Intracavity laser spectroscopy for real-time trace gas detection
08/01/2000
Markus Wolperdinger, Jörg Kutzner, Robert Mellish, George H. Atkinson, Innovative Lasers Corp., Tucson, Arizona
overview
Although no technology should be viewed as a panacea, intracavity laser spectroscopy trace gas sensors do provide a significant advance in measurement capabilities. They are particularly applicable for demanding, industry-leading applications in semiconductor manufacturing such as in situ gas monitoring, where rapid decision making is required during processing. Here, the performance parameters of this new commercial technology are discussed within the context of other currently available nonoptical and optical sensors, showing the advantages of these sensors for high-sensitivity, trace gas detection.
The viability of future advances in semiconductor-manufacturing processes is often viewed in terms of the availability of the appropriate trace gas detection technology. In addition, the capability of integrating trace gas detection technology into highly complex semiconductor-manufacturing protocols — as off-line, troubleshooting equipment and real-time (sec) in situ gas monitoring that permits rapid decision making during processing — is widely seen as an enabling factor in the industry's next technology level.
Nonoptical methodologies
Trace gas detection instrumentation, based on several different nonoptical methodologies, is currently available commercially. These methodologies — reversible quartz crystal microbalances, amperometric hygrometers, ion mobility spectrometers (IMS), mass spectrometers (APIMS and RGA), and gas chromatographs — all require direct physical contact between the gas sample under analysis and the sensor itself. Such contact often results in
- slow response times;
- hysteresis effects due to irreversible contamination of the sensor;
- continuously changing or slowly drifting system calibration;
- limited selectivity; and
- chemical incompatibility between the sensor and reactive (corrosive) gas samples, resulting in sensor deterioration.
 |
In addition, data obtained with these methodologies exhibit limited dynamic ranges (i.e., 10-100) and, with the exception of atmospheric pressure ionization mass spectroscopy (APIMS) and IMS, insufficient detection sensitivity (~10ppb). Although efforts to improve the performance of these instruments are ongoing, it appears unlikely that the inherent limitations associated with these methodologies can be successfully overcome and, therefore, it is also unlikely that any of these methodologies will provide complete solutions for applications in next-generation trace gas detection.
APIMS is one of the most sophisticated and expensive (>$300,000) analytical tools for trace gas detection in the 10ppt-to-10ppb range. Normally, APIMS systems are only practical for centralized, stand-alone measurements required for the certification of a new fabrication facility. APIMS instruments require extensive infrastructure and highly skilled operators to achieve their specified performance. APIMS, however, does provide real-time sub-ppb water measurements for inert gases, but cannot be used directly with corrosive or reactive gas samples.
Small, less-expensive mass spectroscopy instruments — residual gas analyzers (RGA) — can be configured as distributed sensor systems, but they can only operate at low pressures (<10-4torr).
Optical methodologies
The inherent advantages of optical methodologies, based on requiring only optical access to the gas sample (i.e., contact-free measurements of trace gases, resistance to corrosive or reactive environments, rapid response times in seconds, in situ monitoring, etc.), have led to the anticipation that they would overcome the limitations for trace gas detection found with conventional, nonoptical techniques. Since optical methodologies are widely viewed as the most promising candidates for next-generation trace gas detection technology in semiconductor manufacturing, several new optical techniques are currently under development.
Although these optical techniques use both absorption and fluorescence detection, the performance capabilities required for ultra-high-purity trace gas detection for semiconductor applications make absorption the method of choice. (Fluorescence emission detection is currently used for specific tasks such as endpoint detection and exhaust gas monitoring.)
Aside from those based on intracavity laser spectroscopy (ILS), two of the most prominent optical methodologies currently under development use tunable diode laser (TDL) spectroscopy and cavity ring-down spectroscopy (CRDS).
In many TDL-based instruments, the narrow bandwidth wavelength of a diode laser is tuned across the absorption line of the gas of interest to measure the change in laser intensity. The gas concentration can be determined from the resultant derivative signal following an extended data analysis. The selectivity derives from the precise overlap of the diode laser wavelength with the gas species absorption.
In CRDS, the wavelength of picosecond laser pulses is tuned to overlap absorption from the gas species of interest before the laser pulses are optically injected into the passive (not a laser) resonator cavity comprised of two >99.999% reflective mirrors. The laser pulse reflects between the two mirrors while a small percentage of the radiation leaks out of the passive cavity on each round trip. The concentration of the species is derived from the measured decay time of pulse intensities, a property that reflects the degree of absorption.
Although some performance parameters of TDL and CRDS instrumentation are promising, other crucial system parameters (e.g., sample volume, response time, complexity of the mirror system, and detection selectivity and sensitivity) do not meet the requirements of next-generation trace gas detection in semiconductor manufacturing (see table).
