MIME Chemical Vapor Microsensors

A.W. Snow
Chemistry Division

H. Wohltjen and N.L. Jarvis
Microsensor Systems, Inc.

A nanocluster metal-insulator-metal ensemble (MIME) chemical vapor sensor is a solidstate sensor composed of nanometer-size gold particles encapsulated by a monomolecular layer of alkanethiol surfactant deposited as a thin film on an interdigital microelectrode. The principle by which this sensor operates is that vapors reversibly absorb into the organic monolayer, which causes a large modulation in the electrical conductivity of the film. The tunneling current through the monolayer between gold particle contacts is extremely sensitive to very small amounts of monolayer swelling and dielectric alteration caused by absorption of vapor molecules. The nanometer scale of the particle domains and correspondingly large surface area translate into a very large vapor sensitivity range, extending to sub-ppm concentrations. Selectivity of the sensor is regulated by incorporation of chemical functionalities at the terminal structure of the alkanethiol surfactant or substitution of the entire alkane structure. The current focus of research is in mapping the selectivity and sensitivity of sensor elements made by incorporating these functionalities into the shell of the nanocluster. Targeted applications include detection of chemical warfare agents and explosives, and residual life indication of carbon filters and protective clothing.

INTRODUCTION

Detecting hazardous chemical vapors with highly miniaturized analytical devices is a present and future capability that is assuming an increasing importance in many DOD and civilian scenarios. These scenarios involve chemical weapons, concealed explosives, volatile organics in breathing air, residual life indication of gas mask filters and protective clothing, and electronic nose functions (identification of unknown substances by odor). Until 25 years ago, detection of hazardous chemicals relied on large-scale laboratory configured instrumentation such as mass and optical spectroscopies, gas chromatography, and to a smaller extent on colorimetric schemes using wet chemistry. Since then, the impetus has been toward reducing instrumentation size and field deployment of analytical instrumentation. Also during this time, sensors as minimal-component, highly miniaturized, stand-alone devices emerged. These devices generally incorporated a chemically active surface interfaced to an electronic substrate. Examples include metal oxide semiconductor (MOS) devices, miniature electrochemical cells, surface acoustic wave (SAW) devices, and chemiresistors. For vapor sensing, these devices rely on a partitioning of an analyte vapor between the gas phase and that sorbed onto/into the chemically active surface. The chemically active surface then acts as a transducer. A property change caused by sorption of a vapor generates an electronic signal that may be processed into analytical information about the concentration of the analyte vapor. NRL has been active in microsensor research since 1981 with SAW and chemiresisor devices. Features that govern the sensitivity of these devices are the transduction mechanism, vapor partitioning, and the surface area to volume ratio of the chemically active adsorbent coating. The first feature is determined by the coupling between the electronic substrate and the chemically active surface or coating and is varied mostly by the substrate design. The second feature is determined by the chemical interaction between the analyte vapor and the adsorbent surface's chemical structure and is varied by the design and synthesis of this component. Over the past 20 years, much research at NRL and elsewhere has focused on optimizing these features. The third feature (surface to volume ratio of the chemically responsive coating) has received very little attention since the thickness of the chemically active coatings were optimized to be thick enough for good sensitivity but thin enough for a fast response. A nanoscale materials approach changed this radically. A new type of chemical microsensor with distinct advantages derived from nanometer-scale material domains emerged. The objective of the current research is to understand the operating principles of this new sensor and exploit its advantages in practical applications.

