Sensors Magazine Online - Agust 2000 - Bringing Nondispersive IR Spectroscopic Gas Sensors to the Mass Market

There’s a new component architecture for mass-market gas sensors based on photonic bandgap and MEMS fabrication technologies. This combination provides high-quality gas and chemical sensors at consumer prices.

Photo 1. The Kyoto accord will lower the levels for automobile emissions by 2007. Regardless of the engine technology (e.g., gasoline, diesel, or fuel cell), there will be a need for a MEMS IR gas sensor. Such a sensor could also be used to monitor fluids and detect chemicals that indicate the need to either replace fluids or repair the auto.

Two groups of gas sensor technologies are competing in two distinct market segments—affordable but less reliable chemical-reaction sensors for consumer markets and reliable but expensive IR spectroscopic sensors for industrial, laboratory, and medical instrumentation markets. High-volume mass-market applications are limited to carbon monoxide (CO) detectors and onboard automotive emissions sensors (see Photo 1). But because of reliability problems with electrochemical CO detectors, there’s a hesitancy to apply these sensors in other high-volume applications.

Natural gas leak detection, noninvasive blood glucose monitoring, and home and personal air quality monitors need small, efficient, accurate, sensitive, reliable, and inexpensive sensors. Connecting an array of these sensors to wireless networks (which are starting to proliferate) creates many other applications. For example, city emissions regulators could monitor emissions sources from their desks.

A new component architecture for mass-market gas sensors is being developed that combines MEMS technology with nondispersive IR (NDIR) spectroscopy. These new photonic bandgap and MEMS fabrication technologies will simplify the component technology to provide high-quality gas and chemical sensors that can be priced for the consumer market.

IR Gas Sensors

Figure 1. Most gases have a unique IR absorption spectra for conclusive identification and analysis.

NDIR sensors deliver high accuracy and fast response time. And because they don’t require contact with the environment, they’re more reliable. Typically, these sensors detect a specific gas (or set of gases) and are unaffected by other gases.

NDIR gas sensors consist of several discrete components, making them more complex and more expensive than electrochemical sensors. Development efforts for the discrete component IR gas sensors are focused on incrementally reducing the cost, size, and power consumption of the components.

NDIR sensors are based on the fact that most gases have unique IR absorption signatures in the 2–14 µm region. Figure 1 presents the absorption spectra of several of these compounds. Notice that each species has a unique characteristic wavelength. The uniqueness of each gas absorption spectra enables conclusive identification and quantification of chemicals in liquid and gas phase mixtures.

A simple NDIR sensor consists of an IR light source; a sample compartment (of known optical length); an optical filter; and an IR detector, with its associated electronics. The optical filter allows the detector to selectively monitor only the part of the IR spectrum specifically affected by the gas of interest.

As shown in Figure 2, as the gas of interest passes between the powered IR source and the detector, the absorption spectra change along with the detector’s output. The electronics can be designed to detect when the gas reaches a dangerous level and then set off an alarm when the concentration is too high.

The simple single NDIR gas sensor in Figure 2 can be modified to measure multiple gases by adding additional optical filters and detectors, along with the optics for directing the IR energy. Most multigas NDIR-based sensors are limited to three or four spectral channels to maintain good SNR while dividing the light among multiple channels.

For some applications (e.g., combustible gases), there is a partial overlap in the spectral lines. For this situation, a measurement is made of the interfering gas’s concentration at another spectral point. The concentration is then used to adjust the measurement of the gas of concern.

Figure 3. Ion Optic’s current pulsed IR source product line uses foil filaments with a surface texture that improves IR emissions.

Another contributor to measurement error are changes in the level of overall IR energy in the system. These might be due to source illumination variation, light scattering from particulates, or degradation of the optics (typically seen as film buildup). Because the level of IR energy falling on the detector is important to the measurement, many instrument manufacturers measure the IR energy at a reference wavelength that is unaffected by the gas. Then as the IR energy changes, the measurement can be corrected. CO concentration in air, for example, can be reliably determined by measuring attenuation at 4.65 µm and at a reference wavelength of, say, 3.9 µm.

