Maximizing hydrogen pumping speed in cryopumps without compromising safety

Brian Thompson and Michael Eacobacci, Helix Technology

The harsh environment in which ion implant processes take place requires consistent vacuum conditions to maximize tool availability, throughput, and product yield. Ion implanters achieve maximum productivity only when optimal operating conditions, such as the ion beam's interaction with the vacuum environment, are maintained. To improve throughput and quality, users of ion implanters—particularly in high-beam-power processes in which high outgassing loads are generated—must increase high-vacuum pumping speed.

Because hydrogen liberated from photoresist is the dominant gas species during the implant process and is a contributor to high outgassing pressures, OEMs continue to focus on improving their systems' hydrogen pumping speed.

In this environment, cryopump designs must deliver the highest hydrogen pumping speed while adhering to all safety guidelines for hydrogen and other gas species. This article, based on work performed by Helix Technology (Mansfield, MA) and Atmel (San Jose), addresses the potentially competing demands for improved hydrogen pumping speeds and enhanced safety protocols in ion implantation vacuum systems.

Ion Implantation Processing Conditions

Gas Composition. An understanding of implantation process gas composition is essential for achieving the desired level of vacuum performance. The measurements illustrated in Figure 1, which shows the results of several residual-gas analyses (RGAs) from medium- and high-current implanters, correlate well with earlier work.^1,2 The relative composition of the gas shows that the fraction of the gas that is hydrogen increases with higher beam power. The percentage composition of the three main by-product gases from the implant process is plotted as a function of beam power. (Plotting the same data against beam energy or beam current results in a weaker correlation.) These data indicate that lower-power beams (mostly from medium-current implants) tend to generate a rather evenly distributed gas mixture, with no single constituent accounting for much more than 50% of the mixture. As the beam power grows—for example, during high-energy and high-current implants—hydrogen quickly begins to dominate the outgassing mixture.

Figure 1: Percent composition of photoresist outgassing for different implant recipes. The fraction of hydrogen gas increases with higher beam power.

Process Effects of Gas Pressure. One of the primary vacuum-related concerns associated with high-energy implanters is energy contamination. When very-high-speed ions interact with gases in the beamline, an extra electron may be stripped off the dopant before it passes through the final accelerating field. This ion is then delivered to the wafer at a higher energy than was originally intended. Uncontrolled pressure changes can also change the implant angle or reduce uniformity.

Low-energy ion beams can suffer from "blooming"—the gradual increase in width of the beam resulting from its positive space charge. To combat blooming, ion implant tools often add electrons to the beam space in a controlled manner. That is most often achieved by adding an easily ionized gas (xenon, for example) to the vacuum space. Electrons liberated from the flood-gas molecules help keep the space charge of the beam low, thereby helping transport the beam to the target.

Given their very large charge-exchange cross section, xenon atoms also interact directly with beam ions, often causing neutralization of dopant atoms. The photoresist-derived molecules can also cause this type of interaction. The dopant atoms may be impacted before final optics, so that these atoms reach the wafer at the wrong energy or do not reach it at all. The dopant atoms can also be neutralized after the final optics, in which case they will not be counted by the dose measurement system. In either case, implant quality is degraded. When pressure exceeds a given critical threshold, many implant control systems put the process on hold to give pumps time to catch up, adversely affecting tool throughput.

The issues facing medium-current implanters are similar to those that affect high-current and high-energy ones. Furthermore, scanned- or ribbon-beam medium-current tools use more-complicated beam optics to ensure parallel beam incidence on the wafer. If beam ions are neutralized by the residual gas within the electrostatic or magnetic collimating structure, incorrect implant angles can result.

The Role of Cryopumps in Ion Implant Processes

How Cryopumps Work. Cryogenic high-vacuum pumps, or cryopumps, create high vacuum by freezing molecules of air onto cryogenically cooled surfaces inside the pump. The vacuum typically ranges from 10^–4 to 10^–9 Torr. Cryopumps have no moving parts that are exposed to vacuum. Because they require no oil, they are are inherently clean and cannot contaminate products created in the vacuum they produce. Nothing but cryogenically cooled surfaces are exposed to the process chamber.

Cryopumps usually consist of two internal stages, or pump areas, which freeze or capture specific gas species at set cryogenic temperatures. The stages are connected to a sealed cryogenic refrigerator run by an external compressor. The refrigerator cools the stages. In some pumps, stage temperatures are adjustable, so that the pump can be tuned to target specific gas species for evacuation from the process chamber.