MIME SENSOR CONCEPT

The new sensor is based on a metal nanocluster encapsulated by a single layer of organic molecules. Figure 1 illustrates the materials concept. A single cluster, depicted in the upper left corner, is composed of a gold core encapsulated by a shell of alkanethiol molecules. The gold core may range from 1 to 5 nm in diameter, and the electronic properties of this material originate from it. The alkanethiol molecules of the shell are bonded to the surface of the gold core by a gold-to-sulfur bond. This organic shell forms a very thin insulating barrier. Its thickness, which may vary from 0.4 to 1.0 nm, has an enormous effect on electron tunneling between adjacent clusters. The organic shell also imparts an organic character to the cluster that promotes solubility in organic solvents such that a cluster as much as 90% gold by weight will dissolve in toluene. This solubility makes processing these clusters into thin films very facile. The sensor is fabricated by depositing a film of these clusters onto a micron-scale interdigital electrode substrate. When connected to a small bias (50 to 500 mV), a nanoamp current flows through the film. Exposure to vapors causes very large changes in the conductivity of the film. This results from a sorption of the vapor into the very thin shell, and the consequent swelling of the shell results in a small but very significant increase in the distance between cores of adjacent metal clusters. The tunneling current is extremely sensitive to the distance between cores. A final feature of significance in Fig. 1 is the packing of the clusters in the film. Being spherically shaped clusters, any type of packing will have nanometer-scale voids within its matrix. The size differential between a typical vapor molecule, such as toluene, and a nanocluster is approximately a factor of 10. As such, this network of voids in the cluster matrix provides rapid ingress and egress of vapors, much more so than the slow diffusion into polymer films used on typical microsensors. This provides a pathway for a much faster response and recovery for sensors based on a metal cluster ensemble. Combined with the cluster ensemble's fast kinetics for sorption and desorption is an extremely large surface to volume ratio that translates into a highly enhanced sensitivity for this MIME sensor. The MIME sensor derives its name from the metal-insulator-metal ensemble character of the cluster film. The critical features in its design are the dimensions of the core and shell and the chemical composition of the molecules composing the shell. These features are described in the following paragraphs.

FIGURE 1
The metal-insulator-metal ensemble (MIME) sensor concept. A micron-scale interdigital electrode is coated with a film of alkanethiol stabilized gold nanoclusters and exposed to toluene vapor. The toluene adsorbs into and in between the alkanethiol monolayer shell, and the consequent swelling causes an increase in the separation distance between gold cores and a reduction of electron tunneling between them.
Fig2 Image

FIGURE 2
The alkanethiol stabilized cluster synthesis. After reduction of the gold ions, the competitive processes of gold particle growth and alkanethiol surface complexation determine the size of the gold nanocluster.

CLUSTER SYNTHESIS AND CHARACTERIZATION

The dimensions of the core and shell of the cluster are determined by the conditions of its synthesis. Figure 2 illustrates the alkanethiol-gold cluster synthesis. The two critical reagents are the gold chloride and the alkanethiol. They are suspended in a common medium, and a reducing agent, typically NaBH4, is added. The trivalent gold is reduced to neutral gold, and the gold atoms aggregate to form a particle nucleus. The gold particle grows by the addition of gold atoms and smaller particles. As a competing process, the alkanethiol reacts with the neutral gold surface to form a sulfur-to-gold bond. The gold particle growth is terminated when its surface is encapsulated by complexation with the alkanethiol. The relative rates of gold particle growth and alkanethiol surface complexation are dependent on the concentrations of the respective gold chloride and alkanethiol reagents. Thus, the molar ratio of these reagents is a simple way to regulate the core size of the cluster. Typically, this ratio ranges from 1:3 to 8:1 and causes the corresponding core diameters to vary from 1 to 5 nm. The shell thickness of the cluster is determined by the molecular chain length of the alkanethiol selected as the reagent. The number of carbon atoms in the chain length typically ranges from 4 to 16 and generates a corresponding shell thickness ranging from 0.4 to 1.0 nm. A matrix of these clusters has been synthesized as is depicted in Table 1, where the varying core and shell relative sizes are represented pictorially by concentric circles.

As a shorthand to designate individual clusters, the general abbreviation Au:Cn(X:Y) is used. X:Y is the gold chloride:alkanethiol synthesis stoichiometric ratio that correlates with core size, and n is the number of carbon atoms in the alkanethiol chain that correlates with shell thickness. The number beneath each cluster in Table 1 is the corresponding bulk DC electrical conductivity. An appreciation for the respective effects of the shell thickness and core size on the electrical conductivity can be obtained by examining the magnitude of conductivity variation down the column headed (1:1) and across the row headed C12, For the former, the variation is on the order of 107, which illustrates why such minute swelling effects on the shell have such dramatic effects on the conductivity.

