A new era for high-current, low-energy ion implantation
11/01/2000
Thomas M. Parrill, Michael S. Ameen, Michael Graf, Richard Mazzola, Axcelis Technologies Inc., Beverly, Massachusetts
overview
For high-current ion implantation, energy reduction is the critical technology process trend, and it presents fundamental process and equipment design challenges. Effectively transporting low-energy ions is only one aspect of the challenge, though, because new process equipment must work in new fabs with higher demands for process control, productivity, and reliability.
 Figure 1. A high-current beamline must be specifically designed for low-energy ion transportation. Several critical areas, discussed in the text, are highlighted. |
A new generation of wafer process equipment is needed to manufacture ICs incorporating devices below the 180nm technology node. Junction depth (xj) is one of the critical components of technology node scaling, and it is driving the continued development of high-current, low-energy ion implanters. Published roadmaps call for drain extension xj to decrease steadily below 60nm for the last phase of the 180nm node, which includes 100nm microprocessor gate lengths [1]. Such junctions may be formed with ion implantation and annealing by reducing both the as-implanted dopant depth and the thermal diffusion during activation. It is known that, for p-type junctions, this approach requires ultralow-energy (<1keV) ion implantation and "spike" rapid thermal anneals (essentially zero seconds at maximum temperature) [2].
In this article, we first review the basic issues associated with delivering low-energy ions to silicon wafers. We then discuss three major fab needs that significantly influence the design considerations of low-energy implanters: process control, productivity, and reliability.
Delivering and controlling low-energy ions
As the importance of high ion implantation beam current at low energy grows, finding solutions to the fundamental issues of space charge blow-up and beam instabilities becomes more challenging and more critical. These fundamental challenges exist throughout the entire beamline, from the ion source extraction to the final drift region near the wafer (Fig. 1).
The ion current density (Jmax) extracted from a charged particle source has fundamental limitations described by the Child-Langmuir equation [3]. The basic dependence of the maximum extracted ion current density on charge (q), mass (m), extraction potential (f), and extraction gap (d) in a planar geometry is the following:
 |
null
As f decreases (a requirement for low-energy implantation), the maximum allowable current density decreases. Practical limitations on physical electrode design and electrical breakdown limit the extent to which reductions in extraction voltage can be counteracted by reductions in extraction gap dimension (d). In addition, since there are practical limitations to the cross-sectional area available to transport the beam, the maximum useful beam current that can be extracted from the ion source decreases strongly as the extraction potential decreases. In implanter design, this fundamental space charge effect can be addressed by allowing larger beam sizes or by implementing deceleration technology ("decel"), which effectively increases the extraction potential. Designing implanters for larger beams, however, has implications for other implant parameters, as discussed below. With decel, care must also be taken to minimize energy contamination caused by neutralization before or in the decelerating field.
Beam size is limited by defining apertures along critical points in the beamline. An increase in beam size, due to space charge blow-up or deliberate beamline design, has other fundamental performance consequences. The resolving slit located at the focal point of the beamline is one such critical point related to the system mass resolution, which helps to define and control beam purity at the wafer. The ability of the implanter to filter out energetic species with mass-to-charge ratios similar to the desired implant species is a defining parameter of any implanter.
Implementing a series of selectable resolving apertures (rather than using one fixed aperture) allows a reasonable compromise between beam current and mass resolution over a wide variety of high-current applications. Slits must be sufficiently narrow to filter out near-neighbor contaminants, including C+ when implanting B+ (mass 12 vs. mass 11), BF+ when implanting P+ (mass 30 vs. mass 31), and CF2+ when implanting BF2+ (mass 50 vs. mass 49). Slits must also be as wide as practical to transport the highest possible beam current — the relationship of mass resolving slit width to low-energy beam current is approximately linear. Sets of selectable resolving apertures span a range of sizes that vary by about a factor of two to provide mass resolution (m/Dm, where Dm is the full width half maximum) that exceeds 20 even at the lowest mass and energy. Dedicating slits to different species has the added benefit of reducing sputtered cross-contamination on wafer surfaces.
 |
In drift regions of the beamline, an ion beam with high space charge experiences a strong diverging force due to the relatively high density of similarly charged particles in the beam. This diverging force is constant for a given space charge density, regardless of the velocity of the beam, so the net effect on beam size over a given length of travel will be much greater for slowly moving (i.e., low-energy, high-mass) ions. The obvious mitigation is to neutralize as much space charge as possible and to minimize the total distance the ions must travel.
