Infrared Technology for Chamber Clean Endpoint Detection

Chemical Vapor Deposition (CVD) and Atomic Layer Deposition (ALD) processes deposit material on all surfaces in a process chamber. Over time, the thickness of these deposits increases to the point that material begins to delaminate, producing gas-phase particulates that negatively impact process yield. Remote and in situ chemical etching processes are used to periodically remove these deposits from chamber walls, maintaining chamber cleanliness.

It is important that the exact time at which the cleaning process ends (the “endpoint”) be optimized so that only the CVD/ALD deposits are removed, with no etching of the chamber and component surfaces. Over-etching reduces the lifetime of chamber components, increases baseline particulate levels, and significantly increases cost of ownership of the system. In addition, since chamber cleaning processes are energy intensive and use large amounts of potentially toxic gases, the reductions in energy and gas consumption achieved through the exact determination of the cleaning process endpoint can significantly reduce material costs and reduce the carbon footprint of semiconductor fabs. Lowered gas usage also reduces the load on the pollution abatement systems in these facilities.

Chamber Clean

The etching processes used for chamber cleaning employ atomic fluorine, atomic oxygen, or other reactive species that are typically produced in a plasma discharge. These react with CVD/ALD deposits on the chamber components, producing volatile by-products such as SiF₄ that are pumped away. Endpoint times for a cleaning process have historically been determined after an initial optimization and then kept constant in the cleaning recipe. Small changes in process parameters such as the purity and dissociation efficiency of the precursor gas, vacuum pumping characteristics and geometry, and the temperature of the process chamber can affect the efficiency of the cleaning process and therefore the actual time required to reach the endpoint of the clean. The use of fixed process endpoints thus risks the occurrence of over- or under-etch in the cleaning process.

Endpoint detectors that use real-time measurements of the concentration of etch by-products in the system exhaust can provide a precise determination of the cleaning process endpoint. Such detectors monitor, for example, SiF₄ levels to determine exactly when the concentration drops below a defined threshold and the cleaning process can be deemed to be finished. They are also highly beneficial for troubleshooting, even in highly reproducible chamber clean processes, as a suddenly out-of-control process can be extremely difficult and time consuming to troubleshoot without the additional insight provided by an endpoint detector. Data on reactant and by-product concentration profiles throughout the full clean can also provide valuable insights for process optimization.

Different sensing methodologies, including Optical Emission Spectroscopy (OES), plasma impedance measurements, or infrared absorbance measurements can be used for endpoint detection. However, infrared spectroscopy endpoint detectors have advantages in that they offer high accuracy and repeatability, and their measurements are not affected by variations in process and plasma conditions.

Infrared Sensors for Endpoint Detection

Etch process endpoints can be detected using the infrared optical absorbance data for by-products such as SiF₄. Most gases absorb infrared radiation, and the strength of this absorption is proportional to the concentration of the gas and the length of the optical path through the gas. The strength of the absorption (“infrared absorbance”) is calculated as the log of the ratio of the optical intensity determined either 1) with and without the sample gas present or 2) using the intensity of light in a region where the gas absorbs and another region where it is does not absorb. The absorbance is wavelength dependent with spectral characteristics that are unique to each gas. Since the measured absorbance changes with the concentration of the gas, the signal from an infrared detector can be used for endpoint determination. The sensitivity of infrared absorption measurements ensures the detection of the low concentrations that exist under endpoint conditions.

Figure 1. Basic components of an infrared absorbance-based system for endpoint detection. Note that the filter can be placed before or after the sample gas cell.

Figure 1 shows the basic components of an infrared absorbance-based endpoint detector. A broadband infrared light beam created by an IR source traverses a gas cell through which the sample gas is flowing. Its intensity is reduced with the reduction dependent on the concentration of the gas present. The beam goes through a filter that selects those wavelengths specific to the gas being measured after which it impinges on a detector that measures the light intensity. Periodically, a “zero” signal is collected while flowing inert gases such as N₂ or Ar or with the gas cell under high vacuum. This step eliminates instrument effects and hardware drift and ensures that the value determined for the absorbance depends only on the sample gas.

