| AVS 55th International Symposium & Exhibition | |
| Electronic Materials and Processing | Wednesday Sessions |
| Session EM-WeM |
| Session: | High-K Oxides and High Mobility Substrates |
| Presenter: | A. Herrera-Gomez, Cinvestav-Unidad Queretaro, Mexico |
| Authors: | M.D. Morales-Acosta, Cinvestav-Unidad Queretaro, Mexico A. Herrera-Gomez, Cinvestav-Unidad Queretaro, Mexico F.S. Aguirre-Tostado, The University of Texas at Dallas J. Kim, The University of Texas at Dallas R.M. Wallace, The University of Texas at Dallas |
| Correspondent: | Click to Email |
The interfacial layer that is formed as hafnium oxide is deposited on silicon affects the performance of hafnium-based C-MOS devices. Although the composition of the interfacial layer could be evaluated from electrical measurements,1 a more direct approach is desirable. The ideal technique for this type of studies is ARXPS because of its unparallel chemical and depth resolution. We performed high resolution ARXPS studies on ALD HfO2 film grown on H-terminated and 1nm SiO2-terminated Si(001) surfaces employing 5 and 30 cycles (H2O and TEMA-Hf). The deconvolution of the peaks was done robustly. Since surface potentials could also cause “chemical” shifts on the binding energies, the physical origin of the different peaks was assigned not only from the peak position but also from the detailed analysis of the take-off angle dependence of the peak area. It was possible to learn about the early stages of the HfO2/SiO2 interface formation since the thickness of the hafnium oxide layer for the samples with 5 ALD cycles were less than one monolayer. Although the 30 cycle ALD growth resulted in stoichiometric 2nm HfO2 films for both surfaces, in the early stages the growth of the HfO2 was more efficient for the SiO2-terminated sample. The first ALD cycles in the H-terminated samples caused the formation of 1.5ML of oxidized silicon. The Si 2p binding energy was 102.9eV, 0.4eV smaller than for the SiO2-terminated samples. Our analysis shows that this difference could be associated to the known dependence of the Si 2p binding energy on the oxide growing process, and not to the formation of a hafnium silicate layer for the H-terminated surfaces. As reported elsewhere,2 there is a component (Hf*) with Hf 4f binding energy 0.6eV higher than for hafnium oxide. The analysis strongly suggests that Hf* could be associated to the first layer of HfO2 in contact with the SiO2, and not to a hafnium silicate layer. We concluded that the ALD process produced an abrupt SiO2/HfO2 interface, even for the H-terminated samples. A clear description of the self-consistent ARXPS analysis will be presented. This work was supported by the Semiconductor Research Corporation and the Texas Enterprise Fund.
1S. K. Dey, A. Das, M. Tsai, D. Gu, M. Floyd, R. W. Carpenter, H. De Waard, C. Werkhoven, and S. Marcus. J. Appl. Phys. 95, p. 5042 (2004).
2N. Barrett, O. Renault, J.F. Damlencourt, and F. Martin. J. Appl. Phys. 96, p. 6362 (2004).