Charge neutralization is partially accomplished via plasma electron flood (PEF) technology that serves the dual purposes of low-energy beam transport enhancement and wafer-surface charge control. Well-designed PEF systems provide critical control of beam transport in the drift region near the wafer. Experiments have shown low-energy beam current enhancements >50% using a well-designed PEF system. Figure 2, for instance, shows 2keV B+ beam profiles measured at the wafer position using an in-situ profile monitor [4]. Beam blow-up and the associated loss of beam current is evident when plasma flood control is absent.
Process control
In ion implantation, dose control is the primary concern for process control — the dopant must be delivered to the wafer repeatably and uniformly. While technology for controlling implant dose was developed some time ago [5], low energy presents special challenges. As the energy of the implant approaches the sputtering threshold, effects such as dopant sputtering and surface erosion must be considered. It has been demonstrated that ~10% of the implanted B dose (at a nominal dose of 1 x 1015cm-2) may be lost due to self-sputtering effects at 500eV and ~20% at 200eV [6]. Other near-surface effects, such as native oxide thickness and surface preparation, also become significant contributors. At 200eV, for example, over 40% of the B dose may be lost due to the combined effects of sputtering and native oxide removal [6].
Photoresist outgassing must be managed, since it is related to both dose control and energy contamination via charge-exchange with ambient gas atoms that neutralize energetic ions. Fortunately, the phenomenon is well understood — the primary system variable is pressure, which is readily measured. The effect, which may be described as a ratio of the measured current (Im) to the actual current prior to neutralization (Io), is
 |
null
where P is the pressure and K is a factor related to the neutralization cross section of the species being implanted in the outgassing ambient. The actual dose to the wafer consists of measured and neutralized species, both of which must be accounted for. Improved control is therefore afforded by a dosimetry system capable of compensating for the effect. However, energy contamination cannot be compensated for since the deeper implanted species potentially defeats junction depth, the parameter that requires low energy.
Energy contamination can be monitored, though. Since space charge forces act only on charged particles, any loss of useful beam current at the final implant energy does not necessarily correspond to an equivalent loss of neutral dopant atom flux. These neutrals may arrive at a higher energy, depending on the position in the beamline where charge exchange occurred. Monitoring beam transmission throughout the beamline therefore provides a mechanism to identify situations where risk of higher energy contamination is present. A situation with insufficient space charge control near the wafer could lead to such a condition. With automated monitoring and tuning parameters available to correct these potential occurrences, a high degree of confidence that low-energy contamination is below tolerable limits is provided.
Another important process control issue is charge neutralization for gate oxide integrity. As gate length shrinks and junction depth decreases, gate dielectric thickness also decreases, placing new demands on charge control technology to avoid damage to sensitive devices. As oxide thickness decreases, effects such as direct tunneling through the oxide increase, leading to soft breakdown mechanisms that are difficult to control and measure.
Stress induced leakage current has been shown to increase, and sensitivity to dielectric field strength has been proposed as the cause of damage in ultrathin oxides [7].
Low-energy beams are by definition not as rigid as higher-energy beams and therefore are less likely to introduce high values of electrical stress on the wafer. Ultrathin oxides are more susceptible to damage by the beam electric field, however [7]. These materials, along with related nitrides and oxide-nitride-oxide stacks, are damaged by Fowler-Nordheim-type current flow induced by nonuniform plasmas in contact with the wafer. Spatially effective ion beam neutralization is therefore paramount for achieving desired device yields.
 Figure 3. Plasma electron flood (PEF) a) schematic with circuit and b) three-dimensional cutaway view. |
PEF technology, besides aiding beam transmission, has proven to be a highly effective method to meet the charging demands of advanced devices [8]. The technology, illustrated in Fig. 3, provides a self-contained supply of electrons generated near the wafer surface. Electrons couple to the beam through a "plasma bridge," which acts as a low impedance electron source that can respond rapidly to potential gradient changes during processing. The temperature of electrons (Te) generated with a PEF system is very low, on the order of several eV, if proper setup is used. Lowering Te reduces the Fowler-Nordheim current at the wafer. The table shows the primary PEF setup parameters and effects of those parameters. Although PEF operation may be tuned to accommodate special device requirements, a typical setup is valid over a wide range of ion beam conditions.