FIGURE 3
Response of Au:C8(1:1) MIME sensor to five 60-s exposure-purge cycles of toluene at high and low vapor concentrations. The right axis indicates the current change in the MIME device, and the left axis is the sensor response when the current change is converted to a frequency.

MIME SENSOR RESPONSE TO VAPORS

Figure 3 shows the response of an Au:C8(1:1) MIME sensor to five 60-s exposure-purge cycles of toluene at high and low vapor concentrations. The sensor response is a measured current change (right axis) that is electronically converted to a frequency (left axis) via precision current-to-voltage and voltage/ frequency converters to allow data acquisition over a wide dynamic range using a computerized frequency counter. The toluene vapor causes a very large and rapid decrease in conductance of the sensor. Greater than 90% of the signal response occurs within 1 s of its 30-s exposure. The recovery is equally rapid and complete. The lower portion of Fig. 3 indicates that detection limits well below 1 ppm are achievable.

The MIME sensor response to toluene displays a dependece on both the core and shell dimensions of its cluster component. When the matrix of clusters depicted in Table 1 is investigated, optimum sensitivities are found for both the core diameter and the shell thickness in the midrange of the matrix. Clearly two effects operate in each case. In the latter, a thicker shell requires more sorbed vapor to achieve an amount of swelling comparable to that of a thinner shell, while a thinner shell has less of an organic character to solvate the incoming toluene. In the former case, it is not clear why a variation from 1 to 5 nm would pass through an optimum in sensor sensitivity to toluene.
Fig4 Image

FIGURE 4
Vapor response isotherms of the Au:C8(1:1) MIME sensor to toluene, water, and 1-propanol vapors based on 15°C vapor pressure. The inset displays the toluene response down to a 2.7-ppm concentration.

Figure 4 shows the dependence of the Au:C8(1:1) MIME sensor response to toluene vapor concentration along with responses to 1-propanol and water vapors. The inset depicts this response at the very low concentrations for toluene. Sensor response is expressed as the change in conductance normalized to the baseline (purge conditions) conductance, and vapor concentration is expressed as fractions of that corresponding to the vapor pressure of the pure liquid at 15°C. Since toluene, 1-propanol, and water have nearly the same vapor pressures, the vapor concentrations of each are similarly comparable. The response to toluene vapor is very large and deviates slightly from linearity, with the slope of the curve becoming greater at the low end of the concentration range. This sensor is remarkably insensitive to water vapor, even at high concentrations. The lack of moisture sensitivity is of great importance for practical environmental sensing applications. Propanol vapor produces an initially unexpected result of a conductance increase at high concentrations. Propanol is a very polar organic vapor. Its sorption into the cluster shell increases the dielectric character between cores through which electrons tunnel. The observed conductance increase is believed to result from this increased dielectric character promoting charge transport between clusters.

While this Au:C8(1:1) MIME sensor can easily discriminate toluene vapor from water or alcohol vapors, it also responds to less polar organics to varying degrees. The selectivity of a MIME sensor for a particular coating-vapor combination is determined by the vapor's partition coefficient between the gas and coating phases. The partition coefficient is dependent on both physical properties of the vapor and its chemical interactions with the coating. Gas chromatography and early research with polymer-coated SAW sensors¹ demonstrate that the partition coefficient K (ratio of vapor absorbed in the coating to that remaining in the gas phase) is inversely dependent on the physical vapor pressure of the analyte vapor p2 and on the chemical activity coefficient for dissolution of the vapor in the coating γ as follows:

K = (RT)/(M₁p₂γ). (1)

where R is the gas constant, T is the temperature, and M1 is the molecular weight of a coating molecule (M1 becomes an undefined constant when the size differential between coating and vapor molecules becomes large). This indicates that sensors respond more strongly to analyte vapors with low vapor pressures than to those with higher vapor pressures. When the Au:C8(1:1) MIME sensor is challenged with a homologous series of alkane vapors (pentane through dodecane) in which the chemical interaction is constant, the intensity of the sensor response displays a quantitative inverse dependence to vapor pressure, as expected from Eq. (1). For this reason, when we wish to investigate chemical interactions as a guide for enhancing sensor selectivity, care must be taken to either select vapors with very similar vapor pressures (as was done in Fig. 3) or to invoke a vapor pressure compensating factor. In designing sensor coatings with vapor selectivities, the approach is to incorporate chemical structures that will have an interaction with a target vapor. Chemical interactions are quantified by an activity coefficient γ, which appears in Eq. (1). The activity coefficient is a number between 1 and 0 having a value close to 1 for nearly "ideal systems" where no chemical interactions are occurring, and progressing to smaller numbers as chemical interactions of increasing strength occur. Reversible chemical interactions range from weak van der Waals forces, to dipole attractions, to hydrogen bonding, to strong acid-base and charge-transfer interactions.

Various models and schemes have been developed to correlate molecular structures with magnitudes of interactions. Qualitatively they are useful, but unfortunately attempts to quantify these models result in multi-parameter equations of 3 to 8 variables that sacrifice utility for precision. A reliable approach is to build a database with some systematic variations in structure. Table 2 depicts some thiol ligand molecules having varied chemical functionality that are used for this purpose. These include a variety of alkanethiols with different terminal functional groups, a case where an ethyleneoxide chain is substituted for the alkane chain, and a variety of aromatically functionalized thiols. The corresponding clusters have been prepared, and a database exists. Some results are depicted in bar graph form in Figs. 5 and 6 for sensor responses to a variety of vapors at a concentration corresponding to one tenth of their vapor pressure. (All vapors except DMMP were chosen for a narrow range of vapor pressures.)

FIGURE 5
Vapor response pattern to DMMP (left) and toluene (right) for an array of seven MIME sensors with cluster coatings composed of different alkanethiols.

Figure 5 is a MIME vapor response pattern for an array of seven different MIME coatings responding to the same vapor. Dimethyl methylphosphonate (DMMP) is a standard phosphanate nerve agent stimulant and displays a significantly different pattern than toluene. It is these vapor response patterns that allow a sensor array to make identifications.

Figure 6 is a MIME coating response pattern displaying the response profile of one coating to seven vapors. It is clear that the fluoroalcohol functionalized cluster coating responds most strongly to basic and acidic vapors. The octanethiol cluster coating displays a particularly interesting pattern in that the polar vapors (nitromethane, 1-propanol, piperidine, and acetic acid) are displaying a response in which the relative resistance is decreasing (conductance increasing) as was observed for the propanol example in Fig. 4. The swelling mechanism described earlier does not accommodate a conductance increase. This clearly indicates a change in the transduction mechanism.

FIGURE 6
MIME sensor coating response profile to seven different vapors by a fluoroalcohol functionalized MIME sensor (left) and an octanethiol MIME sensor (right).

The correlation of vapor polarity with a MIME sensor response in the direction of increasing conductance indicates that a change in the dielectric character of the medium between the metal cluster cores can influence tunneling current in a direction opposite to that resulting from the swelling mechanism. This dielectric effect is not as noticeable in a cluster with a more polar shell such as the fluoroalcohol functionalized cluster in Fig. 6.

The charge transport through granular metal films has been studied since the 1960s, and several models and mechanisms have been described. One that is particularly simple and parameterizes the cluster core and shell dimensions and the dielectric constant of the intercore medium is:²

where σ is the conductivity, E_a is the activation energy, R is the gas constant, T is temperature, e is an electron charge, r is the radius of the cluster core, s is the spacing between cluster core surfaces, e is the dielectric constant of the medium between cluster core surfaces, and ξ₀ is the vacuum dielectric constant. An increase in the dielectric constant of the cluster shell has the effect of decreasing the activation energy for conductivity which, in turn will, increase the conductivity between clusters. This model is consistent with our observations of the sorption of very polar vapors resulting in conductivity increases.

