Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://danp.sinp.msu.ru/Articles_GSIB/nimb237_235_reviewGCIB_SConductors.pdf Äàòà èçìåíåíèÿ: Fri Oct 7 22:33:38 2005 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 22:42:28 2012 Êîäèðîâêà:

Nuclear Instruments and Methods in Physics Research B 237 (2005) 235­239 www.elsevier.com/locate/nimb

Gas cluster ion beams for wafer processing
M.E. Mack
*
Epion Corporation, 37 Manning Road, Billerica, MA 01821, USA Available online 28 June 2005

Abstract Gas cluster ion beams (GCIB) represent a powerful new tool for wafer processing. This paper will review the development status of GCIB equipment as pertains to semiconductor applications. ñ 2005 Elsevier B.V. All rights reserved.
PACS: 36.40.þc; 36.40.Wa Keywords: Cluster ion beams; GCIB; Equipment for GCIB

1. Introduction Gas cluster ion beams (GCIB) have been demonstrated to be valuable for surface smoothing and etching [1]. Commercial GCIB equipment has been developed that allows corrective etching of wafer surfaces to obtain smooth surfaces with a consistent and uniform thickness profile [2]. This feature has resulted in the use of GCIB for SAW device production [3]. Most recently GCIB has been used to dope silicon with boron [3,4] and to form low defect germanium­silicon layers [3,5]. Commercial 300 mm GCIB processing equipment is available at the 500 lA level and currents
*

Tel.: +1 978 215 6206; fax: +1 978 670 9119. E-mail address: mmack@epion.com

up to 1 mA have been achieved in the laboratory [7]. While the currents are low compared to most ion implant equipment, the atomic flux levels of such beams are in the range of one to several amperes making them well suited for directed beam chemistry, deposition, etching and surface modifications. The design of the Epion GCIB equipment, the only such systems currently in production, has been described in the literature [7,8]. Beam lines are similar to those used in ion implanters but are modified to account for differences such as the high gas flow in the beam, itself, and the high magnetic rigidity of the clusters. In the past several years, the equipment has evolved largely to address production worthiness issues. It is the purpose of this paper to report on the significant progress that has occurred in these critical areas.

0168-583X/$ - see front matter ñ 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.nimb.2004.12.139


236

M.E. Mack / Nucl. Instr. and Meth. in Phys. Res. B 237 (2005) 235­239

2. Cluster formation Clusters are formed by an adiabatic expansion of the cluster gas through a supersonic nozzle into a vacuum. With typical nozzle stagnation pressures of $5000 Torr, exit temperatures of $1 K would be expected. Actual cluster temperature is higher, since the atoms give up heat on nucleation. In many cases binary and tertiary mixes of gases are used for processing. For example, NF3 and O2 are often combined for etching while diborane, B2H6, is combined with argon for boron doping [4,6] or may be combined with germane and argon to produce a boron doped germanium infusion [6]. In all these mixtures the etchant or deposition gas is a minor fraction of the total gas flow. In such cases RGA studies have shown that the gases do not necessarily nucleate at the same flow rate. With germane and argon increasing the flow rate results in germane nucleating first with the argon nucleation following only at higher flow rates. Thus, by varying the flow rate the composition of the clusters formed can be varied.

Fig. 1. Cutaway of the self-neutralizing ionizer.

3. Ionizer development for metals reduction Clusters leaving the nozzle and entering the ionizer will travel with roughly the sound speed characteristic of the gas entering the nozzle. For typical cluster sizes (2000­15 000 atoms) this corresponds to a kinetic energy of 130­1000 eV. At these energies any departure from space charge neutrality within the ionizer will result in a rapid blow up of the jet with a significant loss of beam current. For this reason, one of the first ionizers developed at Epion was the so-called self-neutralizing ionizer (Fig. 1). As with all such ionizers clusters are ionized by electron impact [7]. In this design electrons extracted from thermionic filaments pass through the jet and then strike the opposite beam forming electrode to produce low energy secondary electrons. These electrons help ensure that the ionized jet remains space charge neutral. This design proved very effective and with modifications achieved over 1000 lA argon cluster beams. A major limitation of the self-neutralizing ionizer is that gases evolved from clusters during the

