Characterizing and abating ion implant process effluents

Josep Arnó, Joseph Sweeney, Paul Marganski, and Brian Kingston, ATMI; and Steven Roberge and M. Christopher Dolan, Axcelis Technologies

Ion implant processes utilize materials that are potentially toxic to humans and damaging to the environment. Conventional dopants include arsenic, phosphorus, and boron, many precursors of which are delivered as gases in their halogenated or hydrogenated forms. Arsine (AsH₃), phosphine (PH₃), and boron trifluoride (BF₃), which are highly toxic (hydrides) and corrosive (halogens), are among the most common of the precursor gases. Table I indicates the threshold limit values (TLVs) and immediate-danger-to-life-and-health (IDLH) levels of these gases.

In order to ensure safe operations, many IC manufacturers use sub-atmospheric-pressure gas-delivery sources and implement other measures to minimize personnel exposure to implant dopants.

Only small fractions of the dopants introduced into the tool are implanted onto wafers. The bulk of the source materials are either deposited as solids on the walls of the source and beam-line chambers or are exhausted through various pumps. A typical ion implanter accommodates a number of turbo pumps, backed by roughing pumps that evacuate chambers in a continuous mode. In addition, cryogenic, or cryo, pumps collect condensable materials, which are vented intermittently during cryoregeneration processes.

Ion implantation tools present a unique emissions-abatement challenge: scrubbing low flow rates of highly toxic materials that are diluted with relatively high flow rates of nitrogen (N₂) from the roughing and cryo pumps. Four traditional abatement methods have the potential to meet this challenge. Water scrubbers using injected chemical enhancers can achieve the efficiency necessary to remove implant effluent materials. However, such systems' cost of ownership and relatively large footprint, as well as the need to dispose of arsenic-containing liquids, limit their practical use. Thermal and plasma abatement methods simply dissociate materials into solid form, thereby generating hazardous dust. In addition, the electrical or other fuel costs of such tools can be prohibitive, considering the relatively small volume of materials being vented from implanters. Thus, of the four possible methods, dry scrubbers seem best suited for this application.

Dopant Name TLV Concentration Value (ppm) IDLH Concentration Value (ppm) Arsine (AsH3) 0.05 3 Phosphine (PH3) 0.3 50 Boron trifluoride (BF3) 1a 25

Table I: Threshold limit values and immediate-danger-to-life-and-health values of selected dopant gases.

Whichever method is adopted, the optimal performance of the abatement control equipment relies on a profound understanding of the fate of the materials introduced into the tool as a function of time and type of pump. Few studies have been published describing the nature and concentration of gas-phase species emitted during ion implantation. One characterized the process effluents of a high-current implanter during implantation of AsH₃, PH₃, BF₃, and silicon tetrafluoride (SiF₄) by measuring the concentration of gas-phase species from roughing and cryo pumps using Fourier transform infrared (FTIR) spectroscopic techniques.¹ Another study, by other researchers, used the same method to analyze the exhaust of the roughing pump of a high-energy implanter during implantation of AsH₃ and PH₃.²

Figure 1: Schematic of the 200-mm ion implanter's chambers and pumping system, including the dry scrubbers installed on top of the tool.

The research project described in this article combined the comprehensive characterization of gas-phase species at the exhaust of all roughing and cryo pumps of a 200-mm implanter with the performance testing of point-of-exhaust scrubbers installed after each pump. The gas-phase analyses were performed at the Axcelis Technologies development laboratories in Beverly, MA, during standard implant processes using AsH₃, PH₃, and BF₃ sources. Measurements were taken with an FTIR spectrometer operating in a quantitative, continuous, in-line mode.

The efficiencies and pressure drops of the abatement tools were characterized during test implantation runs using the three common dopant species and during cryoregeneration. A comparison between inlet and outlet concentrations provided information about the effectiveness of the scrubbers. Pressure-drop and scrubbing efficiency also were examined periodically for one year to validate long-term operation of the abatement technique. This work describes gas-phase emissions only; it does not account for solid-phase by-products.

