A Novel Lightweight Low-Power Dual-Beam Ozone Photometer Utilizing Solid-State Optoelectronics

Lars E. Kalnajs Laboratory for Atmospheric and Space Physics, University of Colorado, Boulder, Colorado

Search for other papers by Lars E. Kalnajs in
Current site
Google Scholar
PubMed
Close
and
Linnea M. Avallone Laboratory for Atmospheric and Space Physics, University of Colorado, Boulder, Colorado

Search for other papers by Linnea M. Avallone in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Recent advances in semiconductor materials and fabrication techniques have allowed the development of light-emitting diodes (LEDs) with wavelengths extending down into the UV-C region (λ < 280 nm). A new ozone photometer has been developed utilizing these novel light sources. The application of solid-state technology to the proven dual-beam UV absorption technique has improved instrument performance while reducing power consumption and weight compared to existing instrumentation. The newly developed instrument is expected to have an accuracy of 1% at surface level pressure, a resolution better than 1 ppb, and measurement rates up to 1 Hz over the range of ozone mixing ratios encountered from the earth’s surface to the middle stratosphere. Size, weight, and power consumption have also been significantly reduced, with a mass of 3 kg and a power consumption of less than 5 W. Initial development is focused on an instrument suitable for measurements from autonomous platforms and in harsh environments; however, the technology is highly adaptable to other applications.

Corresponding author address: Lars E. Kalnajs, Laboratory for Atmospheric and Space Physics, University of Colorado, 1234 Innovation Dr., Boulder, CO 80303. Email: kalnajs@colorado.edu

Abstract

Recent advances in semiconductor materials and fabrication techniques have allowed the development of light-emitting diodes (LEDs) with wavelengths extending down into the UV-C region (λ < 280 nm). A new ozone photometer has been developed utilizing these novel light sources. The application of solid-state technology to the proven dual-beam UV absorption technique has improved instrument performance while reducing power consumption and weight compared to existing instrumentation. The newly developed instrument is expected to have an accuracy of 1% at surface level pressure, a resolution better than 1 ppb, and measurement rates up to 1 Hz over the range of ozone mixing ratios encountered from the earth’s surface to the middle stratosphere. Size, weight, and power consumption have also been significantly reduced, with a mass of 3 kg and a power consumption of less than 5 W. Initial development is focused on an instrument suitable for measurements from autonomous platforms and in harsh environments; however, the technology is highly adaptable to other applications.

Corresponding author address: Lars E. Kalnajs, Laboratory for Atmospheric and Space Physics, University of Colorado, 1234 Innovation Dr., Boulder, CO 80303. Email: kalnajs@colorado.edu

1. Introduction

Ozone is one of the most important gases present in our environment. It is critical to all life as a filter of harmful solar ultraviolet (UV) light in the stratosphere and it is a significant component of anthropogenic pollution and a health hazard in the troposphere. Accurate and reliable measurements of ozone concentrations are vital to understanding the protective layer of stratospheric ozone, diagnosing its depletion, and monitoring its subsequent recovery. Ozone is also the benchmark species for quantifying urban air pollution, and accurate measurements of ozone levels related to anthropogenic pollution are critical from both regulatory and health perspectives.

Although there are several techniques and technologies in existence for the measurement of ozone, there is no single method that is suitable for all measurement locations and requirements. In particular, there is no single instrument capable of making high-resolution measurements over a long time scale (months or longer) in a lightweight (under 5 kg) and low-power (under 5 W) form factor. The instrument described herein is designed to meet the requirements for long-duration autonomous ozone monitoring on a remote and battery- or renewable energy–powered platform. The intended application is long-duration stratospheric ballooning; however, the design is equally applicable to other platforms such as uncrewed aerial vehicles (UAVs) or to ground-based autonomous measurement stations. Additional design features allow the instrument to be adaptable to situations that require a response rate as fast as 10 Hz.

a. Background

Because of its highly reactive and unstable nature and large temporal and spatial variability, ozone presents a significant measurement challenge. There are three commonly used techniques for in situ detection and quantification of atmospheric ozone: detection of chemiluminescent radiation resulting from the reaction of ozone with nitric oxide (or similar chemiluminescent processes); measurement of the current produced from the reaction of ozone with iodide in a liquid potassium iodide (KI) concentration cell; and absorption of the ultraviolet mercury-vapor (Hg-vapor) line at 254 nm. Each method has a particular set of advantages and drawbacks for the determination of ozone.

Chemiluminescence is typically used when there is a requirement for high temporal resolution (e.g., Eastman and Stedman 1977). There are two chemiluminescent reactions that are used for the determination of ozone: the reaction between ozone and nitric oxide (NO), which produces an electronically excited state of NO2 that emits light in the infrared (830 nm), and the reaction between ozone and ethylene, which produces photons near 400 nm. The latter is the U.S. federal reference method for the determination of ozone (Title 40, CFR Part 50, Appendix D; available online at http://ecfr.gpoaccess.gov). Although these and other chemiluminescent techniques offer fast time response and high sensitivity, they have the significant drawback that they require an expendable and toxic reagent and, in the case of the nitric oxide technique, a cryogen to cool the infrared detector. In addition, chemiluminescent techniques do not offer an absolute measurement of ozone and require careful and frequent laboratory calibration to obtain accurate results.

