for existing sensors and are yet to be achieved [5,9].

2. Discussion

In this review, we provide a brief summary of important chemical threats to homeland security, further focus on modern concepts in chemical sensing, and examine the origins of the most significant unmet needs in existing chemical sensors. We analyze the opportunities, specific requirements, and challenges for wireless chemical sensors and WSNs. We further provide details of a new approach for selective chemical sensing that relies on a sensing material with different response mechanisms to different species of interest applied onto a transducer that has a multi-variable signal transduction ability to detect their independent changes. This sensing approach was demonstrated utilizing battery-free passive radio-frequency identification (RFID) tags that were adapted for chemical sensing. We illustrate the performance of the RFID sensors developed in measurements of toxic industrial materials (TIMs), humidity-independent detection of toxic vapors, and detection of chemical agent simulants, explosives, and strong oxidizers.

2.1. Threats

Every day, nations face an existential threat from the intersection of terrorism and weapons of mass destruction. Chemical agents and TIMs are among those compounds that homeland security experts expect to be utilized in future terrorist attacks [6]. The following sub-sections summarize the characteristics of some chemical agents, TIMs, and existing and emerging explosive materials.

2.1.1. Chemical agents [10–13]

Toxic chemical substances intended to kill, seriously injure, or seriously incapacitate people through their physiological effects are known as chemical agents (CAs) or chemical-warfare agents (CWAs). Over 50 different chemicals in liquid, gas or solid form were used and stockpiled as CAs during the twentieth century. CAs can be organized into several categories (that can slightly vary in different literature sources) according to the manner in which they affect the human body. Nerve agents disrupt the mechanism by which nerves transfer messages to organs. Blister (vesicant) agents cause severe skin, eye and mucosal pain and irritation. Pulmonary (choking) agents attack lung tissue, primarily causing pulmonary edema. Blood agents prevent the body from utilizing oxygen. Representative examples of CAs are presented in Table 1 [10–13]. Other types of less lethal CAs include riot-control agents (e.g., pepper spray with capsaicin as an active ingredient, tear gas with ortho-chlorobenzylidene-malononitrile or chloroacetophenone as an active ingredient) and incapacitating agents (e.g., 3-quinuclidinyl benzilate, fentanyl-based Kolokol-1). The time-weighted average (TWA) exposure limits for CAs are ~10⁻⁷ –10⁻⁵ ppm for nerve agents, ~10⁻⁴ –10⁻¹ ppm for blister (vesicant) agents, ~10⁻¹ –10¹ ppm for pulmonary (choking) agents, and ~10⁻¹ –10¹ ppm for blood agents [13].

Table 1.

Categories of chemical agents [10–13]. The TWA exposure limits for different categories of chemical agents are ~10⁻⁷ –10⁻⁵ ppm for nerve agents, ~10⁻⁴ –10⁻¹ ppm for blister (vesicant) agents, ~10⁻¹ – 10¹ ppm for pulmonary (choking) agents, and ~10⁻¹ – 10¹ ppm for blood agents [13]

Nerve agents	Blister (Vesicant) agents	Pulmonary (choking) agents	Blood agents
GA - Tabun	HD - Sulfur Mustard	CG - Phosgene	CK - Cyanogen chloride
GB - Sarin	HN - Nitrogen Mustard	DP - Diphosgene	AC - Hydrogen cyanide
GD - Soman	L - Lewisite	Cl - Chlorine	SA – Arsine
GF - Cyclosarin	MD – Methyldichloroarsine	PS - Chloropicrin	KCN - Potassium cyanide
VX - Methylphosphonothioic acid	PD – Phenyldichloroarsine	DM - Adamsite	NaCN - Sodium cyanide
Novichok	ED - Ethyldichloroarsine	BCME - Bis(chloromethyl) ether
		NH3 – Anhydrous ammonia

2.1.2. Toxic industrial materials [10,14]

TIMs are industrial chemicals, other than CAs, that also have harmful effects on humans. TIMs are also often referred to as toxic industrial chemicals (TICs). They have a LCt₅₀ value (lethal concentration for 50% of the population multiplied by exposure time) less than 100 g-min/m³ in any mammalian species and are produced in quantities >30 ton/year at a given production facility [10]. Although they are not as lethal as the highly toxic CAs, their ability to make a significant impact on the population is related to the amount of TIMs that can be released during a terrorist attack.

TIMs are ranked in three categories with respect to their Hazard Index (HI) ranking, indicating their relative importance. A high HI indicates a widely produced, stored or transported TIM that has high toxicity and is air-borne. A medium HI indicates a TIM that may rank high in some categories but lower in others (e.g., number of producers, physical state, or toxicity). A low HI indicates that this TIM is unlikely to be a hazard unless specific operational factors indicate otherwise. Table 2 [10] summarizes TIMs by their HI.

Table 2.