One important parameter is Iarc, the current used to drive the plasma in the PEF. This current directly determines the plasma electron density (Ne). Figure 4 shows CHARM charge monitor results for implants over a range of Iarc values for As and BF2 beams. Since some devices are more sensitive to one polarity of charge, this allows a window of latitude for process optimization.
Shallow junction process control therefore requires proper ion implant dosimetry, energy contamination control, and charge control. In ion implantation systems, uniformity control via two-dimensional mechanical scanning eliminates the need to tune beam profiles iteratively and therefore reduces setup time prior to implant, compared to a hybrid (ion beam plus mechanical) scanning approach. The common implementation of 2D mechanical scanning — the spinning disk — is typically termed "batch" while the hybrid approach is "single-wafer." The term "batch" in this context is somewhat misleading since it typically connotes equipment, such as vertical furnaces or wet benches, which process several cassettes (lots) of wafers simultaneously. Multilot batch systems present queuing problems for manufacturing fab personnel, who must sometimes decide whether to wait for additional lots to arrive (to form a full batch), or to go ahead and process the available lots. "Batch" implanters are more properly termed multiwafer tools, since the batch size is a fraction of the typical lot size (25 wafers). Multilot queuing is therefore not an issue for implant. Setup time is an important factor, however, favoring multiwafer implanters with 2D mechanical scanning. This example illustrates how the method of across-wafer dose control affects productivity via setup time, and how both of these factors are related to the implanter architecture.
Productivity
High implant productivity for volume IC manufacturing is not limited to increasing beam current at low energy. Improvements in beam utilization can have an equivalent effect that is often overlooked. One effective way to improve efficiency is the use of active dose-control, executed so that disk scanning (in and out of the ion beam) reverses as soon as the wafer edges are out of the beam. Without this, the disk must scan additional distance to make sure there is no possibility of the disk reversing while the beam is still on the wafers, a situation that would result in a large dose error. Figure 5 illustrates the concept. When the ion beam is "small" (Fig. 5a), the disk scan length and total effective implant area are reduced compared to the "large" beam (Fig. 5b) case. This situation is easily encountered in a system that switches from 80keV As+ to 1keV B+, where space charge differences produce different beam sizes. For a typical disk size, the difference in effective implant area for a 30mm beam is approximately 18% lower than for an 80mm beam. For a nominal dose of 1 x 1015cm-2 and 3mA beam current, this area reduction is equivalent to a 20% increase in beam current.
 |
Figure 4. CHARM a) positive and b) negative potentials as a function of Iarc.
As process time decreases, increases in the beam current result in diminishing throughput and cycle time returns. The platform overhead therefore becomes increasingly significant. In this case, efficient platform design should maximize the amount of wafer handling performed in parallel with implant and minimize the amount in the critical path between implants. The salient figure of merit is the maximum implant time at the system mechanical limit, with larger values indicating a more efficient platform. An added benefit of increasing the implant time is process control "insurance." Better control is available for longer implants, when all other factors are equal.
Reliability
With increasing pressure to decrease process development and qualification time (driven by the need to reduce overall time to market and time to revenue), fabs do not have the luxury of spending significant time to "break-in" new process technology. Also, development centers or pilot lines for advanced process development are now typically embedded within manufacturing fabs, and process equipment must support manufacturing requirements as well as run advanced processes in order to maximize capital asset use. To accommodate this trend, equipment suppliers must address reliability issues before shipping early systems to customers by incorporating proven technology into new designs and by instituting rigorous reliability programs for new equipment products or platforms.
One tactic is extrapolating the implanter end-user environment to lower level systems, assemblies, and components for testing. Accelerated test methodologies such as HALT (highly accelerated life testing) and HASS (highly accelerated stress screening) can be developed and subsequently passed onto subassembly suppliers for execution. This process shortens the time required to qualify new designs while improving the identification of infant failure modes and, in some cases, longer-term failure modes.