Among the limitations of TDL detection is the need to lower the total sample pressure below that found typically in gas lines and chambers. This low-pressure restriction exists because most gaseous absorption bandwidths are too broad at higher pressures to be effectively scanned by the wavelength range available from TDL (i.e., reduced pressures narrow the absorption bandwidths). Two important disadvantages of CRDS are the intrinsically small dynamic range over which trace gas concentrations can be measured and the need for pulsed lasers and modulated continuous wave (CW) lasers of well-defined duration.
 Figure 3. ILS-M150 calibration curve for 2-20ppb moisture in nitrogen. |
An important limitation shared by both TDL and CRDS involves the need to maximize the optical path length for the gas sample to increase detection sensitivity. In TDL, multireflection cells typically provide a folded optical path up to 50-100 times the physical dimension of the gas cell (i.e., typically 0.5-1.0m). As a consequence, sample cells must have high reflectivity, multipass mirrors that require precise alignment and are usually exposed to the sample gas. Multipass cells also have liter-size volumes that slow gas exchange rates, especially at low concentrations, and consequently, ensure slow minutes-to-hours response times to changing contamination levels. In CRDS, >99.9% mirror reflection is used to increase the absorption path length. The maximum detection enhancement reported of a 100km effective path length using a 0.5m physical cell can only be achieved with state-of-the-art super-mirrors (>99.999% reflectivity) [1]. Super-mirrors are not available for many wavelengths (i.e., for gas contaminants) and must be exposed to the sample gas detected, often resulting in deterioration requiring realignment and cleaning. More typically for CRDS, 30m absorption path lengths can be achieved with >99.5% reflectivity mirrors and a 10cm cell gas [2].
Strategic goals for gas sensors
The decision to use any gas sensor technology in semiconductor manufacturing, especially for in situ applications on wafer-processing tools within lithography chambers or in plasma-etching tools, often depends on its capability to provide an idealized set of performance characteristics. It must be automated; have a real-time response time of seconds; be nonperturbing; have high sensitivity with 1-5ppb to 50ppt detection; and have measurement conditions where no sample preparation or modification is required.
Such sensor performance should include updating confirmation to ensure that readings reflect an accurate calibration, and should not require expert operational personnel.
 Figure 4. ILS-M150 calibration curve for 2-4000ppb moisture in nitrogen. |
None of the optical or nonoptical sensor technologies currently available meet all of these desired criteria. Current state-of-the-art technology can only verify the purity of process gases laboriously and with high cost through off-line measurements and slip-stream sampling methods. There is little direct access to reactive chambers and fewer still measurements during processing such as in plasma etching. Although optical methodologies under development (e.g., TDL or CRDS) have promising aspects, all have to overcome several extremely limiting, system-inherent, shortcomings before they will be able to meet next-generation requirements.
Principles of ILS
The principles of ILS are significantly different than those underlying conventional absorption spectroscopy (e.g., TDL and CRDS), and the performance of ILS gas sensors overcomes most limitations characteristic of a sensor using those methods (Fig. 1).
In the conventional approaches, the Beer-Lambert law defines the relationship between the intensity of the external radiation source (laser, lamp, etc.) prior to its entering the sample (I0), and its intensity after exiting the sample (I). Aside from producing I0, the radiation source is not actively involved in the absorption measurement. The gas species detected is identified from the specific wavelengths absorbed (i.e., its absorption spectrum) and its concentration quantitatively determined by independent calibration data and from the Beer-Lambert law. For a given gaseous species and wavelength, the detection sensitivity is determined by the path length over which absorption occurs. Detection sensitivity can only be increased by increasing the optical path length within the sample, which is often achieved via a multipass gas cell. Even the typical maximum of 40m (e.g., 80 passes within a 0.5m long cell) requires both higher intensity radiation sources and the measurement of differences between increasingly larger numbers (I0 and I). The detection sensitivities for sensors of many gases of interest using conventional absorption are typically limited to >100ppb. Some commercial sensors now operate in the <10ppb range, but only with longer path lengths obtained by larger cells and more demanding optical systems, as discussed below.
