New Designs in High-Current Ion Implanters

		At a Glance
	High-current ion implantation energy requirements are moving to lower energies much faster than the industry had originally expected. To meet these challenges and maintain throughput, equipment manufacturers have redesigned ion sources, beamline architectures and batch processing systems. In addition to these topics, thermal processes and control of particulates are discussed.

The industry has predicted the demise of ion implantation at 10 keV, below 5 keV and then below 1 keV. Today, R&D labs such as Philips Materials Analysis Group (Eindhoven, the Netherlands) are evaluating 200 eV implantation systems for production worthiness for the next eight  to 10 years. While most full-production, high-current implant equipment operate at 15 keV and above, predictions are that within one to two years, 3-5 keV production systems will be coming on-line.

 

High-beam-current ion implanters (lead photo) capable of generating currents up to 25 mA are required predominantly for 1x10¹⁴ and 1x10¹⁵ atoms/cm² shallow junction formation, only 10's of nanometers deep. To meet these challenges and maintain throughput, equipment manufacturers have redesigned ion sources, beamline architecture and batch processing techniques and have developed single-wafer systems.

Ion sources

High-current implants, traditionally batch processes, rotate wafer-filled disks through the ion beam. (Source: Eaton)

Three types of implant systems are typically found in the fab: high current, medium current and high energy. High-current implanters are optimized for lowest cost-of-ownership (COO) for high-dose implants and can generate up to 10 times more beam current than medium-range implanters. While there is considerable overlap among implant types in terms of application, the economics dictate which system is used (Fig. 1). Available high-current, low-energy capable systems can operate from 200 eV (Fig. 2) to as high as 120 keV and implant all production species.

 

1. The dose/energy map illustrates present (solid lines) and evolving (broken lines) ion implant applications. (Source: Eaton)

2. TEM of a 100 nm line shows a 200 eV boron implant, with major doping <20 nm deep, using xR LEAP and RTP XE Centura systems. (Source: Applied Materials)

Ion sources can be filament-based or indirectly heated by a cathode. Filament- and cathode-based technologies use resistive heating to generate electrons that in turn generate ions. A third method employed by Eaton (Beverly, Mass.) is an RF-based source technology.  Though radio frequency ion sources have been around for years, they were never used for ion implantation in high-volume production environments. An RF source excites the molecules in a magnetically confined environment, generating a cooler, quiescent plasma with higher extractable beam currents. The result is greater beam current and source lifetime.

Once the ions are generated, efficient beam transport is essential in delivering high-beam currents to the wafer. Critical parameters include energy distribution in the beam, ion selectivity and beam diameter. Beam current is lost, for example, by space charge effects. In traditional implanters, the ion beam is extracted from the source by the extraction electrode. The initial beam energy is determined by the potential difference between the electrode and the ion source. At different extraction energies, the distance between the source and the extraction electrode varies. The extracted beam current and the initial shape of the beam are influenced by the position of the extraction electrode.

Traditional implanters tune a beam by moving mechanical assemblies in three axes: in and out, back and forth and up and down. Performed prior to transporting the beam, positioning as well as temperature settling times increase tune time. One solution provided by Eaton is to eliminate moving parts and replace them with a static assembly. Ions are extracted through a self-registered, stationary pentode unit. Tune times for changes in energy, current and transitions between species have been reduced by a factor of 3 to 4 relative to the traditional arc discharge/variable extraction technology, said David Duff, high-current product line manager at Eaton.

Deceleration of the beam, or decel, is typically performed to increase beam current for low-energy beams. The beam can be extracted from the source at a relatively high energy, providing that good extraction current is available. This higher energy beam can then be decelerated down to the desired energy without loss of beam current. An electrode set, further down the beam line, decelerates the ions from the initial extracted energy to the desired final energy. A potential drawback is energy contamination, which can result in a tail of ions implanted at the higher extraction energy, putting device parametrics at risk. All suppliers are implementing methods to delay the use of decel by overcoming space charge limitations to efficiently collect and transport the beam as it was created in the source.

The University of Surrey (Surrey, UK) and Applied Materials (Santa Clara, Calif.) collaborated to develop a patented differential lens (decel) capability to replace quadrupole interchange technology for its xR LEAP implanter. This unique differential lens, combined with a short beamline and open optics
architecture, minimizes beam loss because of space charge so that a high beam density is transferred to the wafer. The beam is focused at the crossover just prior to the wafer surface, and very low beam energies, below 1  keV, can be produced with beam currents as high as 2.5 mA.

Batch vs. serial processes

Traditionally, high-current implants employ batch processing to maintain low COO and, according to Peter Bealo, SHC 80 product manager at Varian (Palo Alto, Calif.), because technology has not been available to sufficiently cool individual wafers. Batch processing requires placing up to 13 wafers on a solid aluminum disk and rotating it 1000-1200 rpm to effectively average the beam power over a very large area. The average power per wafer and therefore heating are minimized.

3. Further reductions in throughput (~4%) may be seen with batch high-current implanters because of the use of dummy wafers. (Source: Varian)

For single-wafer implants, Varian has developed a sophisticated beamline architecture that forms a rectangular or ribbon beam, 200 and 300 mm wide and 40 to 50 mm in height. The beam has a uniform horizontal profile and, unlike traditional nonuniform beams, requires no averaging. The beam is stationary and wider than the wafer. The wafer is scanned vertically through the beam at 1 cycle per second. To shape the beam, an extraction system provides a divergent beam, the angle of which is matched to a 90 analyzing magnet. Once focused for good mass resolution, the beam passes through a mass resolving aperture, diverges again and enters a 70magnet for additional focusing. The beam emerges horizontally uniform and parallel at the wafer surface.

