Issue



High-purity valves evolve to meet future demands


12/01/2002







by Kirk Mikkelsen, Don Ware, Mark Caulfield and Jim Kegley

The design and material property of high-purity valves have improved over the years to provide tighter liquid chemical process control and increased cleanliness

In the semiconductor industry, there are a number of products working behind the scenes to ensure material integrity. These products range from wafer carriers and shipping boxes to the tubing and valves that handle the critical fluids used in semiconductor processes—all greatly impacting a semiconductor manufacturer's throughput, efficiencies and costs.

One key material's integrity management product is the high-purity valve, which is used in all facets of the semiconductor industry—from chemical distribution to process tools. The valve is responsible for controlling critical fluids by stopping, diverting and metering liquid flow. However, to fully understand the major role this product plays, it is important to look at the valve's history, evolution and future.

High-purity past

The high-purity valve has been around for several decades but has made substantial changes since it was first introduced, particularly in design and material property. When these valves were first introduced to the semiconductor market, they were fully machined products constructed of Teflon (PTFE). The product, with features acquired from the metal valve industry, was not highly reliable.

In addition, the semiconductor process tools incorporated crude fluid handling systems and the need to improve the valve became apparent—particularly as wet chemical process tools began to evolve in the early to mid 1970s. Thus, driving demand for a new sophisticated, chemical-resistant and reliable valve.


A Weibull analysis can find the life cycle reliability of a valve. The solid straight line is the best-fit regression line, while the curved line represents the 95 percent confidence that the failures experienced by the subset of valves tested represents a full population.
Click here to enlarge image

null

One of the key issues with early high-purity valves was the material of construction. Early PTFE was susceptible to creep and cold flow, which negatively affected the valves' reliability and longevity. Teflon perfluoroalkoxy (PFA) improved creep and cold flow issues and had the ability to be injection molded—characteristics that would improve the product's reliability.

Injection molding provided several benefits to the new valves, including a reduction in leakage and improved cleanliness at a lower cost. Since that time, injection molding of valves and fittings has become the standard.

However, new processing methods of PTFE and modified PTFE are once again bringing machined products back as viable options. This allows designers of high-purity valves to select the best material for the given application, taking full advantage of material properties.

Improving with time

The microelectronics industry—which has been the primary driver behind the development of high-purity valves—is continually creating new challenges.

Wafer line widths continue to decrease and circuit complexity continues to increase, requiring significant changes to the role and definition of high-purity valves. The industry changes are demanding tighter liquid chemical process control and increased cleanliness. This is driven partially due to tighter geometry but also the increased costs of specialized high-purity chemicals and a need to tightly control consumption. The purity levels of liquid chemicals continue to be pushed to lower and lower levels, driving costs higher and higher.

Other challenges include industry standards—such as Semiconductor Equipment and Materials International (SEMI; San Jose, CA) F57—that are being developed in an attempt to create minimum performance requirements for liquid chemical handling components. The standard's prime focus is on attributes such as surface finish, particle performance and extractible cation/anion levels.


A Weibull plot can be used to demonstrate Weibayes analysis. According to the plot, the first of the six valves would fail on average at 1,400,000 cycles.
Click here to enlarge image

null

Fab safety is also playing an increasing part in the changes required for high-purity valves. The physical complexity and costs of capital equipment and fabs have increased dramatically. As a result, product reliability and purity are more paramount than ever.

Reliability is important to the semiconductor industry because it enables a longer life cycle, less periodic maintenance and increased resistance to its harsh environment. By continually improving and enhancing the valve's technology, it can now combat many problems with leaks and durability, as well as withstand extreme pressure and temperature conditions. Today, the valve also has lower maintenance costs and is safer in managing caustic chemicals.

Purity (or cleanliness) is another area that has been significantly improved over the valve's evolution. By experimenting with the materials used to construct the high-purity valve, levels of contamination due to trace metals are down significantly.