Experimental Setup and Equipment

For this study, 40 X 20 roughing pumps from Ebara (Tokyo) were used. The roughing and cryo pumps installed on an Axcelis high-energy 200-mm GSD-VHE ion implanter were equipped with zero-footprint Novasafe dry scrubbers from ATMI (Danbury, CT). Figure 1 is a schematic of the implanter chambers, pumps, and scrubbers, and the photograph in Figure 2 shows three of the small scrubbers installed on top of the implanter tool. Drawings showing exterior and interior views of a scrubber are provided in Figure 3. Measuring 30 cm high and 25 cm in diameter, each scrubber is filled with two layers of chemisorption media with the capacity to abate hydride and acid implant dopants. Sampling ports installed at the inlet and outlet of the scrubbers were used to collect gas streams into a Model I2000 FTIR spectrometer from Midac (Irvine, CA) and a portable gas monitor, respectively.

Figure 2: The ion implanter with three scrubbers installed on top.

The spectrometer used to measure the implanter effluent gases at the scrubber inlet included a multipass (10-m path length) gas cell to quantitatively measure the concentrations of IR-active materials in situ in a continuous mode. The instrument was calibrated against the implant dopants AsH₃, PH₃, and BF₃. Full-spectra (from 600 to 4500 cm^–1) measurements were collected periodically (every 4–16 seconds depending on the specific experiment) at a 0.5-cm^–1 resolution. Samples were drawn continuously at an approximately 4-std L/min flow rate using a Teflon diaphragm pump.

The detection limits of the portable gas monitor used to analyze the scrubber outlet stream were 15 ppb, 32 ppb, and 0.6 ppm for AsH₃, PH₃, and hydrogen fluoride (HF), respectively. The monitor's output signal was logged into a computer to enable continuous monitoring of each scrubber's effluent stream. In some instances, the FTIR spectrometer was also used to analyze the scrubber exhaust, to corroborate the measurements collected using the portable monitor. Absolute pressures before and after the scrubber were measured using pressure transducers and were logged at 1-second intervals in order to provide time-dependent pressure characterization. Also, a 0–1-in.H₂O pressure differential magnehelic gauge provided precise pressure-drop information across the scrubber.

Figure 3: External and internal three-dimensional views of a point-of-exhaust dry scrubber.

Implanter Effluent Characterization

Roughing-Pump Emissions. The effluent gases from all three implanter roughing pumps were analyzed during ion implantation using AsH₃ and PH₃ from ATMI SDS sources and high-pressure BF₃ dopant gas. The materials evacuated from the tool were diluted with nominal 10.2-, 11-, and 6-std L/min nitrogen flows contributed by the ballasts of the RP1, RP4, and RP2 dry pumps, respectively (see Figure 1). Three standard recipes covering a wide range of tool operation parameters were tested for each gas source used. Prior to running implant processes, the analyzer was calibrated in situ by measuring the concentration of the gases emitted from the source chamber pump (RP1) at different flow rates with the beam off.

Figure 4: Source chamber emissions through roughing pump RP1 during calibration and implantation of AsH3.

During arsine implantation, small yet measurable levels of AsH₃ were detected at the exhaust of the source-chamber roughing pump, as seen in Figure 4. (Respective parameters for beam energy, beam current, and gas flow for the three recipes used in this testing were 200 keV, 300 µA, and 1.8 std cm³/min; 2 MeV, 600 µA, and 1.5 std cm³/min; and 90 keV, 11 mA, and 2.2 std cm³/min.) As observed in previous studies, the unstable beam conditions occurring during the beam-tuning stages resulted in concentration spikes.^1,2 Once the ion beam stabilized, AsH₃ concentrations of up to 1.8 ppm (more than 35 times arsine's TLV) were detected. Levels of emitted AsH₃ increased with beam energy and flow rate.

Figure 5: Source chamber emissions through roughing pump RP1 during calibration and implantation of PH3.

After implantation using arsine, the tool was purged with an argon beam and tests using phosphine began. The recipes used in that testing were 200 keV, 400 µA, and 2.5 std cm³/min; 1 MeV, 550 µA, and 2.8 std cm³/min; and 90 keV, 1 mA, and 1.9 std cm³/min. As had occurred with AsH₃, some PH₃ exited the tool through the source pump during implantation, as shown in Figure 5. Up to 20 ppm of the gas (corresponding to more than 65 times phosphine's TLV level) were detected initially, but emissions later stabilized near 10 ppm.

