System Detects, Classifies Wafer-Edge Defects

Edge defectivity has become a critical device manufacturing problem, yet from many device makers' perspective, there is a lack of inspection capability for the wafer's edge. Several solutions provide excellent results; however, they seem mostly targeted at wafer manufacturers whose prime concern is chips, cracks and other physical damage on the wafer.

The IC manufacturer is interested in production-related issues on the wafer's edge — often quite different from actual silicon damage. The need is for inspection specifically targeted at film-related issues resulting from process integration steps, such as deposition, CMP and etch. These can be delaminated films or residues, possibly defects on a resist film or maybe a metal film, which produce difficult-to-control wafer-edge defect sources. These varied defect sources at the wafer edge pose singular inspection challenges in imaging, defect source identification, and automated classification. Currently, there are two wafer-edge inspection approaches: One is a line scan laser that produces a scatter signal and uses thresholding to pick up signal high points, then returns to these points and takes pictures with a camera. Although proven in wafer manufacturing, it is designed to find scattering defects such as chips, cracks and particles. Film defects and residues tend not to significantly scatter.

The other approach, CCD imaging, operates much like a digital microscope. Its drawback is two-fold. One limitation is that the capability to image and detect defects is dependent on the wafer edge's curvature and shape. At the magnification level necessary to see 2-5 µm defects, the depth of focus (DoF) severely limits the field of view in focus — out-of-focus defects are difficult to identify, complicating analysis. Although not a problem on a flat surface, it can be on a curved edge.

A major CCD-based system limitation for use in process control is the lack of accurate, reliable automatic defect classification (ADC). CCD technology can provide working pictures of a wafer's edge, but so far the lack of robust ADC has prevented production use for process control. Edge defectivity data must be accurate, reliable and quantified to become useful in process monitoring.

KLA-Tencor (San Jose) has introduced an edge inspection system that promises to meet these process-monitoring needs. Its VisEdge system uses a polarized solid-state laser source that focuses to a small (~5 µm) spot on the edge. From there, the scattered and reflected light are collected, as well as phase change and beam deflection (caused by topographic features), using multiple detection sensors.

The individual detection channels capture unique characteristics of the wafer edge and associated defect sources. For example, the reflected (or specular) and phase channels are very sensitive to films and residues and produce clear images of delamination and flakes, while the scattered light detector is most sensitive to particles, chips and cracks, etc. The laser-based imaging system is used to scan the wafer edge and produce a continuous 360° view of it at high resolution. Simultaneous acquisition of the multi-channel images allows comparison of the signals from these sensors, providing the capability to view defects in multiple ways and allowing classification with higher accuracy. The technology offers the combined capabilities of a scanning scatterometer, reflectometer, ellipsometer and profilometer; and offers the advantage of identifying unique defect sources along the wafer's edge.

Typical results can highlight a defect such as edge-film delamination. This can occur where a film may have been deposited over something like a chemical mechanical planarization (CMP) residue. When the wafer goes through subsequent processing, such as a thermal cycle, the residue underneath the film might outgas, causing delamination. Shown in the Figure is the output of all four detectors at a specific wafer region with film “blisters” that have either flaked off (open) or have film over them (closed), which can then be classified using the system's ADC software with very high purity. This is crucial, because the wafer's edge is noisy; there are many film edges, incomplete film removal, as well as considerable variation.

With the application of multi-sensor imaging, it can be seen that specular and phase images are particularly sensitive to wafer-edge film defects. It also provides sufficient information to distinguish between open and closed blisters. (Source: KLA-Tencor)

Once the defect is identified and classified and its location determined, actionable data for process control is generated. This data can drive source identification by correlating wafer-front surface defects to edge sources.

Find more information on inspection, measurement and test.