Vacuum technology: Forces of change keep reshaping vacuum systems in ion implant
04/01/2004
 |
Since they were introduced into semiconductor processing in the 1970s, ion implanters have been characterized by change. An explosion of implanter architectures has occurred in the last decade. Vacuum systems serving these tools have been undergoing their own transformation to keep up with process requirements and growing cost pressures in fabs. Increases in implanter throughput, greater tool flexibility, and the ability to handle new implant species and gases are reshaping vacuum system designs for improvements in pump speeds and performance as well as real-time closed-loop control with integrated "intelligence." The pace of change certainly is not slowing.
Integrated circuit designers requiring unprecedented line densities and higher levels of device performance are putting tremendous pressure on ion-implanter suppliers to push the capabilities of tools while driving down the overall cost of operation. Among the major ion-implanter design goals today are increased tool flexibility, greater process control, tighter dose uniformity, and higher throughput — including faster processing times with improved yields, shorter cycle times, and extended tool uptime.
The growing list of new implanter requirements does not end there. Device manufacturers are also calling for single-tool designs that handle multiple types of implant recipes for deep-well, shallow-junction, and other wafer applications. Systems also must handle new implant species, which require implanters and integrated vacuum-pumping apparatus to cope efficiently and effectively with evolving process-gas loads.
Constantly increasing yield requirements are exerting more pressure for contamination control from all sources, including materials used in tool construction. Energized plasma gases now used in ion implantation demonstrate that different vacuum-pumping schema must be employed within the source beam line and target chamber areas to ensure optimal system performance. Implant designs also must address more stringent safety requirements worldwide, as well as new power-conservation guidelines that affect vacuum-pumping systems, which must run continuously while other tool functions are inactive and are often responsible for a large part of tool-level power consumption.
Getting to where we are now
The mounting pressures on tool design are orders of magnitude beyond what creators of the earliest implanters could have imagined. The first implanters were ion accelerators with radiation chambers, used to subject an assortment of test samples to ion beams. When ion-implantation applications related to semiconductor wafers were developed in the 1960s and 1970s, wafer diameters were about 50mm or smaller, and in most cases the beam was scanned over the surface of the wafer using two dimensions of electrostatic deflection. The process was much like TV, in which an electron beam is scanned over the screen.
As wafer diameters grew, this design proved unable to deliver the necessary level of quality. Consequently, ion-implanter designers created tools using a one-dimensional scanning system with a collimator lens that bends the beam back to normal incidence before it strikes the wafer. The wafer is then scanned mechanically in the other dimension to dose the entire surface, shown schematically in Fig. 1a.
 Figure 1. a) How generic batch ion implanters work compared to b) typical single-wafer implanters. |
Unfortunately, systems employing this flavor of electrostatic scanning are limited in their maximum beam current since strong electric fields prevent neutralization of the space charge by electrons; this beam current limit is typically about 1mA. This limitation led to dedicated tools that did not deflect the beam at all. With the first high-current implanters, multiple wafers could be mounted on a disk in such a way that the entire surface of all the wafers could be passed under the beam (see Fig. 1b). There were two types of systems:
- Electrostatically scanned medium-current systems
- Spinning disk high-current tools
These tools represent the vast majority of the current installed base of ion implanters. But from a vacuum-design perspective, they are very different. Medium-current tools use low-current beams, and they do not require space charge neutralization. As a result, they propagate well in a near-perfect vacuum.
When electrostatic deflection is utilized, any residual gas left in situ will lead to ionization and may contribute to electrical breakdown of high voltages used in the process. When residual gas is left in the accelerator or collimator regions, its presence can lead to energy shifts, angle changes, and dosimetry errors.
Clearly, for medium-current ion implanters to perform ideally, the lower the pressure, the better. The challenge is to pump the gas coming from the wafer and the source as effectively as possible and in particular to prevent endstation gas fluctuations from passing up the beam line. Because of this, designs for future implanters focus on getting as much pumping speed as possible at the endstation and limiting the conductance from the endstation to the beam line. Moreover, process quality in future systems relies on maintaining an acceptable pumping speed throughout the interval between pump maintenance cycles. Ideally designed cryopumps deliver constant pumping speed until their capacity is reached.
High-current ion implanters are far different from their medium-current cousins. If you attempt to send a high-current ion beam through a perfect vacuum, the electric repulsion between the ions themselves will cause the beam to "bloom." The beam assumes a shape similar to a trumpet as its diameter increases, due to its self-repulsion. This is definitely not a desirable design feature.
To transport the beam through the beam line to the wafer, high-current ion implanters inject a "plasma gas" — such as argon or xenon — or water to generate a neutralizing charge of electrons. This aspect of the process is important to the design of the vacuum system because "the lower the pressure, the better" no longer strictly applies. The ideal pressure is typically around 10-5torr.
On the other hand, it is still essential to efficiently pump gases coming from the wafer — hydrogen, hydrocarbons, carbon monoxide, and others — and it's critical to maintain a constant vacuum environment for stable and predictable beam operation. Some cryopumps are now designed precisely for this type of application. They can enable users to select first stage temperature, which typically prevents xenon "hang up," explained later.
Flexibility is critically important
There is an ongoing debate about whether single- or batch-wafer processing is "best." Single-wafer tools have advantages: reduced cycle time, greater flexibility of process parameters (such as implant angle), less product at risk in case of tool failure, and no throughput penalty for odd wafer-lot sizes. Batch tools may have lower cost of ownership for routine high-dose implants.
Nonetheless, a key point here is that in many cases, throughput is limited by pumping speed, regardless of tool type. This limit typically comes about because continued increases in beam-current capability run into limits imposed by the vacuum system, where process quality or uniformity may be compromised by variations in gas pressure. When cryopumps deliver consistent vacuum performance, it is not only possible but likely for the ion-implantation system to deliver greater useful throughput.
A renaissance in implanter design
The last decade has seen an explosion of new implanter architectures. Although each has its own peculiar vacuum requirements, new architectures often require higher beam currents and lower minimum energy for throughput.
SIMOX batch implanters use a spinning disk to mount wafers, and then use a magnetic scan to send the beam across the wafers. These are specialized implants (for separation by implantation of oxygen — SIMOX — wafers), performed at high temperature with very high oxygen beam currents. There is no photoresist or other volatile material on the wafer, so the only gas load is produced by the beam and beam neutralization scheme. Turbopumps are used exclusively to pump the oxygen.
High-current scanned-beam serial implanters have the basic architecture of a medium-current tool but use electromagnets to deflect the beam, enabling them to transport even higher-current beams. These tools require background gas for neutralizing plasma; gas pressure must be uniform throughout the volume swept by the scanning beam.
High-current ribbon-beam implanters use a stationary beam that is uniform in one dimension. The wafer is mechanically scanned in the other dimension. Because the beam is wide, its current density is much lower than other designs with the same throughput; lower background gas pressure is sufficient to neutralize the beam.
Plasma immersion systems implant the entire wafer in a stationary position by immersing it in a uniform plasma. These systems operate at much higher pressure than high-current tools, with pressures in the millitorr range. These systems are also typically turbopumped.
 Figure 2. The evolving and changing process applications for implanters. |
What is critically important both now and into the future is tool flexibility. It is not enough today for an ion implanter to simply do something well. It is critical that the ion implanter have fast cycle times not only for the initially targeted process, but for future processes as well. Competitive cost pressures require tools to be easily and quickly configurable for evolving and changing process requirements (see Fig. 2).
High-energy concerns
When ion-implantation systems generate ion beams with a higher energy, new things start to happen in the vacuum environment. Photoresist outgassing reaches peak pressures rapidly and then drops off dramatically due to carbonization. This causes a distinct difference in the vacuum during the implant and makes uniformity more difficult with short-duration implants. Because customers may have to add xenon, argon, or other gases to create a plasma in which the beam can operate, they will need to deal with those gases as well as increased outgassing that results from the process.
Changing and evolving ion-implantation processes introduce new gases, such as indium, oxygen, and hydrogen, which also must be dealt with in the vacuum environment.
Although hydrogen is commonly assumed to be the major constituent of outgassing during ion implantation into wafers that have photoresist on their surfaces, serious consideration must be given to the other gases caused by ion implantation. When an ion beam impinges on the polymeric photoresist, there is significant molecular reconstruction, releasing a wide range of chemicals. Hydrogen is a common atomic component of the polymers, and this forms H2 gas, which then evolves from the surface. Other gases observed include atmospheric gases that are dissolved in the photoresist. Figure 3 shows both a high-energy and a low-energy B+ implant on the same implanter. The 200keV implant shows typical hydrogen-rich outgassing.
 Figure 3. Peak actual pressure and partial pressures for high-energy and low-energy B+ implants into 300mm photoresist wafers on the same implanter. |
However, on the 20keV implant, the carbon monoxide/nitrogen peak (mass 28) is the dominant gas load followed by water and then hydrogen, which accounts for <10% of the total. This demonstrates that outgassing composition can change significantly over the wide energy range of today's implanter. Gases such as xenon, acetylene, and carbon dioxide add a unique dimension to the implanter vacuum performance. Those gases have a high ionization constant (a desirable quality for a flood gas in the case of xenon), but their pumping speed can vary with the temperature of the first-stage cryogenic array. In standard cryopumps, first-stage array temperatures may fluctuate in response to heat loads, cleanliness of the arrays, and aging refrigerator performance. Since many of these gases are actively cryogenically pumped on the first stage of the cryopump, maintaining a constant first-stage temperature will add consistency to the vacuum performance of the implanter.
Xenon pressure "hang up" in particular is a challenge for cryopumps. The earlier generation of cryopumps maintained the first-stage temperature at about 65K. At this temperature, xenon has an equilibrium vapor pressure of 10-5torr, while at the second-stage temperature of 13K, this drops to <10-15torr. Under these circumstances, some of the xenon is cryo-adsorbed onto the first-stage surface, while most of it condenses onto the second stage as a low vapor-pressure solid. When the xenon flow stops and one would expect the pressure of the chamber to drop to the usual base pressure (well below 10-5torr), the xenon adsorbed on the first stage liberates and maintains a significant partial pressure in the chamber. After a few minutes, the first-stage xenon is exhausted — it has all been recondensed on the second stage — and the pressure does eventually drop to the expected base-pressure value. This process is colloquially known as "hang up" and can lead to excessively long pump downtimes.
The solution to hang up is to increase the first-stage temperature to 80–100K. This prevents any significant cryo-adsorption on the first stage, as at this temperature the equilibrium pressure is about 0.1torr. Then, with no xenon stored on the first stage, the pressure will drop quickly upon the cessation of gas flow. This example illustrates why it is so important to be able to tune the operating point of the cryopump in light of the specific requirements of the application.
System-productivity trends
The increased use of different gases and higher-energy beams creates environments that should bring us back to a core fact of ion-implantation life: Gases must be strictly minimized except where needed, regardless of the gas species, and conductances must be limited to the beam line itself. This requires ongoing technical advances designed to control the temperature of the cryo-condensing arrays and with it, the presence of gases and flexibility of the ion-implantation system itself.
For example, first-stage temperature control in the On-Board IS system enhances the system's ability to optimally pump gases such as argon and xenon, which improves overall tool uptime and productivity.
Implant deposit buildup can be a significant problem in ion-implanter pumps. Increasingly, implanter designers are devoting serious attention to mitigating this problem or preventing it from having any deleterious effects on ion-implantation processes. Pumps designed to withstand the effects of buildup offer more consistent vacuum performance for longer periods of time.
Past-generation 250mm pumps have been prone to gradually increasing first- and second-stage temperatures, both between regenerations and over a long period of time. The change in temperature between regenerations occurs by a slightly different mechanism than the change in temperature that happens over a longer period.
Between regenerations, implant deposits and water collect on the first stage. These condensates immediately change the radiative load on the first stage by altering its ability to reflect heat from the surrounding environment. The change can be quite dramatic — a factor of 4 to 1. Changes to the surface properties of the first stage can increase the radiative load that the pump dissipates through the pump orifice by a factor of four. This increase in load can impact many 250mm pumps' first-stage refrigeration capacity, causing first-stage warming. This warming of the first stage in previous 250mm pumps can readily result in warming of the second stage. Changes in temperature of either stage can affect pumping performance; variable pumping performance between regenerations means that the implant process is not truly consistent. The solution is to regenerate the cryopump more frequently, but this reduces uptime and productivity without really solving the problem.
Over a long period of time, first-stage warming continues to be related to the buildup of deposits from the implant process. Even with a very aggressive regeneration routine, some implant deposits stay on the first- and second-stage cryoarrays; the regeneration process is not able to remove them. These deposits will increase the load on the first stage of the cryopump described previously, but now it can happen between regenerations. After many months of service, the problem can actually hamper the cryopump's ability to reach its out-of-the-box first- and second-stage temperatures immediately after a regeneration. Consequently, the implant-related deposit problem can also contribute to long-term drift of the first- and second-stage cryopump temperatures, leading to long-term drift in the system vacuum performance.
Cryogenic refrigeration systems can handle implant deposit buildup with almost 50% more cooling capacity than anything previously available. For example, the incremental cooling capacity of Helix's On-Board IS 250F cryopump keeps the cryogenic surfaces at constant temperatures even with a 4× load buildup through the pump orifice. The result is that the cryopump can remain in service considerably longer before it needs to be serviced.
 Figure 4. Traditional cryopump control. |
As tool control-system architectures evolve to a distributed control methodology that complements modular designs, the cryopump control systems must evolve as well. Illustrated in Fig. 4 is a traditional cryopump control system with limited or no real-time closed-loop control. An OEM's system-design engineer is responsible for integrating the cryopump controls and, more important, developing the algorithms required to regenerate the cryopump and monitor the second-stage temperature.
When system engineers require real-time closed-loop control for time-critical information, they seek alternative design approaches that have sufficient bandwidth and response time to satisfy their requirement. In most cases this includes purchasing high-speed analog-to-digital converters and protocol translators to transmit the time-critical information to the system controller. The net result is an increase in development budgets and added hardware. In addition, because the cryopump control system was developed for a specific tool, the level of cryopump control system reuse is minimal. Therefore, tool-specific cryopump control systems are often not easily integrated into new development initiatives or easily adapted to new applications.
 Figure 5. Adaptive control architecture for implant cryopumps, such as the On-Board IS system. |
A solution is adaptive control systems shown in Fig. 5. The OEM's system engineer integrates a set of high-level software commands in the system controller, which communicates directly to the cryopump controller to provide real-time control and data acquisition of time-critical measurements. This is accomplished by a distributed cryopump control architecture in which the cryopump controller monitors key pump parameters, compressor helium supply and return pressures, and AC line voltage and frequency. The information is applied to an adaptive algorithm in the pump controller that adjusts in real time to the pump's consumption of helium by varying the speed (rpm) of the cryopump's drive motor. The entire system can adapt to changing heat loads of the vacuum system and variations in AC line voltage and frequencies without sacrificing vacuum performance.
Back to the future
Looking ahead, the newest ion-implantation vacuum system designs must address a growing list of factors and new requirements into account as device manufacturers continue to push tool capabilities. Among the major trends are:
- Lower base pressures and higher operating pressures in endstations, and lower pressures in beamlines.
- Far greater flexibility/programmability to enable ion-implantation systems to accommodate evolutionary changes in implant processes and gas species.
- Rapid pressure reading availability for real-time closed-loop control.
- Optimized pumping of gas species that could cause surface defects.
- Quicker recovery times to base pressure for incremental uptime/productivity.
- Higher system reliability — less downtime for maintenance operations.
- Lot-to-lot and wafer-to-wafer vacuum consistency.
- Long-term high-productivity vacuum performance (overcoming the accumulation of implant-related byproducts).
- Proper management
emoval of potentially hazardous gases and condensed solids. - Increasing vacuum performance without compromising energy efficiency.
- Integrated vacuum subsystems — both integral to and optimized for the process tool and essential to its consistent performance and productivity, not "one from column A and two from column B" components added incidentally to the system.
Frank Sinclair received his MS at Cambridge U. and his PhD at London U. He is VP and CTO at Helix Technology Corp., 9 Hampshire St., Mansfield, MA 02048; ph 508/337-5139, e-mail [email protected].
Michael J. Eacobacci, Jr., holds a BSME and a MS in materials science, and is a senior technologist at the CTI-Cryogenics unit of Helix.