Toxic industrial materials (TIMs) listed by Hazard Index (HI). The LCt₅₀ values of TIMs are < 100 g-min/m³ [10]

High	Medium	Low
Ammonia	Acetone cyanohydrin	Allyl isothiocyanate
Arsine	Acrolein	Arsenic trichloride
Boron trichloride	Acrylonitrile	Bromine
Boron trifluoride	Allyl alcohol	Bromine chloride
Carbon disulfide	Allylamine	Bromine pentafluoride
Chlorine	Allyl chlorocarbonate	Bromine trifluoride
Diborane	Boron tribromide	Carbonyl fluoride
Ethylene oxide	Carbon monoxide	Chlorine pentafluoride
Fluorine	Carbonyl sulfide	Chlorine trifluoride
Formaldehyde	Chloroacetone	Chloroacetaldehyde
Hydrogen bromide	Chloroacetonitrile	Chloroacetyl chloride
Hydrogen chloride	Chlorosulfonic acid	Crotonaldehyde
Hydrogen cyanide	Diketene	Cyanogen chloride
Hydrogen fluoride	1,2-Dimethylhydrazine	Dimethyl sulfate
Hydrogen sulfide	Ethylene dibromide	Diphenylmethane-4,4′-diisocyanate
Nitric acid, fuming	Hydrogen selenide	Ethyl chloroformate
Phosgene	Methanesulfonyl chloride	Ethyl chlorothioformate
Phosphorus trichloride	Methyl bromide	Ethyl phosphonothioic dichloride
Sulfur dioxide	Methyl chloroformate	Ethyl phosphonic dichloride
Sulfuric acid	Methyl chlorosilane	Ethyleneimine
Tungsten hexafluoride	Methyl hydrazine	Hexachlorocyclopentadiene
	Methyl isocyanate	Hydrogen iodide
	Methyl mercaptan	Iron pentacarbonyl
	Nitrogen dioxide	Isobutyl chloroformate
	Phosphine	Isopropyl chloroformate
	Phosphorus oxychloride	Isopropyl isocyanate
	Phosphorus pentafluoride	n-Butyl chloroformate
	Selenium hexafluoride	n-Butyl isocyanate
	Silicon tetrafluoride	Nitric oxide
	Stibine	n-Propyl chloroformate
	Sulfur trioxide	Parathion
	Sulfuryl chloride	Perchloromethyl mercaptan
	Sulfuryl fluoride	sec-Butyl chloroformate
	Tellurium hexafluoride	tert-Butyl isocyanate
	n-Octyl mercaptan	Tetraethyl lead
	Titanium tetrachloride	Tetraethyl pyrophosphate
	Trichloroacetyl chloride	Tetramethyl lead
	Trifluoroacetyl chloride	Toluene 2,4-diisocyanate
		Toluene 2,6-diisocyanate

Many TIMs are toxic-by-inhalation (TIH) gases. The top TIH gas of all highly hazardous chemical processes is ammonia, which is involved in over 30% of all highly hazardous chemical processes [14].

2.1.3. Explosives, taggants, and oxidizers [15–17]

Explosives are energetic materials, which are chemically unstable and can lead to an explosion, causing sudden expansion and decompose to produce a large amount of heat and change in pressure, along with loud noise and flash. Low explosives undergo deflagration at a rate of <3000 m/s, whereas high explosives undergo denotation at a rate of 3000–9000 m/s. Table 3 lists some examples of low and high explosives. Explosives are also categorized as commercial and military explosives. Commercial explosives (e.g., dynamite, slurry explosives and emulsion explosives) are mostly used for mining or blasting applications.

Table 3.

Explosives, taggants, and oxidizers and their reported vapor pressures [15–17]

Low explosives	Vapor pressure (@ 25°C)
Black powder	Not applicable
Black powder substitute	Not applicable
Smokeless powder	Not applicable
High explosives
2,4,6-trinitrotoluene (TNT)	9.9e−4 Pa (~10 ppb)
Pentaerythritol tetranitrate (PETN)	1.9e−6 Pa (~18 ppt)
Cyclotrimethylene trinitramine or RDX	6.3e−7 Pa (~6 ppt)
Cyclotetramethylene tetranitramine or HMX	Not available (~ sub ppt)
Taggants
2,3-dimethyl-2,3-dinitrobutane (DMNB)	0.28 Pa (~2.8 ppm)
Ethylene glycol dinitrate (EGDN)	6.4 Pa (63 ppm)
ortho-mononitrotoluene (o-MNT)	(~ppm)
para-mononitrotoluene (p-MNT)	(~ppm)
Oxidizers
Ammonium nitrate	0.0013 Pa (13 ppb)
Urea nitrate	8.8e−7 Pa (8.7 ppt)
Potassium chlorate	Not available (~ sub ppt)
Sodium chlorate	Not available (~ ppt)
Potassium perchlorate	Not available (~ sub ppt)
Potassium nitrate	Not available (~ ppt)
Ammonium perchlorate	Not available (~ sub ppt)
Potassium permanganate	Not available (~ sub ppt)

