In pristine fabrication processes such as front-end semi, several trace gas species must be completely purged from the ambient air. This includes airborne molecular contaminants like NH3 that can impact yields, as well as toxic species such as H2S, HCl, and HF, which can cause safety issues and potentially damage expensive optics. Scrubbers are effectively used to eliminate several of these species, but the purity of the filtered air must then be verified and monitored by trace gas analysis instruments. A new tool —WS-CRDS (wavelengthscanned cavity ring down spectroscopy) — is the first technology to combine high speed data sampling with sensitivity to the parts per trillion level, in a hands-free platform that does not require frequent calibration. This article reviews this new technology, compares it to traditional methods, and presents data for NH3, HF, and H2S.
Nearly every small molecule (e.g., H2O, H2S, NH3) has a unique near-infrared absorption spectrum. At sub-atmospheric pressure, this consists of a series of narrow, well-resolved sharp lines, each at a characteristic wavelength. Because these lines are well-spaced and their wavelength is wellknown, the concentration of any species can be determined by measuring the strength of this absorption, i.e. the height of a specific absorption peak.
However, in conventional infrared spectrometers, trace gases provide far too little absorption to measure, typically limiting sensitivity to the parts per million at best. And in NDIR (nondispersive infrared) instruments, the measurements often have the additional problem of cross-talk between certain gas species. Alternatively, more sensitive methods based on ions (mass spectrometry), wet chemistry, and/or separation science (gas chromatography) are slow, require frequent calibration, and a skilled operator. Furthermore, the instruments are often bulky and expensive (up to several hundred thousand dollars) limiting them to lab-based measurements rather than real-time monitoring applications. WS-CRDS combines the simplicity and real-time speed of infrared spectroscopy with the sensitivity and precision of these other technologies, offering a LDL (lower detection limit) and sensitivity at the parts per trillion level.
WS-CRDS utilizes infrared absorption but avoids the low signal/noise limitation of traditional IR instrumentation in several ways. First, it creates a long (up to tens of kilometers) effective path in the compact sample chamber, so that the light absorption is much more detectable. Second, it performs absorption measurements in a way that is independent of any fluctuations in the light source (laser) intensity. And third, WS-CRDS scans the entire absorption line with a narrow-line laser, so even if there is another gas component present with a close overlapping line, the target molecule measurement is unaffected.
The principles of WS-CRDS are outlined in Figure 1. Light from a wavelength-tunable laser diode enters the sample cavity which contains three mirrors, which have exceptionally high reflectivity (>99.999%). When the laser is on, the cavity fills with circulating laser light. Some light intentionally leaks out of the mirrors as it cycles around the cavity. This leakage is proportional to the intensity of light in the cavity, and is measured by a photodetector. The electrical signal at the photodetector is therefore proportional to the instantaneous intensity within the cavity.
When the signal from the detector reaches a steady state condition, the laser is abruptly switched off. The light already with the cavity continues to bounce between the mirrors. But because the mirrors do not have 100% reflectivity, the light intensity inside the cavity slowly leaks out and therefore steadily decays to zero in an exponential fashion. This decay is followed in real-time by the photodetector and is determined solely by the reflectivity of the mirrors. Even with a cavity of only 25 cm in length, the average distance that any photon effectively travels within the cavity can be over 20 kilometers. Now if the cavity contains a gas species that absorbs even weakly, this introduces a second mechanism that drains the intracavity intensity. This results in a shortened decay time, from which the instrument calculates the sample absorbance and hence concentration. Moreover, the decay rate is independent of the initial intracavity intensity and therefore completely independent of fluctuations in the laser intensity. Using this approach, many trace gas species can be quantitatively measured with sensitivities at ten parts per trillion and lower detection limits down to 30 parts per trillion. Plus the high dynamic range of this technology allows WS-CRDS to provide up to four significant figures of accuracy and to measure concentrations above the parts per million level.
The latest generation of WS-CRDS instruments are compact, stand-alone tools that can fit in a standard 19 inch rack. And because this measurement technology involves no moving parts, these instruments are extremely rugged, and can go months between recalibration. This makes them suitable for remote operation through an Ethernet port. With a maximum sampling rate of 1 Hz, the on-board processor allows both high-speed sampling as well as averaging over longer time intervals for very high sensitivity. Data manipulation as well as logging and archiving are all supported. Also, in addition to single species instruments, models are now available with several integrated laser sources allowing them to measure four or more species at once, with no cross-talk whatsoever (see Sidebar).
AMMONIA IN SEMICON FAB
The full complement of WS-CRDS advantages can best be understood in the context of three different applications: NH3, HF, and H2S. In the semiconductor industry, it is well known that parts-per-billion levels of ammonia exposure during the photolithography process can lead to yield loss and unscheduled equipment downtime. Ammonia is emitted into wafer fab air by various semiconductor processes including CVD, wafer cleaning, coater tracks, and CMP, as well as from outdoor air. In some cases it can alter chemically amplified resists and lead to “T-topping,” a destructive distortion of the wafer features that can cause the chip yield to plummet. Furthermore, ammonia is also photoreactive and can deposit on optical surfaces of lithography systems, causing haze and leading to expensive and unpredictable downtime.
