PROCESS EQUIPMENT—VACUUM PUMPS

Developing an ultraclean pumping system for low-pressure, high-gas-flow applications

Tadahiro Ohmi, Kazuhide Ino, and Katsuyuki Sekine, Tohoku University; and Tadashi Shibata, University of Tokyo

The semiconductor industry is entering a new manufacturing era by using 300-mm wafers based on 0.25-µm-and-beyond design rules. Under this regime, single-wafer processes will become the norm in order to maintain process uniformity on the entire wafer surface and to establish as close to perfectly reliable semiconductor manufacturing as possible. Therefore, high-speed processes such as film growth and etching with processing rates >1 µm/min will be required to maintain productivity of >60 wafers/hr. High-speed processing is inevitably accompanied by a huge amount of reaction-gas generation from large wafer surfaces, typically a few hundred standard cubic centimeters per minute for plasma-enhanced chemical vapor deposition (PECVD) and reactive ion etching (RIE). In order to obtain high-quality processes, source gas concentrations on the wafer surface must be greater than the reaction gas concentration by at least one order of magnitude, resulting in a high source-gas flow rate of more than several thousand standard cubic centimeters per minute. Therefore, a new pumping system which can handle a large gas flow needs to be developed.

This article presents a newly developed high-pumping-speed turbomolecular pump (TMP) and a gradational leadscrew backing pump.^1—3 Since the backflow of impurities from the outlet line to the chamber in the presence of a high gas flow is also a concern, the impurity back-diffusion under high gas flow was investigated. With reference to the results of that study, guiding principles will be presented for the design of a vacuum pumping system for use in low-pressure, high-gas-flow processing.

Experimental Setup

Figure 1: Experimental setup used to evaluate pumps.

A diagram of the experimental setup is shown in Figure 1. The system was designed to carry out various plasma processing experiments (including RIE, PECVD, and sputtering), which represent the features of typical low-pressure semiconductor processing systems. The volume of the processing chamber was approximately 37,700 cm³. The vacuum pumping system was composed of a magnetically suspended TMP, and a dry backing pump (PDR-090A; Ulvac, Kanagawa, Japan) with a catalog-listed pumping speed of 1800 L/min. Four different TMPs, listed in Table I, were evaluated and compared. A conductance adjustment valve was placed at the inlet of the backing pump, and the effective pumping speed of the backing pump was varied.

TMP Model No.	Manufacturer	Catalog-Listed Pumping Speed for N2 and H2 (L/sec)	Catalog-Listed H2 Compression Ratio for Zero Flow
STP-H1000C	Seiko Seiki	1000, 800	1 x 103
DMS2000AJ-XH	Daikin Industries	1430, 880	2 x 104
STP-H2002C	Seiko Seiki	2100, 1600	2 x 104
STP-A2202C	Seiko Seiki	2200, 1700	2 x 104

Table I: List of evaluated turbomolecular pumps.

A simulated primary process gas was fed into the chamber, while another gas was injected into the outlet of the TMP as a simulated impurity gas. The partial pressure of the latter gas that back-diffused into the chamber was measured by a quadrupole mass spectrometer (QMS) (AQA200; Anelva, Tokyo) connected to the processing chamber through a very-small-conductance needle valve. In this manner, the impurity back-diffusion phenomenon could be experimentally investigated by measuring the impurity level, which was defined as the ratio of the partial pressure of back-diffused impurity gas to the total gas pressure in the chamber. The QMS was evacuated by a TMP (H200C; Seiko Seiki, Chiba, Japan) and a dry pump (Drytel 31; Alcatel, Annecy, France) to keep the measurement pressure <10^—6 torr.

Results and Discussion

TMP Pumping Characteristics under High Gas Flows. Figure 2 shows the relationship between the measured chamber pressure and the nitrogen gas-flow rate for three pumping speeds of the backing pump. The TMP used was the STP-H2002C. As the gas-flow rate increases in the range below ~700 std cm³/min, the chamber pressure linearly increases according to the relation f = 78.9 PS, where f is the gas-flow rate in std cm³/min, P is the chamber pressure (torr), and S is the effective TMP pumping speed (L/sec).

Figure 2: Chamber pressure versus nitrogen gas-flow rate for three pumping speeds of the backing pump. An STP-H2002C TMP was used.

The effective pumping speed of the TMP was calculated from the data shown in Figure 2 and is plotted in Figure 3. The value of S for the low gas-flow rate is ~700 L/sec, a reduction from the catalog specification of 2100 L/sec because of the elbow connecting the TMP and the chamber, and the wire netting at the inlet of the TMP.

Figure 3: Effective pumping speed of TMP corresponding to Figure 2 versus nitrogen gas-flow rate for three pumping speeds of the backing pump.