Studies of stratospheric ozone levels are generally performed with the potassium iodide concentration cell, which is commonly referred to as the electrochemical concentration cell (ECC; Komhyr 1967), which is similar in principle to the Brewer–Mast instrument (Brewer and Milford 1960). The ECC technique introduces ambient air into a potassium iodide solution that forms half of an electrochemical cell. The other half of the cell contains a reference potassium iodide solution of known concentration. Ozone introduced into the sample cell increases the concentration of free iodine (I2) and induces an electrical potential between the cells. This potential is proportional to the amount of ozone introduced to the sample cell. The primary advantage of the ECC technique is the low weight (∼600 g), minimal power consumption (∼2 W), and low cost (approximately $600 U.S. as of 2009) of the instrumentation. It is the technique of choice for stratospheric soundings because the instrument can be launched on a small helium balloon and does not need to be recovered. The drawback to the ECC technique is the use of a liquid reagent that is prone to freezing and evaporation and the need for premeasurement calibration. Its use is largely limited to atmospheric soundings with durations of 2–4 h.

The most common technique for routine ozone monitoring, which is the one applied here, is ultraviolet absorption. This technique is a direct application of the Beer–Lambert law [Eq. (1)], where the ozone number density n can be determined from the change in transmitted light I/Io, the pathlength l, and the wavelength-specific ozone absorption cross section σ at 254 nm. The UV light source used in almost all instruments of this type up to now has been the low-pressure mercury-vapor lamp that has an atomic emission line at 253.7 nm. The ozone absorption cross section at this wavelength has been extensively researched in laboratory studies and is well understood at room temperature (Rinsland et al. 2003) as well as over a significant range of atmospherically relevant temperatures (Malicet et al. 1995; Voigt et al. 2001). Early implementations of this technique used a single absorption cell containing the air to be analyzed and a solar-blind UV detector at the opposite end of the cell to measure UV transmission (an example of this design is the 1003-AH by Dasibi Environmental Corporation). The cell alternately contains sample air and ozone-free air. In this configuration, the maximum measurement rate is twice the time taken to switch from ambient to ozone-free air, which in practice is 20–30 s.

A refinement of this method is the dual-cell technique that is implemented in the Thermo Environmental Instruments model 49 (TEI 49) and similar instruments. These instruments use two absorption cells; at any given instant, one cell contains air scrubbed of ozone and the other contains ambient air. The cells are illuminated by the same Hg-vapor lamp and the impact of variations in lamp intensity (which can be significant) can be minimized by taking the ratio of the intensities measured at the ends of the two cells. Ozone-free air is generated internal to the instrument from ambient air that is passed through an ozone-specific scrubber, commonly MnO2. The use of an ozone-specific scrubber is advantageous because any species besides ozone that has a significant absorption cross section at 254 nm is not removed by the scrubber and will appear in both cells, thus not affecting the ratio of transmission through the two cells. By alternating which cell receives scrubbed air and which receives sample air, the impact of variations in detector response and lamp emission patterns can be minimized. Typically, the maximum measurement rate is equal to the time required to alternate cells, which in practice is on the order of 10 s. Dual-cell UV absorption is the most commonly used technique for environmental and air quality ozone monitoring and is an approved U.S. federal equivalent method (Title 40, CFR Part 53) for the determination of ozone. The dual-cell absorption method is also commonly used as transfer standard; a detailed discussion of the sources of error and the relative uncertainty associated with the method can be found in Viallon et al. (2006).

The main drawbacks to this technique arise from the use of Hg-vapor lamps. There is a significant power and weight penalty to thermally stabilize the optical bench and to generate the high voltages necessary to produce plasma in the Hg-vapor lamp. Additionally, because of the nature of the plasma lamp, there can be significant variability in lamp output both in terms of overall intensity and the emission geometry, which can affect the ratio between the cells:
i1520-0426-27-5-869-e1

b. Motivation and design goals

Although the gamut of distinct requirements for specific ozone measurements has largely been covered by one or more of the measurement techniques and technologies discussed above, there is a pressing need to address a subset of these requirements with a single instrument. For many applications, an ideal instrument would be capable of making fast, accurate, and precise measurements autonomously under nonideal conditions for extended periods of time while consuming minimal power in a compact and lightweight form factor. Only some of these requirements can currently be met by any existing instrument at any one time. The motivation to develop a new UV absorption ozone instrument is to achieve a larger subset of these design goals and to produce a versatile instrument with operating parameters that can be reconfigured to suit a variety of applications.

The specific application that this instrument is designed to meet is long-duration ballooning in the Antarctic stratosphere. The critical parameters for this application are low mass, low power consumption, high reliability, and good precision. Instrument mass directly affects the size and cost of the balloon as well as the maximum achievable altitude. Particularly for operations in the wintertime polar stratosphere, power is limited by the amount of available solar radiation for photovoltaic collectors. Finally, the instrument must be able to function under the harsh conditions of the polar stratosphere for periods of up to six months, necessitating a robust design as well as redundancy in case of a single component failure.