The first stage of the pump, the primary pumping surface of which is the inlet array, is generally operated at temperatures between 65 and 100 K. Its main function is to pump or capture water vapor. The second stage consists of a series of metal pumping surfaces, which are arranged in patterns designed for particular applications. Generally operated at temperatures ranging from 10 to 20 K, this stage can pump gases such as nitrogen and argon. The metal pumping surfaces are partially covered with charcoal granules. Gases such as hydrogen and helium, which cannot be frozen at typical second-stage temperatures, are adsorbed by the charcoal granules and thereby removed from the vacuum chamber. A schematic drawing of a cryopump is pictured in Figure 2.

Figure 2: Schematic drawing of a cryopump's internal arrays. DIAGRAM COURTESY OF HELIX TECHNOLOGY

Safety Implications of Increasing Cryopump Hydrogen Capacity. It is important to differentiate between the hydrogen pumping speed required to improve process control and the cryopump hydrogen capacity that enables long regeneration intervals. The hydrogen speed of a cryopump is measured using a speed dome constructed to the industry-standard specification ISO 1608-1:1993(E). Hydrogen capacity relates to the quantity of hydrogen the pump can "cryoadsorb" before its effective speed drops to 50% of its peak speed.³ Capacity is dictated by the total amount of cryoadsorbant present in the pump.

Although hydrogen speed and hydrogen capacity are independent functions, there is a practical reason for the widely held perception that they are linked. In most applications, absolute capacity, which is defined by the amount of cryoadsorbant present in the cryopump, is never reached because vacuum processes require a certain base pressure before production can start. Base pressure climbs as the cryoadsorbant in a cryopump fills.

The constant-flow-rate hydrogen-capacity curve in Figure 3 is based on a standard speed test specified in ISO 1608-1:1993(E). Derived from Helix Technology's On-Board IS 320FE cryopump, it shows pumping speed as a function of the total amount of hydrogen pumped and illustrates the
effects of filling the capture sites of the cryoadsorbant. As the gas accumulates and gradually saturates the cryoadsorption capacity of the pump's arrays, pressure increases and the effective speed decreases.

Figure 3: Hydrogen capacity chart for the IS 320FE and the 10F cryopumps. (Data from Helix Technology)

A key objective when designing a cryopump is to achieve very high hydrogen pumping speed. To achieve that objective, the overall conductance of hydrogen to the cryoadsorbant and the cryoadsorbant's hydrogen capture probability must be improved without increasing hydrogen capacity. While simply adding cryoadsorbant to the vacuum space would increase the hydrogen speed, it would also increase the absolute hydrogen capacity of the pump, raising the hazardous-chemical potential energy stored inside. Therefore, when developing the IS 320FE pump, engineers focused on maximizing conductance to the cryoadsorbant, maximizing the capture probability of the cryoadsorbant, and providing the optimal temperature for hydrogen capture at the cryoadsorbant.

Optimizing a Cryopump Design

Maximizing Conductance. Maximizing the conductance of hydrogen to the second-stage array and the cryoadsorbant there were investigated in parallel. Careful analyses of various vacuum-space designs using Monte Carlo calculation methods were conducted, enabling the investigators to optimize the size of the second-stage array for a radiation shield of a particular size. A design was developed that conducts hydrogen to the cryopump in the most optimized fashion and that also minimizes the probability that a hydrogen molecule will reflect back out of the cryopump once it reaches the inside of the radiation shield.

Figure 4 shows the results of optimizing a traditional array with a standard hydrogen capture probability of 40%. In this analysis, the diameter of the second-stage array was varied. As the diameter of the array was increased in size, its capture area and speed increased. At the same time, because a larger array occupied a greater fraction of the space available, the conductance of hydrogen to the sides and back of the array was reduced. A balanced approach resulted in an optimum design point, but it did not achieve the targeted gains in overall hydrogen pumping speed. Nevertheless, significant increases in the capture probability of the second-stage array were achieved in part by increasing hydrogen conductance to the hydrogen-capturing cryoadsorbant in the array.

Figure 4: Second-stage array optimization for standard capture probability. Optimum system performance is achieved with a design that balances an increase in the diameter of the second-stage array with the necessity of reducing conductance of hydrogen to the sides and back of the array.

The Monte Carlo calculation in Figure 5 illustrates the thermal aspects of an array design with a high hydrogen capture probability of >80%. As the array increased in size, the pumping speed increased. However, a larger array also resulted in greater heat load. In this design, the heat load that is absorbed by the second-stage array increased dramatically near the region where the optimum design point was reached. However, doubling the refrigeration requirements to handle the heat load is prohibitively expensive. System optimization must balance between increasing the capture probability of hydrogen molecules and increasing the amount of refrigeration required. Hence, the optimum design represents a balance between the peak shown in Figure 4 and the practical limit dictated by the cost of refrigeration. The correct balance was calculated with a higher first-strike capture probability than used in Figure 4, shifting the peak pumping speed to a higher diameter.

