Ion Implantation Goes Beyond Traditional Parameters

For years, implant was dominated by hardware design and practically considered a commodity product. Today, it is central to device performance changes. The rubber meets the road at the transistor shrink, and implant is a key part of it. OEMs are focusing on integration, cooperative processes with those involved in activation, and gate formation. Where implant once was an orphan fab process done by different groups, it is now being rapidly integrated into the main line of transistor formation (Fig. 1).

"The recent ITRS roadmap revision formalizes what several leading device manufacturers had already discovered — there's a strong preference for extending mainstream ion implantation technology," said Thomas Parrill, high-current product line manager for Axcelis Technologies (Beverly, Mass.). "While junction depth must still be reduced for each shrink, there has been a 180° shift in the extension sheet requirement. Instead of decreasing along with junction depth, the sheet resistance targets now increase. The industry has recognized the new requirements are the preferred way to manufacture high-performance transistors, and the effect of extension resistance on drive current is not as severe as previously thought."

"It's clear that a key challenge for ion implant is transistor scaling, which has obvious implications," said Norma Riley, chief marketing officer for the Parametric and Conductive Implants Group at Applied Materials (Santa Clara, Calif.). "Implant will continue to extend its capabilities and be a vital part of device formation. Changes resulting, for example, from the emerging use of SOI (silicon on insulator) will involve new emerging implant applications used in the formation of these substrates."

"Some companies tout the 'one-machine' concept, where one single ion implanter processes all ion species, doses and energies," said Sarko Cherekdjian, senior product manager for Therma-Wave (Fremont, Calif.). "This fictitious Holy Grail has been pursued since 1986. Ion implanters can perform a multitude of implant process recipes, but trade-offs are inevitable in beam purity, repeatability or throughput. This is because the implanter architecture governs the process specifications achieved. For example, if you want a tight low-dose implant, you must meet the process requirements for a good medium-current implant — beam purity with excellent dose uniformity and repeatability. Using a high-current batch ion implanter without neutral trap or beam filter for low-dose implants will compromise the ion beam purity and repeatability. Implant users must realize the consequences of using the wrong ion implanter architecture for the desired implementation step."

	1. Ion implant technology continues to reinvent itself to meet the requirements demanded by continuing shrinks, delaying the introduction of alternative methods such as plasma ion immersion and laser doping. (Source: Axcelis Technologies)

Riley views implant as a still-growing market segment. "Implant was solely dominated by productivity issues. Now other technical challenges are emerging, requiring doping precision and accuracy. There's concern over energy and doping accuracies, as well as the overriding worry over productivity."

Applied divides the market between parametric and conductive applications, focusing on a device technology rather than a hardware segmentation. "In the conductive implant arena, transistor engineering is dramatically scaling the source/drain (S/D) structures, resulting in junctions becoming extremely shallow," Riley said. "Consequently, requirements are for abrupt junctions with higher doping levels, but lower energy." (Fig. 2) This makes hardware design increasingly difficult, but the technology is coping, and improvements in beam current and low-energy performance continue. OEMs face issues of solid solubility, which require the focus to shift from the implant to the activation side. As long as platform manufacturers can provide low energies with good productivity and accuracy, the problem then is one of solid solubility and shallow junction formation.

"Although we're getting requests for lower energies, manufacturing will most likely occur at the 500 eV level," Riley said. "You can be reasonably productive there, particularly with the hardware improvements taking place. We're working with our counterparts in the Thermal Product Group, exploring new methods of dopant activation and looking for ways to increase device performance. For example, can we apply new information methodologies such as data — feedforward, feedback — between the implanter and the annealing system, that can be used for in situ optimization of the junction formation, making transistor performance more repeatable?"

Most ion implant critical challenges originate in transistor scaling. "For parametric doping, there's the accuracy required in the halo implant formation," Riley explained. "As devices shrink, the accuracy of placing the halo and the source/drain extension becomes more important. We believe that transistor scaling will continue pushing doping layers shallower — implant systems must become much more precise as we move to <130 nm device nodes."

2. The increasingly demanding architectures of new devices require unprecedented control to insure performance. As measured by Boxer Cross, actual data from implant systems show that even small energy errors have considerable effect on junction depth. (Source: Applied Materials)

Plasma ion immersion (PII) is considered by many to be in the cards; however, it is not without its warts. There are three areas where there are issues with PII. One is dosimetry accuracy. Conventional implant dosimetry does not work well for PII. Ion implant's precision is taken for granted — the dose is precisely defined, not by time or cutoff — the dose in silicon is actually measured as it goes on. This is not a simple matter with PII. The second area pertains to tilt requirements. Many implants today are tilted to get the dosage under the gate. Here, PII is inefficient and would be limited to just a few implants, reducing its flexibility. The third area also pertains to precision: PII makes mass resolution impossible because, when a plasma is produced, everything is in the plasma. Work is ongoing to minimize that effect, but mass analysis in a beam-line implanter is very precise, and that precision would have to be surrendered. Beam technology innovations are expected to provide some of PII's benefits.