The chosen design is similar to the dual-cell UV absorption technique of Proffitt and McLaughlin (1983), which is also used in the widely available TEI 49 ozone analyzer and subsequent models (Thermo Scientific Corporation, Franklin, Massachusetts). The UV absorption technique was chosen because it is an absolute measurement that requires infrequent calibration during operation. Significant changes have been made to the basic design to increase reliability and to decrease weight and power consumption while maintaining or improving the accuracy, precision, and time response over existing UV absorption instruments. One of the most significant changes is replacing the single low-pressure mercury-vapor UV light source with a pair of solid-state UV light-emitting diodes (LEDs) with the same center wavelength. The flow path has been redesigned to improve reliability and provide complete redundancy. The phototube UV detectors and detector optics have been redesigned around silicon carbide (SiC) photodiodes to provide greater resolution in a smaller and more robust form factor. The electronics and control system has been updated with higher-speed and higher-resolution data acquisition and field-reconfigurable software-based instrument diagnostics and control.

2. Instrument description

a. Light source

The most novel and significant change to the dual-beam UV absorption ozone photometer described previously (Proffitt and McLaughlin 1983) is the substitution of two solid-state LEDs in place of the customary Hg-vapor lamp. Recent advances in materials science have facilitated the reliable production of light-emitting and light-detecting devices in the near-UV part of the spectrum based on wide band gap group III nitride materials (AlGaN; Anceau et al. 2005). Among these devices are LEDs with emission wavelengths as short as 250 nm. Two LEDs with a peak emission at 254 nm are substituted for the traditional Hg-vapor lamp to illuminate the sample and reference cells (see Fig. 1). Each LED has an integrated feedback photodiode and thermistor within the sealed LED package and is a variant of the UVTOP-255 UV LED produced by Sensor Electronic Technology (Columbia, South Carolina). The feedback diode is operated in photovoltaic mode to give constant monitoring of light output without the need for external beam splitters or additional optics. This integrated photodiode provides a reference value that can be used to correct for variability in light intensity, which is one of the major limitations in Hg-vapor lamp–based instruments. The output of the photodiode is immediately amplified by an ultra-low-noise difet instrumentation amplifier [operational amplifier 129 (OPA129)] enclosed with each UV LED in a shielded housing. The UV LED package also includes a UV-transmissive hemispherical lens, which focuses the output into a beam with a half-maximum width of 5°. This directed output dramatically increases the effective usable photon flux of the light source when compared to the essentially isotropic radiation pattern of a Hg-vapor lamp.

It is difficult to meaningfully compare the power output of Hg-vapor lamps and UV LEDs because of the differences in spectral widths and radiation geometries. A more practical comparison is between the relative output intensities observed at the opposite end of the absorption cell as viewed with an appropriate UV detector, which is a SiC photodiode in this case. The effective intensity from a single LED driven with a nominal continuous wave (CW) current of 20 mA is approximately 600 nW mm−2 compared to the effective intensity at the detector from a standard Hg-vapor lamp and the optical assembly from a TEI 49C of 50 nW mm−2. Short-term lamp variability is a major limiting factor for the determination of ozone at a rate faster than the cell switching rate. The LED has a 1σ noise level of 50 ppm at 10 Hz over a period of 100 s (Fig. 2) compared to ∼500 ppm for a typical Hg-vapor lamp.

The spectral power distribution of four sample LEDs is shown in Fig. 3. The spectral full-width half-maximums of the emission peaks range from 10.8 to 11.9 nm, with the peak emission wavelength between 253.8 and 254.4 nm. Examination of the absorption cross section of various gases that are present at significant abundances in the stratosphere has shown none with an absorption feature that would fall in the 240–260-nm range without appearing at the Hg emission line at 253.7 nm. Thus, the broader emission wavelength of the UV LEDs compared to the narrow Hg-vapor line does not add any significant potential for interference.

It should also be noted that the ozone absorption cross section is consistently large around 254 nm, changing by less than 10% between 244 and 264 nm. However, the absorption cross section needs to be treated somewhat differently than with the Hg-vapor lamp, which produces a narrow bandwidth atomic emission line. The effective ozone absorption cross section is the integrated product of the spectral emission of the UV LED with the ozone absorption cross section as a function of wavelength. The product of the normalized emission from the four LEDs in Fig. 3 and the ozone cross section at 293 K (Burrows et al. 1998) is shown in Fig. 4. It should be noted that the variability in the integrated effective cross section for the four different devices falls within 0.5% of the mean, and thus device-to-device variability is not a significant source of uncertainty in the effective ozone cross section.

Because of the large expected variation in operating temperature for the instrument and the lack of thermal control, the temperature dependence of the ozone absorption cross section must also be considered. There are several published empirical models for the temperature dependence of the ozone absorption cross section (Bass and Paur 1981; Burrows et al. 1998; Voigt et al. 2001). These models strive to capture the variation in the ozone cross-section continuum from 240 to 790 nm. By comparing the product of these models and the average UV LED emission spectrum with the high-resolution temperature-dependent data of Burrows et al. (1998), it was determined that a simple linear fit described the temperature variation of the effective ozone cross section from 203 to 293 K over the relatively small range of relevant wavelengths. The linear fit represents the temperature dependence of the integrated effective absorption cross section to better than 0.2% from 203 to 293 K. This represents a smaller level of uncertainty than that of the original spectral data used to calculate the temperature dependence, which is estimated to be between 0.7% and 2% (Orphal 2003). The combined uncertainty from measurements of the ozone absorption cross section, device-to-device variability, and temperature dependence are estimated to be 1%.