High explosives [e.g., trinitrotoluene (TNT), pentaerythritol tetranitrate (PETN), cyclotrimethylene trinitramine or RDX, and cyclotetramethylene tetranitramine or HMX] are named as military explosives because they were commonly used by military forces worldwide [15]. Plastic explosives (Composition C-4, Detasheet, and Semtex A/H) are also commonly used in the military. Plastic explosives are malleable and very stable, but they detonate to produce a large amount of energy. Such characteristics also make them ideal for use by terrorist groups, since the explosives can be easily shaped to conceal within luggage or a vehicle. The energetic materials in plastic explosives are RDX and/or PETN, which both have very low vapor pressure. Hence, plastic explosives legally manufactured after 1991 are required to be doped with taggants (volatile chemical markers) to allow successful detection by vapor detectors. A list of commercially available taggants is provided in Table 3. Improvised explosive devices (IEDs) constructed by terrorists often contain home-made explosives (HMEs) rather than conventional high explosives due to ease of access to unrestricted starting materials. Table 3 also provides examples of strong oxidizers that are contained in HMEs [15–17].

2.2. Challenges in chemical sensing of homeland security threats

In order to serve homeland security effectively in detecting known and emerging threats, new sensing technologies must exceed previous technologies in sensitivity, response speed, selectivity, and long-term stability [1]. While improvements in sensor sensitivity and response speed were demonstrated using new transducer designs and large-surface area sensing nanomaterials [5,8], high selectivity of chemical detection and adequate long-term stability of sensors are the most significant challenges [5,9], due to aspects summarized below.

2.2.1. Selectivity

Gas-induced response of sensors should be considered on several levels, ranging from response to almost all gases and vapors, to response to only a specific gas, and finally to response to the gas of interest in the presence of gaseous interferences and uncontrolled operating conditions. Most materials demonstrated for sensing are sensitive to not only target analytes, but also a wide variety of gases and vapors. The presence of gaseous interferences starts to complicate sensor measurements significantly degrading the accuracy of sensor performance. Variable or uncontrolled operating conditions, especially temperature fluctuations, further reduce the accuracy of sensor performance.

There are at least two reasons for low selectivity of individual sensing materials. One reason relates to the required performance where a sensor needs to be (1) highly selective while being (2) fully reversible. These two requirements are in conflict, because reversibility of sensor response is achieved via low energy of interactions between the analyte and the sensing film, while high selectivity of sensor response is achieved via high energy of interactions between the analyte gas and the sensing film. As a result, often individual sensors that have good reversibility are not selective and cannot detect minute concentrations of analyte in the presence of elevated levels of interferences. Another reason is the fundamental nature of adsorption and absorption interactions between vapors and materials that does not provide molecular-recognition selectivity similar to that in biomolecular interactions [18].

Despite of the diversity of sensing materials and their response mechanisms [9], no fully selective gas-sensing materials exist. A common approach to address the problem of poor selectivity of individual sensors is to build an array of partially selective sensors and to process the array response using multivariate analysis [19]. In such sensor arrays, individual transducers are coated with sensing materials, where one or more response per sensing material (e.g., resistance, current, capacitance, work function, mass, temperature, optical thickness) is measured. Although using identical transducers in an array simplifies the fabrication process, combining transducers based on different principles or employing transducers that measure more than one property of a sensing film [20] could improve array performance through coupling. Minimizing the number of sensors in an array is attractive, as it simplifies device fabrication and sensing-material deposition and reduces data-processing noise [9].

Real-world environments are much more complex than a controlled laboratory setting, so it is critical that sensor prototypes developed for homeland security applications are not disappointingly affected by environmental factors (e.g., variable levels of water vapor in air, temperature fluctuation and background interferences). Water vapor, which is the most abundant and high-concentration interference in ambient air, at times can be up to 95% of its saturated H₂O vapor pressure, Other potential application-specific chemical interferences that add to the complexity of the background [21] include ubiquitous fluids and common industrial solvents, typically at the 0.01–0.1 P₀ level.

2.2.2. Long-term stability

Requirements for the long-term stability of a sensor may vary several orders of magnitude depending on the application. For example, a sensor to be used in first-responder applications can be a one-time disposable sensor without need for long-term stability, whereas sensors in a WSNs must be stable over extended periods of time. Understandably, numerous published literature results do not report details on the long-term stability or instability of sensing materials. This situation arises from the nature of projects that may focus on only the initial discoveries of sensing materials {e.g., sensitivity [22,23], response/recovery times [24], or new transduction principles for gas sensors [25]}. Nevertheless, it is critical to bring forward early the possible challenges in materials stability.

Some researchers provide information about materials instabilities [26–31], providing the sensing community with important data and knowledge that truly facilitate decision-making in the broad area of sensor research. Some of the best reported results on long-term stability with room-temperature-operated sensors include sensors with dielectric polymers with 36–46 months’ stability. For example, silicone block polyimide sensing polymer had no detectable differences in the amount and the type of oxygen on the surface of fresh and 36-month-old films, as evaluated using X-ray photoelectron spectroscopy [32]. Another sensing material in the form of polymerized particles of diethyl ester of p-phenylenediacrylic acid also showed good vapor response with only 20% signal decrease over 46 months [33].

