B. Rogers*, R. Whitten, J.D. Adams
Nevada Nanotech Systems, Inc., 661 Sierra Rose Drive, Reno, NV, USA 89511
Unattended sensing applications necessitate robust, compact, low-cost, low-power sensor units. The microcantilever- based Self-Sensing Array (SSA) technology developed by Nevada Nanotech Systems, Inc. (NNTS) is a strong candidate for such units. SSA technology is expected to provide the selectivity, sensitivity, durability, low cost, and low power needed for unattended sensors and sensor networks. The sensor employs a variety of sensor coatings and the ability to analyze the electrical and thermal properties of molecules on the cantilevers. This so-called Lab-on-a-TipTM technology could lead to enhanced chemical identification capabilities of the trace detection platform.
Keywords: microcantilever, array, unattended, multidimensional, sensor
The broad spectrum of substances and physical quantities detectable with microcantilevers includes chemical and biological materials, radiation, heat, magnetic susceptibility, force, and attogram mass changes. These sensors operate by detecting changes in resonance response or deflection caused by mass loading, surface stress, or changes in damping conditions. Often this is accomplished by functionalizing cantilevers with specialized coatings. For instance, coated microcantilevers have been used to detect plastic explosives (PETN and RDX) at low parts-per-trillion concentrations in air.1 In addition to sensitivity and versatility, they also have advantages in compactness, cost, and scalability. These advantages could be beneficial in environmental, industrial, medical and homeland security trace detection applications.
We have designed, built and tested a prototype microcantilever sensor system for measuring the concentrations of unlawful or hazardous materials. The main tasks were building and testing a prototype detector against several potential threat agents. There are three categories of material this system is designed to detect—chemical, biological and explosive. To test the effectiveness of the system, we have tested seven chemicals—including toxic industrial chemicals ammonium hydroxide, toluene diisocyanate (2,4), formaldehyde, and allyl alcohol. We also used microcantilever sensors to identify Serratia marcescens, a biological agent analog to the plague, from among other bacteria. Finally, we detected three explosive vapors: trinitrotoluene (TNT), pentaerythritol tetranitrate (PETN) and cyclotrimethylene trinitramine (RDX). With the completion of these demonstrations, we find this technology to be promising for unattended sensing applications, some of which are shown schematically in Figure 1.
The advantages of microcantilevers derive from their minute size, low-power operation, and simplicity. However the systems required to operate microcantilevers have in past implementations been bulky, power-hungry and complex. The various schemes used to monitor cantilever motion include external laser detection,2,3 integrated piezoresistive4,5 detection, and integrated capacitive sensing.6 Laser systems are large, burdensome to align, and limit the number of cantilevers that can be monitored.7 Piezoresistive systems require many electrical connections for large arrays. Capacitive systems rely on a very small sensing gap that can present fabrication, operation, and coating difficulties.
Piezoelectric8,9,10 cantilevers do not require external optics or actuators and have the inherent strength of self- sensing and integrated actuation, meaning that the actuation signal can also be monitored as a sensor signal, and each element can be actuated independently and directly. The self-sensing method therefore replaces the cumbersome systems conventionally used to monitor cantilever deflection and enables compact, simple, and scalable cantilever sensing applications. Sensor element power consumption is measured in nanowatts.11
Chemical selectivity advantages can be achieved using cantilever arrays. Arrays provide more useful and redundant information due to the multiplicity of individual cantilever responses and the ability to selectively coat cantilevers and create differential signals.12,13,14 SSAs containing microcantilever sensors can be configured electrically into parallel and series groups. This means that the electrical input and output leads of the array will be the only two contacts needed to address the array. Figure 2 shows the SSA in a flow cell, an SEM image of an SSA, and data generated using an earlier prototype array for driving and sensing purposes. Small changes in fabricated length from one element to the next allow separation of resonance frequencies for use in a frequency-sweep method.