High-pressure BF₃ was the last dopant selected for characterization. In this case, the three beam energy, beam current, and gas flow process recipes were 1 MeV, 1 mA, and 2.8 std cm³/min; 200 keV, 500 µA, and 3.0 std cm³/min; and 90 keV, 1 mA, and 2.6 std cm³/min. In contrast to the hydride gases, a large percentage of the BF₃ gas delivered into the tool was evacuated through roughing pump RP1, as illustrated in Figure 6. Concentrations approaching 300 ppm (95% of the total volume of delivered gas) were detected. A nearly constant level of 10–15 ppm of HF was also measured during BF₃ calibration and implantation. The ratio of BF₃ emitted through this pump to the gas-delivery rate decreased with increasing beam energy.

Figure 6: Source chamber emissions through roughing pump RP1 during calibration and implantation of BF3.

The striking difference between hydride and acid emissions through the source pump is consistent with previously reported studies.¹ This disparity is independent of the type of implanter (low- or high-current), suggesting that there is a fundamental relationship between the bond energies of the individual dopant gases and the levels of gas emission.

Tests using the same source gases and implant processes were also performed with the analyzer installed after the beam-line chamber pump (RP4). As shown in Figure 7, the concentrations of source materials measured at the exhaust of this pump were typically one to two orders of magnitude lower than those detected after RP1. The implantation of BF₃ resulted in concentrations of approximately 16 ppm of source materials exiting through the beam-line pump. This value is ~20 times lower than the BF₃ volumes released through the source chamber pump. Implantation using AsH₃ showed an initial release of 600 ppb of the parent material, with subsequent effluent levels falling below the detection limits of the analyzer (~100 ppb). PH₃ implant recipes also resulted in an initial surge in effluent level to 650 ppb), which stabilized to ~350 ppb.

Figure 7: Beam-line chamber emissions through roughing pump RP4 during implantation of BF3, AsH3, and PH3.

In addition, characterization attempts at the exhaust of the RP2 roughing pump that serviced the process chamber during implantation using the various gas-flow recipes failed to measure any dopants or hazardous by-products. The emissions data for roughing pumps RP1 and RP4 are summarized in Table II.

Test No. Implant Conditions and Dopant Delivery Information Emissions Data Measured at the Exhaust of Pump Source Utilization by Tool (%)b Gas Source Gas Flow Rate (std cm3/min) Beam Current (mA) Beam Energy (MeV) Source Gas in Mass Rate (mg/min) RP1 (Source) Pump RP4 (Beam-line) Pump Avg. Concen. (ppm) Total Mass Emitted (mg)a Avg. Concen. (ppm) Total Mass Emitted (mg)a 0 AsH3 1.8 Beam auto tune 5.85 2.6 0.666 Not measured 98.58 1 AsH3 1.8 0.3 0.2 5.85 0.4 0.0829 0.0215 3.3 X 10–3 99.72 2 AsH3 1.5 0.6 2.0 4.88 0.163 0.0441 0.12 0.0282 99.88 3 AsH3 2.2 11.0 0.09 7.15 1.11 0.158 0.0166 2.62 X 10–3 99.35 4 PH3 2.5 0.4 0. 3.54 12.2 1.07 0.35 0.035 94.84 5 PH3 2.8 0.55 1.0 3.97 9.5 0.697 0.38 0.0181 96.41 6 PH3 1.9 1.0 0.09 2.67 6.5 0.189 0.38 0.017 93.14 7 BF3 2.8 1.0 1.0 7.91 256.4 20.9 6.2 1.78 8.11 8 BF3 3.0 0.5 0.2 8,48 289.6 39.6 7.5 1.89 3.32 9 BF3 2.6 1.0 0.09 7.34 258.4 21.2 7.9 1.99 3.09 aMass totalized (integrated) over each test time period. bThis value was computed by comparing the combined RP1 and RP4 emissions with the dopant delivery rate.

