Issue



Monitoring particles in process chemicals


04/01/2007







Considerations for selecting and operating a particle counter

By Ross Bryant and Roger Carlone, Particle Measuring Systems

High particle levels in process chemicals have been shown to directly affect product quality. The need for continued reductions in particle size and quantity in cleanroom environments is well established. The International Technology Roadmap for Semiconductors (ITRS), sponsored by the United States Semiconductor Industry Association (SIA), defines the need for persistent decreases in particle levels in process gases, water, chemicals, and air in order to meet future generation technology nodes. The ability to accurately measure these contaminants is the first step in developing a comprehensive particle control plan.

Principles of operation

Although numerous techniques are available to monitor and characterize particulates, most require significant off-line analysis and interpretation. This usually reduces the number of measured samples, resulting in fewer data points and time-delayed results. Liquid particle counters (LPCs) are widely used for continuous on-line monitoring of process chemicals. Proper selection and implementation of an LPC are important for ensuring data integrity.

Click here to enlarge image

Although the design, performance and specifications of LPCs may differ, a few key principles remain the same. First, particle counters do not directly count particles. Particle counters, based on light scattering, are designed to measure the equivalent optical size of particles as referenced against a calibration standard. In most cases, that standard is a polystyrene latex sphere (PSL). These are readily available and National Institute of Standards and Technology (NIST) size traceable.

Laser light, with wavelengths usually between 600 and 800 nm, is used to illuminate the sample cell or capillary. As individual particles traverse the laser beam, light scattering occurs because of the index of refraction difference between the transport media and the particle. The amount of scatter depends on many factors, but is largely dependent on the particle size, shape and index of refraction. Scattered light is then collected and focused onto a photodetector, where the signal is converted into an electronic pulse (see Fig.1).

Click here to enlarge image

The dashed lines (see Fig. 2) represent the voltage values or channels defined for each particle size during calibration with PSLs. Each pulse represents an individual particle, while the magnitude of the pulse determines its size by comparing the PSL equivalent amount of scatter.

Selecting the right instrument

There are a few key things to consider when selecting a particle counter for your application:

  • sample volume
  • resolution
  • sensitivity

Sample volume: LPCs can be classified as either volumetric or non-volumetric. The terms refer to the sample percentage illuminated by the laser beam in the sampling region or capillary during operation.

In volumetric counters, the beam is shaped such that its intensity is mostly uniform across the capillary. Because the energy intensity profile of a laser is Gaussian-shaped, the beam span in volumetric counters is significantly wider than the capillary. This has the effect of producing a uniform beam so that the traversing particle is exposed to the same amount of light energy regardless of where the particle is located in the flow path, improving resolution. Since 100 percent of the capillary is uniformly illuminated, all particles larger than the sensitivity limit of the instrument are counted. State-of-the-art volumetric counters can detect particles as small as 0.1 μm. These are used to monitor process baths, for incoming quality control checks, as well as for some process chemicals.


Figure 3. The Liquistat is an example of a volumetric particle counter.
Click here to enlarge image

In order to increase the sensitivity of an LPC, the background light scatter from the capillary must be eliminated. By removing the capillary, the laser beam is then focused and shaped in such a way that only a very intense portion of the beam illuminates a very small portion of the total flow, greatly improving the signal-to-noise ratio. The result is a significant improvement in instrument sensitivity. State-of-the-art non-volumetric counters can detect particles as small as 0.05 μm; however, the decrease in sample volume greatly reduces resolution. Non-volumetric counters are typically only used to monitor DI water and some extremely clean process chemicals. These counters tend to produce poorer instrument-to-instrument matching and should only be compared using total cumulative counts.

Resolution: Resolution is defined as the ability to distinguish particles of various sizes. For example, an instrument with very good resolution, a spectrometer, would be able to resolve the difference between a 0.8 μm and 0.9 μm particle. Instruments with average resolution would consider the particles to be of similar size. Volumetric counters provide great resolution and, hence, offer a large number of user-selectable size channels. In contrast, many non-volumetric counters offer a few very broad size channels that are manufacturer-defined.