Figure 5: Second-stage array optimization for high capture probability. Overall pump design optimization must balance between increasing the capture probability for hydrogen molecules and increasing the heat load, and therefore, the amount of refrigeration required.

Maximizing the Hydrogen Capture Probability and Optimizing Temperature. Several factors contributed to overall improved capture probability of the cryoadsorbant:

• The temperature gradient between the cryogenic refrigerator second-stage heat station and the cryoadsorbant surface was minimized. To minimize the temperature gradient, the thermal conduction paths were sufficient for heat from the cryoadsorbant to be transferred to the second-stage heat station effectively. In addition, proper materials and methods for promoting a minimal temperature gradient while providing high-quality, durable adhesion were selected to adhere the cryoadsorbant to the arrays.

• The shape of the cryoadsorbant surface was microscopically rough. A rough cryoadsorbant surface benefits hydrogen pumping speed by maximizing the intrinsic capture probability. It also fosters secondary impacts of hydrogen molecules not adsorbed on initial impact. A rough cryoadsorbant surface also produces a much greater surface area for adsorption, thus increasing the saturation threshold.

• The temperature of the cryoadsorbant was maintained so that there was a balance between the capture probability at the surface and hydrogen diffusion across the cryoadsorbant bed. While second-stage temperature has very little effect on hydrogen capture probability in traditional cryoadsorbants, it has a significant effect on hydrogen diffusion away from the capture sites, as demonstrated through controlled testing.⁴ An optimum cryo-pump system controls second-stage refrigerator temperature at 13.5 K ± 0.5 K while maintaining the initial pumping speed as the pump fills with hydrogen.

Comparing Cryopump Designs. The foregoing analytical processes and design methods have led to considerable improvements in cryo-pump hydrogen pumping speed. For example, Figure 6, which shows hydrogen pumping speed as a function of pressure, demonstrates that the IS 250FE cryopump can attain a pumping speed of 7000 L/sec, an increase of 55% over the IS 250F. That increased speed occurs under true molecular-flow conditions and is consistent across several decades of pressure below the 10^–4 Torr range. The enhanced pump has the same hydrogen capacity as the standard model, demonstrating that additional cryoadsorbant is not required to improve hydrogen pumping speed significantly. This feature ensures that the pump adheres to safe design practices. Also shown in Figure 6, the IS 320FE 320-mm cryopump attains a hydrogen pumping speed of >12,000 L/sec. As shown in Figure 3, that pump holds enough cryoadsorbant to retain approximately 32 std L of hydrogen.

Figure 6: Comparison of hydrogen pumping speed versus pressure for three cryopumps. (Data from Helix Technology)

Power Loss Considerations. From a safety perspective, it is critical to properly manage cryoadsorbed hydrogen and condensed gases during short-term power losses. In case of a long-term power loss, the potentially reactive gases trapped in a cryopump must be safely vented. Therefore, cryopumps should feature normally open purge and exhaust valves.

End-users do not want to be forced to implement long, 2–3-hour regeneration cycles in the event of short power outages lasting as little as 30 seconds. Significant tool throughput gains can be realized by insulating end-users from negative effects caused by very brief power losses. Without such a mechanism, two or three 30-second power losses per week can lead to as many as 10 hours of unanticipated machine downtime and an attendant loss of productivity.

The cryopumps discussed here have an integrated backup power source (an uninterruptible power supply system) that holds the vacuum-space purge valve closed for a carefully selected 2-minute safety period. Consequently, the system guarantees high tool uptime, even in environments where short-term power losses are common. The time during which the vacuum-space purge valve is closed is much longer than typical short-term power losses. However, it is not so long that significant amounts of condensed gases can be liberated from second-stage surfaces, as can occur during extended power losses.

Demonstrating Ion Implant Tool Performance

To test the hydrogen pumping speed of the IS 320FE cryo-pump, the unit was installed on an EHPi-500 ion implanter from Varian (Gloucester, MA) and run under real-world conditions at Atmel in a semiconductor production environment. The RGA ion current data in Figure 7 show that for unbaked photoresist wafers, a substantial improvement in implanter performance was achieved after substituting an IS 320FE pump for the existing 10F model. The new-technology pump reduced hydrogen partial pressure by 43%, which not only improves implant uniformity but also increases the total tool throughput.