Sensor material stability is often evaluated as change in a baseline material property or vapor sensitivity over time. Unfortunately, analysis of the stability of the sensor-response pattern to several expected analyte and interference vapors is reported only rarely [27,31.34.35] because of the time and the resource required to generate such data and the detailed knowledge needed about gas composition in an intended end-user application. However, information on stability has the most value for assessing practical sensor implementation [31,35] and appropriate multivariate data-analysis algorithms that are most immune to long-term drift effects [34].

2.3. Specific requirements for wireless chemical sensors and networks

In wireless sensors, electronic transducers are spatially and galvanically separated from their associated read-out/display components [9]. The benefits of wireless sensors, compared to traditional tethered sensors, include the non-obtrusive nature of their installations, higher nodal densities, and lower installation costs. These attractive features of wireless sensors facilitate their development for numerous applications, including those important to homeland security.

2.3.1. Wireless sensors

A typical wireless gas sensor has three highly interdependent subsystems that include a sensing transducer with its associated means of power, a sensing material deposited onto the transducer, and a data transmitter. The data transmitter may be a separate component to which the sensor is connected, or the sensor itself can be the data transmitter [36]. The sensing material can be a separate component or it can also act as the transducer [5,18]. Wireless sensors communicate with a read-out/display component (sensor reader) or between each other via a WSN. The communication distance depends on several parameters including the power stored on-board the sensor or delivered to the sensor, the power needed for the sensor to operate at a predetermined signal-to-noise ratio, the amount of environmental clutter that reduces sensor performance, and the communication protocol. The key requirements for wireless gas sensors include minimal or no power needs for the transducer and sample-introduction components and the ability to correct sensor response adequately for fluctuations in ambient temperature and chemical interferences.

Depending on the sensor design and mode of operation, wireless sensors can be categorized as active and passive. Active wireless sensors have an on-board power supply (e.g., a battery, a supercapacitor, or an energy harvester) and can transmit signals up to several hundreds of meters. Passive wireless sensors lack an on-board power supply and receive their power from the electromagnetic field generated by the sensor reader. Although active wireless sensors communicate over greater distances than passive sensors, their larger size and the need to maintain the power source are possible limitations in some applications. Without the need for a battery, the life of a passive sensor is limited mainly by the stability of the sensing film and transducer. Thus, while the communication read range of passive sensors is shorter than that of active sensors, passive sensors are attractive because they are long-lasting, cost-effective, and inconspicuous devices. Power to passive sensors can be provided by inductive or capacitive coupling [37] with the delivered power dependent on operation frequency, sensor-antenna size, pick-up coil size, impedance-matching conditions, and the power of the sensor reader. Short-range wireless communication with small sensors and low power readers (proximity communication) can be implemented where it is important to minimize the effects of ambient clutter.

Most of the electronic transducers that have been developed have also been implemented in wireless sensors by implementing wireless excitation or direct attachment to the analog input of a specialized integrated circuit (IC) chip for further digital data transmission of sensor signal [9]. Table 4 gives examples of companies that manufacture passive and active wireless sensors.

Table 4.

Examples of companies that are manufacturing passive and active wireless sensors

Sensor type	Detected parameters	Company
Active	Volatile organic compounds, NH3, HCl, H2, CO2	Esensors, Inc., Amherst, NY, USA www.eesensors.com
Active	Volatile organic compounds, O2, CO, H2S	RAE Systems, Inc., San Jose, CA, USA www.raesystems.com
Active	CO, CO2	Acme Engineering Prod. Ltd, Montreal, Quebec, Canada www.acmeprod.com
Active	Ammonia, Chlorine, Chlorine Dioxide, Carbon Monoxide, Hydrogen, Hydrogen Chloride, Hydrogen Cyanide, Hydrogen Fluoride, Hydrogen Sulfide, Nitric Oxide, Nitrogen Dioxide, Oxygen, Phosgene, Phosphine, Sulfur Dioxide	Gastronics, Inc., Bedford Heights, OH, USA www.gastronics.com
Active	Temperature and humidity (as sensor array plugged into the headphone jack of a smartphone)	X-info Wieland Sacher GmbH, Munich, Germany www.app-solut.com
Passive	LF, HF, UHF sensors for pressure, temperature, humidity	Phase IV Engineering, Inc., Boulder, CO, USA www.phaseivengr.com
Passive	Chipless RFID sensing tags for humidity	TagSense Inc., Cambridge, MA, USA www.tagsense.com
Passive	UHF sensors for temperature	Schneider Electric SA, France www.schneider-electric.com
Passive	RFID sensors for product security and authentication	Schreiner LogiData, Munich, Germany www.schreiner-logidata.com
Passive	NFC sensors for humidity/temperature	eProvenance, Bordeaux, France, www.eprovenance.com
Passive	NFC sensors for humidity	Microsensys GmbH, Erfurt, Germany www.microsensys.de/