 Figure 5. ILS-M150 system response to moisture level changes in the low-ppb concentration range. |
In ILS, the laser itself is the detector and the significant enhancements in detection sensitivities derive from the competition between the gain and loss properties within the optical resonator of a multimode homogeneously broadened laser resonator. The critical aspect of such a competition is associated with the dynamics by which the laser resonator reaches threshold conditions required to produce radiation. ILS is not simply an expanded multipass cell. Since the losses within the laser resonator, including the absorption of the gas sample, are present before the laser reaches threshold conditions (i.e., before radiation is produced from the laser), the gain properties of the multimode laser can adjust to the presence of optical losses (e.g., gas absorption). These adjustments are manifested in a variety of ways, including the tuning of wavelengths at which the laser operates (i.e., energy within the resonator shifts away from the wavelengths at which the gas absorbs to those wavelengths where the gas does not absorb). Thus, the spectral output (i.e., wavelength-resolved spectrum) of the ILS laser has superimposed on it the absorption spectrum of the intracavity gas sample.
The competition between the gain and loss properties of the ILS laser results in large increases in the detection sensitivity; the optical path length equivalent to that required for conventional absorption measurement approaches1000km, while the physical dimension of the gas cell remains a few centimeters. In addition, all the absorption information is on the laser beam exiting the ILS laser, thereby making it easy to detect and even to transmit it over optical fibers.
These general principles of the ILS methodology are illustrated by an idealized optical resonator commonly found in lasers comprised of two or more highly reflective mirrors (Fig. 2). Such an active optical cavity contains a medium (e.g., gas, liquid, or solid state) in which optical gain is produced by appropriate excitation (e.g., by a small, commercially available semiconductor diode laser). Typically, the coating of one mirror is only partially reflective, thereby permitting some light to escape from the laser cavity. The "intracavity region" between the two reflective mirror surfaces containing the gain medium defines the laser resonator or cavity. The spectral output of an ILS laser is normally broad and without sharp, distinctive spectral features (Fig. 2a). The spectral output of the laser can be altered by different optical elements (e.g., filters), but more important, by placing gaseous molecules, atoms, ions, or radicals inside the optical resonator in the intracavity region (Fig. 2b). When the gain medium and the optical elements of the resonator are held unchanged, these intracavity optical gaseous species become the dominant factors determining the resultant laser output (e.g., spectral output, power, polarization, phase, and temporal characteristics).
The spectral properties of the intracavity absorber, and therefore of the species being detected, can be readily identified from the wavelength-resolved output of the ILS laser. The laser signal is typically detected with a multichannel diode array detector after dispersion by an appropriate spectrometer. Other simpler and less costly detection methods to quantitatively determine gas concentrations have also been developed.
Advantages of ILS for trace gas detection
The new detection methodologies embodied in ILS offer a unique set of measurement advantages:
- detection sensitivity enhanced several orders of magnitude to the sub-ppb range;
- trace gas detection within multicomponent gas samples;
- access to corrosive-reactive gas samples;
- quantitative concentration data greater than three orders of magnitude in dynamic range (e.g., 1ppb to 5ppm);
- real-time response times in seconds;
- small volume, small length (i.e., centimeter) gas cells to facilitate rapid gas exchanges at low concentrations of trace contaminants (e.g., water); and
- digital control automation, including programmable measurement protocols and time-stamped data acquisition.
ILS trace gas sensors retain the advantages associated with conventional absorption spectroscopy, while incorporating superior quantitative detection sensitivity, rapid response times, small volume gas cells, and access to corrosive-reactive samples. This combination of operational advantages is unique from any of the optical or nonoptical measurement technologies currently available commercially or under development.
Practical trace gas detection
Although the advanced optical technologies associated with ILS have been actively investigated in research laboratories for more than three decades, they have only recently become commercially available [3]. The earliest versions of these sensors focus on the detection of moisture, ammonia, and carbon dioxide in a variety of inert and corrosive carrier gas samples for single-digit ppb to ppm concentrations (>103 dynamic range in a single instrument). Other trace gases can be targeted and both lower (e.g., 50ppt) and higher (e.g., >100ppm) concentration ranges can be measured with different ILS instrumentation.
Among the most challenging ultra-high-purity applications in semiconductor manufacturing is monitoring water (moisture) at the ppb or lower level. Data below from commercial ILS instrumentation illustrates general performance characteristics for semiconductor manufacturing, among other fields of use. The data are analyzed by on-board microelectronics to obtain water concentrations that are displayed on a liquid crystal screen. An RS232 interface provides the ability to control the instrument remotely through an external computer or laboratory management system and to data-log the results for months.
The two most important parameters for any trace gas sensor are the lower limit of detection (LOD) achievable and the dynamic range over which gas concentrations are quantitatively measurable.
 |
With the exception of APIMS, the LOD of other commercial optical and nonoptical trace moisture sensors is limited to the low, double-digit ppb range. Generally, these limitations stem from restrictions specific to each methodology and from the gas delivery and handling within each instrument. In some cases, improved LODs (i.e., single-digit ppb) have been described as a result of extensive data averaging or the extrapolation of measured concentrations. A typical series of measurements for our trace moisture sensor shows that the 2-20ppb range is readily accessible with response time <1min (Fig. 3). Longer averaging times reduce the error to <150ppt. These are all direct measurements in 0.5-1liter/min flowing gas through 1/8in. gas lines with no sample preparation or mathematical extrapolation.
