Stage Technology Confronts New Demands

The wafer stage has generally been viewed as just part of a more sophisticated system. However, with the coming of larger substrates and increasingly complex processes, it has taken center stage (no pun intended), as new demands raise its complexity, requiring new capabilities and materials never before used in our industry, to meet evolving specifications (Fig. 1).
 
According to Stephen Kohnle, business development manager, semiconductor and flat panel display, at Danaher Precision Systems (Salem, N.H.), cleanroom compatibility requirements are getting increasingly stringent. "Equipment OEMs must meet strict cleanroom requirements at ISO 3, 2 and even ISO 1 levels.¹ For the stage manufacturer this creates a challenge in ensuring that components meet those requirements, and in the selection of raw materials and components to build the stage." Cleanliness requirements must be considered, as well as the stage's overall performance, and even materials' resonant frequencies.

Jim Fraser, product manager for Ferrotec (Nashua, N.H.), points out that wafer transport has existed since processing started taking place in vacuum. "Wafer transport is mechatronic; it's something that moves up and down to pick a wafer and goes around and around to move it from one place to another. Whenever mechanical motion must occur in a vacuum, two questions arise: what's the contamination potential, and where the prime mover will reside. Will it be in vacuum with the wafers, or outside? In the latter case, mechanical motion must be transferred from the outside world into the clean, high-vacuum environment."

Wafer stage requirements

Scott Jordan, director of NanoAutomation technologies at Polytec PI (Tustin, Calif.), views wafer stage technology as having to meet four demands, depending on the application. "The first is to provide extraordinary, interferometric-class resolution and repeatability across the stage's full travel — 300 mm and sometimes more," he said. "To accomplish it, often interferometers or the latest in advanced glass scale techniques is used. This is a brute force approach to motion control, using extreme built-in metrology. While this method has better feedback resolution in a conventional stack arrangement and a familiar architecture, its disadvantages lie in the areas of cost and the control itself, because with high feedback rates, high speed controls become necessary."

1. Wafer technology has not stopped evolving since its emergence. Today’s systems incorporate full modal analysis and other capabilities. (Source: Danaher Motion)

This approach forces adjustments in control philosophy. This is expensive, because support, training, and software must be changed and updated. Faster microprocessors and feedback are required, while cabling issues are more critical because a pulse rate higher than what some of these instruments can tolerate originates from the digital feedback. The method is also more susceptible to generating EMI or getting it from other sources, sometimes causing channel crosstalk, which becomes increasingly worse with a high pulse rate feedback.

"The second option is to combine lower-resolution, long-travel motion control with fast, on-the-fly, high-resolution motion control, which performs quick, effective corrective actions," Jordan said. "Rapid adaptive optics techniques allow the optical portion to be modulated to compensate for errors originating in the moving portion. The disadvantage is that the optical or lithographic technique must be adaptable to this approach. It isn't possible to just do optical fine corrections on any machine, because the optics have to be specifically designed. Also, there's a cost issue, since the user is paying for an additional layer of mechanics and controls. Sometimes this can be balanced by savings from not having to integrate ultra-resolution metrology into the full 300 mm stage."

The third approach, according to Jordan, is the coarse/fine approach where lower resolution is used, perhaps on a high-speed translation stage, to move the work piece while, in parallel, a piezoelectric nanopositioner performs either very high-resolution terminal positioning or continuously executes mid-motion corrective operations. "This set-up has been used for a long time to level wafers, where high resolution is necessary but not much is needed in the way of travel, and high speed and stiffness may also be required. It's increasingly done for transverse and in-plane rotations, as well as tip/tilt, because it's a cost-effective way to get nanometer resolutions called for next-generation lithography and inspection devices."

This approach also allows faster time to market for the tool, because little changes when upgrading a tool designed five or 10 years ago to work at 50 or 30 nm linewidth. "This can be done by adding the additional layer of nanometer motion control on top of what's already there, without ripping apart the whole design," Jordan said. "While it means another control layer to worry about, this isn't crucial." The main disadvantage is that although this is more cost-effective than trying to provide this level of nanometer scale metrology throughout the travel of a 300-mm translation stage, getting down to 30 nm linewidths requires an investment. Also, the approach does not provide nanometer-scale repeatability over the mechanism's entire travel range — bidirectional repeatability is still no better than the weakest link in the chain: the coarse positioning stage. This method provides resolution, but not necessarily repeatability. This is solved by using pattern recognition of the wafer itself and performing an alignment after each coarse motion step.

"The fourth option is parallel kinematics — multiple actuators operating on a work piece in parallel, instead of a stack of stages," Jordan said. "This works well in certain packaging applications, as very high-resolution, high-speed piezoelectric-class nanopositioners, where a single work piece is increasingly actuated by several actuators working in parallel. This allows, for instance, the x actuator to compensate for errors occurring during the y actuator's motion. Since it's supported in parallel by various actuators, it also makes the work piece a much stiffer mechanism than can be achieved by stacking up the X stage on top of the Y stage on top of the Z stage, etc. It also provides faster settling (faster damping) for motorized actuators, as well as compensating for parasitic motions on the fly."

Another advantage is that, depending on the configuration, virtual coordinate system transformations may be performed. For example, in a hexapod configuration it is possible to rotate around practically any point in space. The controllers are designed for this; as with a conventional rotation stage, the rotation point is not fixed at a rotary bearing's center. The controller's firmware defines it. The disadvantage is configuration-contingent. "For example, we make a motorized hexapod device, which provides all six degrees of freedom over a travel range of several inches with high resolution down in the submicron range," said Jordan. "However, travel's fairly limited compared to a 300 mm X-Y stage — limited travel can be an issue with this approach."

Unlike vibration, which can be eliminated, there exist fundamental bandwidth issues in system positioning, when a process is performed on the fly. The goal is to eliminate following error, the tendency for the servo-driven mechanism to always lag slightly behind from where it is supposed to be. According to Jordan, presently, there are workarounds for servos and amplifiers' limited bandwidths and the rest of the elements in the servo chain — technologies are available to address following errors and eliminate recoil-driven motion errors occurring beyond the feedback loop.

Peter Gise, senior marketing manager at Nanometrics (Milpitas, Calif.), does not believe that the technology will meet near-future requirements. "Due to cost, size, and performance needs, a suitable wafer stage technology to meet next-generation needs, especially those of integrated, 300 mm metrology, doesn't exist," he said. "In addition to the necessity of low initial costs to OEMs, low-cost ownership for end-users and high throughput, which are particularly important in today's economy, there are the requirements of accuracy, precision, and reliability that must be addressed. High accuracy and precision capabilities are critical to reduce the number of templates as well as the total search time needed to perform pattern recognition.

Gise added that 300-mm wafer automation now uses edge-gripping positioning and handling to avoid potential wafer damage and particle contamination. "As a result, we opted to design our own high-throughput edge-gripping wafer stage technology. We're one of the few equipment suppliers that manufactures its own X-Y-Z-R-θ stages."

"An understanding of the application for which the stage is built is crucial," said Danaher's Kohnle. "This requires thorough comprehension of what's the OEM's customer-specific application for which the stage will be used, what performance level is needed, what are the payloads and weights, and the speed at which the machine must operate. As higher levels (submicron) of precision and accuracy are contemplated, matters like thermal stability become important. Air temperature must be controlled; otherwise, the different materials' expansion coefficients will affect the stage's servo stability."

There is a move away from steel ways of cross roller bearing stages, toward air bearings for higher levels of precision, since this architecture lacks mechanical contact and is a cleaner design. Also, the advantage of air bearing stages is that error motion is repeatable; as it can be mapped out, whereas a mechanical stage changes as it wears. Aluminum and ceramics were used for air bearings; however, with the levels of performance and cleanliness required today, new materials — perhaps even advanced composites — must be considered (Fig. 2).

2. Air bearings are well suited to semiconductor manufacturing applications. They are designed to float on a thin film of pressurized air while being simultaneously pulled down by vacuum under the bearing. The vacuum force pre-loads the air bearing lands, giving the air film high stiffness without added mass. This configuration suits requirements for high acceleration and fast settling times, eliminates compound errors, minimizes angular and thermal errors, reduces total stage height, and simplifies driving at the center of mass. (Source: New Machine Components Inc.)

Some performance levels may be specified without fully understanding whether they are truly needed for the application. Is 0.25 µm accuracy over 300-mm wafer travel necessary? "Sometimes a natural frequency is specified, say, 200 Hz, at the top of the stage," Kohnle said. "Do you need it or just think you do? We go through each specification with the customer, and carefully question him about what performance requirements are really important. Specific details of the application are critical in this discussion. What really comes into play is the additional hardware our customer will ultimately mount or install on the stage. Understanding this at the beginning of the design cycle is critical as we define the design, select motors, feedback options, and motion controller of the stage, as well as how it'll be tested to meet the performance demands. Designing for volume manufacturing is key. We don't want to create a 'science project,' but a high-performance stage that can be built in volume production as demanded by our OEM customers."

Cabling is the equation's other side — it can impact stage performance. "We'll design a stage that achieves high performance levels — straightness of travel flatness and accuracy," Kohnle said. "When the customer puts in his cables (which may be too heavy and were not discussed before), stage performance is degraded by added inertial forces that weren't considered, creating a problem we must resolve either through the servo-loop, or through a different cable arrangement scheme."

With materials, manufacturers face challenges where the stage must be both light and very stiff. Some materials lack the stiffness needed for high-end stage construction. "For example, we've used ceramics in the design a stage for a memory repair application, where an acceleration of 2 g and a fairly aggressive duty cycle with 1 m/sec velocity were required — aluminum was out," Kohnle said. "For other designs we've used composites such as silicon carbide for applications such as lithography, which requires high stiffness and lightweight construction. These materials have inherent benefits, but also disadvantages — cost being one. When you change materials it may impact cleanroom compatibility; for example, if you have a stage crash, what does the material do? Is it brittle? At the high-end stage applications the need for advanced materials is evident."

According to Ferrotec's Fraser, their main expertise is manufacturing rotary shaft seals using magnetic liquid sealing. "Our exposure is fairly broad in wafer transport and any mechanical motion that occurs when processing wafers in vacuum. Sometimes, we make seals for robotic devices. Typically, a robot arm used in vacuum to pick up a wafer from a cassette and place it in a processing environment has its mechanical motion produced out in the atmosphere — a shaft rotates through the chamber wall between vacuum and atmosphere. Multiple shafts are required if there are wrist and arm and shoulder motions, or wrist and elbow and shoulder motions."

Even if the entire robot arm lives in vacuum, it generally has an atmosphere environment inside. Any time articulated motion is required, it must be accomplished acknowledging the fact that the chamber is a vacuum (anywhere from 10^-3 to 10^-8 Torr) while the arm structure's interior is at atmosphere (760 Torr) to contain motors requiring cooling and wiring, and perhaps even devices needing water cooling. "The farther the prime mover is from where the action occurs, the more accuracy suffers," said Fraser. "Installing the prime mover where the action takes place would simplify achieving accuracy, but complicate achieving vacuum quality and cleanliness."

Another technology, magnetic coupling, uses exterior and interior magnets, with a solid chamber wall between them. When the outside magnet moves, the inside one follows. "Because this is a soft instead of a solid coupling, there's a lost motion disadvantage, but it avoids wall penetration," said Fraser. "However, now there's a support structure for bearings that must hold the rotational and translating axes inside the vacuum, and must be lubricated, giving rise to potential sources of particulate contamination, etc." To preserve accuracy and shorten the prime-mover-to-end-effector distance, motors can be placed closer to the wafers. If a three-axis robot arm has its movers inside, then three motors and three encoders must be part of the moving structure. However, the parts' inertia is considerable, and can affect throughput by slowing down motion or sagging the arm.

Maintaining accuracy depends on several factors, a principal one being size. The bigger the gadget is, the harder this is, because there are longer distances to span and small angular errors matter. The move from 200 to 300 mm wafers has given the mechatronics business considerable problems in terms of accuracy. This typically translates into challenges for bearing stiffness and the stiffness of the arm elements themselves. This drives the industry towards very stiff, very compact bearings, which benefit the business of those who make tapered roller bearings and cross roller bearings, as opposed to the most commonly applied ball bearings. Because of the stiffness requirement, all these bearings have to be made of steel.

Wafer stage technology will continue to evolve. For the future, nanoimprint will impact it. This is the embossing or printing of nanoscale structures using a template pressed on to a substrate to form a pattern, or a stylus to write it on the substrate. This is useful for making masks or for forming structures on wafers in lieu of conventional lithography. In many cases it could allows IDMs to form finer structures quicker, easier, and more economically, than with conventional lithography.