To avoid these potential problems, both semiconductor and equipment manufacturers need to thoroughly characterize the photolithography area under process conditions to minimize the risk from airborne ammonia exposure. Some equipment manufacturers even require periodic audits of amine levels in the lithography area to maintain warranty status. Chemical scrubbers or filters are often used in these critical locations to remove airborne ammonia contamination, but their lifetime and coverage cannot offer complete protection.
Ammonia measurements can be performed using a variety of methods, such as ion mobility spectroscopy, chemiluminescence, or IC (ion chromatography). But all of these older methods have significant limitations, either in measurement time, sensitivity, accuracy, or ease-of-use, that prevent more widespread deployment. Fortunately, WS-CRDS provides a combination of high sensitivity, wide dynamic range, and very linear response.
In a collaborative project with Horiba (Kyoto, Japan), we compared WS-CRDS with a commercial ion chromatograph over the range 0-10 ppbv.1 We found excellent agreement between the two methods. However, the IC measurement accuracy was found to be very dependent on the pH of the impinger water used to trap the ammonia. Obviously, this potential complication is not an issue for WS-CRDS.
In this study, we also demonstrated the use of WS-CRDS to investigate the onset of ammonia breakthrough in a filter used to scrub ammonia from clean room air in semicon fab. Here we used two identical WS-CRDS analyzers to monitor the filter input and output. The input ammonia concentration was randomly stepped over the 0-900 ppbv range as shown in Figure 2. Here the orange dashed line highlights the critical 1 ppbv level, above which disruption of the photolithography process may occur. This particular filter was rated to absorb approximately 250 grams of ammonia, corresponding to four years at a 20 ppbv input level, termed 4 YE (year equivalents). Yet the first signs of filter fatigue in these tests occurred at only 1 YE, clearly demonstrating the critical importance of actively and continuously monitoring of the ammonia level.
MONITORING H2S WITHOUT CROSS-TALK
Detecting trace levels of the poisonous gas H2S is a challenge in semicon, petrochemical, and numerous other industries. It’s an excellent example of the species discrimination capabilities of WS-CRDS compared to other more traditional spectroscopic methods. Specifically, because sulfur is a second row element, H2S can be monitored using UV fluorescence. But several other gaseous sulfur compounds (e.g., COS, CS2, SO2) that are often present also produce UV fluorescence that overlaps with that from H2S. So whenever there is the possibility of the presence of trace levels of any these other species, UV fluorescence cannot be relied upon to monitor the H2S level.
In contrast, there are several absorption lines of H2S in the near infrared with no overlap or even proximity from spectral lines associated with either sulfur compounds (COS, CS2, SO2), water, carbon dioxide, or nitrogen oxides — the latter being a potential issue in petrochemicals. The high spectral resolution of a WS-CRDS instrument thus allows it to make trace H2S measurements that are completely immune to crosstalk from other species, even if these species are present in very high concentrations that are rapidly changing. In the latest H2S analyzers based on WSCRDS the lower detection limit is 3 ppbv and the instrument precision is 5 ppbv (in ten seconds) or 1 ppbv (with five minutes of signal averaging).
HF IN SEMICONDUCTOR AND ALUMINUM SMELTING
HF is another gas species that must be monitored in the semiconductor industry, primarily for health and safety reasons. HF is a toxic and corrosive chemical that is used directly as a wet chemistry etching agent. It can also be formed by leaks of BF3 used for ion implantation. And in a very different application, HF must be monitored in the aluminum smelting industry where it is actively produced as a by-product from the reduction cells.
These applications require a HF monitor with ppbv sensitivity or better. Also, the monitor has to be corrosionresistant over a long period (months and years). In addition, the smelting industry requires removal (and monitoring) of HF from the extended headspace over a large-scale “potroom” production environment rather than in a controlled cleanroom space. So any analyzer must be capable of automated remote operation in potroom roof areas with limited accessibility. The latest generation of trace-HF analyzers based on WS-CRDS now exceeds all these requirements (Table 1). For example, with only five minutes of signal averaging, these instruments deliver a precision of 10 pptv and a lower detection limit of 30 pptv (at 3 sigma). And with only ten seconds of data averaging, the precision is still 50 pptv. So even the smallest leaks of HF can be detected and quantified immediately.
Another potential issue with HF is long-term corrosion. In these new WS-CRDS instruments, all the gas handling and sampling hardware use materials and surface treatments to eliminate any chance of corrosion due to the effects of HF. Furthermore, the critical mirrors are particularly immune to HF corrosion as their reflective surfaces are made from fluoride materials which are naturally inert to reaction with HF.
In conclusion, the technique of WS-CRDS offers a robust way to measure several important trace gas species in real-time with the same high sensitivity available from techniques such as IC. Moreover, WS-CRDS is the first method that delivers this capability
- Eric Crosson, Katsumi Nishimura, Yuhei Sakaguchi, Chris W. Rella, and Edward Wahl, “Real-time ultra-sensitive ambient ammonia monitor for advanced lithography,” SPIE Proceedings, Volume 6349, October 2006.