Process Equipment Control

Investigating medium-current tools for indium implantation applications

Gregory Luckman, Michel Bruehwiler, and Robert Rathmell, Axcelis Technologies

Studies show that contamination prevention features permit two implanters to be used with indium, antimony, and various other species.

As the semiconductor industry advances into deep-submicron device geometries, many manufacturers of advanced ICs are expressing interest in adopting indium as an alternative to boron for channel and source/drain engineering applications. Many sub-0.18-µm transistors require a low dopant concentration in the channel region to enhance charge-carrier mobility and a higher dopant concentration below the source/drain extension junctions to suppress punchthrough, improve subthreshold slope, and reduce I^_off current. Low-diffusivity heavy dopants such as indium and antimony can produce the supersteep retrograde profiles that meet these requirements.

Indium and boron experience approximately the same excess transient enhanced diffusion above their respective intrinsic diffusivities.¹ However, the as-implanted distribution of indium is steeper than that of boron. Because the intrinsic diffusivity of indium is no larger than that of boron, the narrower as-implanted indium profile results in narrower annealed doping profiles, as shown in Figure 1. Indium is an acceptor dopant in silicon and, as such, has already been used successfully in a number of semiconductor applications, including retrograde p-tubs, source/drain halo regions for NMOSFETs, buried channels for polysilicon gate PMOSFETs, and base regions for NPN (negative-positive-negative) bipolar transistors.^2–5

Figure 1: Comparison of as-implanted and annealed indium and boron doping profiles. Annealing was performed at 1000°C for 10 seconds.

Although indium provides some important performance advantages, there are concerns that must be addressed before a fab switches to its use. Indium implantation requires the use of highly hygroscopic indium trichloride (InCl^₃) as the source material. Indium remaining in an implanter's source region or other area following processing could lead to energetic cross-contamination if the tool is subsequently used for implanting other species. To prevent such contamination, two medium-current implanters, the 300-mm MC3 and the 200-mm 8250^_HT from Axcelis Technologies (Beverly, MA), were designed to incorporate an energy filtration system and a vaporizer cooling system. After a discussion of the sources of energetic indium in implanters, this article focuses on these contamination prevention features and reports on the manufacturer's in-house studies.

Sources of Energetic Indium

During the implantation process, ions of the extracted species acquire energy as they travel from the source to the wafer surface. A singly charged indium ion in either of the two tools that are the focus of this article, for example, will acquire a maximum energy of 250 keV in three stages: first, 40 keV during its extraction from the source; second, 68 keV as it passes through the parallelizing lens; and third, 142 keV as it moves through the acceleration tube. If the tool is then used to implant another species, such as arsenic, boron, or phosphorus, any indium remaining in the source region could be ionized and extracted by the 40-kV potential used to extract ions of the subsequent species. To prevent such cross-contamination, the tools include an analyzing magnet that filters 40-keV singly charged ions (or 80-keV doubly charged ions, etc.) at the desired mass when it is tuned for the mass of subsequently implanted species.

However, if residual indium is also present in the area between the analyzing magnet and the parallelizing lens, 40-keV ions of another species could sputter it off of surfaces within the beam line. Positively charged indium produced by such sputtering will most likely possess a relatively low energy level, but as it travels through the parallelizing lens and acceleration tube during a subsequent 250-keV implant, it can achieve an energy of 210 keV. Generally, subsequent ⁷⁵As⁺ ions are more efficient at sputtering residual ¹¹⁵In than are ¹¹B⁺ or ³¹P⁺ ions.

In addition, some ¹¹⁵In⁺ generated from the source may not be eliminated by the analyzing magnet. If the indium is ionized partway across the extraction gap, it may have the same magnetic rigidity as a lower-mass ion with the full extraction energy. The energy of this indium is a function of the extraction energy and dopant used in the recipe run immediately after indium implantation. For a ⁷⁵As⁺ beam with 40-keV extraction energy, ¹¹⁵In⁺ with an energy of ~26.1 keV will have the same magnetic rigidity as the 40-keV arsenic. The locations at which the indium can acquire 26.1 keV constitute an equipotential surface within the extraction gap and do not correspond to any physical surface on which substantial indium could be absorbed. Nevertheless, for a 250-keV As⁺ implant, 26.1-keV indium will remain in the implanter beam and, after passing through the parallelizing lens and acceleration column, will achieve an energy of 236.1 keV (250 – 40 + 26.1 keV).

An analogous scenario is observed in 40-keV ¹¹B⁺ and ³¹P⁺ ion extraction. In these processes, the energies required for ¹¹⁵In⁺ to remain in the beam beyond the analyzing magnet are 3.8 and 10.8 keV, respectively. Thus, for 250-keV ¹¹B⁺ or ³¹P⁺ implants, the undesired indium will achieve energy levels of 213.8 and 220.8 keV, respectively.