ionization processes [7] produce an elevated internal pressure. With corrosive gases, particularly NF3 in O2, this results in attack of the various ionizer parts particularly the filaments. Filament life is shortened and unacceptable metals contamination is produced on wafers being processed. To overcome these shortcomings the reflex ionizer was developed. In this ionizer the beam forming and repeller electrodes are replaced by graphite rods. Fig. 2 illustrates the reflex ionizer and shows a simulation of electron trajectories from one of the three filaments. In the figure the inner row of rods is at ionizer potential, the filament is biased at þ165 V and the second ring at þ325 V, both with respect to ionizer potential. The reflex design causes recir-

Fig. 2. Cross section of reflex ionizer showing electron trajectories.


M.E. Mack / Nucl. Instr. and Meth. in Phys. Res. B 237 (2005) 235­239

237

culation of the ionizing electrons through the cluster jet. Note also that the electrons eventually strike the innermost rods where they will again produce secondary electrons to ensure neutrality of the jet. Finally, it can be seen from the figure that the conductance for gases out of the ionizer is much improved yielding lower internal pressures.

Fig. 3. PEG ionizer showing multi-aperture electron extraction.

The reflex ionizer does give excellent metals performance with less reactive gases but with corrosive gases contamination can still be high. For this reason one additional ion source was developed. This ionizer uses multi-aperture extraction from an argon plasma as the source for the electrons in a scheme very similar to the reflex ionizer for recirculation of the electrons. This ionizer is shown in Fig. 3. Like the filament in the reflex ionizer the body of the plasma electron gun (PEG) is biased negatively with respect to ionizer potential. Electrons are extracted by an accel­decel extraction using a positive electrode adjacent to the PEG body. This arrangement has the advantage that the filament used to generate the argon plasma is shielded from the corrosive gases by the argon. While some back diffusion of corrosive gas into the PEG arc chamber does occur the partial pressure of the this gas within the arc chamber is low and does not measurably contribute to erosion of the filament. To further safeguard against metals contamination all parts of the electron extraction and charge confinement (reflex) assembly are constructed of graphite. Fig. 4 compares the performance of the three ionizers in terms of metals contamination on the wafer with etchant gases. The data for the self-neutralizing (SN) ionizer was obtained by TXRF while VPD ICPMS was used for the reflex and

Fig. 4. Evolution of metals contamination.


238

M.E. Mack / Nucl. Instr. and Meth. in Phys. Res. B 237 (2005) 235­239
87 86
Thickness (nm) Uniformity (%)

PEG ionizers. The improvement with the PEG ionizer is striking. With the PEG metals performance with less corrosive gases mixtures is the same or better than with O2­NF3. Although not readily apparent from Fig. 4 a considerable reduction in tungsten contamination occurs with the PEG ionizer, indicating that the corrosive gases are not attacking the filament within the arc chamber. As a result a long source life would be expected and this is the case. Life tests were run using the PEG with 5% NF3/oxygen for 400 h and with 5% GeH4/argon for 350 h. In both cases the source continued to run normally after test. In each case the ionizer was eventually shut down to evaluate wear and in both cases no wear was evident on the filament or the arc chamber.

0.60

Ge Film Thickness (nm)

85 84 83 82 81

0.50

0.40

0.30
+1%

80
Mean = 79.2 nm

79 78 77 0 2 4 6 8 10 12 14 16
-1%

0.20

0.10

Wafer Number

Fig. 5. Germanium deposition repeatability and uniformity.

4. Charge control Charge control in the PEG beam line is accomplished using a plasma electron flood (PEF). Charge control is simplified by virtue of the relatively low currents and by the fact that each charge corresponds to many atoms that are to be transported and deposited. Charging is much reduced over ion implantation at the same dopant dose where each atom conveys a charge. In the case of the boron infusion positive and negative charging of less than 2 V for a boron dose of 1E15/cm2 is observed with CHARM 2 monitors.

source gas to form the Ge bearing clusters. The clusters were accelerated through a 30 kV accel and the wafers were exposed for a dose of 5E14/ cm2. The germanium thickness was measured by ellipsometry. The data spans one day of operation. Long-term dose variation data still must be determined so that in this regard dose control represents a work in progress.

