 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
Characterizing effluents from PECVD, plasma etch processesRobert G. Ridgeway, Peter J. Maroulis, and Richard V. Pearce, Air Products and Chemicals There has been increasing interest in the type, quantity, and environmental impact of gaseous emissions from the manufacturing of semiconductor materials over the past several years. Production facilities must control their emissions of hazardous air pollutants (HAPS) and, more recently, perfluorinated compounds (PFCs) to regulated or agreed-on levels. This is typically achieved using postprocess treatment methodologies to abate or capture unreacted process gases and process by-products. The types of process emissions include diverse chemical species, including reactive, corrosive gases (hydrogen chloride [HCl], fluorine [F2]); pyrophoric gases (silane [SiH4]); and inert gases (PFCs, e.g., tetrafluoromethane [CF4], hexafluoroethane [C2F6]), all of which are used in various wafer processing steps or formed as process by-products. As IC manufacturers increase both production capacity and the complexity of manufactured devices, more wafer processing steps requiring the use of more process gases will result in increased emissions. In the study reported in this article, emissions from two distinctively different processes were sampled at a major semiconductor manufacturing facility to identify and quantify process by- products, quantify unreacted process gases, and determine utilization efficiencies of these gases. These processes included plasma-enhanced chemical vapor deposition (PECVD) of silicon oxide and plasma etching of polysilicon. Quadrupole mass spectrometry (QMS) was used as the analytical method in the study because of its high sensitivity, large linear dynamic range, speed of response, and ability to withstand the reactive, corrosive nature of the effluent streams being analyzed. On-site calibrations of QMS response to several of the species identified in the effluent were performed to enhance the accuracy of reported data. A significant objective of this study was to determine QMS's effectiveness as a tool for carrying out quantitative effluent analyses. The information gained during this study will be applied to understanding what additional technology is required to improve the reliability of process emissions data. Experimental Methods All sampling was performed nonintrusively downstream of the process tool vacuum system. A closed-loop sampling system, illustrated in Figure 1, drew process exhaust gas through the area of the QMS housing the source. The excess sample gas (that which did not flow into the QMS source) was combined with QMS pump exhaust and safely vented from the work area. A sample pump with a throttling valve maintained sampling pressure at 720 torr. This represented a pressure drop of 30 torr through the sample system, resulting in a sample flow of approximately 0.5 std L/min.
The QMS was operated in both the analog and selective ion monitoring (SIM) modes. The analog mode obtained a complete mass spectrum from 0 to 200 amu. These spectra aided in the qualitative identification of process by-products. The SIM mode monitored selected ions of interest. These SIM profiles helped quantify unreacted process gases and process by-products. Calibration of the QMS response to compounds of interest is critical if quantitative results are desired. The sampling of process exhaust, which contains many reactive, corrosive species, can cause changes in QMS sensitivity. The tool must be calibrated frequently to ensure accurate quantitative results. The dynamic dilution method was used to directly calibrate the QMS response to several compounds. The response factors of by-products for which no standards were available were calculated based on the response factors obtained for other compounds at the time of sampling and the ratios of physical parameters associated with ionizing and detecting molecules. These parameters included total ionization cross sections, standard fragmentation patterns, and natural isotopic abundance of atoms contained in the molecules. Experimental Results PECVD of Silicon Oxide. To study this process silicon oxide was deposited on wafers using tetraethylorthosilicate (TEOS) and oxygen. Cleaning of residual silicon oxide from chamber walls was done using a combination of C2F6, O2, and nitrogen trifluoride (NF3). Effluent from two different clean recipes was monitored. The recipes consisted of a two-step clean using 900 std cm3/min each of C2F6 and O2 for the first step, followed by the addition of 75 std cm3/min of NF3 to C2F6 and O2 during the second step. The second recipe was a one-step clean using 600 std cm3/min each of C2F6 and O2 and 75 std cm3/min of NF3. The primary focus of this analysis was on effluent resulting from the chamber cleans, since the chemistry used for the cleans results in HAPS and PFC emissions. Process gases such as NF3, C2F6, and O2 were detected in the effluent, as were such major by-products as silicon tetrafluoride (SiF4), carbonyl fluoride (COF2), hydrofluoric acid (HF), F2, and CF4. During deposition, TEOS was seen initially during the stabilization step, but the QMS signal attributable to TEOS disappeared 20 seconds after deposition began.
Figure 2: C2F6 emissions from one-step chamber clean.