LF, low frequency; HF, high frequency; UHF, ultrahigh frequency; NFC, near field communication

Both active and passive wireless sensors can contain read/write or read-only memory in the form of IC memory chips or interdigital reflected power lines. This feature adds an RFID capability if the operation frequency and the communication protocols meet the regulatory requirements [37,38]. RFID devices (wireless labels or tags with read/write or read-only memory) have been recognized as a disruptive technology and are widely used in diverse applications ranging from asset tracking, detection of unauthorized opening of containers, and automatic identification of animals [37,38]. Typical operating frequencies of RFID sensors are 125–135 kHz (LF, low frequency tags), 13.56 MHz (HF, high frequency tags), 868–956 MHz (UHF, ultrahigh frequency tags), and 2.45 GHz (microwave tags) [37]. Recent developments of sensors communicating with smartphones are based on the near-field communication (NFC) protocol.

Fig. 1 illustrates the operating principle of a typical digital RFID sensor with two representative sensing examples [39]. A digital RFID sensor contains a special IC chip with an analog input that accepts a capacitance or a resistance transducer (Fig. 1A). Such IC chips typically have a one-bit sensor resolution providing a threshold (ON-OFF) measurement capability. Higher resolution of sensing is also available (such as 8–12 bit) with a significant expense of the cost of the IC chips. Such IC chips are available for operation at LF, HF, and UHF frequencies. Fig. 1B and Fig. 1C illustrate operation of such LF and UHF RFID sensors where conventional analog chemicapacitor transducers were attached to an analog input of custom-fabricated IC chips of these RFID sensors. Similar to the operation of tethered individual transducers, individual wireless sensors were responding to different concentrations of vapors but were unable to discriminate between different types of vapors of similar dielectric constant.

Figure 1.

Operation of digital RFID sensors with specialized IC chips containing an analog input. (A) Operating principle of a digital RFID sensor with a conventional analog transducer attached to an analog input of IC chip of RFID sensor. (B) Detection of water (H₂O), toluene (Tol), and tetrahydrofuran (THF) vapors using an LF Digital RFID sensor with an attached chemicapacitor. Concentrations of vapors: 0.18, 0.36, 0.53, and 0.71 P/P₀. (C) Detection of water vapor using an UHF Digital RFID sensor with an attached chemicapacitor. Concentrations of water vapor: 0.04, 0.09, 0.13, 0.18, 0.22, and 0.27 P/P₀ (Reprinted with permission from [39]).

The significant attractive feature of digital sensors is that they provide simple data transmission from attached analog sensors for numerous applications, where selectivity of sensing is not an issue (e.g., in humidity or temperature measurements) and where relatively high sensor cost ($5–50) is acceptable. However, such single wireless sensors cannot discriminate between vapors that produce similar responses. Thus, as expected, attachment of conventional transducers that have insufficient selectivity to an analog input of digital RFID devices does not improve sensing selectivity. Further, digital read-out of these sensors limits the dynamic range of measurements and achievable signal-to-noise ratio.

2.3.2. Wireless sensor networks

In a WSN, individual sensors are typically arranged into wireless sensing nodes with the key hardware (long-lifetime battery or energy-harvesting source, simple signal-conditioning components, low-power processor) and software (small memory, computational capacity, high modularity) requirements for individual nodes [40]. For homeland security applications, the requirements for individual gas sensors in a WSN include high selectivity and high sensitivity toward target analytes.

At present, challenges of WSNs for gas sensing include power consumption of individual sensors and handling of massive heterogeneous data from the WSN. The inadequate long-term stability of many research prototypes of gas sensors further prevents their reliable applications in WSNs. Thus, examples of sensors in WSNs include commercially-available gas and VOC sensors [41–43] that often consume significant power (30–50 mW and up to 400 mW per sensor) [43,44]. Low power consumption in commercially-available systems was reported using only simple humidity sensors [41,42].

The opportunities for WSNs with gas-sensing nodes originate from the synergistic combination of new data-generation and processing concepts with new sensor-integration concepts. Sensors arranged as networks can significantly benefit from novel data-generation and data-processing concepts currently unavailable for individual sensors. Three main aspects of these advantages can be summarized as:

1. the ability for efficient sensors communications;

2. improvement of detection accuracy through data fusion; and,

3. opportunities for automatic re-calibration of individual sensors on the network.

Ideally, in a wireless sensing node, it will be attractive to minimize the number of sensors in an array or even to have only a single, highly accurate, stable sensor because of the obvious benefit of minimizing power need. In addition, similar to conventional tethered sensor systems, minimizing the number of wireless sensors in an array is attractive because it simplifies data analysis, reduces data-processing noise, and simplifies sensor-material deposition and device fabrication [9].