When the nitrogen gas-flow rate is >700 std cm³/min, however, the chamber pressure deviates widely from the linear dependence, approaching the respective pumping characteristics of the backing pump. The deviation occurs at almost the same gas-flow rate of ~700 std cm³/min, namely at the same inlet pressure of the TMP for all pumping speeds of the backing pump. Once the degradation occurs, however, the TMP pumping speed is strongly dependent on the pumping speed of the backing pump, namely the outlet pressure as shown in Figures 2 and 3. The most significant degradation in TMP performance is observed with the backing pump at the lowest pumping speed of 250 L/min. Thus, the TMP's pumping ability is severely degraded under a large gas throughput. A faster pumping speed for the backing pump is a preferred way of suppressing the degradation as much as possible, although it cannot be completely prevented. In order to examine why this degradation occurs, the pumping characteristics of two TMPs, STP-H1000C and STP-H2002C, were compared. The pumping speed and the compression ratio of these TMPs are different, as Table I indicates. Figure 4 shows pumping speed versus the backing pressure for the two TMPs. It is apparent that the degradation in a TMP's performance is not determined by its outlet pressure. It is also apparent that, as previously mentioned, once degradation occurs TMP pumping speed is strongly dependent on the outlet pressure.

Figure 4: Effective pumping speed versus backing pressure for the STP-H1000C and STP-H2002C TMPs, respectively.

Figure 5: Effective pumping speed versus chamber pressure for the STP-H1000C and STP-H2002C TMPs, respectively.

Figure 5 compares the pumping speed of these TMPs as a function of the chamber pressure for three pumping speeds of the backing pump. A sudden drop in pumping speed was observed at almost the same inlet pressure of ~10 mtorr for the two different TMPs. The mean-free path of nitrogen at 10 mtorr is 5.0 mm, which is close to the characteristic dimension of the entrance of these TMPs. This suggests that as the inlet pressure increases, the gas flow in the front blades as well as the rear blades of the TMP change from molecular to transition, and the pumping speed consequently begins to decrease. It seems reasonable to conclude that the pumping degradation occurs at the inlet pressure where the blades closest to the chamber leave the molecular flow region and go into the transition flow region, because TMPs are designed to operate and work only under molecular flow conditions.⁴ Therefore, the blades nearest the inlet need to be designed in such a way as to maintain the pumping speed under gas flow as high as that at low inlet pressures.

In order to improve the TMP performance under high gas flow, a characteristic dimension in the pump, particularly at the entrance (such as the distance between each blade vane or the distance between the rotor blade and the stator blade), was shortened to increase Knudsen's number. This enables us to maintain the molecular flow region at higher pressures and to improve the TMP performance under a high gas flow. Figures 6 and 7 compare the pumping characteristics of the four TMPs. A newly developed TMP, STP-A2202C, has blades that were improved by shortening the characteristic distance compared to STP-H2002C. The newly developed TMP features greatly improved pump performance under a large gas flow and can endure a gas flow as great as ~1500 std cm³/min while maintaining the acceptable pumping speed.

Figure 6: Chamber pressure versus nitrogen gas-flow rates for four TMPs.

Figure 7: Effective pumping speed of four TMPs as a function of nitrogen gas-flow rate.

Impurity Back-Diffusion through TMP under High Gas Flows. Figure 8 shows the impurity level, which is defined as the ratio of the partial pressure of helium (the simulated impurity gas) to the total gas pressure in the chamber, as a function of the nitrogen (the simulated process gas) gas-flow rate for three different pumping speeds of the backing pump. The helium challenge gas was injected into the outlet of the STP-H2002C pump at a constant flow rate of 400 std cm³/min. The outlet pressure is shown in Figure 9 for various helium gas-flow rates. Figure 8 demonstrates that the back-diffusion through a TMP can be suppressed significantly by using a higher gas-flow rate from the inlet. This phenomenon is interpreted as follows. At a certain nitrogen gas-flow rate as low as several standard cubic centimeters per minute, a thread groove rotor/stator located under the turbine blades⁵ or the blades closest to the outlet would be in the transitional gas-flow region—as expected from the outlet pressure data in Figure 9—and molecular collisions could not be avoided. Under such conditions, the impurity gas molecules have a chance to be dragged out by the stream of nitrogen gas molecules through molecular collisions. This could explain the decrease in the impurity levels when the nitrogen gas-flow rate is increased. It is important to note that the impurity (helium) back-stream is completely suppressed even in a high-gas-flow condition where degradation in the TMP performance occurs because of the back-diffusion of the nitrogen process gas caused by the molecular collisions in the front blades, as previously discussed. Figure 8 also indicates that the impurity levels are lowest in the whole range of the nitrogen gas-flow rate for the backing pump having a maximum pumping speed of 1800 L/min, indicating that using a high-speed backing pump is essential to suppressing the back-diffusion of impurities to the chamber.