The current NNTS prototype utilizes a signal generator to sweep an array using only two leads. The benchtop prototype system, including the IC on the custom PCB, is shown in Figure 3. A schematic representation of the system design is shown in Figure 4.
The sample handling sub-system design shown within the dashed box in Figure 4 includes sample acquisition, filtering, preconcentration, sample chamber, and a fan pump to move sample through the system. The preconcentrator is discussed in the next section. The sample acquisition design includes a tubular air intake to the preconcentrator with a wire-mesh-covered inlet to filter out debris. Sample is collected in the preconcentrator. When thermally desorbed, this sample passes down a pipe approximately 5 cm from the outlet of the preconcentrator into the sample chamber.
The inner dimensions of this chamber are approximately 5 mm × 5 mm × 2 mm. Air can be routed directly through the silicon chip containing the SSAs. The volume of this sample chamber is two orders of magnitude smaller than the previous design (see Figure 5). This reduction in volume translates to significantly faster sensor “rise-time” because, with the prior chamber design it took time for low-flow-rate, low-concentration vapor to fill the chamber and create an equilibrium concentration to which the sensor responds. Compare the data shown in Figure 6. One curve shows data collected by a polymer-coated SSA in the reduced volume flow cell; the other curve shows data from a prior cantilever design, also coated with the same polymer, in the larger Plexiglas chamber.
In an attempt to develop an acceptable method of comparing the false alarm rates (and selectivity) of various chemical analyzers, Bonne et al. 15 developed the idea of orthogonal channel capacity (OCC). OCC is analogous to the peak capacity of a gas chromatograph (GC) or mass spectrometer (MS) and represents the number of linearly independent measurements, n, that any instrument, not just a GC or MS, is capable of making. The more orthogonal, or linearly independent, measurements an instrument is capable of making, the better the expected selectivity and the lower the expected false positive rate.
To see how various analyzers compare using this metric, first consider that the peak capacity of a GC can be estimated at 200 and the peak capacity of a MS can be estimated at 300 (see Ref. 15). This would give a total OCC for a GC-MS of (200)(300) = 60,000. For comparison, an array of polymer coated surface acoustic wave (SAW) devices or standard microcantilevers would have an OCC of approximately 5 (due to the number of orthogonal polymer analyte interactions), a micro-GC with a nonselective detector such as a TCD† or FID‡ would have an OCC of approximately 50 (less peaks for the micro-GC than the standard GC), and a MGC-MGC-CID-MDID-IT-MS§ would have an OCC of approximately 180,000.
Bonne et al. don’t include the common ion mobility spectrometer (IMS) in their comparison table, but they do estimate the OCC for an IMS to be 5-10, and a National Academy of Sciences report on airport passenger screening,16 using a similar but different methodology, claims that the IMS has 10 times less informing power** than a MS. This would map to an OCC of approximately 30 for an IMS without a preseparation stage. Combining the two reports gives a range of 5-30 for the OCC of an IMS detector, so for the purpose of this discussion, let’s assume that the OCC of an IMS analyzer is 20. Based on all possible sensing materials, transducers, geometries, modulation parameters and parameter shapes, Gopel17 estimates that the chemical sensor feature space could ultimately reach 1021 features. To improve upon an OCC of 180,000, it would therefore be necessary to build an instrument having approximately 2×105 orthogonal features, a fraction of the total number possible.
Table 1 shows OCC estimates for the NNTS patent pending†† LT-SSA analyzer components. Note that additional channel capacity could be possible from certain components, including capacitive and thermogravimetric measurements. The Thermally Desorbed Preconcentrator Tube is paired with the LT-SSA on this list so as to compare with other instruments that can be paired with a GC. Reported values take into account measurement time constraints, the fact that multiple measurements are being made with a single coating type, and the fact that some coatings may not be applicable to every type of measurement.