When these energetic indium–generation mechanisms occur, the undesired indium is either removed by the analyzing magnet or exits the acceleration tube with lower energy than a subsequent 250-keV implant species such as arsenic. If implant recipes with lower energy than 250 keV follow indium operation, similar mechanisms will also apply. However, as implant energy decreases, the acceleration tube, extraction electrode, and parallelizing lens voltages also decrease. The result is that at lower energies, any residual indium in the beam will have energy that is a lower percentage of the implanted species' energy than it would have for a 250-keV recipe.

Contamination Prevention Features

In addition to the tunable analyzing magnet, the two medium-current implanters incorporate other cross-contamination prevention features that permit their use with more than one species.

Energy Filtration System. An electrostatic filter known as the angular energy filter, or AEF, deflects the implanter beam to eliminate contaminants before they reach the wafer. Ions with the correct energy are deflected by 15°, while those with the wrong energy-to-charge ratio strike the filter's slits and are eliminated, as indicated in Figure 2. (The operation of the filter has been described in detail elsewhere.⁶) In most cases, the residual indium ions that enter the filter have a lower energy level than the ions of the subsequent implant species and are eliminated in the same manner as energy contaminants are
in the implanters' other applications.

Figure 2: Side view of the species filtration system showing the energy slits, neutral-particle trap, and contamination dumps.

The energy filtration system also eliminates neutral particles that may be created by sputtering or by interactions between the ions in the beam and photoresist gases released during the implant. These neutral particles cannot be detected by an implanter's dose measurement system, yet must be eliminated to prevent dose errors. Furthermore, the filter eliminates energetic contamination for lower-energy implants in the acceleration or deceleration mode of operation. In the acceleration mode, low-energy contaminants in high-energy implants are deflected and steered into low-energy contamination dumps; in deceleration mode, potential high-energy contaminants are deflected into high-energy contamination dumps. Both of these dumps are depicted in Figure 2, as is the trap for neutral particles.

Vaporizer Cooling System. In ion implantation tools, indium beams are generated using solid InCl^₃ at a vaporizer temperature between 200° and 300°C. The temperature of the vaporizer, which is located adjacent to the ion source arc chamber, is controlled by a software algorithm–regulated nitrogen cooling system. If implantation of another species is performed with solid InCl^₃ present in the vaporizer, passive heating of the vaporizer caused by its proximity to the superheated arc chamber may be enough to cause the indium to sublimate. The result would be the contamination of the arc chamber and the other implant species. To prevent such contamination, the two medium-current implanters include active vaporizer cooling (AVC) capability in the ion source module.

The effectiveness of the vaporizer cooling system has been verified by comparing the steady-state temperature of a vaporizer with and without the cooling feature during high-power arc chamber operation for triply charged phosphorus implantation.⁷ For these measurements, the arc chamber was operated at 6.0 A and 120 V. When the cooling system was turned off, the vaporizer temperature was approximately 200°C; when it was activated, the steady-state temperature of the vaporizer decreased to <100°C, well below the sublimation temperature of InCl_³.

In addition to preventing indium sublimation, the cooling system enables the implanters to be used with vaporizers for both indium and antimony. Research indicates that for PMOSFETs, highly angled antimony ion implantation is effective in controlling short-channel effects with high drive current and low junction leakage for <0.1-µm devices.⁸ With the implanters' cooling capability, indium trichloride and antimony trioxide can be left in source vaporizers at the same time. During antimony generation, the high vaporizer temperature of up to 500°C needed to produce antimony beams will not heat the cooled, idle indium vaporizer to the critical sublimation temperature of InCl_³. In addition, both indium and antimony vaporizers can be left in place when the tools are implanting other species. Figure 3 depicts the location of the vaporizers relative to the arc chamber.

Experimental Data

To verify and quantify the effectiveness of the implanters' AEF and AVC features, even under extreme conditions, the manufacturer performed extensive studies using 200- and 300-mm bare silicon wafers. In these tests, an indium dose of up to 2 x 10¹⁶ ions/cm² was implanted onto dummy wafers followed by a short argon purge before implants were performed with singly charged arsenic, boron, or phosphorus. Secondary ion mass spectroscopy (SIMS) was then used to measure the indium levels in the implants.

Figure 3: Three-dimensional view of the implanters' ion source showing vaporizers for antimony and indium, which can be left in place when species such as phosphorus and boron are being implanted.

