Characterizing defects in processed wafers at semiconductor fabs is a key step toward identifying and eliminating their sources. In the submicron or micron range, the FIB technique is an accepted way to expose and cross section defects for direct observation and analysis and to prepare the defects for characterization by SEM, TEM and other techniques. The most common method of attack is to cross-section defects with FIB and image them with FE SEM (see sidebar for Glossary of Terms).
One of the most popular instruments in labs that support the semiconductor fab is the combined FIB/FE SEM workstation. Integrating the two techniques in one system speeds up and improves the overall process of cross sectioning and observation/characterization. And in today’s megafabs, saving time to characterize a defect that has the production line on hold is financially significant.
The DualBeamTM 820 from FEI Co. in our lab at AMD combines fully functional FIB micromachining and microdeposition capabilities with a high-resolution FE SEM/EDS system without compromising the capabilities of either technique.
Defect Characterization Process
Our laboratory supports the fab by figuring out the nature and source of defects that have been found through defect inspection or electrical test to affect device yield–also known as “killer” defects.
In a typical defect review sequence, an inspection system scans a wafer for defects and discovers the location, size and coordinates. A human operator, sometimes using a SEM, classifies the defects and tags the original defect file, attaching an image of the defect and verifying its size, location and chemistry.
Engineers then check for correlations. If a systematic problem arises, wafers are pulled for more detailed characterization by cross sectioning in a FIB/FE SEM workstation, then characterized and analyzed by SEM and EDS. Process, test, material and yield engineers use the characterization data to determine the source of defects.
Process-design geometries are rapidly shrinking into the deep submicron range, demanding a need for state-of-the-art analytical systems. Defect monitoring tools detect particle defects less than 0.1 µm. However, because inspections are only done at certain steps of the process, some killer defects may go undetected until electrical tests are performed.
It is our job to work with fab and integration engineers as equal members of the defect-reduction team to provide detailed information, as quickly as possible, on the physical and chemical nature of the defect. We look at many kinds of defects:
Glass flakes.
Metal bridging.
Misprocess (missed steps or a bad recipe).
Hot spots (can be caused by high resistance from an open or from foreign matter).
Metal voids.
Various kinds of particles.
SEM and TEM micrographs reveal voids, shorts and wafer misprocessing, provided the sample is cross-sectioned in the correct location. High-resolution SEM images are extremely helpful in diagnosing defects in process layers.
Intensity differences (gray-scale contrast) in layers and at interfaces can be related to composition and structure and can offer clues on how to proceed with further analysis. The FIB/FE SEM workstation is particularly useful here because a series of high-quality images can be made during the cross-sectioning process to provide a 3-D characterization of the defect.
IC yield is rarely degraded by human-born contamination. Fabs today run under ultra clean-room conditions using robotics, minimizing unnecessary human handling. As a result, most defects are generated by either misprocessing or equipment contamination. For example, if EDS identifies a particle as iron or chromium, it is usually the result of chamber wear or abrasion.
In addition to the FIB, SEM and EDS, the most important tool we have is a high level of interaction and information transfer with the defect-reduction teams. The days of sending off a sample and waiting for an answer are over.
Today, we are in close and constant contact with the fab, and we need to know as much about the sample process history as possible. For example, was a key process gas bottle changed right before the problem started, did the problem crop up right after routine maintenance, or was a new valve seat just installed?
Armed with these clues and information from the FIB, SEM and EDS, we can help the fab identify most of the defects and particles that kill yield. Sometimes we also turn to other materials-characterization techniques, such as AES and SIMS.
Some problems are solved in a few hours–and most within one day. However, problems that involve unknown complex interactions may require longer. Anything we can do to decrease the time required to identify the defect source is important.
Improving the Process
A stand-alone FIB system also offers sample imaging using ion-induced secondary electron images which are adequate for monitoring the cross-sectioning process and useful for understanding the nature of the defect or particle. These images sometimes show different contrast than electron-induced SEM images, especially between metals and dielectrics. This contrast complements normal SEM imaging.
Imaging in a stand-alone FIB workstation, however, requires that the sample be tilted and the ion beam directly bombard the surface to cause the emission of secondary electrons. This process not only takes time and requires sample movement, it also leads to the implantation of gallium from the ion beam and alteration of the face of the cross section or the surfaces of the TEM thin section.
During FIB cross sectioning and sample thinning, the ion beam is essentially parallel to the face of the sample and implantation and alteration problems do not occur. By using the SEM portion of the FIB/FE SEM workstation to observe the cross-sectioning process in real time, the negative effects of observing with the ion-beam are avoided. It is especially important not to alter either the composition or morphology of thin sections prepared for TEM observation.
Speeding Up the Process
Combining FIB with FE SEM also provides answers faster and makes running samples less tedious. Turnaround time for a typical analysis is decreased by 15% to 30% over using separate FIB and FE SEM systems. Eliminated are the time-consuming cycles of sample loading and unloading and feature location, plus sample tilting for imaging and cross sectioning in separate FIB and FE SEM systems.
Although the FIB images are more surface sensitive than SEM images and the information is complementary, the ultimate resolution of the FE SEM is higher than FIB images. In addition, the FE SEM beam is nondestructive relative to the FIB and generates X-rays for elemental analysis.
Defect Review FIB/SEM Tool
With its 8-inch wafer capability and extremely accurate stage navigation, our system is highly effective as a defect-review SEM. When it is coupled with software that translates defect coordinate files from inspection systems, an operator can routinely drive the stage to any defect for SEM imaging and analysis.
