Defect/Yield Analysis

Moving physical failure analysis from the lab to the fab

Rainer Weiland and Wolfgang Tittes, Infineon Technologies; and Jim Jackman, FEI

A contamination risk assessment has revealed that dual-beam instruments can be safely integrated into production lines, thereby shortening analytical and process development cycles.

Semiconductor fabrication is a highly complex operation that may include hundreds of interactive processes, each of which must function within a tightly defined operational window. When running optimally, these processes can create value at an astonishing rate, often measured in millions of dollars per day. A single out-of-control process, however, can cause economic losses of similar proportions. In high-volume production, the time needed to detect and correct a defect excursion is a primary determinant of a company's economic success. During process development, the time devoted to problem solving may play an even more significant role because it affects time to market for new products and may ultimately control a company's ability to acquire and retain market share.

The collection of techniques and tools that are used to detect defect excursions, diagnose their root sources, and direct corrective action can be referred to as physical failure analysis (PFA). Traditionally, PFA has been performed outside of the fab in a laboratory filled with complex analytical equipment. When a defect excursion cannot be resolved by standard in-fab inspection and review procedures, samples of the defective product wafers are taken from the fab to the lab, from whence they never return. This transfer introduces costly delays into the analysis cycle and requires the sacrifice of an entire wafer to characterize defects whose root cause can usually be determined by analyzing only a few isolated devices.

Driven by the escalating costs of production delays and the increasing value of in-process wafers, some fabs are now moving PFA on-line. Dual-beam instruments that combine a focused ion beam (FIB) system and a scanning electron microscope (SEM) are the key to this transition from traditional laboratory analysis. As semiconductor devices continue to shrink in size and grow in structural complexity, their analysis requires a technique with both high spatial resolution, to determine three-dimensional structural morphology, and the ability to look below the surface, to examine buried defects. Dual-beam instruments provide those capabilities: using high-energy gallium (Ga) ions, the FIB can cut precisely located cross sections, and the SEM can then generate high-resolution images and compositional analyses of the revealed features.

In principle, dual-beam in-line analysis should require the sacrifice of only the cross-sectioned devices, allowing the remainder of the devices on the wafer to continue through the production process. Thus, it has the potential to significantly reduce both analytical cycle times and scrap costs. In practice, however, the acceptance of dual-beam analysis in the fab has been limited by concerns about the contamination of neighboring die that might occur during the FIB milling process. After a brief review of the benefits of dual-beam systems, this article describes a contamination risk assessment study carried out by Infineon Technologies (Dresden, Germany). It was found that, with proper procedures, GA contamination can be confined within the analyzed die, and the rest of the wafer can be processed successfully to create fully functional devices.

Instrument Capabilities

Dual-beam analysis combines FIB sample preparation and high-resolution SEM imaging and analysis capabilities in a single tool. SEMs are already widely accepted in the semiconductor fab; it is the combination of their capabilities with a FIB system's ability to add and remove material from a wafer that gives dual-beam instruments their power, allowing analysts to extract detailed three-dimensional information from cross-sectional views of designated features.

Dual-beam instruments offer a number of other benefits for in-line analysis applications:

• In addition to cutting cross sections, the FIB can enhance the revealed surface to bring out subtle detail. Such enhancement involves the introduction of minute quantities of specialized gases that preferentially etch specific types of materials. This process creates surface topography and, thereby, SEM contrast at interfaces between materials that etch at different rates. Ion-assisted gaseous etching may be used to reveal detail not visible in "native" cross sections. In Figure 1, for example, the right-hand side of the image has been etched with an oxide-specific gas to reveal layers not visible in the untreated left-hand side.

Figure 1: Ion-assisted gaseous etching may be used to reveal detail not visible in "native" cross sections. In this example, the right-hand side of the image has been etched with an oxide-specific gas to reveal layers not visible in the untreated left-hand side.

• The material deposition capability of a FIB system also plays an important role in in-line PFA by allowing analysts to refill the pits created during the cross-sectioning process with insulating material. Left unfilled, such pits pose a contamination risk by serving as collection points for extraneous material—particularly during chemical-mechanical polishing.
• When combined with a high-accuracy stage, the scanning electron beam in a dual-beam instrument provides the navigation capability that is essential to analyzing defects on product wafers. Using this beam for navigation takes advantage of its high resolution and reduces contamination risks by limiting the use of the FIB to only the targeted die.
• Although beyond the scope of this article, another, potentially very powerful, option offered by in-line dual-beam systems is sample extraction. Dual-beam analysis alone cannot always provide a complete solution to a defect excursion problem. Some cases may require the use of other, laboratory-based techniques such as Auger electron spectroscopy, secondary ion mass spectroscopy (SIMS), transmission electron microscopy, or atomic force microscopy. In these cases, a dual-beam instrument can be used to excise a small sample from a die for laboratory analysis, permitting the sampled wafer to remain in the production line (see Figure 2).

