Low Resistivity Contact Methodologies for Silicon, Silicon Germanium and Silicon Carbon Source/Drain Junctions of Nanoscale CMOS Integrated Circuits.

Abstract

State-of-the-art p-channel metal oxide semiconductor field effect transistors (MOSFETs) employ Si(1-x)Ge(x) source/drain junctions to induce uniaxial compressive strain in the channel region in order to achieve hole mobility enhancement. It is also know that the elec- tron mobility can be enhanced if the MOSFET channel is under uniaxial tension, which can be realized by replacing Si(1-x)Ge(x) with Si(1-y)C(y) epitaxial layers in recessed source/drain regions of n-channel MOSFETs.

This dissertation focuses on epitaxy of Si(1-y)C(y) layers and low resistivity contacts on Si, Si(1-x)Ge(x), and Si(1-y)C(y) alloys. While these contacts are of particular importance for future MOSFETs, other devices based on these semiconductors can also benefit from the results presented in this dissertation. The experimental work on Si(1-y)C(y) epitaxiy focused on understanding the impact of various process parameters on carbon incorporation, substitutionality, growth rate, phosphorus incorporation and activation in order to achieve low resistivity Si(1-y)C(y) films with high substitutional carbon levels. It was shown, for the first time, that phosphorus lev- els above 1.3x10^(21) cm^(-3) can be achieved with 1.2% fully substitutional carbon in epitaxial layers. Specific contact resistivity (C) on strained Si(1-x)Ge(x) layers was evaluated using the existent results from the band structure calculations. Previous work on this topic mainly focused on barrier height and the doping density at the interface. In this work, the impact of the tunneling effective mass on specific contact resistivity was calculated for the first time for strained Si(1-x)Ge(x) alloys. It was shown that due to the exponential dependence of contact resistivity on this parameter tunneling effective mass may have a strong impact on contact resistivity. This is especially important for strained alloys in which the tunneling effective mass is dependent on the strain. The contact resistivity was found to decrease with Ge concentration due to the smaller tunneling effective mass in strained Si(1-x)Ge(x). These calculations can also be extended to Si(1-y)C(y) junctions when better models for the Si(1-y)C(y) band structure are available.

Two different metallization schemes have been considered. In the first approach, two band edge silicides are used to achieve low-resistivity contacts on complimentary MOSFETs. For this purpose, experiments on band edge silicides including PtSiGe, NiSiC and ErSiC were conducted. The impact of Ge and C on silicide formation and the barrier height at the interface was investigated. Barrier height values around 0.3 eV were achieved with PtSiGe and ErSiC contacts formed on p-Si(1-x)Ge(x)and n-Si(1-y)C(y), respectively. Due to the exponential dependence of contact resistivity on barrier height, this barrier height is low enough to yield contact resistivity figures below 10^(-8) Ohm-cm^(2) even with modest doping levels. On the other hand, smaller barrier height values will be needed for Schottky barrier MOSFETs.

It is more desirable to use a single metal contact metal on both p- and n-channel MOSFETs, which requires tuning of the barrier height. Impurity implantation was considered as a means to achieve barrier height tuning. Extremely small barrier height values below 0.1 eV were obtained by sulfur segregation for the silicides of Pt, Ni and NiPt on n-type Si and Si(1-y)C(y). Indium segregation was used for the first time to lower the hole barrier height to obtain barrier height values below 0.2 eV on p-Si. The results provide several approaches that can be used to form low resistivity contacts. We believe that the knowledge gained from this work is expected to have a significant impact on choosing the most effective and economical approach to form low-resistivity contacts in CMOS manufacturing.