Ge is a promising candidate to replace Si for beyond scaling devices because of
its narrow band gap, high mobility and low dopant activation temperatures. However,
unlike Si, it is difficult to grow an insulating oxide comparable to SiO2/Si on Ge.
Therefore, a lack of thermodynamic stability at the high-k/Ge interface hampers the
development of Ge metal-oxide-semiconductor (MOS) devices due to the desorption
of GeO. The typical GeO2/Ge system has been extensively investigated since the
native oxide on Ge is thermodynamically unstable. However, note that below 700ºC
GeO desorption is not derived from the decomposition of GeO itself but from the
reaction with Ge substrate. For GeO2/Ge, GeO desorbs at around 550oC whereas no
GeO desorption can be observed from the GeO2/SiO2/Si structure. The desorption of
GeO from high-k/ Ge system deteriorates the surface and interface quality, which
leads to the degradation of the electrical characteristics.
Therefore, suppressing the GeO desorption is essentially important to improve the
interface quality so as to realize the Ge-based devices for future application.To crack
the hard nut of this issue, various methods have been used for Ge passivation
such as ozone treatment, high-pressure oxidation,oxides capping, surface nitridation,
rare earth introduction, and atomic oxygen radical treatment.
At the present stage, by applying the above mentioned techniques, Ge-pMOSFET and
Ge-nMOSFET with remarkable electric characteristics have been reported in which
the GeO desorption could be considerably suppressed and the interface trap density
(Dit) has been reduced down to quite low values (mid 1E10~1E11cm-2/eV) . Although
currently some successes have been achieved through process optimization and
technical evolution, much effort is still needed for its industrial application. To realize
further interface control of Ge-based devices, the reactions among high-k/Ge stacks
need to be systematically understood.
Refs: K. Kita et al, JJAP, (2008);S.K. Wang et al, SISC (2009);
S. Wang et al, SSDM(2009); K. Kita et al, IEDM (2009);