2.4. Towards a technical solution – multi-variable wireless sensors

An attractive technical solution to the insufficient selectivity of existing sensors is to implement a new sensing concept for selective sensing that requires only a single sensor rather than a sensor array [36,45–48]. By applying a carefully-selected sensing material onto the resonant antenna of the RFID tag and measuring the impedance of the antenna, the impedance-spectrum response is correlated to concentration of a chemical compound of interest in the presence of high levels of background interferences. The digital data is also written into and read from the IC memory chip of the RFID tag. This IC memory chip can also store sensor calibrations and user-defined information.

This new sensing principle requires only a single sensor, but it involves combining a sensing material with different response mechanisms to different vapors and with a transducer that has a multi-variable signal-transduction ability to detect their independent changes. The action of several response mechanisms in a single sensing film to different vapors was achieved leading to independent detection of these responses with a single sensor, and correction for variable ambient conditions.

2.4.1. Operating principle

The operating principle of multivariable RFID sensors is illustrated in Fig. 2 (A and B). Reading and writing of digital information into the RFID sensor and measurement of impedance of the RFID sensor antenna are performed via mutual inductance coupling between the RFID-sensor antenna and the pick-up coil of a reader. A conventional digital RFID reader acquires digital data from an IC memory chip on the RFID sensor. This digital data has a unique factory-programmed serial number (chip ID) and user-written data about the properties of the sensor (e.g., calibration curves for different conditions) and the object to which the sensor is attached (e.g., fabrication and expiry dates). The origin of response of RFID sensor to chemical parameters is described in Fig. 2A. Upon reading the RFID sensor with a pick-up coil, the electromagnetic field generated in the sensing region extends out from the plane of the sensing region and is affected by the dielectric property of the ambient environment. When the resonant antenna of the RFID sensor or a complementary sensing region [47] is coated with a sensing film, the analyte-induced changes in the dielectric and dimensional properties of the sensing film affect the impedance of the sensor circuit through the changes in film resistance and capacitance. Such changes facilitate diversity in response of individual RFID sensors and provide the opportunity to replace an array of conventional sensors with a single vapor-selective RFID sensor.

Figure 2.

Operating principle of passive battery-free RFID sensors with the multivariable signal transduction. (A) Origin of response of RFID sensors to chemical parameters via a sensing film deposited onto the resonant antenna. (B) Measured impedance spectrum (real part Z_re and imaginary part Z_im of impedance) and representative parameters for multivariate analysis. (C) RFID sensor with an antenna structure specifically developed for sensing application. (D) RFID tags from different manufacturers adapted for chemical sensing (Reprinted with permission from [45,46])

For selective vapor quantification using individual RFID sensors, impedance spectra of the resonant antenna are measured and several parameters from the spectra are calculated (Fig. 2B). Examples of calculated parameters include the frequency of the maximum of the real part of the impedance F_p, the magnitude of the real part of the impedance Z_p, the resonant frequency of the imaginary part of the impedance F₁, and the anti-resonant frequency of the imaginary part of the impedance F₂, signal magnitude Z₁ at the resonant frequency of the imaginary part of the impedance F₁, and signal magnitude Z₂ at the anti-resonant frequency of the imaginary part of the impedance F₂. Other spectral parameters may be also measured from the impedance spectra. Upon appropriate selection of a sensing film, the film-coated RFID sensor has different responses for each tested analyte or the analyte and interferences. Examples of RFID tags adapted for sensing are presented in Fig. 2 (C and D) [45,46].

By applying multivariate analysis of the full impedance spectra or the calculated parameters, quantification of analytes and their mixtures with interferences is performed with individual RFID sensors. The multivariate analysis tool described here is principal components analysis (PCA), a robust unsupervised pattern-recognition technique [19]. PCA projects the data set onto a subspace of lower dimensionality with collinearity removed. PCA achieves this objective by explaining the variance of the data matrix in terms of the weighted sums of the original variables (e.g., F_p, F₁, F₂, and Z_p) or complete impedance spectra with no significant loss of information. These weighted sums of the original variables are called principal components (PCs).

In applications detailed in sub-sections 2.4.2–2.4.5, several classes of sensing materials were used from diverse classes of materials {e.g., dielectric polymers [36,49], conjugated polymers [47], single-wall carbon nanotubes [39], and monolayer-protected metal nanoparticles (NPs) [47,50]}. These materials were selected because of their diversity of vapor-response mechanisms.

2.4.2. Interference-independent detection of vapors of interest

Water vapor is the most common interference in practical applications of sensors. Using the knowledge obtained in the design of the sensing materials for RFID sensors, several types of new conjugated polymers as sensing materials were synthesized and determination of toxic vapors was performed down to 900 ppb limits of detection (LODs) and significantly suppressed humidity effects. In a representative example, a conjugated polymer poly(fluorene)-diphenylpropane [47] was applied onto an RFID sensor for vapor-response testing. It was found that, from conservative estimates based on the multivariable sensor response, the detection of toxic gases can be performed in the presence of 27,000-fold more concentrated water vapor. Measurements over an extended period of time (45 h) were also performed using a dielectric polymer poly(etherurethane) to evaluate effects of even higher humidity levels, up to 76% RH, where magnitude of response of the sensor also did not change to the analyte at high humidity of the carrier gas [49].

The ability of RFID sensors to operate in the presence of variable ambient humidity and to reject effects of ambient humidity was evaluated in detail using toluene as a model analyte and octanethiol-capped AuNPs [47] as a sensing material. While individual measured parameters were affected by relative humidity, as demonstrated previously [9], multivariate analysis results, as shown in a PCA scores plot versus experimental time (see Fig. 3A), fully corrected for the humidity effects [39].

Figure 3.

Demonstration of humidity-independent operation using a single sensor with the multivariable signal transduction: (A) Plot of PC₁ vs. PC₂ vs. time illustrating sensor selectivity. Concentrations of toluene vapor: 89, 160, and 261 ppm; RH levels: 0, 10, 20, and 40 %RH. (B and C) Examples of sensor response Z_p at 0 and 40 %RH, respectively. (D) Correlation between actual and predicted toluene concentrations in the presence of different amounts of water vapor (0, 10, 20, and 40 %RH) (Reprinted with permission from [39]).

Critical to the sensor performance, sensing material applied onto the RFID antenna had the same magnitude of response to the model analyte vapor in the presence of different humidity levels [see Fig. 3 (B and C)]. Full correction of the toluene response at different humidity levels was done using multivariate analysis of multiple responses from the single sensor. The resulting multivariate calibration curves at variable RH were identical and provide a new capability to quantify vapors at different humidity levels. The correlation plot between actual and predicted toluene concentrations in the presence of different amounts of water vapor (0, 10, 20, and 40 %RH) is illustrated in Fig. 3D. The calculated standard error or predicted toluene concentration (3 ppm) was unaffected by the variation in humidity in the range 0–40% RH.

2.4.3. Detection of CWA simulants

Effects of different vapors on the response of the developed sensors were evaluated by selecting a difficult combination of vapors {i.e. methanol (MeOH), ethanol (EtOH), water (H₂O), and acetonitrile (ACN) [36]}. ACN was selected as a simulant for blood CWAs. A dielectric polymer tetrafluoroethylene and sulfonyl fluoride vinyl ether copolymer [36] were selected as sensing materials. Responses from different analytes were indistinguishable if an RFID sensor measured only a single parameter. By applying a PCA model developed on the multi-variable response of this single sensor, discrimination of some vapors was achieved. Fig. 4A illustrates that three out of four vapors were resolved using a selected sensing film and the read-out from the RFID sensor [36]. ACN vapor was clearly distinguished from H₂O, MeOH, and EtOH vapors.

Figure 4.

PCA score plots demonstrate selective detection of CWA simulants using a single RFID sensor with the multi-variable signal transduction and different sensing materials. (A) Use of tetrafluoroethylene and sulfonyl fluoride vinyl ether copolymer for detection of ACN. Vapors: (1) H₂O, (2) EtOH, (3) MeOH, and (4) ACN, concentrations of each vapor: 0, 0.02, 0.04, 0.07, 0.10, 0.15, and 0.20 P/P_o, three replicates per concentration. (B) Use of polyaminobenzene sulfonic acid functionalized SWNTs for detection of ACN. Vapors: (1) H₂O, (2) ACN, (3) dichloromethane, and (4) chloroform, concentrations of each vapor: 0, 0.0178, 0.0356, and 0.0533 P/P_o. (C) Use of peptide-capped Au nanoparticles for detection of ACN and MeS. Vapors: (1) acetonitrile, (2) dichloromethane, (3) methyl salicylate, (4) ethanol, (5) toluene, (6) 1-pentanol, (7) chloroform, (8) salicylaldehyde. Concentrations of each vapor: 0, 0.044 and 0.089 P/P_o. Water vapor background is 0.18 P/P_o (Reprinted with permission from [36,39,50]).

Detection of the CWA simulant (ACN) and interferences (H₂O, dichloromethane, and chloroform) was also performed using polyaminobenzene sulfonic acid-functionalized single-walled carbon nanotubes (SWCNTs) as a sensing material [39]. Fig. 4B demonstrates that PCA modeling of the multi-variable response of this single sensor resulted in good discrimination of these vapors [39].

Metal NPs with a monolayer of peptides (AYSSGAPPMPPF-capped AuNPs) were also explored as a sensing material for RFID vapor sensors [50]. These peptide-functionalized NPs were recently developed [51]. In these experiments, ACN and methyl salicylate were used as CWA simulants and dichloromethane, ethanol, toluene, pentanol, chloroform, and salicylaldehyde vapors as interferences. Discrimination of most of these model vapors was achieved using PCA of the multi-variable response of this single sensor, as shown in Fig. 4C [50]. These developed sensors illustrated in Fig. 4 had good vapor-discrimination ability, as evidenced by the relatively high contributions to the second PC of the developed PCA models in the range 8–32%.