The flux from the UV LED is also temperature dependent, with a negative intensity–temperature coefficient, operating with higher efficiency at low temperatures. Although many commercial UV photometers employ active temperature control of the light source, it is not necessary with the UV LEDs, which can significantly reduce power consumption. Using the integrated feedback photodiode, it is possible to account for the temperature coefficient in data processing or, if desired, to control the drive current to the LED to maintain a constant output that is independent of temperature. An additional consideration is the stability of spectral power distribution of the UV LEDs with temperature variations. Figure 5 demonstrates that the wavelength of peak spectral power and the full-width half-maximum spectral width do not significantly vary for temperatures ranging from −40° to +40°C.

A potentially serious limitation of the first generation of deep-UV LEDs is their relatively short lifetime. The manufacturer estimates that the useful lifetime of the UVTOP-255 is between 300 and 500 h at a nominal drive current of 20 mA. Although it is expected that future generations of these devices will have significantly longer lifetimes, this limitation has been addressed in the current instrument design. Unlike Hg-vapor lamps, UV LEDs can be driven in a pulsed mode at frequencies in excess of 1 kHz, which increases both the lifetime of the LEDs and their peak intensity. The drive frequency is limited by the detector analog-to-digital sampling rate; however, frequencies of 10 Hz with a duty cycle of 0.05 are possible, which will extend the usable lifetime of the UV LEDs to between 5000 and 10 000 h.

b. Detectors

The two primary UV detectors in this instrument are SiC photodiodes operating in photovoltaic mode. A high-gain, low-noise transimpedance amplifier is integrated into the detector housing to minimize analog noise. Similar to the light source, the photodiodes have an integrated UV-transmissive lens that increases the effective sensitive area from 1 to 11 mm2 without introducing additional electrical noise or added parasitic capacitance. The half-maximum detector field of view is approximately 5°, and sensitivity at 254 nm is approximately 0.13 nV nW−1 mm−2. The high performance amplifier, radio frequency (RF) shielding, and a low-noise 24-bit delta–sigma analog-to-digital converter give a total detector electrical noise level of 2 μV (1σ at 1 Hz), which corresponds to a maximum theoretical ozone measurement resolution of 0.1 ppb at 1 Hz. For comparison, the phototube-based detector system used in the TEI 49 (TEI part 8592) has a maximum theoretical ozone resolution of 1 ppb at 1 Hz, limited by the 100-kHz voltage-to-frequency converter.

c. Data acquisition and control system

Data acquisition and instrument control are accomplished with an embedded Advanced RISC Machine (ARM9) single-board computer (SBC; TS-7260, Technologic Systems). This SBC has sufficient computing power to run a reduced version of a desktop operating system, in this case TS-Linux, for maximum software and interface configurability with a minimal footprint and power consumption (<0.25 W under standard load). In addition to using a low-power clock-scalable processor (ARM EP9302, Cirrus Logic), the TS-7260 SBC can de-energize unused subsystems [e.g., universal serial bus (USB), Ethernet, serial] through software control to further reduce power consumption. The operating code is stored in 32 MB of onboard flash memory, and additional data storage and development code are contained on USB flash memory drives. The user interface is a remote terminal connected via Ethernet or RS-485 serial connection. The main board provides a PC-104 bus for communication with the data acquisition boards (Fig. 6).

Analog signals from the primary detectors and integrated LED-monitoring photodiodes are digitized using a high-resolution, low-noise 24-bit Σ–Δ primary analog-to-digital converter (PADC; MPC-624, Micro/Sys Corporation, based on the LT2440 converter, Linear Technologies). The 24-bit converter has a maximum achievable resolution of 60 parts per trillion (ppt) and a practical total electrical noise level better than 500 ppt when combined with the detector system described previously. The Σ–Δ architecture allows for software-definable sampling rate and resolution. At the highest resolution [24.6 effective number of bits (ENOB) or 3.9 ppt theoretical ADC resolution], the maximum sample rate is 6.875 Hz, which yields a maximum ozone measurement rate of 1 Hz in continuous mode and 0.1 Hz in low-power pulsed mode. The fastest sampling rate is 3.52 kHz, which corresponds to 17 ENOB or a resolution of 8 ppm and a maximum ozone measurement rate of approximately 10 Hz, which is limited by the flow system. The primary analog-to-digital converter is also designed for low-power operation, and it can be instructed to enter a low-power sleep state between measurements at lower data rates.

Thermodynamic parameters necessary for calculation of ozone mixing ratio (pressure and temperature) are measured using solid-state piezoresistive pressure transducers (MPX2102 freescale semiconductor; 1 per cell) and 2-kΩ precision thermistors (2 per cell). The specified accuracy of the MPX2102 is 1.5% with a precision of 0.5%; however, with individual calibration, accuracies of better than 0.5% can be achieved. The thermistors were calibrated in a thermal chamber against a National Institute of Standards and Technology (NIST)–traceable thermometer (Dostmann P600). After a four-point calibration function is applied, the residual error in determining the air temperature is less than 0.5° over the range of −60° to +60°C. The flow through the instrument is determined using a solid-state differential pressure gauge (MPX5010 freescale semiconductor; labeled “dP” in Fig. 1) measuring the pressure drop across a slight restriction on the instrument outlet. The differential pressure gauge uses significantly less power (35 mW) than commercially available thermal or paddle-wheel flow meters, has no moving parts, and has been shown to work down to −40°C. The flow measurement using the differential pressure gauges has a resolution of approximately 20 cm3 min−1 over a range of 0–1000 cm3 min−1. This resolution is sufficient for this application, because flow is a purely diagnostic quantity not directly used in the calculation of mixing ratio. Analog signals from these transducers are digitized by a secondary 8-channel 16-bit analog-to-digital converter (TS-ADC16, Technologic Systems). The 16-bit converter (25-ppm resolution) is sufficient to match the precision of the attached transducers.

