After HF cleaning of 4º off-axis 4H-SiC (0001) wafers covered with 5 μm-thick n-type epitaxial layers (ND ~ 1×1016 cm-3), ~30 nm-thick SiO2 layers were thermally-grown in dry oxygen at 1300ºC. As the post-oxidation annealing techniques, the annealing in NO:N2 =1:2 ambient at 1150ºC for 2hrs (NO-POA) and the additional annealing in H2O:O2=9:1 ambient at 800ºC (LT-H2O-POA) for various duration time, were sequentially conducted. Finally Au electrode was deposited to form a MOS capacitor. For the NO-POA process of SiC after thermal growth of SiO2, it has been clarified that nitrogen atoms are introduced mostly in the first few monolayers of SiC, and remains even after removal of SiO2 by chemical etching completely [5]. We evaluated the effects of our NO-POA and LT-H2O-POA on the amount of nitrogen introduced at the interface, from the relative change of core level XPS intensity ratio of N1s to the substrate component of Si2p, after the complete removal of SiO2 layer. It was found that the long-time LT-H2O-POA results in a significant reduction of nitrogen density at the interface after the growth of several MLs of SiO2, due to the consumption of nitrogen-introduced SiC surface layers for the oxidation by H2O. However, it would be noteworthy that most of the nitrogen still remains at the interface after a short-time LT-H2O-POA to cause the growth of limited amount of SiO2 around 1 ML.
Next we evaluated the interface state density (Dit) of SiO2/SiC MOS capacitors fabricated with different duration time of LT-H2O-POA after NO-POA, by using conductance method. We found Dit was significantly reduced by short-time LT-H2O-POA, and minimized when only 1 ML of SiO2 was grown during the H2O-POA. The Dit value at the energy level at 0.2 eV below the conduction band edge of SiC was reduced below 5×1010 cm-2eV-1, which is less than half of the value observed just after the NO-POA (without H2O-POA). This would be the minimum Dit ever reported for 4H-SiC (0001) MOS capacitor at this energy level. On the other hand, LT-H2O-POA with excess duration time resulted in an increase of Dit. After the long-time LT-H2O-POA to cause ~1 nm SiO2 growth, the Dit value became even larger than the one just after the NO-POA. This deterioration would be caused by the partial lost of passivating nitrogen at the interface by the growth of new oxide layers, as indicated by the XPS analysis. One of the methods to evaluate the density of such slow traps in SiC MOS capacitors is the characterization of the hysteresis of CV curve observed for a voltage sweep just after the ejection of trapped charges by UV irradiation [6]. We observed that the slow trap density also significantly decreased by short-time LT-H2O-POA when the additional growth of SiO2 was limited to ~1 ML, but gradually increased by a long-time one, which is a quite similar behavior as Dit change by H2O-POA.
These results show that the roles of nitrogen introduction and oxidation by H2O on SiO2/SiC interface defect annihilation are different from each other, and the interface with minimized Dit and less near-interface oxide traps is achievable by the sequential combination of those two POA processes, when the amount of additional oxidation by H2O is limited to ~ 1ML of SiO2.
References
[1] J. Rozen et al., IEEE Trans. Electron Dev. 58, 3808 (2011).
[2] D. J. Lichtenwalner et al., Appl. Phys. Lett. 105, 182107 (2014).
[3] H. Hirai and K. Kita, ICSCRM 2017, Washington DC.
[4] H. Hirai and K. Kita, Appl. Phys. Lett. 110, 152104 (2017).
[5] R. Kosugi et al., Appl. Phys. Lett. 99, 182111 (2011).
[6] H. Yano et al., IEEE Trans. Electron Dev. 46, 504 (1999).