Thermal drift can also contribute to meas urement error. When the IR source runs continuously, it heats up the sample compartment. One way to correct the measurement is to simply measure the ambient temperature and apply an empirically derived correction factor. Another method is to mechanically chop the energy from the IR source using a chopper wheel (or tuning fork). Then a dark reference meas urement can be made when the source’s energy is blocked off. This modulation technique is also useful with pyroelectric detectors because they require a modulation of IR energy to generate a measurement signal.

Chopper wheels and tuning forks can add complexity and increase the size of the sensor. Since 1997, Ion Optics has been making an electrically pulsed IR source (see Figure 3) that allows the NDIR gas sensor designer to eliminate the chopper and its associated electronics. With a thin foil filament, the pulsIR source can achieve >82% intensity modulation (representing 700ºC of temperature modulation) for pulse rates <2 Hz. The company uses ion beams to treat the surface of the foil filaments to improve their IR emission. This treatment allows the IR sources to be run cooler while still providing significant in-band illumination over a desired waveband in the 2–20 µm region. Cooler operation lessens the heating of the sample compartment. Ion Optics has also determined that the output of the IR sources is stable and can be monitored and controlled by electrically monitoring the electrical drive waveform.

An IR Gas Sensor Chip

Figure 4. With the SensorChip, Ion Optics offers a gas sensor that can consist of as few components as a battery, the sensor, and molded plastic optics.

Ion Optics is focusing its efforts on reducing the complexity of the NDIR gas sensor by integrating the optical filter, detector, and IR source on a single chip. The company expects the emitter and detector to each be silicon microbridges. Combining the active elements of the sensor lets the company use inexpensive, molded plastic optics (see Figure 4).

The function of the optical filter will be replaced with surface texturing of the emitter and detector. Using standard lithography practices, Ion Optics will texture the surface with structures similar to photonic bandgap (PBG) designs being used for microwave transmission/reflection filters.

A PBG structure is an artificially engineered periodic dielectric array in which the propagation of electromagnetic waves is governed by band structure-like dispersion [1,2]. When electromagnetic waves with wavelengths on the order of the period of the dielectric array propagate through the structure, the light interacts in a manner analogous to that of electrons in a periodic symmetric array of atoms (or crystal). The structure exhibits allowed and forbidden exten ded states, a reciprocal lattice, Brillouin zones, and Bloch wave functions [1,3]. The photonic bandgap structures, or photonic crystals, have been developed as transmission/reflection filters and low-loss light-bending waveguides and for inhibiting spontaneous emission of light in semiconductors, which can lead to zero-threshold diode lasers [4].


Figure 5. Research from 1985 in photonic bandgap structures showed that the technology can be used to create narrow bandpass filters in the IR region.

In many ways, PBGs are similar to metal grid structures that have been recognized as spectrally selective filters at microwave, millimeter, and far IR wavelengths [5,6,7]. The grids depicted in Figure 5 show metal mesh and crossed dipole structures and their resulting transmission spectra. The transmission curves exhibit a narrow bandpass, which is related to feature size.

There’s also been recent work with lamellar (1D) gratings in silicon that has shown that the IR transmission spectrum of the grating exhibits pass and stop bands, which are related to grating line width and spacing [8].

Figure 6. With narrowband tuned emission and detection,
a SensorChip designed for CO sensing at 4.67 µm can successfully mask
the interference from the nearby CO2 at
4.26 µm.

Most attention given to PBG structures treats them only as filters or waveguides. Ion Optics believes that a PBG structure will also act as a tuned band emitter because according to Kirchoff’s law, absorbance and emittance are equivalent.