A custom-built module provides power regulation and distribution for the instrument as well as digital control over the instrument subsystems. Input power (14–36 VDC) is converted to 5 VDC (CPU, analog system, and transducers) and 12 VDC (LEDs and pumps) using DC–DC converters. The regulated power from the DC–DC converters is distributed to the various subsystems through solid-state relays. These relays are controlled via digital input/output (I/O) from the computer, allowing individual subsystems to be de-energized when not in use to reduce power consumption. This module also provides power and digital control for two miniature three-way solenoid valves that switch sample and reference air between cells. These valves are magnetic latching valves (LHLX0500350B, the Lee Co.) that require current only when switching, which results in an average power consumption of less than 1 mW. The wetted material in these valves (silicone and epoxy resin) is not inert to ozone; however, laboratory tests have shown that, because of the extremely small wetted area, there is no measurable ozone loss to the valve surfaces over a temperature range of −40° to 40°C and flow range of 10–1000 cm3 min−1.

d. Mechanical design

The primary requirements for the first proposed field deployment of the instrument are low-power consumption and operation at low ambient temperature. Thus, the flow system for the prototype instrument was designed to meet these requirements. Pumping is provided by two constant-volume Teflon reciprocating pumps (Ensci Corporation, Boulder, Colorado) operating in parallel. These pumps are designed for use on ECC ozonesondes and have a proven track record for operation in the stratosphere. Laboratory tests have shown that the pumps will operate reliably and will start from a cold-soaked condition at temperatures down to −38°C and can run continuously for periods of at least three months. Each pump provides ∼200 cm3 min−1 of flow with a power consumption of 1 W. The pumps are computer controlled via solid-state relays. Two pumps are included for redundancy as well as to provide higher flow (400 cm3 min−1) for faster time response measurements. Future applications of this instrument could employ a different pumping system chosen specifically for the given situation. To achieve the maximum time resolution (10 Hz) of the instrument, a higher flow rate (∼10 L min−1) would be required.

Each optical absorption cell is constructed of a 50-cm-long, 19-mm-diameter Teflon tube, with a 6-mm internal diameter. These tubes are securely mounted inside an aluminum channel to provide structural rigidity. The thick-walled tube is used to accommodate threaded holes for gas fittings. The internal diameter of the tube matches the lens size of the detector and UV LED. Both ends of each cell have been bored out to accept the body of the detector or UV LED with a silicon O ring, ensuring a gas-tight seal, thus removing the need for UV-transmissive windows. The optical length of the cell is determined by the distance between the detector and LED lenses and is measured to be 48.8 cm. The Teflon cells have matte finish, which minimizes reflections off of the cell walls. The tubing on the inlet side of the absorption cells is Teflon-lined polyvinyl chloride (PVC). All components are cleaned with de-ionized water and methanol and are exposed to high ozone (>1 ppm) to ensure passivation before use. Wall losses to the cells, detectors, tubing, and fittings were measured by introducing large concentrations of ozone to the system at low flow rates (∼100 cm3 s−1) and comparing the measured ozone mixing ratio at the inlet and outlet of the instrument using a TEI 49C. For ozone mixing ratios ranging from 100 to 1000 ppb, the difference between inlet and outlet mixing ratios was smaller than could be resolved with the precision of the TEI 49C and stability of the ozone generator (∼3 ppb).

Two ozone scrubbers containing manganese dioxide screens (MnO2) are employed in the instrument. Because of the low flow rate through the instrument, these scrubbers can be significantly smaller than the standard scrubbers used in commercial ozone instruments. Each scrubber contains approximately 50 cm2 of MnO2-coated copper screening. Testing has shown that the scrubbers can catalyze up to 1 ppm of ozone to molecular oxygen at surface pressure and temperatures down to −50°C. Two scrubbers are used for redundancy against scrubber fouling or solenoid valve failure. The overall dimensions of the instrument are 12 cm × 10 cm × 58 cm, with a mass of 2.98 kg.

e. Data processing and reduction

Ozone number density is calculated by applying the Beer–Lambert law to the four primary analog signals from the detectors and the feedback photodiodes. Depending on the data rate relative to the cell switching period, the data-reduction methods are slightly different. In all cases, the primary and integrated UV LED–monitoring detectors are zero calibrated by taking a measurement for each detector with both UV LEDs off and then subtracting these voltages from each detector on subsequent measurements. When the data rate is the same or lower than the flow switching rate (e.g., a cell switching period of 10 s and a measurement period of 10 s), I [refer to Eq. (1)] is defined as the average of the primary detector counts on the sample cell normalized by the LED feedback detector on the sample cell over an entire switching period. Similarly, Io is the average of the primary detector signal on the ozone-scrubbed cell normalized by the feedback detector on that same cell. With knowledge of the cell length (l = 48.8 cm) and the integrated, temperature-dependent ozone absorption cross section σ over the emission spectrum of the UV LED as a function of temperature, the number density can be calculated using Eq. (2). A linear model for the temperature variation of the ozone cross section (as described in section 2a) is applied during data processing. Furthermore, the mixing ratio can be derived from the ozone number density divided by the air density as determined from the average of the cell pressures and temperatures and the ideal gas law. For data assurance or in the case of a component failure, it is possible to accurately calculate the ozone concentration from each cell individually at a rate of twice the switching period. In this case, I and Io are calculated sequentially from a single cell. However, there is a greater potential from interference from rapidly varying species that absorb at 254 nm using this technique:
i1520-0426-27-5-869-e2