Figure 2: Micrograph showing the dual-beam instrument's sample-extraction capability. An analyst can excise a small sample from a die for off-line laboratory analysis and return the wafer to the production line.

Gallium Contamination Risk Assessment

The electron beam of a dual-beam instrument is known to be noncontaminating, and FEI (Hillsboro, OR) has shown that the general particulate contribution of such systems is <0.5 particles >0.2 µm per wafer pass, with exceedingly low levels of backside contamination. Thus, the primary contamination risks associated with in-line dual-beam analysis are linked to the FIB. The contamination assessment conducted by Infineon focused on two sources of such contamination: Ga ions from the beam, and metal atoms or ions sputtered from elements within the FIB column and subsequently deposited on or implanted into the wafer surface.

Gallium Contamination. Ga acts as a dopant, but because its energy level is above the valence band of silicon, it does not represent a recombination center for electron/hole pairs. As the first step in the study, researchers performed extensive electrolytical metal (Elymat) analyses, which showed that there was no significant change in carrier diffusion length after FIB processing. Based on those measurements, a maximum allowable Ga concentration of ~10¹³ at/cm², an order of magnitude below the Si substrate doping level, was established.

Once this criterion for maximum Ga had been set, three areas at risk for contamination were evaluated: the parts of an analyzed wafer distant from the analysis site, wafers that had been adjacent to analyzed wafers during thermal or wet processes, and production equipment. No contamination was found on wet benches or furnaces, and the only Ga contamination found on parallel-processed wafers was ~10¹² at/cm² on the backside of a wafer that had been adjacent to the analyzed wafer in the furnace. Therefore, the remainder of the study concentrated on the spread of Ga contamination from the analysis site.

Gallium Diffusion. SIMS depth profiling was used to characterize the vertical diffusion properties of Ga in various materials. Figure 3 shows vertical diffusion into bulk silicon as determined by SIMS for different implanted doses. The penetration depth is defined as the distance from the surface at which the Ga concentration reaches 10¹⁵ at/cm². Similar results were observed for oxide and nitride layers, and there was no evidence of interface diffusion or segregation. Based on the assumption that lateral Ga diffusion is of the same order of magnitude as vertical diffusion, it was concluded that neither phenomenon will cause problems on wafer areas beyond the analyzed die.

Figure 3: Vertical diffusion into bulk silicon as determined by SIMS for different implanted doses. The penetration depth is defined as the distance from the surface at which the Ga concentration reaches 1015 at/cm2.

Gallium Spreading from FIB-Deposited Refill Material. Although it is beneficial to use a FIB to refill the pits left by cross-sectioning to prevent their becoming collection points for extraneous material during subsequent processing, the the refill material (TEOS) will consist of up to 30% Ga because of the deposition process and will itself become a source of Ga contamination during subsequent thermal processes. Figure 4 offers dramatic evidence of Ga spreading from the refill material. The next part of the study evaluated this worst-case scenario of Ga spreading as well as the possibility of using a protective SiO₂ layer to contain the risk. Time-of-flight (TOF)-SIMS was used to measure the lateral distribution of Ga after each step in a typical process and SIMS was used to create depth profiles. The seven-step test process consisted of:

1. Depositing a 20-nm SiO₂ protective layer on the Si substrate.

2. Milling a 40 x 60 x 6-µm crater.

3. Refilling the crater with FIB-deposited SiO₂.

4. Cleaning the wafer with alkaline H₂O₂.

5. Annealing the wafer at 1050°C for 30 minutes.

6. Cleaning the wafer using a hydrofluoric acid dip to remove the sacrificial oxide layer.

7. Creating a 33-nm SiO₂ layer through thermal oxidation at 1000°C for 20 minutes.

Figure 4: Lateral Ga surface contamination determined by TOF-SIMS mapping: (a) after refilling a FIB crater and cleaning the wafer surface, and (b) after tempering the refilled wafer at 1050°C for 30 minutes. Comparing the two figures reveals the potential severity of the contamination risk.

The lateral distribution results are presented in Figure 5. After the refilling step and both cleaning steps, the Ga was tightly confined to the immediate vicinity of the crater. After each thermal process, it was more widely distributed, but it dropped below the critical level at distances within 1 mm of the crater. Figure 6 shows the SIMS depth profile taken 0.5 mm from the crater after the annealing step. While the profile shows desorbed Ga diffusing into the surrounding region, its concentration drops below detectable levels within 20 nm of the surface. These data confirm the efficacy of the SiO₂ protective layer.

Figure 5: TOF-SIMS measurements showing both Ga concentrations across a wafer following the refilling of a milled crater and the spread of Ga contamination from the FIB-deposited refill material after each of four subsequent process steps.

Figure 6: SIMS depth profile taken 0.5 mm from a filled crater showing Ga diffused in a 20-nm layer of SiO2 following an annealing step at 1050°C for 30 minutes.