Presently, PII is not as efficient as a clean beam, although for unique applications it works well. For example, if a hip joint must be implanted, and there is a curved shape with an indent, PII will implant at all angles all around the material. With ion implant, it typically is possible to only go in one direction — it will not cover this kind of feature. However, plasma tends to be dirty, is not as uniform, and in high doses is likely to get hot.

PII is under some consideration for hydrogen implants. The problem with this is that the wafer must be very cold because of hydrogen mobility, and contamination of H₂ or anything else that is with the hydrogen being implanted at a certain energy is undesirable. The concept is not meeting with great success.

Basically, PII is the ion source's ion implanter, without the mass analyzer. It may eventually become useful for specific applications because, instead of looking at $1 or $2 for the effective cost of the medium-current-style implant, the cost is a few cents per wafer. Potentially, PII is so inexpensive that it is worthwhile to research whether the device works. Like GaAs, once predicted to replace silicon, PII may excel in niche applications.

PII has been investigated for many years, and is still being researched at universities. Several IC manufacturers are interested in it for many reasons, and most ion implant OEMs have development programs that have produced useful IP. However, there is uncertainty in productizing it because of the continued success in low energy on the mass analyzed platform. Most believe it can still meet the roadmap requirements for a few more generations. Beyond that, everyone faces a possible architecture change, and it is not clear what the best solutions might be.

As smaller device technology node requirements drive the energies for S/D extension, gate doping, and contact implants to increasingly lower energies, traditional beam-line-based ion implanters lose significant levels of productivity because of the physics of beam "blowup" from space charge forces. Although ion implant suppliers have made significant strides in ion beam transport to preserve traditional systems, these tools' throughputs eventually may become too low to be production-worthy.

	3. Shown are the lateral doping profiles of S/D across the channel. Implants were done self-aligned to the gate. As beam incident angle deviates from 0°, the source junction (right) moves away from the gate edge due to implant shadowing effects, detrimentally affecting device characteristics. (Source: Varian Semiconductor Equipment Associates)

Thus, it is not clear whether there is really a place where plasma doping fits in the roadmap. It is an interesting technology that certainly bears investigation, but also one whose time has not yet come. Current technology is fully capable for requirements in the foreseeable future. Beyond that, it is unclear whether it will be the proper response to the radical changes that might occur.

Right now the transition to SOI is taking place, and it is beginning to go mainstream. This will intensify shallow junction work and change the use of some traditional implants, such as well implants. "We believe we're well-positioned with our extended-energy-range tool, which provides very precise, low-dose implant capability over a very wide process regime," said Applied's Riley. "We're continuing to improve low-energy performance and productivity on our conductive implant platforms. SOI will be the next big transition, although the question is whether the infrastructure currently exists to support widespread adoption — much will have to be achieved in a hurry."

Axcelis believes that current implant and RTP technologies should easily take the industry through the 70 nm node. "The ultralow-energy, mass-analyzed implant plus the spike in the RTP should satisfy the roadmap through that technology node," Parrill said.

Like its competitors on the implant side, Axcelis continues to improve its low-energy beam transport capability to increase the throughput for implants at low energy, where the biggest throughput hit takes place. "We've introduced an electron confinement technology capable of reaching considerably higher beam currents than before, sometimes doubling the beam current at low energy," Parrill said. Axcelis overcame the space charge associated with low-energy beams by introducing electrons into the beam line to neutralize the positive charge of the ions.

For some time, there have been attempts to produce implanters with a split personality — low- as well as high-current-capable. As Therma-Wave's Cherekdjian put it, "Trying to make a machine that does all species — boron, arsenic, phosphorus, antimony — and all energies is, in my opinion, presently not practical."

Going from 200 eV all the way to 3 MeV requires drastic changes in ion optics to achieve ion beam efficiency. Presently, there does not appear to be a compromise-free solution that keeps throughput or beam purity from suffering. "Direct-line-of-sight ion implant tooling is an example," Cherekdjian said. "For this toolset, high-energy implants can, and are regularly specified, with doubly and triply charged ions. This is insane, as single-plus, double-plus and neutral-ion contaminants will be present in this implanter architecture. These unwanted ions are vacuum-dependent, and are just waiting to proclaim their presence. Devices sensitive to these contaminants won't survive."