According to Bealo, serial processing throughput higher than that of batch processing (Fig. 3) can be achieved with a patented configuration where two robots are used to handle the wafers. One robot orients the wafer, and the other puts it on the platen. There is no additional translation of the robots themselves, noted Bealo, resulting in less than 4 sec between wafers, a decrease in idle time of ~25% over batch processing. In addition to a larger ion beam, an electrostatic chuck with gas-cooling serves to reduce the amount of heating per square centimeter. Clamping times on the order of 0.1 sec mean that throughput >200 wph can be maintained.

More new features

Implanting source/drain extensions may require high-angle implants under photoresist for shallower depths. The ability to tilt the wafer during implants thus becomes important. High-current batch implanters can generally be tilted 0 However, with serial implants, the electrostatic platen can be tilted 0(Fig. 4) in two axes for implants underneath wafer features and onto sidewalls of trenches.

To make batch processing more efficient, Applied Materials has replaced the standard disk with a spoked wheel (Fig. 5) that has less than 10% of the equivalent area of a solid disk within the beam path, stated Michael Lochner, head of business development, implant division at Applied Materials. The spoked wheel is water-cooled to keep temperatures <50Å. The decreased mass means greater heat dissipation, faster spin-up and spin-down, and tests indicate a factor of 10 decrease in metal contamination, noted Jay Sorochin, director of customer operations at Applied Materials.

300 mm wafers

As bigger die prompt the need for larger wafers in order to sustain productivity, 300 mm tools are coming on-line. Batch systems will require a redesign for 300 mm wafers. Serial machines, according to one end-user, may provide an upgrade path, essentially requiring a larger beam. There are currently three 300 mm high-current implant systems in the field, at International SEMATECH, CNET (division of France telecom) and SELETE (Japan).

Serial implanters are not required for 300 mm implants, though batch machines must overcome two potential problems. A 300 mm wafer contains as many as 2.25 times more die than a 200 mm wafer and costs 10 times as much. The potential for having to scrap a lot is extremely small, though with batch processing it can be very costly. Further costs can be incurred with the need for dummy wafers to fill wafer slots on the implant disk to maintain balance. Dummy wafers, when processed multiple times, can also become a source of particles.

Particles

As device geometries shrink, one of the biggest problems in ion implantation is particles, and higher beam currents can mean more particles, one process engineer noted. Erosion of beamline components and microdischarging are primary contributors to particle contamination in high-current implanters. To combat this, tool manufacturers such as Varian use a patented, novel mass slit design that distributes the intense heat of the high-current beam between the circumferences of two graphite-jacketed cylinders in their batch system. While defining a slit for the beam, the cylinders rotate opposite to the beam path. Not only are fewer particles eroded from the cylinders, but eroded materials have an initial velocity away from the wafer and thus have a low probability of contaminating it.

4. A high-current serial implanter design tilts the wafer over a range of 0Å to +60Å. (Source: Varian)

The combination of the plasma and high temperatures generate particles in  the ion source which make their way to the wafer via electrostatic forces. The low ion density of a large beam means a smaller electrostatic force and, therefore, fewer particles arriving at the wafer. For its ribbon beam, serial system, Varian claims to have only one-third the particles of a batch system.

Materials are a major source of particle generation. Robust testing and studies are under way at all vendor sites, investigating ultralow-wear, vacuum-compatible materials. While equipment required to measure particulates is still being developed, modeling work is ongoing to determine the source of contaminants.

Thermal processes

The effects of transient enhanced diffusion (TED) during post-implant annealing mean that a reduction in boron ion energy may not result in a proportional reduction in junction depth. Thus, with junction depths <50 nm, the need to control TED is increasingly important.

5. The 300 mm patented spoked wheel design is water-cooled to keep temperatures <50Å. (Source: Applied Materials)

Driven by excess interstitials introduced by the implant process itself, TED can move junction depths 10's of nanometers. Recent studies by Aditya Agarwal of Eaton (Beverly, Mass.) and Dave Eaglesham of Lucent Technologies (Murray Hill, N.J.) indicate that lower energy implants result in a faster annihilation of the implantation-induced excess interstitials and hence reduced TED. Reducing the implant energy without a corresponding reduction in dose, however, can lead to an exceedingly high concentration of dopants in the implanted layer, Agarwal noted. If the boron concentration, for example, exceeds a threshold of a few percent, a silicon boride phase may be formed during the anneal, injecting interstitials and further enhancing boron diffusion. A dose of 1x10¹⁵ boron atoms/cm² at 0.5 keV, for example, can produce a junction two times deeper than without this enhancement.

The future of ion implant is rapid thermal annealing (RTA), said Lenny Rubin, principal scientist at Eaton. Precise thermal and ambient control during post-annealing will be essential for producing sufficiently shallow p⁺ source/drain extensions with reproducible junction depths.

Rapid thermal anneals provide better time, temperature and ambient control than traditional furnace anneals. Annealing time, for example, can be controlled to within 1 sec using RTA systems, compared with 30 sec using annealing furnaces, thus opening the possibility for higher temperature anneals. Similarly, the ramp rate of RTAs is considerably faster at 50Å/sec, compared with 15Å/min of furnaces. The nitrogen ambient of RTA systems provide further controls by eliminating potentially detrimental oxygen and water vapor. To study and optimize the relationship between ion implant and RTA for ultrashallow junction formation, Applied Materials' has used the xR LEAP in conjunction with the RTP XE Centura system. To date, they have demonstrated 150Å/sec RTP ramp rates.

In practice, end-users typically quantify the amount of anticipated diffusion and vary the process accordingly. Implants defining channel length place dopants shallow enough to compensate for TED. Even if shallower junctions lessen the effect of TED, little margin for error remains as device geometries shrink.