As the purity of chemicals used in semiconductor processes increase and the geometries in microelectronics decrease, the purity of tubing and valves become essential. The importance of material selection and understanding has become forefront to many in the industry to enable the highest performance in key products. Testing has also become important to educate manufacturers and ensure they are getting the most out of their product.

Reliability testing

One of the more sophisticated ways to analyze capabilities of the high-purity valve is through the use of Weibull analysis. Weibull analysis is a statistical method used to analyze time-based failure data.1, 2

Software packages are available that perform Weibull analysis. The number of samples on test and the times (cycles) for each failure are entered into the software. It then plots the data on a loglog-log plot and finds the best-fit regression line and regression parameters.

Figure 1 shows such a plot for a set of valves containing hydrochloric acid. The solid straight line is the best-fit regression line, while the curved line represents the 95 percent confidence that the failures experienced by the subset of valves tested represents a full population.

To find the life cycle reliability of a valve, choose the threshold reliability. In Figure 1, 95 percent reliability (five percent unreliable) was chosen. The point at which the best-fit regression line crosses the 95 percent reliability line is the cycle-to-failure value where 95 percent of the valves would be expected to survive (on average) if cycled under these conditions. This value does not account for the small sample size used to represent a full population of valves. To take into account the sample size, the cycles-to-failure value should be taken from the point at which the 95 percent confidence line intersects 95 percent reliability.

Weibull analysis works well for situations where a minimum of two, but preferably three or more, failures have occurred. However, there are many situations where decisions have to be made prior to the start of testing or after one or fewer failures have occurred. In these situations, techniques such as Weibayes analysis and the techniques described in SEMI E10, "Specifications for Definition and Measurement of Equipment Reliability, Availability, and Maintainability," can be used. These techniques work well for predicting the length of time a test will require or how many cycles without failure are needed to get the desired lifetime from a product.

In the case of valves, past data can be looked at to determine the slope of the best-fit line in a Weibull analysis for typical failure modes. The slope indicates the probability density of the failures (how widely distributed the failures will be). The steeper (larger number) the slope, the more narrow (more predictable) the range of failures. This information can be used in a Weibayes analysis, in which valves of similar design typically fail in a similar fashion.


The slope of the best fit line has been set beta equal to 3 and estimated test times—time to first failure on lower 95 percent confidence boundary—are calculated for three different desired cycle lives.
Click here to enlarge image

null

Weibayes analysis is a "special case" Weibull analysis with zero or one failure. In this analysis, the slope of the best-fit line is set based on historical failure data, prior experience or engineering knowledge of the physics of failure. By selecting a desired confidence level and reliability level for a full population and number of samples on test, the minimum test time to achieve those goals can be determined.

Figure 2 shows a Weibull plot that will be used to demonstrate Weibayes analysis. In the plot, the desired minimum life is 500,000 cycles and six samples are put on test. A 95 percent reliability and 95 percent confidence are assumed; this still leaves five percent unreliability and five percent chance that samples could fail sooner than predicted. Statistically, the first of the six valves would fail on average at 1,400,000 cycles.

Testing to this value assures one that based on prior product knowledge, 95 percent confidence and reliability can be obtained at the required product rating of 500,000 cycles. If the first failure occurs at less than 1.4 million cycles, the best-fit line will shift to the left and the product rating will have to be adjusted accordingly. Additional testing and a second failure later than predicted will shift the line back to the right, but based on only one failure, the product cannot be guaranteed to meet the confidence and reliability requirements.

Weibayes analysis can be used to create a matrix of predictions based on the number of samples and the desired product life. This is demonstrated in Table 1 where the slope of the best fit line has been set beta equal to 3 and estimated test times—time to first failure on lower 95 percent confidence boundary—are calculated for three different desired cycle lives.

Table 1 also demonstrates the importance of putting more samples on test. The time to achieve confidence in the design of a new valve is significantly shortened by increasing the number of test samples. This information allows project teams to weigh the expense of additional samples to the savings achieved by obtaining results quicker.