Table II: Summary of implanter conditions and emissions data for roughing pumps RP1 and RP4.

Cryo-Pump Regeneration Emissions. The volumes and types of materials trapped inside cryo pumps depend on the time elapsed since the last regeneration, the tool utilization rate, and the flow rate and nature of the dopants being implanted. In this study, the P2 and P9 cryo pumps for the 200-mm implanter were regenerated 30 days before the test data, and the P3 pump was vented 13 days before the testing. Based on the tool operator's estimation that the implanter was typically active for 6 hours a day 5 days a week, the elapsed time was 132 hours for the P2 and P9 pumps and 60 hours for pump P3. These cryo-pump loading periods would correspond to just 5.5 days and 2.5 days in a typical manufacturing tool operating in a 24/7 environment. The volumes and types of dopant gases used since the last regenerations were unknown.

To measure emission levels separately during testing, cryo-pump regenerations were triggered manually one pump at a time. The bulk of the gas streams was composed of dry nitrogen, which is commonly used to vent cryo pumps. N₂ flow rates of 200, 205, and 185 std L/min were estimated from the pressure drop across the scrubbers for the P2, P3, and P9 pumps, respectively. Gas-phase analyses during cryo regeneration detected parts-per-million levels of PH₃ and sulfur hexafluoride in addition to water, CO, and CO₂. The concentration-versus-time curves were integrated to determine the total volume and mass released by each pump. These values are summarized in Table III.

The relatively low levels of hazardous materials detected in this part of the study were not surprising considering the relatively small loading of the cryo pumps and the high N₂ dilution factors. A more representative study of cryo-pump emissions should be performed on an ion implanter that had been loaded to manufacturing-like levels.

Effluent Abatement Tool Efficiency

Conventional dry scrubbers used for abating ion implantation processes can be categorized as end-of-pipe or point-of-use (POU), depending on their size, location, and flow capacity. Typical POU scrubbers are placed adjacent to the implanter and treat the combined effluents of all roughing and cryo pumps. Pressure-drop considerations typically limit the flow through these scrubbers to <15 cu ft/min. On the other hand, end-of-pipe scrubbers are larger, typically installed remotely, and designed to process >100 cu ft/min of flow with a low pressure drop.

The new generation of dry scrubbers investigated in this study were designed to be installed at the exhaust of each separate pump of an implanter. Compared to end-of-pipe and POU systems, these point-of-exhaust scrubbers do not require expensive floor space or routing of exhaust lines. The results of their performance testing are presented below. Relative pressure data are reported in the familiar unit, inches of water (in.H₂O).

Pressure-Drop Performance. Excessive flow constriction through a dry scrubber can result in potentially damaging pressure buildup. Pressurized exhaust lines can trigger pressure switches that shut down pumps or induce leaks of hazardous substances through faulty connections or fittings. Pressure drop is related to the flow rate through the system, among other factors. In an implanter tool, the exhaust of a roughing pump has a typical flow rate of 10–70 std L/min of inert gas ballast. More detrimental to pressure drop, however, are the large volumes of gases that are vented when the tool chambers are evacuated. During these short-lived pumpdowns from atmospheric pressure, pump throughputs can peak at the nominal pump capacity. For example, a 2800-std L/min flow can occur when using a 100-cu ft/min pump. As opposed to roughing pumps, cryo pumps only vent their contents during regeneration, but during those events, nitrogen-purge-gas flow rates can exceed 200 std L/min.

Emission Type P2 Pump P3 Pump P9 Pump Volume (cm3) Mass (mg) Volume (cm3) Mass (mg) Volume (cm3) Mass (mg) CO 4.19 4.89 0.92 1.07 0.17 0.20 CO2 33.3 61 73.6 135 29.3 53.7 PH3 5.8 8.2 10.9 15.5 3.72 5.27 SF6 0.79 4.8 0.02 0.10 0.01 0.08 H2O 1540 1155 3364 2523 1290 967.3

Table III: Summary of cryo-regeneration emissions data.