Figure 8: Impurity level in the chamber as a function of nitrogen gas flow rate for three pumping speeds of the backing pump. The helium challenge gas was injected into the outlet of the TMP at a constant flow rate of 400 std cm³/min.

Figure 9: Outlet pressure versus nitrogen gas-flow rate at various helium challenge gas-flow rates.

The effect of the helium challenge gas-flow rate on the impurity levels is shown in Figure 10 at several nitrogen gas-flow rates. The impurity levels in the chamber are greatly influenced by the nitrogen rather than the helium gas-flow rate. Figure 11 compares the dragging effects of oxygen, hydrogen, argon, and nitrogen gas molecules against helium back-diffusion. No significant differences can be observed for the gases. Therefore, the impurity compression ratio can be improved by feeding any kind of gas into the chamber.

Figure 10: Impurity levels in the chamber as a function of helium challenge gas-flow rates for various nitrogen gas-flow rates.

Figure 11: Impurity levels in the chamber versus gas-flow rate of the simulated process gases oxygen, hydrogen, argon, and nitrogen. Challenge helium gas was injected into the outlet of the TMP at a constant flow rate of 400 std cm³/min.

Figure 12 shows the helium compression ratio of three different TMPs as a function of nitrogen gas-flow rate. The flow rates of simulated impurity gas were the same—400 std cm³/min—and the pumping speed of the backing pump was 1800 L/min. The impurity compression ratio of the TMPs with a different compression ratio at zero flow could be increased to similar values by an adequate amount of gas flow. A proper nitrogen gas flow can increase the impurity compression ratio of TMPs by several orders of magnitude.

Figure 12: The helium compression ratio of three TMPs as a function of nitrogen gas-flow rate.

These results confirm that the impurity levels in the chamber can be made several orders of magnitude lower than those of the best-performing TMP under ultra-high-vacuum (UHV) operation when an adequate gas flow of ~100 std cm³/min is fed into the chamber. This suggests that the chamber should not be maintained under UHV, especially when a wafer is located in it, and that some gases should be fed into it at a flow rate of ~100 std cm³/min.

Figure 13: Schematic of a gradational leadscrew pump.

Gradational Leadscrew Pump. A backing pump is required to have a high pumping speed and to maintain that speed constant from atmospheric pressure to a low pressure of ~10^—3 torr. A newly developed screw pump (Diavac, Tokyo) features the noble structure of a screw with gradational lead and inclination, as shown in Figure 13; the unit can achieve a high pumping speed of 3600 L/min with a small pumping system. By changing the pitch and the angle of the screw from the inlet to the outlet, the compression ratio and the pumping speed can be effectively improved. In order to increase the compression ratio, a small amount of oil is also injected near the outlet and circulated.

Figure 14 shows the pumping speed of a conventional dry pump and the new screw pump as a function of the inlet pressure. The broken lines indicate constant gas-flow rates. As mentioned, the screw pump has a constant speed of 3600 L/min and maintains that speed down to ~10^—3 torr, while the conventional dry pump achieves a lower pumping speed and maintains it to only 1 torr.

Figure 14: Comparative pumping speeds of a conventional dry pump and newly developed screw pump as a function of inlet pressure.

The back-diffusion of oils from the screw pump is a matter of concern.⁶ Figure 15 shows the amount of back-diffused oils from the screw pump (as the impurity ion current) and the pumping speed as functions of inlet pressure. The broken lines indicate constant gas-flow rates. The back-diffusion decreases with increases in the inlet pressure, i.e., the nitrogen flow rate. At several millitorr, the gas flows into the transition flow region. Therefore, back-diffusion can be suppressed by the nitrogen stream through molecular collisions, and the proper gas-flow rate needed to accomplish this is only about 10 std cm³/min. Hence, impurity back-diffusion from both the TMP and the screw pump can be completely suppressed by continuously flowing gas from the chamber. The gas-flow rate for the chamber must always be maintained at >=100 std cm³/min to suppress back-diffusion contamination from the gas pumping system.

Figure 15: Impurity ion current and pumping speed as a function of inlet pressure.

Conclusion

A high-pumping-speed TMP and a backing pump with a gradational leadscrew have been developed. The new TMP was developed by shortening the characteristic distances in the pump, particularly at the entrance—such as the distance of each blade vane and the distance between the rotor blade and the stator blade—to maintain the molecular flow at higher pressures. The new TMP can withstand a gas flow as high as ~1500 std cm³/min while maintaining its pumping speed. The gradational leadscrew pump has a very wide dynamic range of 10^—3 to 760 torr, with a high pumping speed of 3600 L/min.

