Silicon Gate Oxide Thickness Uniformity during HCl Oxidation

Abstract

Systematic oxide thickness variations across individual wafer surfaces, as well as from wafer to wafer within quartz boats were observed during oxidation. These variations, namely, a radial oxide thinning toward the wafer center and a wafer spacing dependence of average oxide thickness, were associated with the use of during “dry” gate oxidation. The “bull's eye” thickness profile amounted to a 5% variation in a 25.0 nm gate oxide for densely packed (5 mm spacing) wafers. The radial variation disappeared when wafers were spaced far apart (50 mm). The oxide thickness variation was quantitatively correlated to the amount of chlorine incorporated in the oxide where the chlorine content was found to be linearly proportional to the oxide thickness. This oxidation nonuniformity could be modeled by the radial depletion of an active intermediate, chlorine‐containing species during oxidation. While oxides grown in or , did not exhibit a radial thickness variation, those grown in did exhibit thickness variations similar to oxidations. The absolute enhancement of the oxidation rate due to the additions of , , , or in an otherwise dry oxidation was also studied, with alone and alone producing a large enhancement in growth rate (35% thicker oxide for 2.25% or 4.5% additions). The enhancement due to addition was more modest (15% for 2.5% addition). Transients during the switchover in gases, e.g., from during loading to for oxidation, apparently did not play a major role in causing the observed nonuniformities. The use of thick‐oxide “dummy” wafers, grown in a hydrogen‐ or chlorine‐containing atmosphere, between bare silicon wafers had a profound effect on improving thickness uniformity. The oxide thickness variation was reduced when wafers with 0.4 μm dry oxides were interspersed between growing gate oxides; however, this improved oxide uniformity was attributed to a larger spacing between oxidizing wafers. Surprisingly however, variations completely disappeared when the dummy wafers were grown in steam, dry , or dry , suggesting that these dummy wafers play an active role in enhancing the supply of reactive intermediates to adjacent wafers.