More improvements to come

Increased functionality is where the high purity valve is moving in the future. There is an industry drive to increase the uptime in the fab to greater than 95 percent and shorten the equipment integration time, creating a demand for more robust products with a longer life cycle. This will bring about many changes with the valve—some currently taking place—such as reducing the number of components within a product, creating an all-plastic valve option, improving flexibility and implementing electronic intelligence into the product.

A trend that is gaining momentum in high-purity liquid chemical handling is closed-loop flow control. The process gas side of the microelectronics industry has used closed-loop for some time now, so the concept is not new. There are many new adaptations of existing technology becoming available to measure different media characteristics of high-purity chemicals such as pressure and flow. Now that there are reliable means of measuring the characteristics of high-purity fluids, valve function is changing to keep pace.

Closed-loop liquid flow control valves are beginning to evolve. Valves are moving away from the standard pneumatic actuation supply and beginning to incorporate stepper motors and electronics to provide precision movement, control and feedback. The valves are linked to measurement devices such as pressure differential, ultrasonic, coriollis, paddle wheel and other flow measurement technologies. Applications for these closed-loop systems include system backpressure control, precision dispense, blending, chemical totalization and others.

Basic valve design has changed as well. Fabs are larger, more automated and employ new chemical sets. These changes have significantly affected valve design.

Large fabs require greater flow capacity, larger orifice products and more efficient plumbing systems to operate. Fluid paths within valves and fittings are being scrutinized to ensure optimum efficiency. Computational fluid dynamic software is employed as a tool to help optimize flow paths within individual components. This analysis allows for increased flow efficiency from components of the same size. For chemicals such as slurries, it means reduced shear and agglomeration, thus providing healthier chemicals to the point-of-use.

Valve designers are also creating large multi-functional manifolds. Rising costs per square foot of fab space are driving tool manufactures to keep equipment footprints to a minimum. Aiding in maintaining footprint, as well as helping to increase tool functionality, valve manifolds have become a critical tool.

Valve manifolds are incorporating surface mount technology, which allows for easy component replacement and servicing. Technology is being created to allow equipment designers the ability to define their own simple manifold configurations.

Manifolds are being created to combine similar functions, to flush slurry lines, to sample chemicals and to precisely mix and blend chemicals with the incorporation of closed-loop flow control devices. Working from customer schematics, valve designers can create complex flow paths and unique functionality in a single manifold. These small manifold free up space for the many other components to be squeezed into the process tools.

There is also continual drive to cleaner and cleaner chemicals coupled with the emergence of chemical mechanical planarization (CMP) slurries. The "Holy Grail" of control valves is to have no moving parts, no pressure drop and high reliability and accuracy even when exposed to harsh chemicals. Some of today's suppliers are coming close, but the industry still has a way to go.

References:

  • 1. G.W. Ireson and C. F. Coombs, Jr., Handbook of Engineering Reliability and Management, McGraw Hill, 1988.
  • 2. D. Kececiouglu, Reliability Engineering Handbook, Vol. 1&2, PTR Prentice Hall, 1991.

Kirk Mikkelsen is the director of Entegris' Technology Characterization Laboratories located in Chaska, MN and San Jose, CA. He has a B.S. in chemical engineering and an M.S. in management of technology from the University of

Don Ware, a test engineer at Entegris' Chaska, MN-based lab, has worked at the company for eight years in the product testing environment and holds a B.S. in mechanical engineering from the University of Minnesota. Ware can be reached at [email protected].

Mark Caulfield, Entegris' applications engineering manager for its fluid handling products, has worked in the semiconductor industry for 13 years and holds a B.S. in industrial technology from Moorhead State University. Caulfield can be reached at [email protected].

Jim Kegley, a technical market specialist for Entegris, has 22 years of experience in the semiconductor industry and holds a technical degree in design engineering. Kegley can be reached at [email protected].