Sputtering from the FIB Column. As Ga ions are being accelerated within a FIB column, they have the potential to sputter atoms and ions from column components, particularly the liner and apertures. These contaminants can in turn be accelerated through the final aperture onto the sample. To investigate this contamination risk, the surface of a processed wafer was analyzed for a number of metals using vapor-phase decomposition-atomic absorption spectroscopy (VPD-AAS) and vapor-phase decomposition-total reflection x-ray fluorescence (VPD-TXRF). Localized concentrations in a region centered on the crater were also determined using TOF-SIMS with a spot size of approximately 50 µm. The results are presented in Table I. In all cases, the VPD values were well below the critical level of 10¹² at/cm². Even the TOF-SIMS values, which would be more sensitive to locally higher concentrations in the vicinity of the crater, were within the 10¹² at/cm² range. Therefore, it was concluded that sputtering and redeposition of material from the ion column was not a significant contamination risk.

Conclusion

Dual-beam systems can enable semiconductor fabs to integrate physical failure analysis into the production flow. A study conducted by Infineon showed that, with proper procedures, it is possible to manage the risks of Ga contamination presented by such systems' FIB. During the testing, the researchers were able to confine Ga to within 1 mm of a FIB-milled and -refilled crater even after multiple thermal processes, and were unable to detect significant levels of any other heavy-metal contamination. Subsequent use of the dual-beam instrument in the fab has confirmed the study's conclusion that it is possible to perform PFA on product wafers without sacrificing any part of the wafer except the analyzed die.

In high-volume production, bringing PFA from the lab into the fab will shorten the analytical cycle, providing faster feedback for corrective action and contributing directly to improved yield. It will also eliminate the need to sacrifice entire wafers for analysis, greatly reducing the costs incurred by wafer losses. Both of these benefits will become increasingly valuable as new processes and paradigms such as 300-mm wafers increase production capacity and the value of wafers in process.

In process development, these same benefits will accelerate the development cycle, enabling manufacturers to bring new products to market sooner and making it easier for them to acquire and retain market share. As was the case for defect reduction and defect review, PFA will be drawn inexorably into the fab by the fundamental economic forces that drive the semiconductor industry.

Acknowledgments

This article is based on a paper that was presented at the 26th International Symposium for Testing and Failure Analysis held November 12–16, 2000, in Bellevue, WA, and was originally published in the symposium's proceedings. Part of the work described herein was supported by the European Community's European Fund for Regional Development and by funding from the German state of Saxony.

The authors would like to thank Helmuth Murrmann, Harald Eggers, Gerhard Rauter, and Johann Harter for making the contamination assessment project possible and for participating in fruitful discussions throughout its duration. They also would like to thank Christian Boit for his initiative throughout the project. Special thanks also are due to Hans Rettenmaier, Alexander Hirsch, Mr. Pomes, and Bernd Ebersberger for their efforts in running many experiments on the dual-beam tools. Finally, the authors wish to thank Gunter Frey of the Fraunhofer Gesellschaft in Erlangen, Germany, for the SIMS and Elymat measurements, and Rolf Treichler from Siemens for performing the TOF-SIMS analyses.

Rainer Weiland, PhD, is senior manager of the physical failure analysis department at Infineon Technologies in Regensburg, Germany. He joined Siemens in 1986 and assumed responsibilities in its semiconductor division in 1995. Weiland received a diploma and PhD in chemistry from the University of Kaiserslautern, Germany, in 1977 and 1983, respectively. His doctoral thesis was in the field of optical spectroscopy of oriented molecules. (Weiland can be reached at +49 89 23425464 or rainer.weiland@infineon.com.)

Wolfgang Tittes, PhD, is senior manager of the physical failure analysis department at Infineon Technologies in Dresden, Germany. In 1995 he joined the Siemens Microelectronics Center in Dresden, where he worked in the defect engineering group. Tittes received a diploma in physics from the University of Stuttgart and the Max-Planck-Institut for Materials Science, Institute for Physics, in Stuttgart, Germany, in 1989. He received a PhD in chemistry from the University of Dortmund and the Institute for Spectrochemistry and Applied Spectroscopy in Dortmund, Germany, in 1994. His thesis was in the field of elemental mass spectroscopy. (Tittes can be reached at wolfgang.tittes@infineon.com.)

Jim Jackman is director of business development in the Structural Diagnostics Group at FEI (Hillsboro, OR). In that role he is responsible for discovering and developing new business opportunities, bringing 300-mm tools to market, implementing full automation for in-fab applications, and managing the application development engineers who provide user support. Jackman was associated with Philips Electron Optics when it was the prime contractor for a series of government-funded R&D projects in Europe. He joined FEI following the merger of the two companies. He received an MS in communication engineering from the Royal Melbourne Institute of Technology in Melbourne, Australia. (Jackman can be reached at 503/640-7562 or jjj@feico.com.)