Table I shows that a series of experimental steps achieved an exceptionally low total indium level of 280 ppm on the 300-mm tool. First, singly charged indium from InCl^₃ was implanted for 2 hours at an energy of 110 keV and a beam current of 720 pµA, amounting to a total indium dose of 1.7 x 10¹⁶ ions/cm². Following a 5-minute argon purge, an As⁺ implant was performed at an energy of 80 keV, a beam current of 2700 pµA, and a dose of 5 x 10¹⁵ ions/cm². Surface and energetic indium levels in the arsenic-implanted wafers were measured using O-leak and dynamic SIMS. As Figure 4 shows, energetic indium levels were near the SIMS detection limit.

Implant Species Energy (keV) Dose (ions/cm2) Analysis Request Surface SIMS (atoms/cm2) % of Dose Total (ppm) As+ 80 5x 1015 Indium 1.41 x 1012 0.028 280

Table I: Total indium levels on arsenic-implanted wafers with the 300-mm implant tool.

In the production environment, the implanted indium dose is generally limited to <3 x 10¹³ ions/cm² because of the species' low solubility in silicon. Indeed, most indium recipes call for a dose between 1 x 10¹² and 2 x 10¹³ ions/cm². The implanted total dose of 1.7 x 10¹⁶ ions/cm² used in this experiment would be equivalent to processing approximately 1600 wafers at an average dose of 1.05 x 10¹³ ions/cm² per wafer, but most fabs generally process only between 25 and 100 wafers before changing to a different dopant. Consequently, the residual indium levels expected on subsequently implanted wafers after indium processing of 100 wafers at 1.05 x 10¹³ ions/cm² are even lower than the levels indicated in Figure 4, certainly low enough so that the same tool could be used to safely implant indium and other species sequentially.

Figure 4: SIMS profiles of 113In+ and 115In+ levels in arsenic-implanted wafers measured after the implanter had been used for 2 hours of indium implantation.

Although the SIMS measurements described above had consistently shown that indium surface levels on subsequently implanted wafers are generally low and smaller than 0.1%, as indicated in Table I, additional, long-term in-house tests of the two implanters' capabilities were conducted. Again, SIMS was used to measure the surface and energetic indium levels generated when another species was implanted immediately after indium implantation.⁷ The main focus of these tests was to determine the contribution of energetic indium as a function of the energy and species of the subsequent implant (B⁺, P⁺, or As⁺). Preliminary test results suggested that energetic indium levels were likely to be higher in 250-keV As⁺ than at any other energy or with any other species. These initial test results also confirmed the hypothesis that indium levels found after extensive indium operation would decay after the implanter was switched to another species. However, this decay was slow enough to allow several wafers to be regarded as "wafers implanted immediately after running indium" for the purposes of subsequent tests.

Measurements of energetic indium levels versus implant energy and species on wafers processed using the 200-mm implanter were obtained as follows. First, 5 x 10¹⁵ ions/ cm² of singly charged arsenic were implanted into prime silicon wafers at 250 keV after the tool had run "indium-free" for an extended period, and SIMS measurements of energetic indium levels on these wafers were obtained. These results established that baseline indium levels on wafers processed with the tool are at or below the SIMS detection limit.

Next, a total indium dose of 2 x 10¹⁶ ions/cm² was implanted into several dummy wafers and the tool was then purged with argon for 20 minutes. Following this purge, the implanter beam was turned for implanting boron, phosphorus, or arsenic, and doses of 5 x 10¹⁵ ions/cm² of the selected species were implanted at a selected energy into a series of prime silicon wafers. Finally, SIMS measurements of indium levels were obtained on the wafers implanted with these other species.

Figure 5 shows two of the resulting SIMS profiles. The graph on the left displays the ¹¹⁵In levels in an 80-keV ⁷⁵As⁺ implant; energetic indium levels in this case were exceptionally low, measuring approximately 0.0005% (5 ppm) of the implanted dose. The graph on the right shows the ¹¹⁵In levels in a 250-keV ⁷⁵As+ implant, which measured 0.0084% (84 ppm) of the implanted arsenic dose. Both graphs are logarithmic—the maximum indium concentration at any depth is at least three orders of magnitude lower than the arsenic concentration at the same depth.

Figure 5: Sample SIMS plots of indium levels in wafers implanted with arsenic at 80 keV (left) and 250 keV (right).

Figure 6 plots energetic indium percentage levels versus arsenic implant energies. The integration of each plot from the relevant SIMS profiles was performed for depths >400Å. For the lower As⁺ energies (<150 keV), the integration is primarily of the tail of the surface indium signal and is therefore likely to be an overestimate of the level of energetic indium. The highest indium level shown, which is below 0.009% (90 ppm) at 250 keV, is unlikely to be seen in a manufacturing environment because arsenic is commonly implanted at energies below 200 keV.

