High-performance nMOSFET with in-situ Phosphorus-doped embedded Si:C (ISPD eSi:C) source-drain stressor (B.Yang, AMD/IBM, 2008 IEDM)
Yang, B.; Takalkar, R.; Ren, Z.; Black, L.; Dube, A.; Weijtmans, J.W.; Li, J.; Johnson, J.B.; Faltermeier, J.; Madan, A.; Zhu, Z.; Turansky, A.; Xia, G.; Chakravarti, A.; Pal, R.; Chan, K.; Reznicek, A.; Adam, T.N.; de Souza, J.P.; Harley, E.C.T.; Greene, B.; Gehring, A.; Cai, M.; Aime, D.; Sun, S.; Meer, H.; Holt, J.; Theodore, D.; Zollner, S.; Grudowski, P.; Sadana, D.; Park, D.-G.; Mocuta, D.; Schepis, D.; Maciejewski, E.; Luning, S.; Pellerin, J.; Leobandung, E., “High-performance nMOSFET with in-situ phosphorus-doped embedded Si:C (ISPD eSi:C) source-drain stressor,” Electron Devices Meeting, 2008. IEDM 2008. IEEE International , vol., no., pp.1-4, 15-17 Dec. 2008
Abstract: For the first time, embedded Si:C (eSi:C) was demonstrated to be a superior nMOSFET stressor compared to SMT or tensile liner (TL) stressors. eSi:C nMOSFET showed higher channel mobility and drive current over our best poly-gate 45 nm-node nMOSFET with SMT and tensile liner stressors. In addition, eSi:C showed better scalability than SMT plus tensile liner stressors from 380 nm to 190 nm poly-pitches.
- eSiC ~1.8% with in-situ P doping improves Ion as much as 32% due to enhanced tensile stress from SiC
- When eSiC S/D becomes thicker more than 30nm from the original surface, effect from tensile stress liner becomes much less.
- eSiC strain effect is more compatible with scaled pitch than SMT or TL
- structural and electrical connectivity at the junction point between eSiC and extension is important to improve eSiC effect.
- Shallow junction with high extension dose makes the connection worse, so that Ion improvement from eSiC is minimal.
To date, only a few groups have
reported positive progress in introducing eSi:C in thickoxide
long-channel nMOSFET (Ion>20%) [1,2], while little
benefit was achieved with thin-oxide short-channel
nMOSFET (Ion<6%) [1,3]. In fact, all reported thin-oxide
short-channel eSi:C nMOSFET suffered Ion degradation
compared to Si control nMOSFET with SMT and tensile
liner stressors [1, 3-5, table-I].
Fig.1 shows the typical cross-sectional TEM images of 45 nmnode
nMOSFET transistors with in-situ phosphorus-doped
(ISPD) epitaxial Si:C embedded in the S/D regions. The defect
density of the ISPD eSi:C is found to be acceptably low after
both epitaxial growth (a) and device fabrication up to first
metal (M1) (b).
Carbon and Phosphorus profiles characterized
by SIMS (Fig.2) showed chemical Carbon atomic
concentration of ~1.9%, and Phosphorus chemical
concentration of ~ 3E20cm-3. An in-line ìXRD technique was
used to monitor the eSi:C film strain (Fig.3) after each
critical processing steps. The effective substitutional C% in
eSi:C was found increased from 1.67% post epitaxial growth
to 1.85% post device fabrication. The increase in eSi:C film
strain was mainly resulted from the high temperature Laser
Annealing (LSA) post EPI growth.
Fig.6 show that eSi:C flush fill
creates over 2× channel stress compared to 1.2 GPa TL, and
>0.5GPa more stress with 30nm overfill. Additionally, eSi:C
overfill reduces the TL effect, indicating TL becomes much
less effective when eSi:C EPI is 30nm-overfilled.
The Ioff –
Ion plots (Fig.10) demonstrate that the 30nm-overfilled
eSi:C+TL resulted in the highest drive current, showing ~
32% Ion benefit over the Si control with NL, and ~9% Ion
benefit over the state-of-the-art 45nm-node Si control with
STM+TL stressors. eSi:C could be superior even compared
to the combination of SMT+TL.
eSi:C showed better scalability (Fig.15), with
only ~1% Ion degradation from 380nm to 190nm pitch (c.f.
Si control, ~5% degradation).
epitaxial-growth-process sensitivity to extension doping
concentrations. Superior structural and electrical connections
between ISPD-eSi:C to nMOSFET extensions (Fig.19) are
easier to establish with device that has low extension doping
concentration (such as our thick-oxide device) than that with
ultra-shallow-junction and extremely high extension doping
concentration (such as our thin-oxide device). Therefore, it
is much more difficult to enable eSi:C with thin-oxide than
that with thick-oxide devices as reported earlier