The goal is to develop an IR thermal emitter with high emissivity over a narrow band of wavelengths (

~ 0.1) and low emissivity everywhere else. Such sources, emitting hundreds of milliwatts of power in-band, are attractive alternatives to IR light-emitting diodes, which have low quantum efficiencies (~2–4%) and only tens of microwatts output power. Ion Optics chose to make the tuned emitters by making nonrandom patterns (i.e., periodic arrays similar to metal mesh and photonic bandgap filters) on the emitter’s surface. Because it needed wavelengths in the range 2–14 m for gas sensing, this implied sizes as small as 0.5 m, which are readily fabricated by lithography on silicon.

By making the half-width of the IR emission similar to the half-width of the gas absorption peaks (i.e.,

< 0.1), tuned band emission and detection greatly im proves NDIR gas sensor sensitivity. Figure 6 illustrates this by showing how an IR emission peak with

<0.1 at 4.65 m to detect CO will limit the interference of CO₂ nearby at 4.26 m. Ambient air contains about 330 ppm of CO₂, with its fundamental absorption almost eight times stronger than CO. If the emission tuned to 4.65 m overlaps the CO₂ peak, fluctuations in the CO₂ concentration (e.g., from exhaled breath) can cause a false CO reading.

Figure 7. As the tuned emission becomes increasingly narrow, the CO2 masking performance approaches the SNR limit of the instrumentation. A design goal of 0.1 emission width will lead to a
2 ppm CO accuracy in the presence of a CO2 change of 10 ppm.

Figure 7 shows that as emission half-width decreases from 0.5 to 0.1, sensitivity to CO improves from 2.5 ppm to 0.9 ppm (pulsIR at 800K, 100 cm path length). The sensitivity to small changes in CO₂ concentration is greatly reduced as overlap is reduced.

Ion Optics selected patterns based on the work of Byrne et al. [8,9] because of the relatively simple relation given and demonstrated between feature size and transmission wavelength. Specifically, the company used the crossed dipole pattern depicted in Figure 5 and built a variety of feature sizes and spacings.

Dupont made the photomasks using direct-write e-beam lithography. The etching and lithography were performed at the Microdevices Laboratory at the Jet Propulsion Laboratory. The basic fabrication sequence was as follows: deposit photoresist; expose pattern on resist; reactive ion etch (RIE) for 5, 10, or 15 min. to form the cross-shaped cavities of different depths with straight side walls; and remove photoresist.


Figure 8. Shown here are scanning electron microscope photos of two photonic bandgap patterns—10 narrow and 10 broad—used in initial experiments on emission tuning.

Five n– and two n+ wafers were processed. One n– and one n+ wafer were coated with 500A chromium (for adhesion), followed by 1000A gold, before lifting off with the removal of the photoresist, leaving gold coating at the bottom of the etched cavities. In another case, an n– wafer was coated with aluminum to suppress the background emissivity. In this case, metal covered the un etched regions as well as the bottom of the cavities.

Chemical Reaction Gas Sensors

Chemical reaction sensors consist of two different technologies: electrochemical and catalytic. Most suppliers offer many sensors, each designed to detect a specific gas, all having the same pin-out to allow for plug-and-play capability.

Electrochemical sensors measure a change in output voltage of the sensing element caused by chemical interaction of the analyte on the sensing element. These sensors require little power and are small, sensitive, and inexpensive. And their response time is in the tens of seconds.

In some designs, the electrolyte material must be refilled (or the device replaced) at regular intervals. But the major drawback of electrochemical sensors is that the reactant material must come in contact with the environment, creating the possibility of a response to other gases in the sample [9]. In addition, if the sensor sees too much of the gas of interest, it can become saturated, causing a shift in the zero-point and requiring recalibration. Finally, temperature and humidity can also cause drift or false alarms.

Despite these drawbacks, electrochemical sensors are used in many applications, in particular, in handheld toxic and combustible gas detectors for industrial health and safety applications. For these applications, simple reconfiguration and easy replacement are compatible with the regular calibration of the products. Electrochemical sensor manufacturers continue to work on improving the specificity and reliability of these devices.