Sensitivity: Sensitivity is defined as the smallest detectable particle size for a given instrument. End users often request the instrument with the highest sensitivity. This is usually impractical for two reasons. First, only a handful of process chemicals are truly clean enough to justify the highest level of detection. Second, incremental improvements in sensitivity require a significant improvement in signal-to-noise ratio, increasing instrument costs.

Data quality assurance

After carefully considering the factors influencing the proper choice of a particle counter, the next step is the implementation process. Experience has shown that this process is often overlooked. Follow the steps outlined below to be sure that the results from an LPC are stable, accurate and repeatable. These recommendations should be viewed in their entirety. Unless all aspects are considered, the data may be suspect.

  1. Sample point
  2. Filtration
  3. Particle size distribution
  4. Flow rate control
  5. Coincidence loss and saturation
  6. Background scattering
  7. Zero counting
  8. Bubbles
  9. Molecules and colors

Sample point: Thought should be given to the particle monitoring location in the process flow. All components that contact the process chemical, such as valves, tubing, filters and pumps, will contribute particles, sometimes significantly. Only high-quality fluid components should be used where particle levels are monitored. Numerous methods exist for monitoring process chemicals in a manufacturing environment.

For continuous online monitoring, LPCs are typically located downstream of most filtration processes and tend to provide the most stable and repeatable results. Whenever upstream components are changed, the system should be given ample time to clean up before particle monitoring is continued.

Off-line batch sampling requires special attention, as sample containers and handling can cause poor data repeatability and reproducibility. It is important to be able to demonstrate clean and repeatable background counts with DI water before the chemical of interest is sampled.

Filtration: It is always recommended that process chemicals be filtered below the sensitivity limit of the particle counter whenever possible. Because particles slightly smaller than the sensitivity limit can affect background scattering, and potentially coincidence, a good manufacturing practice should include filtration well below the sensitivity limit.

Particle size distribution (PSD): Years of particle counting and customer data analyses show that most ambient particle distributions in continuously filtered liquid systems follow a D-3 size distribution (where D = particle diameter) on total cumulative counts. Figure 4 shows a typical particle distribution in liquid systems. In extremely clean DI water systems, the distribution can be steeper (D-4), and in dirty systems, the distribution can be flatter (D-2). It is equally important that the distribution remain consistent from sample to sample.

When analyzing particle data, it is essential to ensure that this type of particle distribution is present. Deviations from this type of ratio should immediately alert the user to a potential problem.

Click here to enlarge image

Flow rate control: All LPCs have a flow rate specification and are calibrated at a precise flow. This means that real-world particles are intended to pass through the particle counter at the calibrated flow rate. If the proper flow rate is not maintained, both sizing and counting accuracy are compromised. The resident time of a particle, as it traverses the laser beam, has an appreciable effect on the amount of light scattered. If the flow rate is set too high, the electronics do not have sufficient time to fully integrate the signal, compromising sizing accuracy. Counting accuracy is also compromised because the resulting signal is now too small to exceed the particle size threshold. Further complicating the issue, the accompanying software is normalizing the data to the calibrated flow rate. Conversely, if the flow rate is too slow, the transit time is increased, and the particle appears to be larger than its actual size. It is critical that each LPC is controlled to the calibrated flow rate.