Using the same instrument, it is also possible to calculate ozone mixing ratios at a faster rate than the cell switching rate for data rates up to 10 Hz. For this type of measurement, the instrument is operated with a larger external pump to achieve a sufficiently fast cell flushing time. Additionally, the UV LEDs are driven in a pulsed mode at 1 Hz with a 0.9 duty cycle. The 100 ms s−1 dark time is necessary to ensure that the detector baselines are not drifting relative to each other, because this is not inherently accounted for by the cell switching. The primary ADC is run at 440 Hz (21.8 ENOB) to gather 10 digital samples per detector per ozone data point. A high sampling rate is necessary to ensure fast channel switching so that high-frequency variations in light output can be effectively removed by normalizing the detectors relative to the LED-integrated detectors. There are several caveats to data rates faster than the cell switching frequency. Although it is possible to account for detector zero drift, detector span drift can only be accounted for at the switching period and may influence measured data. Also, running the LEDs continuously decreases the instrument lifetime by a factor of 10. Finally, the flows required to ensure a sufficient flushing rate for high-speed measurements can introduce turbulence and pressure fluctuations into the measurement cell.

3. Instrument performance

A prototype instrument was tested in the laboratory using a TEI 49 primary standard (49PS) ozone calibrator. The calibrator can generate a known ozone mixing ratio from an ozone-free zero-air source with an accuracy of 3 ppb and a precision and stability of 1 ppb. The ozonated output from the calibrator was plumbed to the prototype instrument using Teflon tubing, and an analog mixing ratio output from the calibrator was logged using a spare analog channel on the secondary ADC. Both the TEI 49PS and the UV LED instrument were allowed to warm up for 30 min prior to the test, and all tubing was conditioned with high ozone (∼300 ppb) to minimize wall losses. Figure 7 compares the instrument (time shifted to account for sampling mismatch and line delays) to the reported mixing ratio from the ozone generator for ozone set-point steps from 300, 200, 100, and 0 ppb with a final step up to 500 ppb. Instrument noise level can be estimated by measuring the variability in the measured mixing ratio relative to the reported output from the TEI 49PS for a fixed set point. However, a portion of this variability may be due to real variations in the ozone output from the 49PS at either submeasurement time scales or subprecision magnitudes. Instrument accuracy was determined by comparing the root-mean-square (RMS) difference between ozone number density calculated using Eq. (1) (where Io is defined as the 10-s mean detector signal for the reference periods before and after a sample period and I is the 10-s mean detector signal for a sample period) and the reported number density from the ozone calibrator. The precision and accuracy for the instrument calculated from these tests are similar to the specified precision and accuracy for the TEI 49PS; thus, these represent an upper bound for these quantities as the variability cannot be attributed to either the generator or instrument. Table 1 shows noise and accuracy figures calculated from a longer-duration comparison to the TEI 49PS as well as the theoretical values calculated from the propagation of individual measurement uncertainties (described in section 2c) through the data reduction algorithm. Because of the limitations that the precision and accuracy of the ozone calibrator place on the measurement, the theoretical values should present a truer picture of the precision and accuracy achievable at standard conditions.

Figure 8 demonstrates an intercomparison between the UV LED–based instrument and an electrochemical ozonesonde (Ensci 2Z) during a simulated atmospheric sounding. The measurements were taken in a thermovacuum chamber with ozone introduced into a Teflon mixing manifold at partial pressures chosen to be representative of real-world midlatitude ozone profiles. The results show a mean absolute error of 0.24-mPa ozone partial pressure, which increases with altitude. This represents a mean percentage error of 4%. The UV LED instrument shows a consistent bias toward lower ozone mixing ratios, which can be partly explained by the slower response time (20 s as opposed to 2 s for the ECC sonde) and frequent positive excursions in the ambient ozone partial pressure. At the altitudes at which the instrument will operate in the upcoming balloon experiment (50–80 hPa), the mean discrepancy between the two instruments is less than 2%, which is within the estimated accuracy of the ozonesonde.

In addition to tests of instrument measurement performance, instrument components were subjected to extended temperature tests in a thermal environmental chamber to assess their suitability for use in the stratosphere. All components were functional for temperatures ranging from +50° to −38°C. The limiting factor for low-temperature operation is the Teflon sampling pump. All electrical and optical components continued to operate down to −50°C. This extended temperature range (without supplemental heating) is a major advantage over Hg-vapor lamp–based instruments, which require heating to 60°C for operation. It is also predicted that operation at lower temperatures will prolong instrument life, because the emission efficiency of the UV LED is 4 times greater at −35°C than the emission efficiency at room temperature, allowing a smaller drive current to be used to achieve the same intensity.

