Dependence of Epitaxial Layer Defect Morphology on Substrate Particle Contamination of Si Epitaxial Wafer

Abstract

Mirror-polished Czochralski (CZ) wafers have been predominantly used for large-scale integrated (LSI) memory device fabrication. In the design rule for 16 Mb dynamic random access memory or for higher levels of integration, it is becoming increasingly important to overcome undesirable effects caused by surface defects originating from CZ grown-in defects 1,2 or near-surface defects generated during the device thermal processes. 3,4 Since the epitaxial (epi) layer of an epi-wafer does not contain those grown-in defects, 5 one can eliminate quite effectively device failure induced by the CZ grown-in defects. Thus the use of epi-wafers is becoming popular, not only for microprocessor and flash memory device applications, but also for advanced memory devices. The gate oxide integrity characteristics of epi-wafers are generally superior to those of mirror-polished CZ wafers. 5 Even in those epi-wafers, however, there still exist characteristic surface and near-surface epi layer defects. Thus it is important to understand the characteristics and generation mechanisms of these epi-defects. Stacking faults and mounds are the typical epi-defects observed in Si epitaxial layers. 6 Crystal disorders like twins are not observed in industrial Si epitaxial wafer production, which involves a clean, dislocation-free substrate with a high degree of crystal perfection. Investigating the exact generation mechanisms of these defects is not in the scope of this study, but intensive discussion has been provided by many researchers. 7-11 The presence of foreign particles on the substrate surface is one of the origins of epi layer stacking faults. 7-9 A mound is an epi surface protrusion produced by abnormal epitax-ial growth, often originating from a foreign particle on the substrate surface. The defect entity of the mound or the stacking fault in the (100) epi layer is often surrounded by four reversed pyramidal planes, (111) oriented, with a common apex where the foreign particle exists at the epi layer/substrate interface as the defect origin. (The defect model structure is discussed in Fig. 5.) Here we first classify epi layer defects into six geometrical types, as explained in the next section. We then investigate a relation between the substrate particle and the epi-defect morphology to understand how the different geometrical types are generated. Experimental Substrate wafers for epitaxial growth were 200 mm diam (100) oriented 8-10 cm p-type CZ-Si wafers. Before epitaxial growth, the substrate wafers were intentionally contaminated by two different methods. One is a method of wafer exposure in a class 10,000 clean room for particle diameter >0.5 m. The other is a method of depositing 1 m diam standard polystyrene microspheres (STADEX SC-101-S particles, JSR Corporation), which are calibration standards to establish a relation between particle size and scattering intensity of a particle inspection system. The polystyrene microspheres were deposited by a particle deposition system (PDS-100, VLSI Standards). Particles on the contaminated wafers were then observed and analyzed by a defect coordinate linkage system consisting of a Hitachi particle counter LS-6500 and Hitachi scanning electron microscope (SEM) S-4160. Elements of particles were analyzed by energy-dis-persive X-ray spectroscopy (EDXs) using SIGMA-III of a KEVEX mounted on a Hitachi S-4160. Before epitaxial deposition, the sample wafers were preheated at 1150C for 1 min in a H 2 ambient to observe particle deformation characteristics by high temperature heat-treatment. Particles at identical wafer locations were observed by SEM before and after the preheat. We then repeated epi layer depositions several times on the intentionally contaminated substrates, where the particle characteristics had already been analyzed. After each step-wise epitaxial deposition, the particle-induced mounds or stacking faults of epi layers were observed at positions identical to those of the substrate particles by the coordinate-linked SEM or by an atomic force microscope (AFM, SPA-360, Seiko Instruments). The accumulated total epi thicknesses were 0.5, 1, 3, 5, 10, and 15 m. The epi-defect structure was analyzed by transmission electron microscopy (TEM) using a (110) cross-sectional sample fabricated by a dicing saw and a focused ion beam (FIB) system. The final thickness of the TEM sample was 0.7 m. Using a JEOL 2000FX transmission electron microscope, an epi-defect was observed under 200 kV accelerating voltage. The elemental analyses were performed by TEM-EDX using a VOYAGER system from Noran Instruments mounted on a Hitachi HF-2000. Results and Discussion Classification of epi-defects.-In epi layers with various epi thicknesses grown on the intentionally particle-contaminated sub-strate wafers, we observed a variety of particle-induced mounds and stacking faults by SEM or AFM. From the defect morphological characteristics, we classified epi defects into six representative categories , as shown in Fig. 1. These defect types are called rock-type (Fig. 1a), piled-type (1b), pyramid-type (1c), terrace-type (1d), L-type (1e), and I-type (1f). To investigate how these characteristic defect types were generated, we first characterized defect originating particles on the starting substrate wafers. We then performed repeated epi layer depositions on the contaminated substrates to observe sequentially how an epi-defect changes at an identical defect position by consecutive depositions. Characterization of substrate particles and epi-defects.-After intentional particle contamination of substrate wafers, coordinate-linked The morphology of an epitaxial layer defect has been studied in relation to the layer thickness and size/material type of a foreign particle on the starting substrate of a Si epitaxial wafer with (100) surface. As expected, the edge length of a square-shaped stacking fault or a crystal disorder on the (100) epitaxial surface becomes longer as the layer thickness increases. It is further revealed that the defect morphology is predominantly determined by the ratio of the substrate particle size, d, to the layer thickness, t. Micro-protrusion defects can appear when the ratio d/t is relatively large. With decrease of the d/t ratio, the defect morphology gradually changes into a square-shaped stacking fault. The morphology dependence on the d/t ratio is also influenced by the material type of the particle on the starting substrate wafer. The particle, which is deformable during epitaxial deposition, can generate a micro-protrusion when d/t is large. However, this is not the case for an undeformable particle like a polystyrene microsphere. The defect morphology thus provides size and material characteristics information of a particle on the starting substrate wafer.) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 129.15.14.53 Downloaded on 2015-01-28 to IP