AVS1996 Session MS+AS-ThM: Analysis and Control of Defects and Contamination

Thursday, October 17, 1996 8:20 AM in Room 201A

Thursday Morning

Start	Invited?	Item
8:20 AM	Invited	MS+AS-ThM-1 Microcontamination Analysis in the Semiconductor Manufacturing Environment P. Gupta (Intel Corporation); S. Tan (Chemtrace Corporation); Z. Pourmotamed, C. Flores, R. McDonald (Intel Corporation) Several analytical methods have been established for the detection of microcontamination on silicon surfaces and in process liquids in the semiconductor manufacturing environment. However, it is becoming increasingly difficult for the current methods of microcontamination analysis to meet today's need for sensitivity and throughput. To meet these newer challenges, several novel methods for trace elemental (both metallic and nonmetallic) analysis in chemicals and surfaces are being introduced. For example, the use of a magnetic sector ICP-MS as an alternative to the more commonly used quadrupole ICP-MS for the measurement of trace metals in corrosive chemicals offers both sensitivity and higher productivity. Furthermore, this technique allows the analysis of phosphorus, sulfur, chlorine at the ppb detection limits in several chemicals and DI water. Previously, the analysis of these nonmetallic species was limited to DI water, using Ion Chromatographic techniques. To meet the fast turn around time of results that's required by high volume semiconductor manufacturing, techniques have been developed that allow trace metal measurement at the ppb level at the manufacturing line. We will be presenting the first ever results, using a technique known as Dried Residue XRF, that show a detection limits of 1 to 5 ppb of transition metals in semiconductor processing chemical such as Hydrofluoric acid and SC2 solutions. These measurements, obtained in less than 5 minutes, by relatively unskilled personnel, show promise in meeting the trace metal analysis demand in a high volume manufacturing fab.
9:00 AM		MS+AS-ThM-3 Application of Spectral Analysis on Atomic Force Microscopy (AFM) Images: A Study on Si (100) Surfaces S. Fang, W. Chen, S. Halepete, C. Helms (Stanford University) In practice, the standard parameters specified for microroughness from AFM images are peak-to-valley and root-mean-square (RMS) roughness. Although both of them are simple and reliable, they only give information along vertical direction and hence can not fully characterize the surface. Surface roughness can have many effects on device properties depending on its spatial wavelength. Using simple statistical descriptions, such as RMS values, will cause ambiguity on surface characterization. This includes atomic level roughness which can affect gate oxide reliability, all the way up to longer wavelength features which lead to "haze". Currently, correlations between the roughness within a particular spectral range and degradation in either GOI defects or reliability are still not well-established. Although spectral analysis has recently been employed to analyze AFM images, a standard method to convert AFM images to power spectral density (PSD) is not available. For example, dimensions of PSD are denoted as either (length\super 4\) or (length\super 3\). In this paper, we will first review the methods of spectral analysis commonly used by various authors. Several examples will then be given to illustrate the advantages of spectral analysis on roughness measurements. For example, we have applied spectral method to analyze AFM data in terms of identification and correction of AFM artifacts in both real and frequency domain. Both sharp and dull dips are used to study the tip effect on AFM data. AFM measurement and data analysis protocol will be discussed based on our results. Such AFM data analysis method is thereafter applied to investigate several Si surfaces with different surface treatments, including Si/SiO\sub 2\ interfaces to examine oxidation effect on Si surface roughness, as well as intentionally roughened Si surfaces to study their correlation by analyzing relevant PSD and normalized PSD curves.
9:40 AM	Invited	MS+AS-ThM-5 New Models and Methods for Analysis of Particles and Defects A. Testoni (Digital Equipment Corporation) As the minimum feature size on silicon wafers in semiconductor products continues to decrease, so does the maximum size of particles, residues or defects that can be tolerated during wafer processing. With semiconductor manufacturing facilities producing devices with minimum geometries approaching 0.305m and advanced development fabs producing structures smaller than 0.255m, the tolerable size of these `killer' defects is less than 0.105m. The location and composition of these small defects is critical information for identifying and eliminating `killers' and therefore increasing yield. This information must be obtained quickly in order to have significant impact on process development and manufacturing improvement work. This is a difficult analysis challenge. This talk will survey the analytical methods available for defect composition identification and show their strengths and weaknesses through real world applications. The talk will then discuss how we have used modeling and two-dimensional simulation of EDXS/WDXS and AES, using our SESAME software, to enhance the analytical ability of these techniques and enable the identification of small features on complex substrates.