Figure 5. Particle Measuring Systems’ “CLS-700T” is an example of a bubble compression system.
Click here to enlarge image

Coincidence loss and saturation: Because they are designed specifically to detect and analyze a single particle at a time, LPCs are often referred to as single particle optical sensors (SPOS). Situations in which more than one particle traverses the laser beam are possible in process chemicals with high levels of particles. In these cases, multiple traversing particles could be counted as a single large particle, or undercounting could occur if these particles are aligned in the beam pathway. This is defined as coincidence loss and varies by instrument type and manufacturer. For example, the specification may be 10 percent coincidence loss at greater than 10,000 particles/mL. This means that 10 percent of the particles are being lost at this concentration. Also, this phenomenon is not linear or predictable enough to make conclusions about the actual total number of counts once the manufacturer’s maximum concentration is reached. This can be seen in the data by looking at differential counts and observing a flattening of the distribution at the smaller size channels. If the concentration becomes too great, the instrument electronics themselves become saturated and can no longer integrate each pulse independently, leading to numerous errors.

Background scattering: Contamination of the capillary walls is the leading cause of excess scattering in LPCs. If the capillary is severely contaminated or scratched, the amount of scatter coming from it can equal the amount of scatter from the smallest detectable particle size. There are typically two ways to recognize whether this is occurring. First, the accompanying software reports the amount of background scatter, and each LPC has a specification for this value. Once the value has been exceeded, the collected data should be considered suspect.

Additionally, this can often be seen in the data. If the amount of scatter from contamination equals the amount of scatter from the smallest detectable particle size, there will be an excessive number of counts in the first size channel. This will shift the PSD further to the left, resulting in a significant departure from the typical D-3 relationship.

Zero counting: A technique known as zero counting is a good troubleshooting tool for process chemical monitoring. DI water is typically much cleaner than process chemicals and can therefore be used to baseline your instrument. Once you have an established baseline, at any point you suspect a problem, you can rerun the zero count test for instrument verification.

Bubbles: Bubbles in liquid systems will scatter light due to the index of refraction contrast between the liquid media and air. Usually, the amount of scatter is significant enough to be recorded as a particle, especially in the larger size channels. Bubbles can come in two forms: naturally occurring bubbles due to dissolved gases and process chemistry, and artificially created bubbles due to poor handling techniques. Issues with naturally occurring bubbles can be exasperated by the action of creating vacuum during syringe sampling. On-line or compression LPCs should be used in these circumstances. Poor handling of viscous chemistries, or chemicals with high levels of surfactants, can generate bubbles. The best method of dealing with this is preventing bubbles from forming in the first place by improving handling practices.

Molecules and colors: Other concerns with chemicals can include very large molecules, fluid color, or miscibility. These phenomena can scatter light, which may further complicate obtaining reasonable results with an LPC, making it that much more important that the appropriate particle counter be used and that the user fully understand the results being generated.

Conclusion

Selecting the appropriate LPC, and generating quality data, is critical for process control. It is also important to match the appropriate LPC sensitivity, type and sampling technique to the chemical of interest. Additionally, users must understand the fundamentals of optical particle counters and the properties of their process chemicals to ensure a valid measurement.

Ross Bryant is the application engineering manager at Particle Measuring Systems. He has more than 10 years of experience in the semiconductor industry related to metrology and materials research. He has a BS in chemical engineering, an MS in materials engineering, and an MBA. He can be contacted at (303) 419-2128 or via e-mail at rbryant@pmeasuring.com.

Roger Carlone is an application engineer at Particle Measuring Systems. He has worked at PMS for more than six years specializing in liquid, part cleanliness testing and slurry applications. He has a BS in chemical engineering from the University of Texas at Austin. He can be contacted at (303) 944-9130 or via e-mail at rcarlone@pmeasuring.com.

References

  1. Nishi, Y., and R. Doering, Handbook of Semiconductor Manufacturing Technology, Marcek Dekker, 2000.
  2. International Technology Roadmap for Semiconductors, www.itrs.net (ITRS 2006).
  3. Carlone, R., “Troubleshooting for Liquid Particle Counting,” Application Note 72, 2006.
  4. Terrell, E., J. Gromala, and D. Beal, “Understanding Liquid Particle Counters,” Particle Measuring Systems, Boulder, CO, 2005.