A second area, applicable at the integrated system level, is the approach to reliability demonstration test (RDT) planning. Including automation considerations along with the expected education and experience level of the fab work force in the testing protocol is critical to uncovering problems and improving reliability.
Figure 6 is a plot of failure rate vs. time for a low-energy implant system level test operated in a simulated production environment at the equipment manufacturing site. The test protocol required that nonexpert operators run the equipment. This additional boundary condition uncovered aspects of the control system for improvement and led to subsequent increases in operational uptime and mean time between failure (MTBF). Because of the protocol, the events that contributed to the system failures included many "soft fails" that could have easily been classified as assists if the machine had been run by an expert operator.
The plot illustrates the period of "infant mortality," a period when the failure rate decreases with time, typically seen at the leading edge of the classic "bathtub curve." An exponential trend line is included to enhance this point. Calculating MTBF during the infant mortality period would not be appropriate since by definition MTBF is the reciprocal of a constant failure rate in a Poisson or exponential distribution. Once constant failure rate is achieved, the next phase of the reliability program can begin.
Conclusion
The need for low-energy ion implantation to form ultrashallow junctions for advanced technology nodes is well known and is driving the next generation of high-current implanters.
To transport low-energy ions effectively, new beamline designs are required, and we discussed some of the design considerations related to extraction, mass resolution, and space charge neutralization. Simply providing more beam current at low energy is not the whole story, however, since new implanters must work in new fabs with stricter requirements for process control, productivity, and reliability. Implanter productivity is strongly affected by platform efficiency, and tool suppliers must develop rigorous methods of testing and improving reliability before shipping systems into fabs where manufacturing is just as important as development.
References
- The International Roadmap for Semiconductors, "Process Integration, Devices, and Structures," Semiconductor Industry Association (SIA), p. 86, 1999.
- A. Agarwal, et al., "Ultra-Shallow Junctions and the Effect of Ramp-up Rate during Spike Anneals in Lamp-Based and Hot-Walled RTP Systems," 1998 Int. Conference on Ion Implant. Tech. Proceedings, pp. 22-25, 1998.
- P. Spädtke, "Beam Formation and Transport," in Handbook of Ion Sources, edited by B. Wolf, p. 365, CRC Press, 1995.
- P. Splinter, et al., "In Situ Beam Profiling by Fast Scan Sampling," 1996 International Conference on Ion Implant. Tech. Proceedings, p. 272, 1996.
- H. Glawischnig, J. O'Connor, F. Sinclair, "Modern Implanter Concepts," Ion Implantation Science and Technology, ed J.F. Ziegler, Ion Implantation Technology Co., p. 375, 2000 Edition.
- A. Agarwal, et al., "Ultra-Shallow Junction Formation using Ion Implantation and Rapid Thermal Annealing: Physical and Practical Limits," presented at the 197th Meeting of the Electrochemical Society, Toronto, Canada, May 14-18, 2000.
- J. McPherson, V. Reddy, K. Banerjee, H. Le, "Comparison of E and 1/E TDDB Models for SiO2 Under Long-term/Low-field Test Conditions," IEDM Tech. Digest, p. 171, 1998.
- M.E. Mack, et al., "Optimized Charge Control for High Current Ion Implantation," 1998 Int. Conf. on Ion Implant. Tech. Proc., p. 486, 1998.
Tom Parrill received his PhD in materials science and engineering from Northwestern University. He is the high current marketing manager at Axcelis Technologies, 108 Cherry Hill Drive, Beverly, MA 01845; ph 978/232-4261, fax 978/232-4200, e-mail [email protected].
Michael Ameen received his PhD in chemistry from UNC Chapel Hill in 1986. He is manager of process technology at the implant systems division of Axcelis Technologies. He has authored more than 50 articles on semiconductor processing and holds eight patents in various processing areas.
Mike Graf received his PhD in applied plasma physics from the Massachusetts Institute of Technology. He is a chief scientist and manager of high current beamline technology at Axcelis Technologies.
Rich Mazzola is an MS candidate in material science at Worcester Polytechnic Institute. He is manager of the implant and thermal reactors design group of Axcelis Technologies.