Effects of Various Pre‐Intrinsic and Phosphorus Diffusion Gettering Treatments Upon Quality of Czochralski Silicon Wafer Surface during a Simulated 4 Megabit Dynamic Random Access Memory Process

Abstract

Effects of various preintrinsic and phosphorus diffusion gettering treatments upon quality of near-surface region in Czochralski silicon wafers are studied during a simulated 4 Mb dynamic random access memory process. Denuded zone depth and bulk microdefect density are determined by synchrotron radiation section topography. Minority carrier lifetime and junction characteristics from test device structures were measured to determine overall gettering efficiency. A two-step thermal anneal cycle before actual device processing resulted in formation of precipitates, dislocations, stacking faults, and a well-defined denuded zone in samples with medium oxygen content only when a low-temperature cycle at 775~ was followed by a high-temperature one at 1150~ The longest minority carrier lifetimes are observed in samples where very few or no defects are visible in the bulk of the wafer in the section topographs. Phosphorus gettering, however, was found to be effective for improving both minority carrier lifetime and junction properties.Gettering is a means of reducing the concentration of metallic impurities in the active areas of a wafer by localizing them in predetermined regions away from those occupied by devices. This is realized by introducing one or more types crystalline defects to control the unwanted defects. The gettering techniques are classified into two large groups: "extrinsic" and "intrinsic." In extrinsic gettering (EG) the controlling defects are introduced externally, usually at the back side of the wafer. EG is achieved by one or more of the following techniques: metal film deposition, solute diffusion such as phosphorus gettering, mechanical damage, ion implantation, laser damage, Si3N4 film deposition, polysilicon deposition, low-current corona discharge, and Ge-doped Si-epitaxy. In intrinsic gettering (IG) the controlling defects are induced in the bulk of the wafer during the thermal process (1). The essence of IG is to control the thermally induced microdefects; that is, microdefects in silicon crystals can act as gettering sites for undesirable impurities when they are formed in the bulk of the wafer (2, 3). Applications of EG to Czochralski (CZ) Si can cause an interaction between EG and IG to occur. It has been reported recently that EG causes an enhancement of oxygen precipitation in the wafer bulk (4, 5), as well as a gettering of oxygen in the damaged regions (6, 7). Oxygen precipitation occurs with the increase in volume. The very large volume change associated with oxygen precipitate (~126%) creates a strain field. If the strain field exceeds the elastic limit of silicon, then dislocations, dislocation loops, and stacking faults are generated to relieve the strain. The degree of the interactions between EG and IG can be determined by the amount of crystal defects formed by various gettering treatments and subsequent thermal cycles.In this investigation we examined three different preinitial oxidation heat-treatments (IG) and phosphorus diffusion gettering (EG) of CZ silicon. We studied the ...

Integrated circuit process control monitoring (PCM) data and wafer yield analysed by using synchrotron x-ray topographic measurements

“…It has successfully been used in the section geometry mode for the observation of oxygen-induced precipitates and stacking faults inside silicon wafers [3], and for investigating simulations of IC process steps [4]. The strength of the method is that it is fast and non-destructive.…”

“…Previously, a brief undocumented analysis by the same authors was undertaken wherein ICs were processed on wafers in the same process, and these also revealed no denuded zone after the processing, even though the denuded zone constantly seems to appear in various process simulations [3,4]. The brief analysis also dismissed the initial findings of stronger wafer surface strain gradient being linked to higher yield because of non-quantitative data and the lack of supporting studies [28] at the time.…”