Figure 6: Energetic indium as a function of arsenic implant energy. Data are integrated from SIMS profiles.

Several SIMS profiles also were obtained to characterize the energetic indium levels on wafers implanted with phosphorus and boron, and the energetic indium levels from SIMS plots for implants of 5 x 10¹⁵ ions/cm² of B⁺, P⁺, and As⁺ at 250 keV were compared. The results indicate that indium levels in boron and phosphorus implants are approximately four times lower than indium levels in arsenic implants.⁷ Finally, an analysis of all the test results described above reveals that for all the tested species, the expected surface plus energetic indium levels are below 0.1% of the subsequently implanted dose, which demonstrates that the two implanters are effective at eliminating indium contamination.

Conclusion

Low-diffusivity dopants such as indium can offer advantages in sub-0.18-µm transistor manufacturing, but the use of such species raises some concerns. Dedicating an implanter to processing indium is inefficient and costly, yet using a single tool for indium and other species can lead to indium contamination on subsequent implants. The medium-current implanters that are the focus of this article were designed to prevent such cross-contamination through the use of an energy filter and a vaporizer cooling system. The results described indicate that they have achieved this goal and can provide manufacturers with the flexibility required in the semiconductor industry.

Acknowledgments

The authors would like to thank Jim Kawski and Andy Ray of Axcelis for contributing to discussions on indium implantation and to Axcelis's medium-current implanter demonstration team, led by Dennis Klesel, for performing countless implants for this study.

References

1. PB Griffin et al., "Indium Transient Enhanced Diffusion," Applied Physics Letters 73, no. 20 (1998): 2986–2988.
2. GG Shahidi et al., "Indium Channel Implant for Improved Short-Channel Behavior of Submicrometer NMOSFETs," IEEE Electron Device Letters 14, no. 8 (1993): 409–411.
3. H Hu et al., "A Study of Deep-Submicron MOSFET Scaling Based on Experiment and Simulation," IEEE Transactions on Electron Devices 42, no. 4 (1995): 669–677.
4. IC Kizilyalli, FA Stevie, and JD Bude, "n⁺-Polysilicon Gate PMOSFETs with Indium-Doped Buried Channels," IEEE Electron Device Letters 17, no. 2 (1996): 46–49.
5. IC Kizilyalli et al., "Silicon NPN Bipolar Transistors with
   Indium-Implanted Base Regions," IEEE Electron Device Letters 18, no. 3 (1997): 120–122.
6. DE Kamenitsa and RD Rathmell, "Beam Energy Purity in the Eaton NV-8200P Ion Implanter," Nuclear Instruments and Methods in Physics Research B96, nos. 1–2 (1995): 13–17.
7. G Luckman and RD Rathmell, "Indium Implantation Process Performance on the 8250^_HT" (paper presented at IIT-2000, Alpbach, Austria, September 2000).
8. K Miyashita et al., "Optimized Halo Structure for 80 nm Gate CMOS Technology with Indium and Antimony High Angle Implantation" in Proceedings of the International Electron Device Meeting (1999), 645–648.

Gregory Luckman, PhD, is principal applications engineer for Axcelis Technologies' medium-current implant platform in Beverly, MA. A member of the American Vacuum Society, he has been active in the semiconductor materials and capital equipment industries in photoresist technology, plasma etch and plasma deposition, and ion implantation. He received a BA in physics from Cornell University in Ithaca, NY, and a PhD in materials science and engineering from the University of Pennsylvania, Philadelphia. (Luckman can be reached at 978/787-9240 or greg.luckman@axcelis.com.)

Michel Bruehwiler is the product marketing manager for Axcelis's medium-current implant platform. Before joining the company in 1997, he held various engineering and engineering management positions for capital equipment manufacturers in Switzerland. He received a BS in mechanical engineering from the University of Applied Sciences in Rapperswil, Switzerland, and an MBA from Babson College in Wellesley, MA. (Bruehwiler can be reached at 978/787-4265 or michel.bruehwiler@axcelis.com.)

Robert Rathmell, PhD, is chief scientist for the medium-current implant platform at Axcelis, where he has been involved in ion physics and ion implanter applications. He spent many years in research and industry building electrostatic particle accelerators for a range of ion energies. Rathmell is a member of the American Physical Society. He received a BS in physics from Purdue University (West Lafayette, IN) and a PhD in experimental nuclear physics from the University of Wisconsin–Madison. (Rathmell can be reached at 978/787-9045 or robert.rathmell@axcelis.com.)

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