4. Conclusions

Advances in semiconductor materials and fabrication, digital data acquisition, and low-power electronics have allowed the refinement of a proven ozone measurement technique to provide new capabilities. Initial testing of UV LEDs with output at a wavelength of 254 nm has shown them to be a suitable replacement for low-pressure mercury-vapor lamps. These LEDs provide greater usable intensity without the need for temperature control and high-voltage power supplies. Additionally, LEDs provide a more stable light source and can be used in pulsed mode to further reduce energy consumption relative to Hg-vapor lamps. The combination of LEDs as a light source with SiC photodiodes and high-resolution, low-power data acquisition and control electronics facilitated the development of an instrument with measurement performance that exceeds that of conventional UV photometers at size and power consumption closer to that of electrochemical concentration cell instruments. The high stability of the LED light source along with an integrated photodiode for intensity monitoring allows for the measurement of ozone at rates previously only achievable with chemiluminescent instruments. This new technology will be highly applicable to measurements from small autonomous platforms such as UAVs and long-duration balloons, as well as for high-precision measurements in harsh remote or space-limited urban environments.

REFERENCES

  • Anceau, S., Lefebvre P. , Suski T. , Konczewicz L. , Hirayama H. , and Aoyagi Y. , 2005: Enhancement of localization and confinement effects in quaternary group-III nitride multi-quantum wells on SiC substrate. Phys. Status Solidi, 202A , 642646.

    • Search Google Scholar
    • Export Citation
  • Bass, A. M., and Paur R. J. , 1981: UV absorption cross-sections for ozone—The temperature-dependence. J. Photochem., 17 , 141.

  • Brewer, A. W., and Milford J. R. , 1960: The Oxford-Kew ozone sonde. Proc. Roy. Soc. London, 256A , 470495.

  • Burrows, J. P., Dehn A. , Deters B. , Himmelmann S. , Richter A. , Voigt S. , and Orphal J. , 1998: Atmospheric remote-sensing reference data from GOME: Part I. Temperature-dependent absorption cross-sections of NO2 in the 231-794 nm range. J. Quant. Spectrosc. Radiat. Transfer, 60 , 10251031.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Eastman, J. A., and Stedman D. H. , 1977: Fast response sensor for ozone eddy-correlation flux measurements. Atmos. Environ., 11 , 12091211.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Komhyr, W. D., 1967: Nonreactive gas sampling pump. Rev. Sci. Instrum., 38 , 981.

  • Malicet, J., Daumont D. , Charbonnier J. , Parisse C. , Chakir A. , and Brion J. , 1995: Ozone UV spectroscopy. II. Absorption cross-sections and temperature dependence. J. Atmos. Chem., 21 , 263273.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Orphal, J., 2003: A critical review of the absorption cross-sections of O3 and NO2 in the ultraviolet and visible. J. Photochem. Photobiol., 157A , 185209.

    • Search Google Scholar
    • Export Citation
  • Proffitt, M. H., and McLaughlin R. J. , 1983: Fast-response dual-beam UV-absorption ozone photometer suitable for use on stratospheric balloons. Rev. Sci. Instrum., 54 , 17191728.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rinsland, C. P., and Coauthors, 2003: Spectroscopic parameters for ozone and its isotopes: Recent measurements, outstanding issues, and prospects for improvements to HITRAN. J. Quant. Spectrosc. Radiat. Transfer, 82 , 207218.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Viallon, J., Moussay P. , Norris J. E. , Guenther F. R. , and Wielgosz R. I. , 2006: A study of systematic biases and measurement uncertainties in ozone mole fraction measurements with the NIST Standard Reference Photometer. Metrologia, 43 , 441450.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Voigt, S., Orphal J. , Bogumil K. , and Burrows J. P. , 2001: The temperature dependence (203-293 K) of the absorption cross sections of O3 in the 230-850 nm region measured by Fourier-transform spectroscopy. J. Photochem. Photobiol., 143A , 19.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Instrument optical and flow diagram.

Citation: Journal of Atmospheric and Oceanic Technology 27, 5; 10.1175/2009JTECHA1362.1

Fig. 2.
Fig. 2.

Normalized UV LED output measured with SiC photodiode at 10 Hz, demonstrating variations in intensity.

Citation: Journal of Atmospheric and Oceanic Technology 27, 5; 10.1175/2009JTECHA1362.1

Fig. 3.
Fig. 3.

Normalized UV LED emission spectra for four representative devices. Overlaid curve shows ozone absorption cross section for comparison.

Citation: Journal of Atmospheric and Oceanic Technology 27, 5; 10.1175/2009JTECHA1362.1

Fig. 4.
Fig. 4.

Effective ozone absorption cross section for four devices. Effective cross section is the convolution of the normalized UV LED emission spectrum and room temperature ozone cross section. The numerical values listed give the effective cross sections at 254.7 nm.

Citation: Journal of Atmospheric and Oceanic Technology 27, 5; 10.1175/2009JTECHA1362.1

Fig. 5.
Fig. 5.

Normalized UV LED emission spectra as a function of operating temperature. Three curves represent −40° (top curve), 23° (middle curve), and 40°C (bottom curve).

