Modeling and Optimization of Silicon Heterostructure MOSFETs

Recent advancements in the CVD growth of silicon germanium and strained silicon layers have created a new class of devices for silicon ULSI technologies. These new materials provide both strain-enhanced mobilities and energy band offsets, features formerly reserved for materials systems such as the III-V compounds. Some silicon heterostructure devices can be compared to III-V HEMTs by considering the silicon dioxide layer analogous to the wide-bandgap layer in a III-V HEMT. The present modeling effort focuses on an n-channel device with a strained silicon channel that is directly beneath the oxide [1]. Strain was introduced by graded SiGe region (5%-30%) and 30% buffer to fully relax the SiGe lattice. The heterojunction at the back of the channel is type II (strained silicon on 30% SiGe), and has a 160 meV conduction band offset that enhances electron confinement near the surface. Strain in the surface channel causes the six-fold degeneracy in the silicon conduction band to split, creating two valleys with lowered energy and four with raised energy. A mobility enhancement is expected due to a reduction in intervalley scattering in the lowered energy valleys.

This work quantifies the mobility enhancement with fitting a three-term analytic model [2] to extracted mobility data taken over the temperature range from 20K to room temperature. The data were taken using the split-CV technique at moderate frequency to avoid channel transmission line effects. Effective field information is computed from measured inversion capacitance and computed depletion capacitance using a Poisson solution that includes Fermi-Dirac statistics and incomplete ionization. Parameter extraction is being performed using a non-linear optimization program [3]. Individual mobility model components (phonon scattering, Coulomb scattering, and surface roughness scattering) are isolated using different ranges of effective field and temperature. The overall trend in the extracted mobility verifies the expected trends towards more enhancement at higher temperatures and less enhancement at lower temperatures. Good fits have been achieved at temperature extremes, however the complex nature of the scattering mechanisms has made predicting temperature dependencies challenging.

[1] J. Welser, J. Hoyt, J.F. Gibbons, 1992 IEDM Technical Digest, pg. 1000.

[2] J. Watt, Stanford University Ph.D. thesis, 1989.

[3] P. Gill, W. Murray, M. Saunders, M. Wright, User's Guide for NPSOL, Technical Report 86-2, Stanford Systems Optimization Laboratory, 1986.

Richard Williams (rqw@gloworm.Stanford.EDU)
AEL 231
Integrated Circuits Laboratory
Stanford University
Stanford, CA 94305-4055