Experimental measurements were taken of the impurity levels in the chamber caused by the back-diffusion of impurities from the gas pumping system and the impurity compression ratio under a high gas flow. It has been shown that back-diffusion contamination from the gas pumping system can be completely suppressed by continuously flowing gas from the chamber. It is extremely important to note that the chamber should never be maintained under UHV, in particular when a wafer is located in it, and that some gases should be fed into it at a flow rate >100 std cm³/min to suppress back-diffusion contamination from the pumping system. In other words, a wafer must be located in a low-pressure ambient with a proper gas flow and not in UHV throughout processing, including wafer transfer.

In high-gas-flow and large-diameter-wafer processing, the design of the connection between the pumping system and the chamber is extremely important, as is improving the pumps to achieve high-gas-flow conductance and, as a result, a higher effective TMP pumping speed. In the semiconductor industry, the pumping system must evacuate a large volume of gases with high uniformity and high pumping speed. Thus, the use of several compact TMPs in a single chamber and a new processing chamber design for such use will be needed for 300-mm and larger wafer production.

Acknowledgments

This article is based on a paper originally presented at the session on contamination-free manufacturing, cosponsored by the Ultra Clean Society, at the 44th annual meeting of the American Vacuum Society, San Jose, October 1997. Used with permission. The authors wish to thank Daikin Industries, Seiko Seiki, and Diavac for providing the pumps. The majority of this work was carried out in the mini-super cleanroom at the Graduate School of Engineering, Tohoku University, Sendai, Japan.

References

1.Ino K, Sekine K, Shibata T, et al., "Improvement of Turbomolecular Pumps for Ultraclean, Low-Pressure and High-Gas-Flow Processing," submitted to Journal of Vacuum Science and Technology.

2.Ohmi T, "New Era of Semiconductor Manufacturing (I)," Ultra Clean Technology, 9, supplement 1, 1997.

3.Ino K, Sekine K, Shibata T, et al., "Suppression of Impurity-Backdiffusion in Vacuum Pumping Systems for Ultraclean Low-Pressure Semiconductor Processing," in Proceedings of the 44th National Symposium of American Vacuum Society, San Jose, AVS, p 7, 1997.

4.For example, see Harris NS, Modern Vacuum Practice, London, McGraw-Hill, chap 8, 1989.

5.Ishizawa T, Miki M, Urano C, et al., "The Pumping Character of the Magnetic Suspension-Type Turbomolecular Pump of High Flow Rate," Journal of Vacuum Science and Technology, A5, 2965, 1987.

6.Tsutsumi Y, Ueda S, Ikegawa M, et al., "Prevention of Oil Vapor Backstreaming in Vacuum Systems by Gas Purge Method," Journal of Vacuum Science and Technology, A8, 2764, 1990.

Tadahiro Ohmi, PhD, is a professor in the Department of Electronics, Graduate School of Engineering, at Tohoku University (Sendai, Japan). He is one of the world's leading researchers in the field of advanced semiconductor manufacturing. His work deals with high-performance ULSI circuits as well as advanced processes, including extensive research on ultraclean technologies. Ohmi has received numerous scientific honors and awards, made some 500 patent applications, and written or cowritten more than 650 technical papers. He holds BS, MS, and PhD degrees in electrical engineering from Tokyo Institute of Technology and is a member of several professional associations, including the Japan Society of Applied Physics, the Electrochemical Society, and IEEE. (Ohmi can be reached at +22 224 2649; fax, +22 224 2549.)

Kazuhide Ino is a doctoral student in electronic engineering at Tohoku University. He is working on a low-energy, large-mass ion bombardment process for metal-gate SOI devices as well as an ultraclean gas pumping system. He has a BS and MS in electronic engineering from Tohoku.

Katsuyuki Sekine is also a doctoral student in electronic engineering at Tohoku University. His research focus is on low-temperature processing using low-energy ion bombardment as well as an ultraclean gas pumping system. He has a BS and MS in electronic engineering from Tohoku.

Tadashi Shibata, PhD, has been a professor in the Department of Information and Communication Engineering at the University of Tokyo since May 1997. From April 1986 until taking the job in Tokyo, he was an associate professor in the Department of Electronic Engineering at Tohoku University, where he was engaged in extensive research in ultraclean technologies. Before that, he worked in various positions at Toshiba from 1974 to 1986. Shibata earned a BS in electronic engineering and MS in materials science from Osaka University and a PhD from the University of Tokyo. He is a member of the Japan Society of Applied Physics, the Institute of Electrical Engineers of Japan, and the IEEE Electron Devices Society, among others.

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit