Track Systems Meet Throughput and Productivity Challenges

		At a Glance
	As the move toward smaller geometries and larger wafers pro becoming more flexible and intelligent. New photoresist chemistries require that these systems be able to simultaneously maintain a variety of thermal budgets and environments, while productivity and yield requirements demand they not only supply, but become one with the stepper.

Photoresist processing systems are among the fab's most complex pieces of equipment. A stepper processes between 60-80 wafers every hour, but it repeats the same steps: loads, aligns, exposes and returns them to the track for development. At any one moment the track could be transferring 15-20 wafers through multiprocess levels, putting resist on some, develop on others and baking as well as chilling them.

Industry-driving issues

Most track manufacturers agree on three principal challenges driving developments in the industry: throughput, thermal management and timing and synchronization issues. Throughput is stepper-driven. As stepper throughput increases, the track must keep up. Stepper throughput has increased significantly over the past three years, and now high-throughput tracks must run anywhere from 75 to 85 wafers/hr, whereas three years ago it was 40-50.

Thermal and timing issues now go beyond bake plate temperature and uniformity. When a wafer goes into the track, it undergoes various processing steps, and the thermal management of each affects ultimate CDs: how long the wafer is exposed to different temperatures, how uniform the thermal budget is and the time consistency between the different process modules. Before, this was not as significant.

The 200-APS is a next-generation photoresist processing sytem designed to provide the advanced capabilities needed to match developments in leading-edge lithography tools in terms of productivity and process performance.

Track technology is at the 0.25 µm level, migrating toward 0.18 µm DUV processing platforms. According to Rick La France, vice president of marketing at Silicon Valley Group (SVG, San Jose, Calif.), "New, sophisticated technology for DUV processing and the rise in productivity are making track platforms increasingly complex, with 20 or even more processing modules to meet throughput requirements, while carrying out involved processing steps" (lead photo).

Advanced technical work is being done by photoresist track companies on wafer transport timing within the system. Advanced computer simulation models view how wafers move through large track systems, to achieve sequences timed within 1 or 2 sec at every step to ensure consistent on-wafer results.

Adaptive process control is fast becoming a necessity. It is impossible to go below 0.25 µm without it, or at least without in situ monitors to report on what is taking place during critical process steps. It is important for wafers to move at the precise moment they must be shifted from one module to the next, from the soft bake to the chills and from the develop to the hard bake, etc.

Resist dispensing and cost

Driven primarily by high DUV resist costs, another area of improvement sought by track suppliers is resist reduction. This is reflected by the rapid evolution from one platform to the next. According to La France, "I surveyed our customers and discovered they were using about 3.5 cc of i-line resist on a 200 mm wafer. Our new platform requires 2.0 cc per 200 mm wafer, almost a 50% improvement. With about 95% of the $2000-a-gallon DUV photoresist going on wafers evaporating or being thrown off in the spin process and down the drain, this is costly and environmentally unsound."

There are two successful approaches to resist reduction. The first is minimizing resist solvent evaporation, enabling the system to put down more uniform wafers with a thinner layer; the second is improving photoresist dispensing technology. If resist and solvent evaporation can be better controlled, resist use can be minimized without sacrificing film quality.

Some suppliers are working on encapsulated coatings -- doing photoresist coating in an enclosed bowl environment. Engineers are considering encapsulated coating and in situ sensitometry in which film quality measurements are made during the process. Another area is adaptive process control schemes that go as far as **tying the stepper into the actual resist thickness to enable better control of the whole lithocell.

1. Fairchild Technologies' PR 800 photoresist processing system has adopted the "local cleanroom" concept. It has a class 0.1 minienvironment that physically isolates alkaline-sensitive and non-sensitive process stations, preventing contamination at the wafer surface during critical process steps. The unit employs filter cascading to control contamination to <1 pbb.

Long term, into the new decade, dry photoresist processing is expected to begin coming on-line. Work is proceeding now leading to the dry application of photoresist and even dry development.

Tightening environmental control

Temperature and humidity control during the application of the photoresist coating is a major challenge. These factors have strong correlations to photoresist uniformity. Environmental control will play a major role in temperature and humidity control. Presently, 0.2% or 0.4% RH is being specced. This will have to be tightened further.

The move is away from manipulating temperature and humidity with remoted units that pipe air into the system over the coating environment. Regardless how good a temperature and humidity control such a system has, the remote environment is not as good as the fab's. Some manufacturers have placed the temperature and humidity control for an i-line platform right on the system above the coater bowl, obtaining better on-system temperature and humidity control. This directly correlates to coating uniformity, which leads to better CD control (Fig. 1).

Thermal management issues in track systems, particularly with 300 mm, are another industry grail. Already, 200 mm track systems have required thermal redesign.

The temperature of the chemicals is also becoming a factor. Resist chemistry must be controlled tighter than ever before, as do developer chemistries. Thermal control is critical not only for the wafer, but for its environment and the chemicals that interact with it. After the wafer has been exposed, particularly in DUV, it goes into a post-exposure bake module where the resist is baked. If hot plate uniformity is not tight, CD inconsistencies can result across the wafer. Specifications in the 0.1°C accuracy range for hot plate uniformity at 130°C are beginning to make their appearance.

2. Clean Track ACT8 is an example of a mature fielded system for the processing of 150 and 200 mm wafers, incorporating high reliabilty and high throughput.

(Source: Tokyo Electron America)

Measuring that kind of thermal uniformity across the wafer is difficult. One approach uses the SensArray wafer, which has calibrated thermocouples attached to it and wires running to the instrumentation, enabling measurements on nine points across the wafer. This method has the double advantage of accuracy and being an industry standard. Although it works well when hot plates require a ±0.5°C accuracy, if higher uniformity is required, difficulties arise. A major disadvantage is that when hot plates are encapsulated with a lid on them, such as small bake ovens, the thermocouple wires make it difficult to use in this type of environment, because the lid must come down on the wires.

The second approach uses a thermal imaging camera, capable of 0.1°C uniformity readings across the wafer; but, again, without the lid. The situation is not ideal, underscoring why instrumentation and measurement are an industry problem.

3. This summary chart puts the complete lithography equipment set in perspective, as far as sources of CD error are concerned. The photoresist used in the simulation and easurement was APEX-E.

(Source: FSI)

Design vs. footprint

As the technology moves to larger wafers, the footprint battle exacerbates. According to an industry executive who requested anonymity, "Systems have become big and complex, but customers want a small footprint. This is a problem for the track industry. We fight each other about who has the smallest footprint, and technology gets lost in the scramble."

This same executive indicated that every time his company reduces a system, internal thermal management issues increase. The reasons become obvious when studies are done to determine where the heat and heat transfers are. Since a track is basically eight machines in one, with handling between them, it is not surprising that thermal problems can arise within it, with all its motors, bake modules and other components

"In our 300 mm approach, we are under tremendous pressure to make the platform as small as possible," he said. "Customers tell us they don't want it any bigger than the 200 mm platform -- it defies the laws of physics." While for steppers the footprint effect of the move to 300 mm is minimal, for tracks it becomes a scaling issue. The industry "guideline" is that a 300 mm system cannot be more than a quarter or a third larger than a 200 mm system. It appears that most suppliers will be in conformance.

Beyond 0.25 µm

Much remains to be done by the track and resist industries in photoresist chemistries to get to resist coatings for 0.25 µm geometries and below. Probably 70% of all DUV systems are running two-layer. In five years, dry processing of some kind will be in use, possibly O₂ plasma. This is expected to take place at around the 0.13 µm level. The SIA's Roadmap shows dry processing down at 0.10 µm for EUV.

Resist thickness also affects final CDs. The smaller wavelengths used during exposure produce a larger interference effect, which means the energy transferred is highly dependent on photoresist thickness. When a wavelength of light comes into the film and is reflected back, either constructive or destructive interference results, depending on thickness. Thus, thickness can vary the sum total of the energy transferred to the film. This is why resist thickness is important across the wafer and from wafer to wafer. If thickness could be truly controlled, considerable savings in time and expensive material would result. Track systems are incorporating means to guarantee wafer-to-wafer thickness, with better temperature and humidity control in the coating chambers and more precise dispensing and spin technologies.

CD imaging control is of prime importance with i-line and DUV exposure tools. Anti-reflective coatings are used to prevent light energy scatter in the film and in top-side application, to protect the resist from chemical contaminants. For CD control, coating and hot plate temperature uniformity are key, but so is the lag between the time that a bake is finished and when it goes on the cool plate. If the system cannot ensure this transfer timing, CDs will vary from wafer to wafer (Fig. 2).

There is a tradeoff between controlling timing within steps and high throughput and flexibility. If the priority is to guarantee that the time for each wafer is the same, some queuing inefficiency will be inevitable. The scheduler must be optimized to do both -- to achieve a balance that provides the quickest throughput, while controlling timing and keeping the stepper busy. In a system where the central robot does the wafer transfer from hot plate to cool plate, if there is not a sophisticated enough scheduler to track when wafers are going to be done baking, an inefficient situation will result where the robot waits for the wafer to come off the hot plate before going to the next operation.