MKS Instruments offers two infrared-based chamber clean endpoint detectors: the Process Sense^™ NDIR End Point Detector and the T-Series IR Gas Analyzer (IGA-T). Both instruments have similar gas cells, IR emitter sources, and IR detectors. However, the Process Sense^™ Detector employs Non-Dispersive Infrared (NDIR) filter optics, while the IGA-T uses Tunable Filter Spectroscopy (TFSTM)-based optics.

Figure 2. Typical installation configuration of the Process Sense Detector (left in the photograph) or IGA-T End Point Detector (right in the photograph).

The Process Sense NDIR End Point Detector is a low-cost SiF₄ or WF₆ detector specifically designed for high volume chamber clean endpoint detection applications. The Process Sense Detector measures the concentration of etch by-products in the chamber exhaust flows with a sensitivity down to mTorr. This low detection limit makes it well-suited for chamber clean applications in silicon oxide, silicon nitride, polysilicon, and other silane or TEOS-based CVD processes.

The Process Sense Detector employs two filters, one that transmits only the bandpass specific to the sample gas and another that provides a reference signal in a region of the IR spectrum where the gas does not absorb. The absorbance is calculated as the log of the ratio of the two signals and is a single number. Figure 2 shows a typical installation configuration, with the sensor mounted in an isolatable bypass. During deposition, the isolation valves are closed, isolating the sensor from process gases. During the chamber cleaning process, the isolation valves are open and etch by-product gases flow through the detector’s sampling cell. When the absorption signal from etch byproduct drops to zero, the chamber clean endpoint is determined.

The T-Series Inline Gas Analyzer (IGA-T) is the next generation of flexible endpoint detectors and has been designed to accommodate multiple possible configurations (filters and detectors) that provide for increased flexibility in selecting the by-product gases to measure as well as enhanced sensitivity.

The IGA-T uses Tunable Filter Spectroscopy that employs a narrow-band Fabry-Perot optical filter to create an infrared absorbance spectrum in the region of interest for the sample gas. The portion of the infrared spectrum sampled is sufficiently wide to capture absorption features for one or several gases, while simultaneously minimizing interferences due to nearby absorptions. One advantage of TFS technology is that it does not need to be re-zero’d as frequently as NDIR sensors. Tunable filter technology is also useful in that it can extract absorbance data for multiple by-products from the measured spectrum. The IGA-T is factory-configured and calibrated for real-time monitoring of one or two by-products such as SiF₄, WF₆, or CO₂. Exhaust line installations are similar to that shown in Figure 2, with the additional requirement of a processing box containing electronics for embedded spectral processing.

Table 1. Comparison of Process Sense Detector and IGA-T Detector.

Both the Process Sense Detector and IGA-T are supplied as complete integration-ready systems with inline or in-process sampling and configurable manifold and port designs that are easily installed and maintained. Table 1 provides a summary of the various features of the Process Sense NDIR End Point Detector and the T-Series Inline Gas Analyzer.

Conclusion

Efficient and complete chamber cleaning processes are critical for the success of CVD/ALD processes. Endpoint detectors that are based on infrared spectroscopy for the measurement of the concentration of the cleaning process by-product gases offer high sensitivity and repeatability and are ideally suited for real-time monitoring of chamber cleaning processes. MKS Instruments’ Process Sense NDIR End Point Detector and IGA-T endpoint detector are easily installed and maintained in a bypass configuration on chamber exhaust lines. These detectors provide accurate and precise information for real-time chamber clean endpoint determinations as well as valuable process insights on the chamber cleaning process that can be used for optimizing process control.

This article was written by Jim Ye, Senior Manager, Product Marketing, Roberto Bosco, Director of Product Marketing, and Sylvie Bosch-Charpenay, Senior Manager of Applications Engineering, MKS Instruments, Inc. For more information, visit here .