Citation: Journal of Atmospheric and Oceanic Technology 27, 5; 10.1175/2009JTECHA1362.1

Fig. 6.
Fig. 6.

Instrument electrical block diagram.

Citation: Journal of Atmospheric and Oceanic Technology 27, 5; 10.1175/2009JTECHA1362.1

Fig. 7.
Fig. 7.

Instrument response (asterisk) to ozone step functions generated by TEI 49PS ozone calibrator (line shows calibrator output). Ozone instrument data have been time shifted to account for lags resulting from tubing and instrument flushing time.

Citation: Journal of Atmospheric and Oceanic Technology 27, 5; 10.1175/2009JTECHA1362.1

Fig. 8.
Fig. 8.

Comparison between UV LED instrument (asterisk) and Ensci 2Z ECC ozonesonde (line) during chamber simulation of ozone sounding profile. Inset shows smoothed absolute difference between measurements as a function of ambient pressure.

Citation: Journal of Atmospheric and Oceanic Technology 27, 5; 10.1175/2009JTECHA1362.1

Table 1.

Instrument noise levels and accuracy based on comparison to TEI 49PS reported mixing ratio at a range of set points and theoretical accuracy determined by propagating sensor errors through the data-reduction algorithm.

Table 1.
Save
  • Anceau, S., Lefebvre P. , Suski T. , Konczewicz L. , Hirayama H. , and Aoyagi Y. , 2005: Enhancement of localization and confinement effects in quaternary group-III nitride multi-quantum wells on SiC substrate. Phys. Status Solidi, 202A , 642646.

    • Search Google Scholar
    • Export Citation
  • Bass, A. M., and Paur R. J. , 1981: UV absorption cross-sections for ozone—The temperature-dependence. J. Photochem., 17 , 141.

  • Brewer, A. W., and Milford J. R. , 1960: The Oxford-Kew ozone sonde. Proc. Roy. Soc. London, 256A , 470495.

  • Burrows, J. P., Dehn A. , Deters B. , Himmelmann S. , Richter A. , Voigt S. , and Orphal J. , 1998: Atmospheric remote-sensing reference data from GOME: Part I. Temperature-dependent absorption cross-sections of NO2 in the 231-794 nm range. J. Quant. Spectrosc. Radiat. Transfer, 60 , 10251031.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Eastman, J. A., and Stedman D. H. , 1977: Fast response sensor for ozone eddy-correlation flux measurements. Atmos. Environ., 11 , 12091211.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Komhyr, W. D., 1967: Nonreactive gas sampling pump. Rev. Sci. Instrum., 38 , 981.

  • Malicet, J., Daumont D. , Charbonnier J. , Parisse C. , Chakir A. , and Brion J. , 1995: Ozone UV spectroscopy. II. Absorption cross-sections and temperature dependence. J. Atmos. Chem., 21 , 263273.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Orphal, J., 2003: A critical review of the absorption cross-sections of O3 and NO2 in the ultraviolet and visible. J. Photochem. Photobiol., 157A , 185209.

    • Search Google Scholar
    • Export Citation
  • Proffitt, M. H., and McLaughlin R. J. , 1983: Fast-response dual-beam UV-absorption ozone photometer suitable for use on stratospheric balloons. Rev. Sci. Instrum., 54 , 17191728.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rinsland, C. P., and Coauthors, 2003: Spectroscopic parameters for ozone and its isotopes: Recent measurements, outstanding issues, and prospects for improvements to HITRAN. J. Quant. Spectrosc. Radiat. Transfer, 82 , 207218.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Viallon, J., Moussay P. , Norris J. E. , Guenther F. R. , and Wielgosz R. I. , 2006: A study of systematic biases and measurement uncertainties in ozone mole fraction measurements with the NIST Standard Reference Photometer. Metrologia, 43 , 441450.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Voigt, S., Orphal J. , Bogumil K. , and Burrows J. P. , 2001: The temperature dependence (203-293 K) of the absorption cross sections of O3 in the 230-850 nm region measured by Fourier-transform spectroscopy. J. Photochem. Photobiol., 143A , 19.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Instrument optical and flow diagram.

  • Fig. 2.

    Normalized UV LED output measured with SiC photodiode at 10 Hz, demonstrating variations in intensity.

  • Fig. 3.

    Normalized UV LED emission spectra for four representative devices. Overlaid curve shows ozone absorption cross section for comparison.

  • Fig. 4.

    Effective ozone absorption cross section for four devices. Effective cross section is the convolution of the normalized UV LED emission spectrum and room temperature ozone cross section. The numerical values listed give the effective cross sections at 254.7 nm.

  • Fig. 5.

    Normalized UV LED emission spectra as a function of operating temperature. Three curves represent −40° (top curve), 23° (middle curve), and 40°C (bottom curve).

  • Fig. 6.

    Instrument electrical block diagram.

  • Fig. 7.

    Instrument response (asterisk) to ozone step functions generated by TEI 49PS ozone calibrator (line shows calibrator output). Ozone instrument data have been time shifted to account for lags resulting from tubing and instrument flushing time.

  • Fig. 8.

    Comparison between UV LED instrument (asterisk) and Ensci 2Z ECC ozonesonde (line) during chamber simulation of ozone sounding profile. Inset shows smoothed absolute difference between measurements as a function of ambient pressure.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 281 73 4
PDF Downloads 356 45 4