Figure 7: Comparison of hydrogen outgassing with two cryopumps. The new pump reduced hydrogen partial pressure by 43%. The multiple dose holds shown in the inset increase wafer process time, thus decreasing tool throughput. (Data courtesy of Atmel)

In addition to decreasing hydrogen pressure, the pump reduced the total peak pressure of the implanter end-station by as much as 33% when unbaked photoresist wafers were processed, as demonstrated in Figure 8. With that reduction, multiple dose holds could be eliminated, enabling the implanter to continue processing wafers without making disruptive stops and starts. As shown in the figure, the 10F pump operated much closer to the implanter's dose-hold limit than the IS 320FE.

Figure 8: Comparison of end-station peak pressure for two cryopumps. The new pump reduced peak pressure by 33%, eliminating multiple dose holds. (Data courtesy of Atmel)

Figure 9 compares the impact of the 10F and IS 320FE pumps on average maximum end-station pressure. The data are based on many sets of ion-implant runs that were performed under various process conditions using both baked and unbaked photoresist wafers at implant energies ranging from 400 to 500 keV. In all cases, end-station pressure improved.

Conclusion

Cryopumping systems will continue to be the best means for performing high-vacuum pumping in ion implantation systems for the foreseeable future. As companies struggle to optimize productivity, new vacuum-system designs must take into account all the gases produced during wafer processing, even as processes and gas-species by-products evolve and change.

Figure 9: Maximum end-station pressure over many wafer runs and processes. Average end-station pressure was reduced by 20%. (Data courtesy of Atmel)

The extremely high conductance of hydrogen through the gate valve and spool-piece assembly that is used in cryopump installations has simplified the task of optimizing hydrogen pumping speed. Hence, the capture probability for hydrogen has been increased to more than 30%. For atmospheric gases that have very low vapor pressure at 13 K, the capture probability is about 35%, and for water it is already close to unity. In contrast, it will be very difficult to achieve further improvements in pumping speeds for water vapor or nitrogen because of the significant conductance limit imposed by the gate valve and spool-piece interface assembly.

Large incremental increases in pumping speed will require significantly new architectures to increase the solid angle subtended by the pumping array at the gas source. Pumps will either have to be larger or be positioned closer to the point where the beam strikes the wafer.⁵

It is expected that the continued extension of distributed intelligence will drive both ongoing safety enhancements and cost-of-ownership improvements. These trends will require closer collaboration between equipment manufacturers and designers of vacuum-pumping subsystems, resulting in further refinements and improvements in both process control and system uptime and productivity.

References

1. TN Horsky, "Photoresist Outgassing in High Energy and High Current Ion Implantation," in Proceedings of the International Conference on Ion Implantation Technology 1998, vol. 1 (Piscataway, NJ: IEEE, 1999), 654–657.

2. N Tokoro et al., "Consideration of Photoresist Outgassing for MeV Ion Implantation and Cryopump Selection," in Proceedings of the International Conference on Ion Implantation Technology 1998, vol. 1 (Piscataway, NJ: IEEE, 1999), 614–617.

3. M Kimo et al., "Recommended Practices for Measuring the Performance and Characteristics of Closed-Loop Gaseous Helium Cryopumps," Journal of Vacuum Science and Technology 17, no. 5 (1999): 3081–3095.

4. PA Lessard, "Cryogenic Adsorption of Noncondensibles in the High Vacuum Regime," Journal of Vacuum and Science and Technology 7, no. 3 (1989): 2373–2376.

5. S Furuya et al., "Development of Cryopump for Ion Implantation Equipment," in Proceedings of the International Conference on Ion Implantation Technology 1998, vol. 1 (Piscataway, NJ: IEEE, 1999), 396–399.

Brian Thompson is a technical product manager for the CTI business group at Helix Technology (Mansfield, MA). He is responsible for the company's On-Board IS cryogenic vacuum systems used in ion implantation processes. He received a BS in mechanical engineering and an MS in computer science from Rensselaer Polytechnic Institute in Troy, NY. (Thompson can be reached at 508/337-5634 or bthompson@helixtechnology.com.)

Michael Eacobacci is a senior technologist at Helix Technology. With 28 years of experience in the vacuum industry, he holds numerous patents in the area of cryogenic vacuum pumps. He received a BS in mechanical engineering and an MS in materials science from Northeastern University in Boston and received training in technology management at Babson College in Wellesley, MA. (Eacobacci can be reached at 508/337-5221 or meacobacci@helixtechnology.com.)