Copper diffusion in silicon



Journal Title

Journal ISSN

Volume Title


Texas Tech University


Diffusion processes have always been one of the major subjects of solid state science, either due to technological interest or pure scientific curiosity. Because of its technological importance in devices [1] Cu diffusion studies are some of the oldest in semiconductor materials, in particular silicon. However, the Cu diffusion problem was not a great concern for researchers. Recently this has changed. Various authors [2,3] have reported that chemo-mechanical polishing introduces some unknown defect, named X defect, into the host material that gave a difflisivhy much larger than that of Cu, which is known to have the fastest difflisivity coefficient in silicon. Following this, several researchers investigated this fast diffusing defect and finally copper was identified [4] as the most likely candidate. Even though Cu was the principal candidate for this fast diffuser in Si, there were numerous unexplained observations. One of these was the observed pre-exponential factor of the diffusion coefficient [4-5]. Later, to explain all the observed data for Cu diffusion in sUicon Mesli et al. [5-6] undertook an extensive investigation. In the course of this investigation, they developed the technique Transient Ion Drift (TID), which is where we began our investigation. This technique, explained in the following chapters of this work, is a derivative of deep level transient spectroscopy (DLTS) and therefore is a simple and fast method for determining the diffusion coefficients of some fast diffusing impurities [6-7]. At the time we thought to apply this technique to some other fast diffusing impurities in silicon such as hydrogen. However, we also thought that we should apply TID first to Cu, both to verify Mesli's findings, and to gain some experience before applying TID to hydrogen. The only difference between our work and Mesli's reported findings [6] was in the silicon samples we used. Mesh reported that the samples they used were doped with Ga, while our samples were doped with B. Yet, using B doped samples caused confusingly different results from those reported in Mesh's original work. As a resuh, we investigated Cu diffusion in silicon [8] in greater detail than originally anticipated. In this work we report our findings on B doped Cu diffiised Si samples using the technique TID, and will also introduce a new model to investigate the same phenomena. The new model has two chief advantages; it can be applied during a TID run and can also be used to check the results obtained by TID and vice versa. We name this technique transient ion diffusion, TIDIF Below, we first give a brief literature review of diffusion and Cu diffusion in Si. Then, TID and its companion technique, TIDIF, will be presented. Following these are the experimental setup, data acquisition, and analysis procedures. Finally we will conclude this report with our results and an extensive discussion.