Figure 3: NF3 emissions from one-step chamber clean. An important objective of this study was to determine the fraction of PFCs converted to reactive by-products and other PFCs during the chamber clean. These PFC utilization efficiencies (UE) were determined from data shown in Figures 2 and 3. The concentration profiles corresponding to PFC emissions with RF power turned off were used to calculate UE for NF3 and C2F6 during chamber cleans. From these data PFC UE values were tabulated and are shown in Table I; these data also included evidence of CF4 as a by-product of the chamber cleans. These results are presented as CF4 formation efficiency (FE) values and are expressed as the volume-based percentage of C2F6 converted to CF4. Table I also presents net carbon-based PFC UE values.
Table I: PFC utilization efficiency (UE) and CF4 formation for plasma chamber cleaning of TEOS/oxide films. The data in Table I show a significant difference in PFC emission factors for the two processes. The presence of NF3 appears to enhance both C2F6 use and CF4 formation. The net carbon-based PFC utilization was higher in the one-step clean. This finding, when coupled with a 33% reduction in C2F6 flow rate, results in a 57% overall reduction in carbon-based PFC emissions, provided that the cleans are of equal duration. Further reductions can be achieved if the clean times can be reduced for the one-step clean through process optimization. NF3 UE values were slightly lower during the two-step clean. Despite higher UE values, annualized NF3 emissions would be higher during the one-step clean since NF3 is used for the entire clean. However, because of low flow rates and high UE values, the emissions increase for NF3 is estimated to be only 1020 lb/year, depending on the number of wafers processed. In addition to CF4, such significant by-products as COF2, SiF4, F2, and HF were observed in process exhaust during chamber cleans. The concentration profiles in the effluent were integrated with total exhaust flow, which was determined from the process pump purge, for each species to obtain total emission volumes per processed wafer. A volume-based fluorine balance was calculated by combining the PFCs consumed in the process with the integrated emissions. Tabulated results for four wafers are shown in Table II.
Table II: Integrated by-product emissions and volume-based fluorine balance for plasma chamber cleaning of TEOS/oxide films using single-step C2F6/O2/NF3 chemistry. All emission values are reported as volumes.
The data contained in Table II indicate that 91.7% of fluo- rine entering the process tool can be accounted for as volatile emissions. The remaining 8.3% of fluorine not quantified in the effluent could result from several factors, including measurement error. Based on data contained in Table II, the precision of these determinations was quite high with relative standard deviations (RSD) of Other factors that could contribute to the unaccounted-for fraction of fluorine include loss of fluorine in condensed phases (such as particles) and loss on surfaces within the system, such as the process foreline, process pump, and exhaust tubing. These effects would reduce the quantity of fluorine exiting the system in the gas phase, thus remaining undetected. Additionally, the presence of any unidentified by-products would reduce the detected amount of fluorine emitted. Polysilicon Etch. The effluent from a polysilicon etch process using fluorine, chlorine, and bromine chemistry for oxide breakthrough and polysilicon etch was monitored. The process gases included C2F6, Cl2, and hydrogen bromide (HBr), all of which yielded a diverse mix of by-products. The polysilicon etch steps consisted of a main etch (ME) and an overetch (OE). Fluorine-based emissions included C2F6, SiF4, CF4, and HF as well as potentially higher-molecular-weight perfluoronated carbon (CxFy) species, which were not observed during PECVD chamber cleans. The duration of oxide breakthrough was only 15 seconds, a small fraction of total processing time. Thus, volume emissions of fluorine-containing species were significantly less than the chlo- rine- and bromine-containing species emitted during ME and OE.
Figure 4: HCl and Cl2 emissions from polysilicon etch process.
Figure 5: HBr emissions from polysilicon etch process. Primary chemical species observed in the effluent during ME and OE included unreacted Cl2 and such by-products as HCl, silicon tetrachloride (SiCl4), bromine (Br2), and bromine monochloride (BrCl). Low levels of HBr were also emitted. Figures 4 through 8 show emission levels for several of these species during ME and OE processes. These data indicate that HBr is converted to Br2 and BrCl with a high degree of efficiency in the process. Formation of BrCl is believed to result from the following reaction:
Evidence supporting this chemical reaction is seen in the profiles of HCl and BrCl, which increase, and that of Cl2, which decreases during OE when the HBr flow rate is increased.
Figure 6: SiCl4 emissions from polysilicon etch process.
Figure 7: Br2 emissions from polysilicon etch process.