6. Particle contamination Particle performance also represents a work in progress. Particle sources as yet have not been identified. Nevertheless, the performance of the wafer handling has been quantified and the results are good (<15 adders >0.16 lm). In addition, the performance of several of the doping and infusion recipes have been quantified and again results are good. With germanium infusion processing (3E13/ cmþ2 GeH4 30 kV 200 mm wafers) average particle adders of <68 >0.16 and <33 >0.2 lm have been achieved.

5. Dose control Just as in ion implantation dose control is accomplished by measuring cluster charge current and integrating over time to determine the net dose of charge. A difference here is that the cluster dose will be lower because the clusters may have a charge state in excess of one [7]. However, in GCIB equipment recipes are well controlled so that processing would be expected to be repeatable. In fact just as in modern ion implanters recipes are software controlled with user defined limits for all key parameters. Fig. 5 shows wafer to wafer dose variation for germanium deposition. In this test 5% GeH4 in argon was used as the

7. Uptime and reliability Uptime and reliability have been evaluated throughout the development of the Epion GCIB tools. For factory based tools availability is typically greater than 90% with an MTBF of upwards of 100 h or more. MTTR is typically 3 h or less. The GCIB equipment is fundamentally robust

Uniformity (%)


M.E. Mack / Nucl. Instr. and Meth. in Phys. Res. B 237 (2005) 235­239

239

and for that reason is reliable. The PEG based 300 mm machine has only recently come on line. It is expected that MTBF and uptime will improve with the latter machine due to the much longer source lifetime of the PEG source. 8. Conclusion

[2] [3]

[4]

The main effort in GCIB in the past several years has been in process development and in improving the production worthiness of equipment. State of the art performance has been achieved in reliability, in metals contamination, and charge control. Initial results in quantifying dose control and particle contamination have been excellent although additional work remains to be done in those areas. References
[1] N. Toyoda, J. Matsuo, I. Yamada, in: E. Ishida, S. Banerjee, S. Mehta, T.C. Smith, M. Current, L. Larsen

[5]

[6]

[7]

[8]

(Eds.), Implantation Technology-96, IEEE, Piscataway, 1997, p. 808. L.P. Allen et al., International SOI Conference, Williamsburg, VA, October 7­11 2002. J. Weldon, R. MacCrimmon, S. Caliendo, Y. Shao, J. Hautala, B. Zide, M. Gwinn, GCIB Process Development for Electronic and Industrial Application, 5th Workshop on Cluster Ion Beam and Advanced Quantum Beam Process, September 30, 2004, Kyoto, Japan. J. Borland, J. Hautala, M. Gwinn, T.G. Tetreault, W. Skinner, Solid State Technology (May) (2004). J.O. Borland, Alternative USJ formation and characterization methods for 45 nm node technology, Nucl. Instr. and Meth. B, these Proceedings, doi:10.1016/j.nimb.2005.04. 106. A. Kirkpatrick, Gas cluster ion beam infusion doping to form ultra shallow junctions and silicon­germanium layers, Nucl. Instr. and Meth. B, these Proceedings. M.E. Mack, R. Becker, M. Gwinn, D.R. Swenson, R.P. Torti, R. Roby, in: R. Brown, T.L. Alford, M. Nastasi, M.C. Vella (Eds.), IIT2002 Proceedings, IEEE, Piscataway, 2003, p. 665. J. Bachand, A. Freytsis, E. Harrington, M. Gwinn, N. Hofmeester, J. Hautala, M.E. Mack, K. Regan, in: R. Brown, T.L. Alford, M. Nastasi, M.C. Vella (Eds.), IIT2002 Proceedings, IEEE, Piscataway, 2003, p. 669.