2.4.4. Detection of TIMs

Newly-developed RFID sensors were also demonstrated for detection of TIMs. In the initial experiments, ammonia was selected as the analyte of choice based on its high HI and its high ranking among toxic-by-inhalation gases [14]. Polyaniline (PANI) was chosen as the sensor material for these experiments because it is a well-studied intrinsically conducting polymer for vapor sensing [5,18]. Fig. 5 (A–D) presents the sensor responses obtained for NH₃ and H₂O [49]. Deprotonation of the film upon NH₃ exposures resulted in the increase in film impedance (Z_p) and the shift of the sensor resonance (F_p, F₁, and F₂) to higher frequencies. By contrast, the formation of hydrogen bonds and swelling of the polymer upon H₂O exposure resulted in the decrease in Z_p and shift of F_p, F₁, and F₂ to lower frequencies. Measurements of multiple output parameters from a single sensor revealed different recovery kinetics of responses Z_p, F₁, F₂, and F_p during experiments with NH₃. Responses Z_p, F₁, and F_p showed a partial recovery from NH₃, while the F₂ response was irreversible. The sensor response to H₂O vapor was reversible, and at least 100-fold weaker than the response to NH₃. Univariate Z_p, F₁, F₂, and F_p calibration curves to NH₃ showed relatively high response at low concentrations.

Figure 5.

Selective analysis of NH₃ and H₂O vapors using a single sensor with the multi-variable signal transduction. (A–D) Sensor responses F₁, F₂, F_p, and Z_p, respectively upon ~ 10 min exposures of sensor to H₂O vapor (630, 1260, 2205, 3150, 4410, and 6300 ppm) and to NH₃ vapor (4, 8, 14, 20, 8, and 40 ppm). Note reversible response to H₂O vapor and non-reversible response to NH₃ vapor. Inset in B, dynamic response to H₂O vapor. Inset in D, univariate calibration curve for NH₃ determinations. (E) Scores plot of PC₁ versus PC₂ demonstrates discrimination between NH₃ and H₂O vapor responses (Reprinted with permission from [49]).

For comparisons with earlier reported sensors, selectivity of the RFID sensors developed was evaluated using PCA. As shown in the PCA scores plot (see Fig. 5E), the action of several vapor-response mechanisms in a single sensing film was independently detected and quantified with a single multi-variable signal transduction RFID sensor. The LOD of the PANI-based RFID sensors was calculated to be 20–80 ppb from individual Z_p, F₁, F₂, and F_p measurements [46]. The PANI-based sensors were further optimized for detection of low concentrations of ammonia and hydrogen sulfide. LODs achieved for NH₃ and H₂S were 0.3 ppb and 6 ppb, respectively.

2.4.5. Detection of explosives and oxidizers

Preliminary detection of conventional explosives and low vapor-pressure oxidizers was also demonstrated using RFID sensors. In these initial experiments, the main emphasis was to evaluate the detection capability of the sensors. Fig. 6 illustrates the successful detection of 0.2 mg of TNT, 0.1 mg of TATP in ACN, and 5 g of NH₄NO₃ using the RFID sensors developed [39]. Sensitivity and selectivity of the RFID sensors for an expanded list of explosives in Table 3 will be evaluated in future studies.

Figure 6.

Examples of detection of (A) TNT (powder form), (B) TATP (dissolved in acetonitrile) and (C) NH₄NO₃ (crystal form) using developed sensors (Reprinted with permission from [39]).

3. Conclusion

Detection of threats of importance to homeland security is needed with enhanced sensitivity for detection of agents below risk levels, and with increased selectivity for operation in presence of uncontrolled variable humidity and for providing minimal false-alarm rates. These capabilities will be difficult to achieve using evolutionary improvements in existing sensing technologies.

Conceptually new technical solutions are needed that should include a synergistic combination of three key sensor system components:

1. sensing material;

2. appropriate transducer; and,

3. appropriate signal-generation and signal-processing techniques.

The technical solution based on wireless RFID sensors provides sensor performance (e.g., response selectivity) that is difficult to achieve even with sensor arrays. Although, in this work, the applications of developed sensors that we demonstrated focused on implementation in homeland security, the selectivity, the sensitivity, the cost-effectiveness, and the non-galvanic nature of these sensors are attractive for other applications (e.g., cold chain management, food safety, pharmaceuticals, warehousing, agriculture, and industrial process monitoring). For these and many other applications, additional benefits of the RFID sensors developed, compared to tethered sensors, include the non-obtrusive nature of their installations, higher nodal densities, and lower installation costs without the need for extensive wiring.

• Brief summary of chemical threats of importance to homeland security

• Origins of the most significant unmet needs in existing chemical sensors

• Specific requirements and challenges for wireless chemical sensors and networks

• New approach to selective wireless chemical sensing

• New sensors for toxic materials, chemical agent simulants, explosives and oxidizers