Catalytic sensors measure a change in resistance caused by an oxidation change on the surface of the sensor. While these sensors are inexpensive, small, and relatively sensitive, they are also less selective (and thus are capable of generating false alarms when responding to gases other than the one of concern). To increase the rate of oxidation and thereby sensitivity, catalytic sensors heat the sensor surface, thus they require more power than a battery typically provides. The sensors must be recalibrated and usually need to be replaced after one or two years.

The false alarms of electrochemical and catalytic sensors have hurt the reputation of CO detector products with recent product recalls of over 1 million units in early 1999. The Gas Research Institute estimates that more than 80% of emergency calls triggered by carbon monoxide sensors are false alarms [9].

Incremental improvements to existing electrochemical and catalytic sensors will doubtless improve both their performance and their public acceptance over the next few years. But these units will always be faced with the fundamental limitations of the technology—consumable sensor materials and limited adaptability of chemical reactions for sensing different species.

Scanning electron microscopy (SEM) confirmed etched cavity depth and verified that the side walls were straight. Reflectance spectra were measured at room temperature over a range of 2.5–25 µm using a MIDAC FTIR spectrometer (0.5 cm^–1 resolution) with a surface reflectance attachment that set the angle of incidence at 45°.

Figure 9. The photonic bandgap patterns with broad line widths show strong emissions peaks with center frequencies dependent on the feature size.

Thermal emittance measurements and additional room temperature reflectance measurements at near normal incidence were made over a range of 3–11 µm using a chalcogenide fiber-coupled MIDAC FTIR spectrometer with MCT detector (0.4 cm^–1 resolution). During the emittance measurements, samples were placed on top of a thin (~1 mm) graphite sheet on a hot plate or heated sample stage and confirmed to be at 300ºC, 400ºC, or 500ºC, with a Type K thermocouple.

SEM measurements (see Figure 8) confirmed that the etch depth was ~2.2 m for samples etched 5 min., about 6.2 µm for samples etched 10 min., and about 9.4 µm for samples etched for 15 min. Actual depths varied ±5%, depending on the width of the structure. For example, features with the N designation (for narrow line width) were generally etched to lesser depths than those designated B (for broad line width).

Thermal emittance measurements showed enhanced emissivity for wavelengths in the range of 5–11 µm (long wavelength limit established by the fiber-coupled FTIR), depending on feature size. For features with narrow line width (labeled N), the enhancement is slight, but for the larger features (labeled B), the enhancement is significant. The company then had controllable band emission with tuning over nearly an octave in the mid-IR spectrum. The relation between feature size and emittance peak is further illustrated in Figure 9.

The data in Figure 10 from a PBG surface structure demonstrate that thermal radiation from a designed textured surface can be concentrated in a narrow band with low values of

The data suggest that the sensitivity of the detector can reach the theoretical limits outlined in Figure 7.

Conclusion

Figure 10. There is a linear correlation between center-to-center pattern spacing and peak emission wavelengths that shows promise for Ion Optics’ development of a MEMS-based IR Gas SensorChip.

Experiments at Ion Optics show early success in being able to texture the surface to create tuned band emission and tuned band detection. By measuring the emission and reflectance of several patterned silicon surfaces, Ion Optics has determined that the peak absorbance wavelength and line width correlate with feature size and spacing, as well as with surface conductivity.

The tuned band emitter development was done under National Science Foundation Small Business Innovation Research (SBIR) grant # DMI-9860975 and National Institute of Standards and Technology (NIST) Advanced Technology Program (ATP) contract # ATP-99-01-2051.

9. S. Akbar, C-C Wang, and L. Wang. June 1996. “Ceramic Materials May Revolutionize Automotive Emissions Control,” Ceramic Industry:32.

Brian R. Kinkade is Vice President of Mar keting and Sales, Ion Optics, Inc., 411 Waverley Oaks Rd., Ste. 144, Waltham, MA 02454; 781-788-8777, x-103, fax 781-788-8811,
bkinkade@ion-optics.com.