Pressure drop across the respective scrubbers was measured under typical process conditions as well as during pumpdowns of the source, beam-line, and process chambers that were performed using the three separate 35-cu ft/min roughing pumps. The pumpdown measurements were repeated twice, with and without the use of an AutoSoft flow-dampening device from MKS Instruments (Andover, MA). Such devices, which are installed in-line before the pump, consist of a flap gate that automatically shuts under excess flow conditions and opens during normal flow conditions. The facilities draw provided a negative pressure with respect to the room environment (measured at the outlet of the scrubber) of –2.14 in.H₂O.

The test results indicated that the pressure drops across the scrubbers during regular process conditions were low, ranging between 0.005 and 0.02 in.H₂O, depending on the pump being used. As expected, pressure drops increased momentarily during pumpdowns because of the brief spikes in gas-flow rates. Without the use of the flow-dampening device during chamber pumpdown, maximum transient pressure drops of 4.2 and 2.7 in.H₂O were measured across the scrubbers installed after the beam-line and source chamber (RP4 and RP1 pumps, respectively). Use of the dampening devices reduced these pressure-drop spikes by approximately an order of magnitude. Figure 8 depicts the time-dependent pressure at the inlet and outlet of the RP4 scrubber during pumpdown of the beamline chamber.

Figure 8: Time-dependent characterization of the pressure drop across the scrubber installed following pump RP4 during pumpdown of the beam-line chamber, with and without a flow-dampening device.

The pressure drop across the scrubber that simultaneously treats the exhaust streams of all three cryo pumps was also characterized. Regeneration of each cryo pump was triggered manually to capture the individual pressure-drop contributions; the resulting pressure drops across the scrubber were similar, averaging 0.37 ± 0.002 in.H₂O. All three cryo pumps were also regenerated simultaneously, resulting in a pressure drop across the scrubber of 0.91 in.H₂O.

Abatement Performance. To evaluate scrubber efficiency, FTIR spectroscopy (for the scrubber inlet) and portable gas monitors (for the scrubber outlet) were used during implant processing using standard recipes and AsH₃, PH₃, and BF₃ sources. Test conditions included the delivery of up to 5 std cm³/min of dopant gases into the scrubber with the ion beam off. This challenge was sustained for at least 10 minutes for each gas. Throughout this part of the study (in tests involving all roughing and cryo pumps), no chemical breakthroughs were detected. Scrubbing efficiencies of >99.995% were achieved during the full hydride challenge (a 310-ppm inlet concentration was reduced to <0.015 ppm at the outlet). Similarly, efficiencies of >99.8% were derived for BF₃ (a 410-ppm inlet concentration was reduced to <0.6 ppm at the outlet). Confirmation of the scrubber efficiency data was obtained by measuring the scrubber exhaust using the FTIR spectrometer while delivering maximum flow rates of AsH₃, PH₃, and BF₃, with the beam off.

Figure 9: Results of periodic pressure-drop characterization of various scrubbers during normal operation, pumpdown, and cryo regeneration.

Long-Term Scrubber Performance. Since their installation, the performance of the scrubbers has been monitored periodically to provide comprehensive data on their capabilities over time. A 10-month summary of the pressure drop across the systems during normal operation, pumpdown, and cryo regeneration is provided in Figure 9. In all of these tests, the roughing pumps were equipped with flow-dampening devices. The periodic measurements of the scrubber exhaust gases revealed no breakthroughs, indicating that the units continued to abate all toxic gases from the implanter.

Additional long-term scrubber performance data have been provided by Atmel (Beverly, MA), where 30 of the units were installed in mid-2000. Similar to Axcelis's installation, all of these scrubbers are placed on top of implanters to abate the effluent gases from roughing and cryo pumps. In this manufacturing environment, the end-user reported that units installed on high-current implanters have a typical lifetime of 6–12 months, while those installed on medium-current tools have a lifetime in excess of 1 year.

Conclusion

Ion implantation processes emit materials that are potentially hazardous to humans and the environment. While the effluent characterization study presented here revealed that measured absolute concentrations of dopant gases are small, the levels relative to their respective safety thresholds were found to be significant. Comparison of this project's results with those of other emission studies suggests that the frequency and intensity of vented gases are independent of the type of tool used (high-energy versus high-current). Instead, gaseous emissions are correlated to the dopant being implanted (hydrides or acid materials), implanting parameters (dopant flow rates, beam energy, and beam current), type of pump (roughing or cryo), and location of the pump within the tool.