This MIME sensor response to polar vapors with an increase in conductance is an added dimension in selectivity compared with other solid-state sensors such as SAW, MOS, and electrochemical cell devices that respond in only one direction. Some vapors of particular interest are very polar and display exceptionally strong MIME responses in the positive direction. Such vapors include those from explosives based on nitrate chemistry. Trinitrotoluene (TNT) is an exceptionally polar compound with an extremely low vapor pressure (8 X 10^-6 torr at 25°C with calculated corresponding concentration of 0.02 ppb by volume or 100 ng/liter). Figure 7 shows the MIME sensor response to this headspace concentration sampled by a short conduction line from a container of TNT. The container is warmed from 25 to 35°C during the 60-s exposure/180-s purge cycles, and the response follows the increase in vapor pressure. The initial 5-s response is very fast, followed by a slightly slower but strongly continuous increase. After 30-s exposure, a purge returns the signal to the baseline in less than 5 s. Other nitrated organics (dinitrotoluene and dinitrobenzene) and explosives (RDX, urea nitrate, ammonium nitrate, and black powders) cause similar responses. The response by this class of compounds is unique by virtue of both its direction and strength. Of the compounds investigated to date, few types of vapors generate responses in the increasing conductance direction and none in this magnitude of strength. A promising application is in the area of explosives detection.

FIGURE 7
MIME sensor response to TNT vapor illustrating the increase in signal intensity as the TNT source is warmed from 25 to 35°C during five purge-exposure cycles.

MIME SENSOR SYSTEMS DEVELOPMENT AND FUTURE WORK

This research was initiated in 1997 as a collaborative effort between NRL and Microsensor Systems, Inc. Over the years, interfacing electronics, sensor housings, vapor sampling, and transport mechanisms have been under development. An initial prototype is an electronic nose configuration known as VaporLab^TM (Fig. 8). Fig8 Image This is a handheld, battery-powered, rapid vapor identification system initially designed for SAW devices and reconfigured for a MIME sensor prototype. Future applications call for more highly miniaturized systems that can be packaged within the volume of a wrist watch or smaller. The MIME sensor fabrication integrates with planar silicon technology. Future embodiments may see the sensor as a component on a chip that could be incorporated into breathing-air lines, inserted into filters in respirators or into protective clothing for residual life indication, attached to ground scanners for land mine detection, mounted on unmanned aerial vehicles, or deployed as an array of remote drop-off sensors for battle space information. Future basic research will investigate sensor fabrication by chemical self-assembly, reducing the sensor substrate size from the micron to the nanometer size, and investigating the frequency and capacitance of the sensor response.

ACKNOWLEDGMENTS

Support from NRL by way of the sabbatical program for advanced research (AWS) and from Microsensor Systems, Inc., for creation of special laboratory space and facilities access over the critical first year of this research are gratefully acknowledged. Sponsorship over subsequent years by ONR, Microsensor Systems, Inc. CRADA, DuPont Fabrics & Separations Systems, DTRA, and the Joint Service Technology Panel for ChemBio Defense is also gratefully acknowledged. Appreciation for technical guidance/assistance with past and on-going work is expressed to Bruce Yost (DuPont), Mario Ancona (NRL), Ed Foos (NRL), and Richard Smardzewski (US Army ChemBio Center).

[Sponsored by ONR and Microsensor Systems, Inc.]

References
¹ A.W. Snow and H. Wohltjen, "Poly(ethylene maleate)- Cyclopentadiene: A Model Reactive Polymer-Vapor System for Evaluation of a SAW Microsensor," Anal. Chem. 56, 1411- 1416 (1984).
^1\2 B. Abeles, P. Sheng, M.D. Coutts and Y. Arie, "Structural and Electrical Properties of Granular Metal Films," Adv. Phys. 24, 407-461 (1975).