(This article is sponsored by The Boston Group)

There is a growing need for detection of hazardous substances (toxic gases, chemicals) in a hostile environment for security and industrial applications. Most current technology for such purpose use optical techniques which are expensive, bulky and need cumbersome manual operation. Hence they are not suitable for use in most hostile environments. Solid State sensors offer a convenient alternative due to their low-cost of manufacturing, portability, and low power consumption. CMOS technology in particular enables integration of the physical sensor, and circuitry for measurement and analysis on the same chip or a package, thereby lowering cost through mass production. Currently, solid-state sensors are heavily used in the industrial and automotive markets.
Such a sensor system consists of an improved solid-state physical sensor and a highly sensitive measurement circuitry built on chip.

Physical Sensor Design

Majority of current sensors have an active film (such as oxide, polymer or metal) coated on the surface. Sensing is realized by monitoring the change of physical and chemical property of the coated films, such as dielectric constant, conductivity, charge, work function or surface potential, upon the adsorption of target molecules by the film. The concentration of the trapped chemical (also called analytes) is measured by monitoring the intensity of the output voltage or current. To achieve high selectivity, the film needs to be prepared with specific receptors that respond only to certain types of analytes. Some of them are listed below:

1) ISFET (Ion Field Effect Transistor):  Unlike MOSFET, the gate of the ISFET is a membrane. The membrane causes a change in the oxide/solution interface potential at exposure to targeted ions. These potential is detected as a change in the drain current of the ISFET. By comparing the drain current of this device to that of a reference FET, which does not react to ions, the property of the solution is determined. Many materials IC manufacturing compatible materials like SiO2, Si3N4, and others have been used for membrane design. 

2) Enzyme FET (ENFET): These sensors operate on principles similar to ISFET. ENFET is a pH sensitive ISFET with a layer of enzyme immobilized on the surface. When contaminants come in contact with the enzyme, they create gases and chemicals that are measured by the ISFET, thus providing a method to detect indirectly (Example, Urea sensor uses urease as an enzyme). In case of ImmunoFET, a layer of antibody is immobilized on the surface to detect a particular antigen. The antigen-antibody reaction alters the charge concentration in the inversion layer of the FET and measured indirectly through the drain current. However such sensors are not reliable due to difficulty of accurate operation due to interference and limited stability of having a membrane on the gate.

3) Semiconductor gas and chemical sensors: These sensors work on the conductivity, capacitance or work function change in the MOSFET that utilizes the target molecules as the oxide or conducting material. One example is the change in the threshold voltage of the transistor due to the trapped analytes, which cause change in the drain current. Metals, metal oxides, salts, polymers have all been investigated as the sensitive material and the operating regime is anywhere from 0 to 200 degrees (example, semi conducting metal oxides like SnO2, TiO2, or porous silicon). These sensors are also popularly called as Taguchi sensors. Other solid-state sensors use either Schottky diodes or FET for gas detection working on the principle of work function measurement.

One major problem with the above sensors is that in an unfamiliar hostile environment, we will need different sensors coated with different materials to cover the entire gamut of hazardous chemicals to be detected, which makes it highly unpractical for implementation. Another main problem with the current solid state sensors is reversibility. Most sensors degrade over time and cannot be reused due to continuous interaction of the target analytes with the sensitive sensor surface. Moreover, selectivity of these sensors is limited since many analytes exhibit similar change in the drain current. And the third problem is that they are highly unsuitable for single particle or molecule detection because the change in the drain current due to adsorption of a single particle results in undetectable change in the drain current.  New research ideas are therefore necessary to create a highly selective and sensitive sensor device.

Measurement Techniques:

One way of improving sensitivity was shown in fig 1, where we use a distributed sensor network to detect ppm and sub-ppm concentrations of the chemicals in the atmosphere. At the same time, we need advanced measurement techniques to be able to distinguish different analytes using a single or few sensors.  Most measurement techniques from these sensors focus on the average and the rms values of the sensor output for detection. However, in sparse environment with sub-ppm concentrations of targeted molecules, these values will be below the noise level and go undetected or may cause false alarms. Moreover we also need a mechanism to increase the applicability of the sensors to recognize more analytes for which sensors have a very partial but not a null response. A new approach to enable such detection is to use the combination of following two techniques:

1)    Higher order spectral analysis allows for high fidelity detection in such sparse environments. Each analyte has a unique set of parameters that can be used as a signature for its detection.

2)    Broadband measurements allow for excitation of the sensor at frequencies different than the typical optical frequencies. Most particles exhibit different polarization effects due to atomic, electronic, molecular, ionic and van der waals interaction. All these interactions are frequency dependent and can be used as signatures for detection of the particles.

 CMOS technology can be used as a backbone for implementations of such measurement schemes. Circuits operating at frequencies as high as 20GHz can now be easily implemented using CMOS technology enabling broadband measurements in that range.  CPU and DSPs are becoming cheaper and are operating at low power making them feasible for integration with the sensor mote to perform higher order statistical analysis on chip. Reducing feature size and increasing scalability of the CMOS designs enable covert deployment of such sensors for remote analysis and detection.

In the future, we anticipate wide-spread use of such solid-state sensors built on compatible CMOS technology for high fidelity measurement, analysis and detection of gases and chemicals for all applications, ranging from industrial, environmental, automotive and security.

|