The performance of dry scrubbers dedicated to abating implantation tool effluent was also studied. Compared with conventional methods of pollution control, these small devices simplify the abatement process by allowing easy installation at the exhaust of each implanter pump without requiring an additional equipment footprint. Periodic monitoring revealed that the units can provide from 6 to 12 months of service, depending on implanter type and uptime. In performance tests, the pressure drop across the devices was typically <1 in.H₂O during normal operation; during pumpdown of the entire tool, short transient pressure drops reached 4 in.H₂O. No chemical breakthroughs occurred during the study, and scrubber efficiences were measured at >99% for the three dopants used in the testing.

Acknowledgments

The authors would like to acknowledge the assistance of Rob Spreng, who operated the ion implanter during the characterization study and the ensuing performance scrubber tests at Axcelis. In addition, special thanks go to Chad Ramirez of Atmel for providing information on the performance of Novasafe scrubbers installed at that facility.

References

1. J Arnó et al., "Hazardous Gases Emitted by Ion Implanters; Characterization and Abatement," Solid State Technology 41, no. 10 (1998): 113–120.

2. B Goolsby et al., "Analysis of Implanter Source-Region Exhaust by FTIR," in Proceedings of the Electrochemical Society, vol. 2001–6 (Pennington, NJ: Electrochemical Society, 2001), 59–67.

Josep Arnó, PhD, is a director of R&D at Advanced Technology and Materials (ATMI) in Danbury, CT, where he is responsible for the development of innovative abatement methods and the safe packaging of semiconductor materials. He has research experience in molecular spectroscopy, computational quantum mechanics, and plasma abatement. Arnó has authored more than 20 publications in the areas of front-end and back-end semiconductor materials handling and processing. He received a PhD in physical chemistry from Texas A&M University in College Station. (Arnó can be reached at 203/207-9304 or jarno@atmi.com.)

Joseph Sweeney is R&D manager for the treatment group of ATMI. He has seven years of experience in the semiconductor industry and has focused primarily on the development of new abatement products. He received bachelor's degrees in chemical engineering and chemistry from the University of Wisconsin in Madison and is working toward a masters degree in mechanical engineering from the Georgia Institute of Technology in Atlanta. (Sweeney can be reached at 203/207-9364 or jsweeney@atmi.com.)

Paul Marganski is a research engineer at ATMI. For the past three years, he has been developing novel methods for improving hazardous materials abatement in ion implant, III-V MOCVD, and metal etch. In addition, he has been working on wet bench exhaust reduction. Prior to joining ATMI in 2000, Marganski worked for the chemical process design group at Cytec Industries in Wallingford, CT. He received a BS in chemical engineering from the University of New Haven in Connecticut. (Marganski can be reached at 203/739-1464 or pmarganski@atmi.com.)

Brian Kingston is marketing manager for the treatment group at ATMI, a position he has held for four years. Previously, he was in customer support at IBM in New Zealand and at Honeywell in the UK. He received a BS in business and is pursuing an MBA in marketing at Phoenix University. (Kingston can be reached at 408/933-3518 or bkingston@atmi.com.)

Steven Roberge is the global environmental health and safety (EHS) and product stewardship manager at Axcelis Technologies (Beverly, MA). He is involved in all aspects of EHS management and product stewardship for ion implanters, rapid thermal processors, and cleaning and curing systems. Roberge has authored several papers and presentations on various aspects of semiconductor EHS and has contributed to the safety chapter of the Ion Implant Science and Technology Handbook. He received a BS in environmental science from the University of Massachusetts in Amherst. (Roberge can be reached at 978/787-9889 or steve.roberge@axcelis.com.)

M. Christopher Dolan is high-energy engineering product manager at Axcelis, where he has worked for eight years. He received a BS in mechanical engineering from Northeastern University in Boston. (Dolan can be reached at 978/787-4000 or chris.dolan@axcelis.com.)