Figure 8: BrCl emissions from polysilicon etch process. Volume-based emission balances relative to process gas input volumes were calculated for Cl and Br during ME and OE processes (see Table III). Emission volumes were determined based on average emission concentrations for each species and a total effluent flow of 30 std L/min, which was primarily composed of nitrogen ballast and purge of the process vacuum pump. Each value in Table III is based on average emissions from four wafers.
Table III: Volume balance calculations for influent and effluent flows of chlorine and bromine during polysilicon etch. Volumes were calculated from the average of the measured concentration and the total volume of effluent. The volume balance calculations for Cl and Br, reported as equivalent standard liters, are as follows:
As in the case of the CVD process discussed above, the loss of chlorine in condensed phases or unidentified chlorine-containing by-products could account for the fraction of chlorine not quantified. Measurement error could also have contributed, but direct calibrations of QMS response to HCl and Cl2 were performed on-site, improving the accuracy of HCl and Cl2 determinations. The volume balance closure obtained for bromine was higher than that of chlorine. More than 99% of HBr used in the process was converted to by-products. Conclusion This article has demonstrated the successful deployment of mass spectrometry to monitor selected semiconductor processing emissions from both PECVD chamber cleaning and polysilicon etching. A uniquely designed QMS system along with an on-site calibration system provided emissions data that could be used to determine PFC utilization efficiencies, quantify reactive by-products, determine HAPS emissions, and estimate volume balances for halogens used during processing. These kinds of data are useful in developing emissions models for the entire fab, which can then be used for environmental permitting. These data can also be helpful in assessing abatement requirements based on the type and quantity of chemical loading. As an example, consider abatement of effluent from the polysilicon etch process. Based on data presented here, system requirements would include the ability to abate chlorine, hydrogen chloride, bromine, and bro- mine monochloride, but the system would not be required to abate significant quantities of hydrogen bromide since the conversion to by-products is almost complete. Studies of this nature will be valuable in developing cost- and performance-effective waste treatment systems, which will include abatement devices and reclaim/recycle systems that will enable IC manufacturers to capture and reuse un-reacted process gases. (Such systems are being developed for selected PFCs.) This study also indicated areas where improvements in measurement capabilities are needed. Since calibration is necessary to obtain accurate results, an expanded range of calibration standards is required to more accurately determine the emissions of all by-products. Alternative ionization schemes, including positive and negative ion chemical ionization, may be called for to identify certain by-products, and QMS techniques used in conjunction with Fourier transform infrared (FTIR) spectroscopy and other techniques may be needed to fully characterize all process by-products. Acknowledgment This article is a revised version of a paper originally presented at the Institute of Environmental Sciences' 42nd Annual Technical Meeting, Orlando, FL, May 1996. Used with permission. Robert G. Ridgeway, PhD, is a senior principal research chemist in the customer applications group at Air Products and Chemicals in Allentown, PA. He is responsible for working with IC manufacturers and process tool OEMs to characterize and quantify semiconductor process by-products and to evaluate the efficiency of process exhaust abatement systems. Ridgeway received his PhD from Drexel University in 1991. He holds one patent and has more than 26 publications in the fields of analytical and atmospheric chemistry. (Ridgeway can be reached at 610/481-5824.) Peter J. Maroulis, PhD, is a research associate and manager of analytical technical services and customer applications in the electronics division of Air Products and Chemicals. He is responsible for working closely with IC manufacturers and OEMs to analyze gas quality entering the process tool and exiting the tool and abatement systems and to investigate process chem- istries in the wafer environment. Maroulis received his PhD from Drexel University in 1980. He has six patents and 25 publications in the areas of trace gas analysis and atmospheric chemistry. (Maroulis can be reached at 610/481-6242.) Richard V. Pearce is a principal research technician in the customer applications group of Air Products and Chemicals. He is responsible for working with IC manufacturers and OEMs to investigate gas purity requirements for device manufacturing. He is also involved with the characterization and quantification of process tool by-products and the evaluation of PRC abatement and recycle/reclaim systems.
MicroHome |
Search | Current Issue | MicroArchives Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com. © 2007 Tom Cheyney |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Figure 1: Closed-loop sampling system used for process tool emissions monitoring.
5 % for all chemicals except CF4. The accuracy of these determinations was verified by direct calibration of the QMS response to several of the species found in the effluent. Since calibration standards were not available for all species, QMS response factors were calculated from the responses obtained for calibrated species and the physical parameters associated with creating and detecting ions in the spectrometer.