Any defect requiring further information through cross sectioning and/or EDS is done in-situ. With the stage accuracy of eucentric tilt and compucentric rotation, in which the feature of interest is kept at the same position, defects are viewed from the best possible perspective.
Locating the Defects and Particles
Driving the FIB/FE SEM stage to a particular submicron defect on a wafer is akin to finding the proverbial needle in the haystack. The capability to transfer defect coordinate files from the various defect inspection systems and navigate the stage with extreme accuracy is imperative for any state-of-the-art analytical tool.
The FIB operator is guided to the location of a particular defect by several techniques. In our laboratory, we use coordinate transfer via both the Kinemate software from Kinetek and CAD Navigation by Merlin’s FrameworkTM from Knights Technology.
The methods of navigating to particular defects in stand-alone and combined workstations include:
Coordinate Transfer–transferring (via a network or floppy disk) the coordinates of a feature, particle or defect from inspection tools such as a defect inspection system or an E-beam prober.
CAD Navigation Software–accessing the IC design data base and navigating to any point on the circuit layout, the schematic, the netlist or the cell hierarchy with the sample automatically performing the same moves with the combined system.
CAD Overlay–registering the die CAD map to the FIB image and overlaying selected circuit details directly onto the FIB image for enhanced navigation to buried features.
Defect Overlay on CAD Layout–displaying defect-size (X-Y values) and coordinates on a CAD die layout, which is useful in identifying killer defects.
Image Registration–registering a set of optical microscope images focused at different levels beneath the surface to the FIB image of the surface to direct the FIB beam to points beneath the surface, even on highly planarized samples. Similarly, the image-registration software can relate the FIB surface image to other microscopy techniques, such as a SEM or emission microscope image.
Two Killer Examples
To understand the benefits of a FIB/FE SEM workstation, let’s look at two typical cases of tracking down yield killer defects. In each case, the problems were uncovered via defect review and electrical test using the capabilities of the combined system.
A Poly-Silicon Defect
Weekly defect monitor wafers were routinely run through the fab to ensure proper equipment operation and to verify that the various process steps, such as lithography, deposition and etch, were working together. After inspection, a defect wafer map and its corresponding SEM file of the defect coordinates were provided to the materials analyst (Figure 1).
The data was downloaded in the FIB/FE SEM system for inspection. SEM imaging of a FIB cross-sectioned defect area revealed the patterned poly silicon and the large nodule defect itself (white circular ring) embedded in the Poly-Silicon 2 layer. Further FIB polishing into the defect showed the large nodule nucleated around a much smaller oxide core, which was subsequently identified as silicon oxide by EDS.
The analysis from start to finish took less than one hour. Based on the results, the defect reduction team traced down sources of oxide particles at or near the poly-deposition step.
Metal Defects
Defect monitoring inspections include several process steps. Each wafer is probed for electrical shorts and opens, and then classified using an electrical yield wafer map (Figure 3).
In this example, several dice in a wafer lot failed during a signature test (Bin 18) around the periphery of the memory array. Because inspection did not include this area, the defects were not detected until final electrical test.
After identifying the defect location area optically, a wafer was submitted to the analysis group for further investigation. A top-view SEM image exhibited an innocuous looking bump in the passivation over the defect area. A FIB cross section of the same area revealed an incomplete etch of the Metal 2 line.
Elemental analysis by EDS determined the defect contained fluorine and oxygen. Fluorine is normally associated with a chamber-cleaning step.
Based on this evidence, it was suspected a fluorine-enriched particle from a chamber wall deposited on the wafer sometime after metal deposition. During the metal-etch process, the defect particle effectively became a metal etch block, creating a metal short.
Acknowledgment
We thank Janet Teshima of FEI’s Applications Laboratory for her assistance with the development of applications on the DualBeam Workstation.
About the Authors
Eugene Delenia, the Senior Materials Engineer at AMD, has been affiliated with the company since 1987. He earned a B.S. degree in materials engineering at San Jose State University.
Bryan Tracy is a member of the Technical Staff at AMD. He holds a B.S. degree in metallurgical and welding engineering from California Polytechnic State University, an M.S. degree in materials science from the University of California and a Ph.D. degree in materials engineering from Rensselaer Polytechnic Institute.
Homi Fatemi, who joined AMD in 1983, is Manager of the Materials Technology Development Department in the Integrated Technology Division. Previously, he was affiliated with Intel and Fairchild, and was an adjunct professor in the School of Engineering and the Leavey School of Business at Santa Clara University. Dr. Fatemi has a Ph.D. degree in materials science and engineering from Stanford University and a M.B.A. degree from Santa Clara University.
Advanced Micro Devices, ITD Materials Technology Development Laboratory, Mail Stop 32, P.O. Box 3453, Sunnyvale, CA 94088-3453, (408) 732-2400.
Glossary of Terms
FIB–Focused ion beam system with precision microcross-sectioning, milling and device-modification capabilities.
FE SEM–Field emission scanning electron microscopy with high-resolution (15Å) imaging.
TEM–Transmission electron microscopy providing information about the atomic structure and defects present in a solid material.
EDS–Energy-dispersive x-ray spectroscopy for elemental analysis of particles and features, integrated into an SEM.
AES–Auger electron spectroscopy providing high spatial resolution elemental surface analysis.
SIMS–Secondary ion mass spectrometry providing ultra-high sensitivity elemental surface analysis and depth profiling.
Copyright 1995 Nelson Publishing Inc.
October 1995