1090
(Invited) High-Resolution Photoemission Study of High-k Dielectric Bilayer Stack on Ge(100)

Tuesday, October 13, 2015: 14:30
103-B (Phoenix Convention Center)
S. Miyazaki (Graduate School of Engineering, Nagoya University) and A. Ohta (Graduate School of Engineering, Nagoya University)
Significant progress in study of high-k gate dielectrics for CMOS devices has provided considerable leverage for the development of high mobility Germanium MOSFETs.  In fact, Ge-channel with effective carrier mobilities superior to those of Si-channel has been reported in recent years[1, 2].  In practical use, one of challenges nowaday is to reduce the capacitance equivalent thickness, CET, with no serious increase in the leakage current and with keeping advantage in carrier mobility.  To gain better understanding of insulating properties of high-k stack on Ge and their reliability issues, a clear knowledge on energy band profiles in MIS structures is of great importance.  In this presentation, we have demonstrated how to determine precisely the energy band profile of a practically-thin high-k delectric stack on Ge (100) by using high-resolution X-ray photoelectron spectroscopy, XPS, taking a bilayer consisting of ultrathin HfO2 and interfacial TaGexOybarrier layer as an example [3].

After wet-chemical cleaning steps of p-type Ge(100) wafers with a resistivity of around 20 Ω∙cm, a ~2 nm-thick TaOx layer was deposited at 400 ºC in a layer-by-layer fashion, in which Tri (tert-butoxy) (tert-butylimido) tantalum, Ta-TTT, was introduced intermittently using a N2 bubbling method.  Subsequently, ~2 nm-thick HfO2 was formed at 280 ºC by an ALD method using tetrakis(ethylmethylamino)hafnium, TEMA-Hf, and O3.  Subsequently, the post deposition annealing, PDA, was carried out at 400 °C in diluted O2 ambience to densify the films.  The analyses of XPS core-line spectra taken after PDA confirmed the formation of the TaGexOylayer [4].  

The analysis of energy loss signals, ELS, of core level photoelectrons enables us to measure the energy bandgap, Eg, of ultrathin dielectrics [5].  In the Eg determination of ~2 nm-thick HfO2 top layer, the ELS were taken under AlKα excitation in surface sensitive measurements at photoelectron take-off angles, PTAs, of 15 and 30° to reduce the contribution of the ELS from underlying TaGexOy layer and Ge substrate.  After subtraction of Hf4s components from the measured ELS, the onset of the O1s ELS was defined by linearly extrapolating the segment of maximum negative slope to the background level.  Thus, the Eg value of the top HfO2 layer was determined to be 6.20eV within an accuracy of ±50meV, irrespective of PTA.  As for the energy bandgap of TaGexOy interfacial layer, to eliminate the ELS originating from the top HfO2 layer, the ELS of O1s photoelectron measured in the surface sensitive condition at a PTA of 15° were subtracted from the ELS measured at a PTA of 90°.   From the onset of the remaining signals, the Eg value of the TaGexOy interfacial layer was determined to be 4.5eV which was also confirmed by the analysis of the sample after complete removal of the top HfO2layer wet-chemically.

The valence band, VB, offset can be determined directly from the analyses of VB spectra.  By subtracting both the substrate component using a reference VB spectrum separately measured for chemically-cleaned Ge substrate and the VB spectrum taken at a PTA of 15° from the VB spectrum taken at PTA of 90°, remaining signals corresponding to the VB spectra originating from TaGexOy interfacial oxide were obtained.  Thus, from the energy differences in valence band edge among deconvoluted spectra, the valence band offset between Ge and top HfO2 was determined to be 3.5eV and the valence band offset between Ge and the interfacial oxide to be 3.8eV

By combination of measured bandgap values and VB offset values, the energy band profile was completed.   Since there exists almost no conduction band offset between TaGeOxand Ge(100), this energy band profile is not fit to n-channel but applicable to p-channel MOSFETs.

Acknowledgements

This work was supported in part by JSPS Core-to-Core Program of International Collaborative Research Center on Atomically Controlled Processing for Ultralarge Scale Integration. The photoemission measurements using synchrotron radiation were carried out at the beam line 46, SPring-8 with the approval of the Japan Synchrotron Radiation Research Institute (JASRI/SPring-8).  The authors would like to thank Dr. E. Ikenaga of JASRI for his support and assistance on the measurements in SPring-8.   Also, a part of this work was supported by the Research Institute for Nano-device and Bio Systems (RNBS), Hiroshima University, Japan.

References

[1] T. Nishimura et al, Appl. Phys. Express 4 (2011) 064201.

[2] R. Zhang et al., IEEE Trans. Elec. Dev. 60 (2013) 927.

[3] A. Ohta et al., ECS Trans. 64(6) (2014) 241.

[4] K. Hashimoto et al., IEICE Trans. on Electronics. E96-C (2013) 674.

[5] S. Miyazaki, J. Vac. Sci. Technol. B 19 (6) (2001) 2212.