Possibly in the future, a compromise solution may be two, instead of three, platforms. This will depend upon the implant requirements from future semiconductor devices. Presently, CMOS implants are requiring lighter doses with lower energies. One machine could be a low/high-dose, high-tilt, low-energy tool below 80 keV. For the second tool, semiconductor manufacturers may still require implant energies in the MeV range. This would then render a broad energy range unnecessary.

According to Cherekdjian, ion implantation overall has not had any true innovations in terms of dose repeatability. "Each technology node raises the dose repeatability bar — you must be even more repeatable — but there haven't been true advances. We're still measuring the charge; we aren't coupling calorimeters with a Faraday cup, for instance."

Ion implant tilt-angle repeatability is another factor today. Its performance slowly degrades with time. The implant tilt angle of an ion implanter requires additional monitoring and must be recalibrated and requalified on a regular basis.

"If you're in the ultrashallow junction arena, a basic tool would address 0.2 keV to 5 keV arsenic and boron implants," Cherekdjian said. "The ion implanter could be the size of a large refrigerator and not a room. Unfortunately, if you require low-energy implants, such as 0.2 keV, you need a large analyzer magnet or a double split analyzer magnet design to achieve the desired ion mass resolution. However, the large bending and manipulation of low-energy ion beams results in poorer beam transmittance — there's always a compromise."

How long before ion implant runs out of steam depends on the situation's physics — the control of ion beam channeling, lateral dopant distribution, etc., Cherekdjian explained. Other technologies will peter out well before ion implant. Besides PII, emerging technologies such as laser doping may phase in, but not any time soon. "Ion implant users love their momentum analyzer, and as yet have found no superior technology," Cherekdjian added.

Tadashi Oka, manager of the overseas sales department for ULVAC Technologies' (Methuen, Mass.) Semiconductor Equipment Group, has his own view of the contest between PII techniques and low-energy, high-current implanters. "Traditional implanter technology is mass-production-proven. However, plasma immersion techniques can lower CoO (cost of ownership) while reducing capital cost and footprint. Even so, repeatability (mostly dose count technique) in plasma immersion is not yet fully understood."

Oka added that medium-implant technology's major problems today are parallelism of the ion beam, contamination (metal/cross/energy), and wide energy range (Sb implant, for instance) for channel engineering.

The drive for smaller junction depths has been well publicized with shrinking device geometries, said John O'Connor, director of strategic marketing at Varian Semiconductor. "Recently, however, the criticality of lateral junction abruptness to enhance device performance has come to the fore. This abruptness requirement mandates shadow-free implantation that can only be addressed at small geometries with single-wafer, parallel-beam implantation."

Varian Semiconductor's Mehta added that doping concentrations associated with halos/pockets have increased with a concurrent reduction in energy, increasingly moving these implants into the high-current implant category. "These implants also require precise angle control at a variety of large tilt angles, from 0 to 60°, but typically 25 to 45°."

O'Connor believes the major technological hurdle faced by implant is achieving precise dopant location control. "This is becoming more critical as device geometries shrink below 130 nm. For ion implantation, this translates into critical angle control of both the wafer in the implanter and also control of the ion beam itself." He noted that the solution is found in single-wafer processing coupled with parallel ion beam implantation. "The angle of both the ion beam and the wafer's physical location are precisely controlled to ensure that all the dopant ions impacting on the silicon strike at the desired angle. This prevents shadowing and encroachment and enables manufacturers to use zero-degree implantation without the deleterious effects of uncontrolled channeling." Process advantages include lateral junction abruptness for S/D extensions and reduction of well isolation spacing. The solution would enable precise placement of the halos/pocket doping at high-tilt angles ranging from 0 to 60°, but typically 25 to 45°.

Varian believes future doping requirements in the low-energy regime can be addressed by pulsed-plasma doping technology, which overcomes many objections to traditional PII approaches. Charge-based dosimetry methods, similar to those of beam-line ion implant, can ensure accurate doping levels. Also, the short duty cycle of the presence of the plasma protects against many PII process-related concerns, such as particle formation in the plasma and metallic contaminants from etching. Issues related to the lack of mass selectivity can be addressed through careful choice of the feed material. The relaxation of the depth/RS conflict in the latest ITRS furthers the demand for throughput at ultralow energies. The ITRS also shows that, while RS requirements for S/D extension have been relaxed, junction depth requirements continue to become more stringent, eventually limiting the throughput of traditional beam-line implanters. Significant efforts are required to continue producing beam-line implanters capable of very low-energy beams. However, the associated risk, cost and loss of throughput of beam-line implant make this impractical. In contrast, as highlighted in the doping roadmap in the 2001 ITRS, the enormous throughput advantages of pulsed-plasma doping will spur the transition of ultralow-energy implant to this new technology.