In-Line Monitor Introduction to Prevent Metallic Contamination in Wet Bench

Abstract

Contamination controls are very important issues in microelectronics. Any wrong substance introduction in process chambers can cause damages to the production line. Therefore, an extensive control is important because every operation in the process flow (also the most insignificant) can become fatal for the correct functioning of a microelectronic device. The aim of this work is to evaluate the impact of small metallic contamination in the range of 1011÷1012 at/cm2 on silicon substrates implanted with different ion species (As, B and P). An important example of failure related to metallic contamination in a wet bench is reported in this work. The problem appears in a particular class of flash memory devices processing. The electrical parametric test shows a wrong gate oxide thickness and Qbd values out of range, confirmed by early breakdown events and anomalous C-V characteristics. The cause of the failure is morphologically identified off-line by using TEM: the cross section shows a wrong gate oxide thickness and an anomalous interface between gate oxide and silicon substrate. It appears clear that the root failure cause is related to the ion implantation (As in this case) and to the cleaning before gate oxide growth. A short process flow was performed and analyzed step by step in order to identify the failure cause. Many different analytical techniques have been used for each step and all of these provide consistent results. In particular TXRF analysis on wafers processed immediately after cleaning do not show any contamination while Cu and Fe contaminants are observed after sample oxidation and As implant. Metallic contaminants are captured by the substrate after it is implanted with As, and the following RCA cleaning is not able to remove them. In addition, the presence of these metallic contaminants induces roughness of the Si surface and the growth of gate oxide is not controlled (faster oxidation). If different substrates are used, e.g. silicon implanted with B or un-implanted, this contamination level is not detected and does not lead to oxide reliability problems. Once the mechanism of metal contaminant interaction with dopant is identified the introduction of an in-line monitoring is possible, thus allowing to prevent the device failure. The short process loop can be considered as a good method to prepare the substrate before TXRF analysis. After this study the monitor has been integrated in the production line controls