The coatings are selected so as to provide as many orthogonal channels as possible—which is to say that each coating has a somewhat unique chemical affinity for certain chemicals (for example, acidic, dipolar, and nonpolar), depending on fundamental interactions between the vapor and the coating (for example, van der Waals interactions, polarity, and Lewis acidity). By using an array of semi-selective polymer coatings and observing the unique pattern of sensor responses corresponding to a particular chemical vapor, selectivity can be enhanced. The interaction of polymer coatings and vapors can be characterized by seven diverse types of interactions: 1) dispersion, 2) polarizability, 3) dipolarity (minimize basicity), 4) dipolarity allowing basicity, 5) basicity minimizing dipolarity, 6) basicity and dipolarity, and 7) hydrogen bond acidity.18 These seven interactions are approximated as 5 orthogonal channels. In addition to polymer coatings, size selective coatings such as zeolites can be applied as well as metal, metal oxide and/or other coatings. For example, a typical array of conductivity sensors includes 8 different metal oxide coatings.19 Other metal coatings include gold and palladium, which have notable binding relationships with mercury and hydrogen, respectively.20 The size selective coating option is approximated as 5 orthogonal channels and an additional 5 channels are approximated for metal, metal oxide and other coatings.
In addition to coatings, the LT-SSA is being developed to take additional orthogonal measurements demonstrated by others previously such as capacitive,21 resistive,22 calorimetric,23 and thermogravimetric24 measurements. In some applications it is expected that measurement time constraints, and the common coatings will limit the channel capacity of these measurements to 3, 2, 2, and 2, respectively. (See Table 1 for channel descriptions.) However it is expected to be possible to increase the channel capacity of these measurements in some cases.
Figure 7 shows an early demonstration of adding impedance and thermogravimetric information to a SSA measurement. The test involved exposing a two-cantilever array (one coated cantilever and one uncoated cantilever) to TNT, ammonia, and acetone. The resonance frequency and impedance were measured during and after exposure and the resonance frequency information from just the uncoated cantilever was used in order to highlight the selectivity added from heat- and impedance-related information. The analyte in all cases absorbs onto the cantilever, causing a resonance shift, then desorbs. During desorption, a short voltage pulse was applied to a piezoresistive heater embedded in the cantilever – performing a simple thermogravimetric analysis. The amount of analyte removed during this heated, accelerated desorption sequence added more separation to the data since the more volatile compounds desorbed much faster than the TNT. It should be noted that this is only a preliminary test and observation of the desorption differences without the use of a heated cantilever could have provided similar results. We imagine that the heated cantilever will allow quicker desorption analysis and provide the potential to look for different melting temperatures between compounds or even an exothermic reaction. Use of the impedance change between sensor electrodes further separated the data and made it possible to distinguish the more volatile analytes.
In addition to the measurements listed above, it is anticipated that the LT-SSA product will use a thermally desorbed preconcentrator tube (TDPCT)25 as both a preconcentrator and a preseparation stage. It is expected that the TDPCT will have a peak capacity of approximately 10, creating 10 orthogonal channels, and while lower than the 50 or 200 channels available from a micro-GC or GC, respectively, could provide cost, size and speed advantages over both types of GC. With that said, if greater selectivity is required, it is conceivable that the LT-SSA could be used in conjunction with a GC.
Combining all of the measurement channels of the proposed LT-SSA, it is expected that the total OCC will eventually be 360 as shown below in Table 2, along with other chemical analyzers for comparison. In conjunction with the TDPCT, the OCC will be 3600, as shown in Table 1 previously, as well as in Table 2 below. It should be noted that the LT-SSA alone is expected to have a higher OCC than both the SAW Array and the IMS analyzer as well as the MS analyzer. The LT-SSA with TDPCT would not have as high an OCC as a GC-MS, but a GC-LT-SSA would have a higher OCC than a GC-MS and it could be possible in the future to develop an advanced LT-SSA that could have a higher OCC than a GC-MS.
The authors thank Ulrich Bonne for helpful discussions and the Defense Advanced Research Projects Agency (DARPA) for supporting this work (Contract No. W31P4Q-06-C-0099).