Although Werner Rust, general manager for Fairchild Technologies (city, state), is proud of the performance of Fairchild tools, he believes track system manufacturers are becoming subsystems to steppers: "Steppers are so expensive that end users must get maximum use out of them. This means that whenever the stepper is ready, the track better be ready to provide the wafer. The track system and software architecture must function closely with stepper software and its ups and downs. If the stepper goes down, the track system should be able to put the wafers into safe holding positions until the stepper is ready again."

He added, "You may have 25 wafers ready to go for a mask set, and now you want to switch to a new mask set to keep the stepper running. This means the track system must start a new recipe -- with new wafers -- while the others are finishing up. The system must be smart enough to handle two different process flows and recipe sequences sequentially, so that when the last wafer of the first batch has to the stepper, the other wafers having had a different sequence of steps and different recipes on given modules will follow right after, without having to wait until the last wafer of the last masking step and then start the new process. We've become stepper slaves."

According to Kevin Kemp, manager of FSI's Research Engineering Department (city, state), the company is working to relate CD budget information to 0.18 µm requirements. Kemp said, "CD variation is getting smaller faster than actual CD size. As the move has proceeded from 1.0 to 0.5 to 0.25 µm, CD budgets and permissible variations in transistor CDs have tightened. With 0.25 µm, CD variation budgets of 20 to 25 nm are being considered. For 0.18 µm, this will fall somewhere around 14 or 15 nm." As CDs get smaller, tolerance windows become much smaller. At 0.18 µm or beyond, good CD control becomes photoresist processing's most important challenge (Fig. 3).

Maintaining the installed base

Although the focus is on 300 mm (and even 400 and 450 mm), existing 150 and 200 mm fabs cannot be ignored.

Dr. Ron Miller, vice president of technology and applications at Integrated Solutions Inc. (ISI, Austin, Texas), recalled, "In the beginning, DUV work was done on small substrates in a research environment. When it actually goes to production, the lion's share of the work will still take place at the 150 and 200 mm level. With specialty applications, even smaller substrates may be used."

A point not made often enough is that photoresist processing is not exclusively round thin substrates -- there are square and triangular substrates and different features relative to thicknesses. Photoresist processing is an area with great latitude for growth and development. There are many opportunities in customization. ISI is a good example of a company that operates as a niche market player and expects to pursue markets like GaAs and SAW devices, possibly thin film head devices.

4. Shown is a specialized 200 mm polymide tool. This 60 wph system is seen here in a through-the-wall bulkhead configuration. The combined two-island system is capable of processing 80-100 wph. (Source: ISI Lithography)

Although 200 mm currently rules, there are a large number of fabs at 100, 125 and 150 mm, operating on mid-1980s technology. The migration from 100 to 125 to 150 mm took eight to 10 years each time, and there is still overlap. Based on this, it may not be until 2008 before 300 mm fabs begin becoming prevalent. With the present price of DRAM and memory, the change may be slow in coming. The Asian economic situation does not improve the prospect.

Miller observed, "Although you must continue developing your technology, for the next five years or so you must also continue supporting the 150 and 200 mm fabs. Not all your customers are going to jump to 300 mm at light speed.'

So regardless of roadmaps, life must go on.

Half in jest, some question the need for 300 mm. Doubtless it is needed for reasons of efficiency; however, efficiency must be balanced against cost-of-ownership of equipment in existing installations right down to the cost per die vs. product lifetime vs. device size and shapes. There are benefits to 300 mm, but it will be unique to certain products. Nobody will produce power devices at an $0.89 cost per die device on 300 mm substrate.

According to Rust, with shrinking geometries, the transition into DUV and new chemically amplified resists has amplified environmental control difficulties for track manufacturers because these resists are far more sensitive toward any minor variation. With these resists, the post-exposure bake itself is now becoming a crucial part of CD control. In the past it was stabilizing the image, but now that is basically what creates the image in some ways because it is through diffusion processes that the image is moved all the way down through the resist, so any variations can change the CDs for 0.25 µm and below geometries (Fig. 4).

These are exciting times for the track industry. It is awakening to technical innovations that would not have even been considered two years ago. The next five years we will see considerable innovation in the track industry. The evolutionary approach that has been going on for the last decade is going to change considerably as the 0.25 µm and 0.18 µm realms are entered. Really new technology will be applied to the track industry and for photoresist applications. It will go beyond just improving existing technology. Productivity will increase as a result.