Problems with Silicon Wafer Stress? Raman Can Help Identify and Prevent Processing Defects

Silicon Wafer Manufacturing Stress and Defects

Problems with Silicon Wafer Stress? Raman Can Help Identify and Prevent Processing Defects

Post by:
  • 2:40AM Feb 21, 2014
  • Comments off

As a silicon wafer manufacturer, you are faced with pressure to increase yield and speed up the manufacturing process—all while maintaining your product’s quality. Altering even minor processing parameters can generate stress on a wafer’s crystal lattice leading to an increased number of damaged or defective devices on a given wafer. Because most strain patterns are invisible to the naked eye, it is difficult and costly to pinpoint how and where they have developed.

As you have likely experienced, many steps within the silicon wafer manufacturing process have the potential to generate stress in the form of interlayer tension, compression, or shear. In general, strain can be introduced by deposition rates, layer thickness variations, temperature profiles and history, and many other variables intended to increase throughput.  Any time a new layer is added during metallization, passivation, and other deposition processes, there will be a compositional mismatch that imparts some degree of strain to the substrate. Because strain is a common issue during the manufacturing process, it should quickly be assessed to differentiate between defective wafers and those with an acceptable degree of strain that will not adversely affect final device operation.

When analyzing silicon wafers, you have likely relied on transmission electron microscopy (TEM) with selected area electron diffraction (SAED) or x-ray diffraction (XRD) techniques to look for irregularities in the silicon crystal structure. These techniques are generally expensive, time-consuming and/or destructive to the sample.  They also do not show relationships between the spatial distribution of the strain, its intensity, and what features the strain is associated with (such as strain associated with metallization deposits). The most comprehensive tool for determining the extent of strain is Raman spectroscopy with enhanced laser capabilities.

Brightfield image of an indent with a Raman border map
Figure 1: Brightfield image of an indent.

Raman spectroscopy was developed in 1928 and has primarily been used for research. The technique, however, has recently come a long way in terms of analytical power and adoption within the scientific community. Raman works particularly well with silicon wafers, semiconductors, and photovoltaics. During a Raman analysis, light strikes the surface of the silicon sample and interacts with its vibrating molecules, causing the light to shift to different wavelengths. These wavelength shifts are analyzed to determine the type of strain. For example, an increase in Raman frequency typically indicates compression while a decrease typically indicates tensile strain. By using multiple laser types, Raman maps can be created within hours to help visualize how deep the strain has occurred in the sample (Figure 1-4) providing invaluable feedback about the severity of damage and likelihood of failure.

Raman Maps Laser Depth Penetration
Figure 2-4: Using a Raman equipped with 3 different excitation wavelengths (surface 473 nm, mid-range – 633 nm, and deep – 785 nm) allows you to probe the entire sample to identify the location and extent of the strain. Not all Raman devices are equipped with all 3 different excitation wavelengths. Figure 2 shows the most intense strain toward the surface and the center of the silicon wafer, becoming less intense as laser depth increases (Figure 4).

When faced with quality issues, defects, or failure analysis of silicon wafers, Raman spectroscopy should be one of the first techniques applied to identify and characterize the strain and to classify its severity. A proactive measure that can help reduce investigation time is to develop a database of historical issues and their associated processing parameters. This information aids in narrowing down which specific processes need to be monitored with routine Raman analysis to ensure high quality wafers. A second layer of investigation will provide even more information about the strain. Additional techniques such as X-ray photoelectron spectroscopy (XPS) or scanning electron microscopy with energy-dispersive spectroscopy (SEM-EDS) complement a Raman analysis by probing further into the sample and evaluating the dimensional and compositional characteristics of device layout and detected residues and contaminates.

Now that you have been introduced to some of the key features of Raman that aid in investigations of silicon wafer stress, it is time to pinpoint the areas for improvement in your manufacturing processes. What previous data do you have about wafer stress or failures that can become part of your knowledge database? How can you regulate processes to ensure consistent results? As demands for high-speed manufacturing of lighter and thinner wafers increase, so too must attention to detail and quality diligence. Taking a proactive approach to wafer quality by incorporating Raman spectroscopy will save time, money, and product—giving you the competitive advantage.

Thank you to Mark Sparrow and Elizabeth Wolff for their contributions to the article.