10:20 AM		MS+AS-ThM-7 Particle Evaluation in the Tungsten Etch Back Process Y. Uritsky, L. Chen (Applied Materials, Inc.); C. Brundle (C.R. Brundle and Associates); S. Zhang, S. Wilson, A. Mak (Applied Materials, Inc.) The Tungsten Etch Back (WEB) process yield can be severely affected by particle generation caused by etching chemistry reactions. Fluorine radicals from the SF\sub 6\ plasma attack chamber hardware directly, but also produce Ti-based by-products, trough reaction with the TiN adhesive layer, which can then interact with chamber hardware and produce particles. Particle reduction program has been developed focused on particle identification and therefore origin determination and elimination. Using a Particle Analyzer System (PAS) comprising SEM/EDS and also a Scanning Auger Microprobe (SAM), we determined that most of the particles detected on bare silicon test wafers were Ti-Al based, indicating that, during etching of device wafers, Ti-based by-products had reacted with the Al-based chamber hardware, subsequently generating particles. Intermittent particle problems classified as carbon-based and sulfur-based were also observed. Finally, on test device wafers, after completion of WEB process, all the particles discussed above were also located; in addition, ring shaped W-based particles were detected. Based on these results, chamber hardware was modified, including the introduction of the Reactor Laminar Flow (RLF) concept, and the WEB process parameters were adjusted. Following this a WEB marathon showed average defect density below 0.05/sm\sub 2\, a very significant improvement, demonstrating the importance of the particle identification procedure in WEB process defect reduction.
10:40 AM		MS+AS-ThM-8 Practical Limits of EDS in Semiconductor Thin Film Analysis P. King, E. Adem (Advanced Micro Devices) Energy dispersive spectroscopy (EDS) is used extensively in defect review and thin film characterization applications common to semiconductor processing. When integrated with the imaging capabilities of scanning electron microscopy (SEM/EDS), its relative ease of operation and rapid analysis often make it the first tool selected when the need arises to analyze a defect or identify an unknown film. EDS is considered a bulk technique and has predictable limitations with respect to the minimum film thickness which must be present if a practical, timely analysis is to be carried out. This study describes the relationships between operating parameters and material properties which affect EDS of extremely thin films. Signal to background ratios of film/substrate systems are measured for increasingly thin films. Examples are provided from commonly encountered, well understood material systems (SiO2 on Si, TiN on Al, TiN on oxide, etc.) and from an unknown fluorine-rich film on an aluminum bond pad. Arguments are presented regarding the point at which the hand-off to more surface sensitive techniques such as Auger electron spectroscopy is appropriate.
11:00 AM		MS+AS-ThM-9 Defect-induced Redeposition in Target Sputtering C. Lo, D. Draper (Materials Research Corporation) Observations on target sputtering found that redeposition was aggravated when structural defects were present. In this study, formation of redeposited nodules on the sputtered targets, including W-Ti, Cr and W-silicide, were investigated. Microcracks, voids and crevices existing on the sputter target surface tended to act as the preferred nucleation sites. A higher rate of redeposition than erosion was also detected adjacent to the defect regions, resulting in a significant growth of the redeposited nodules. It is presumed that a localized electromagnetic field, induced by the structural discontinuity, may force the bombarded species to be trapped at these regions.
11:20 AM		MS+AS-ThM-10 Imaging ESCA: Can You Analyze What You Can't See? J. Hammond, J. Moulder (Physical Electronics, Inc.) For ESCA analyses of small features in failure analysis applications, the analyst is typically faced with samples that are covered with insulating films of carbon, oxidized surface species, and unwanted organic residue. For these samples, the precise alignment and multiple analyses of closely spaced regions demonstrating good and bad product performance is required. For the required analyses, the use of scanning X-ray induced secondary electron images, scanning electron beam produced backscattered electron or Scanning Auger images, or in-situ optical microscope alignment can produce ambiguous or erroneous definition of the surface contaminated areas of analysis. By using an off- line, high resolution optical microscope and digital stage coupled to an automated, high precision, in-vacuum stage and sample introduction system; accurate and precise definitions of multiple surface and sub-surface analyses areas can be obtained. Statistical data will be discussed showing that the standard deviation of sample alignment is less than 10 microns. ESCA analyses from multiple contaminated bond pads will be discussed. ESCA Zalar rotation depth profiles on and off the contaminated bond pads will illustrate the analysis of buried interfaces, the position of which can not be defined by traditional in-situ alignment techniques.
11:40 AM		MS+AS-ThM-11 Secondary Ion Mass Spectrometry of a Copper Polyimide Thin Film Packaging Technology C. Parks (IBM Analytical Services) In its quest for better conduction and higher switching speeds, the computer chip industry has a mid-term goal of incorporating wiring layers of copper and dielectric layers of polymers in high-performance microchips. In microelectronic packaging, IBM currently manufactures a thin film technology which uses chromium-cladded copper and polyimide dielectric. This paper explores Secondary Ion Mass Spectrometry (SIMS) capabilities for analyzing copper-polymer technologies.The paper reviews how the strengths of SIMS (ppm sensitivity and ability to readily profile a 20 um thick polymer stack) can be exploited to address processing and materials issues. Two examples are given in which SIMS is used to test for the presence of, and to quantify, a buried monolayer-coverage adhesion promoter. The presence of trace metals in these samples raises a more general issue of polymer to wiring layer interactions. A sub-topic of polymer/metal interactions, gettering of chlorine impurities, is explored in detail. A chlorine ion implant through a thin Cu/Cr/polyimide stack is used to evaluate SIMS quantification issues, in particular the basic SIMS concept of ion yield saturation. Then gettering of chlorine at metal/polymer interfaces is demonstrated by annealing Cu/Cr/polyimide stacks where the polyimide has a known and intended impurity level.