Studies have suggested that interface state generation under negative-bias temperature (NBT) stress results in positive oxide charge trapping. The latter is ascribed to the trapping of hydrogen species, from Si-H bond dissociation, in the oxide bulk. In this paper, we present evidence from dynamic NBT instability showing no apparent relationship between the two degradation mechanisms. The level of positive oxide trapped charge is shown to remain constant despite a nonnegligible increase of threshold voltage <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">V</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">t</sub> shift due to interface state generation. This observation implies that Si-H bond dissociation did not result in any significant positive oxide trapped charge, and that the latter is due to a different mechanism (e.g., hole trapping at oxygen vacancies). The inference is supported by results from channel hot-hole stress, which is known to dissociate Si-H bonds but did not increase the level of positive oxide trapped charge. Possible reasons for the difference between previous and current studies are discussed. We also examine the observation on the “universal scalability” of <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">V</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">t</sub> drift curves from different stress conditions, as explained by <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">E</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">'</sup> - <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">P</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b</sub> coupling, i.e., Si-H bond dissociation is driven by hole trapping at nearby oxygen vacancies. We revisit an earlier observation that shows that such scalability only holds in the initial stage when positive oxide charge trapping is dominant, and that the scaled <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">V</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">t</sub> drift curves eventually diverge in the later stage. Further evidence shows that the divergence is caused by interface state generation proceeding at a faster rate compared to positive oxide charge trapping. The result confirms that a part of interface degradation proceeds independently of positive oxide charge trapping.
Read full abstract