The trace moisture sensor has a linear response extending over more than three orders of magnitude without compromising other system parameters such as LOD, response time, and long-term stability (Fig. 4).
System response time
The response characteristics of trace gas sensors typically depend on three independent but interrelated factors:
- the intrinsic response characteristics of the sensor methodology;
- the design, construction, and material of wetted internal gas lines and gas-handling components; and
- how the sensor is incorporated into the measurement infrastructure.
These factors contribute in varying degrees to overall system response and, in most cases, are difficult to identify separately.
With our trace gas sensor, the small-volume ILS sample cell (<2cm3) makes it feasible to measure trace moisture changes at <10ppb within a few seconds (Fig. 5). The intrinsic ILS sensor response time is <100msec and, therefore, the overall sensor response time is limited only by delays caused by the gas-handling system. Other choices for the design and construction of the gas-handling systems could result in even faster sensor response times.
The long-term stability of any trace gas sensor also typically depends on three independent, but interrelated factors: the intrinsic stability of the sensor methodology; factors external to the sensor such as room-temperature fluctuations; and verification and adjustment of the sensor calibration.
To address these issues, we have incorporated a NIST-traceable, on-board calibration system. This design is instrumental in achieving an operational stability of <±5% during 8 hr (Fig. 6).
Conclusion
Although ILS techniques have been part of the continuing development of advanced optical detection technologies underway for the past two decades, the design and commercialization of trace gas sensors appropriate for applications in industries such as semiconductor manufacturing have only recently come to fruition. These applications include direct, real-time process control, in situ chamber measurements, and the detection of trace contamination in both inert and corrosive gas samples. The accuracy of gas sensors operating in the ppb or lower concentration ranges depends critically on the availability of equally accurate calibration standards and protocols that ensure proper sensor function when decision-making measurements are made. The ILS sensor technology described here — an emerging laser-based technology derived from intracavity laser spectroscopy — incorporates such automated, dynamic calibration methods.
References
- D. Romanini et al., J. Chem. Phys. 99, 6287, 1993.
- M. Murtz et al., Appl. Phys. B, 1999.
- ILS-M150 from Innovative Lasers Corp. is a trace gas sensor for non-corrosive gases. Other models are available for corrosive gases. These instruments include an optical, laser-based ILS detection system, an intracavity gas sample cell, and a fully integrated, NIST-traceable on-board calibration system to provide a reference gas and automated system verification.
Markus Wolperdinger received his BS and MS in chemistry, and his PhD in physical chemistry from Ludwig-Maximilians-University in Munich, Germany. Wolperdinger is director of product development at Innovative Lasers Corp.
Joörg Kutzner received his PhD from Bielefeld University, Germany, for his work on UV photoablation, nonlinear optics, and laser mass spectroscopy. He is a senior research scientist at Innovative Lasers Corp.
Robert Mellish studied at the Imperial College of Science, Technology, and Medicine in London, England, and received his PhD for research into all-solid-state ultra-fast laser oscillator and amplifier systems. He is a senior research scientist at Innovative Lasers Corp.
George H. Atkinson received his PhD in physical chemistry. He is professor of chemistry and optical sciences and head of the chemistry department at the University of Arizona. In 1992, Atkinson founded and now serves as president and CEO of Innovative Lasers Corp., where he also directs R&D programs. 3280 E. Hemisphere Pl. #114, Tucson, AZ 85706; ph 520/760-6644, fax 520/760-6644, [email protected].
Successful implementation of ILS demands supplier backup
 |
Atkinson's intracavity laser spectroscopy system may fill several needs for the semiconductor-wafer-processing community: the need to rapidly qualify new gas distribution systems in the fab at a reasonable cost; the need to detect and measure contaminants in both inert and corrosive process gases as these gases are actually being used in the fab; and the need to measure process exhaust gases to ensure compliance with clean air and safety regulations. This technology may also enable cost-effective, real-time, in situ process monitoring for fault detection and classification (FDC). As chipmakers continue to improve process tool productivity (wafer-monitoring reduction, early yield excursion detection, run-to-run and real-time process control, and scrap minimization using in situ measurements and FDC), intracavity laser spectroscopy applications may fill even more of these needs. However, to be truly useful, this technology must be packaged so that it is easy to operate, cost-effective, and well supported with applications by the supplier.
Brad Van Eck, project manager, RTP tool development, sensor integration for improved OEE, International Sematech, Austin, TX