(Keynote) Border-Trap Characterization for Ge Gate Stacks Using Deep-Level Transient Spectroscopy

Abstract

1) Introduction To fabricate high-performance Ge MOSFETs, a high-quality gate-stack of Ge is essential. Although a GeO2 interlayer is very effective to improve the interface quality between Ge and gate-stack, the existence of large number of border-traps (BTs) in GeO2 is still a serious issue that degrades the performance of Ge MOSFETs. Therefore, the BT characterization for GeO2/Ge structure is desired. In this work, we intentionally used relatively strong injection to carry out deep-level transient spectroscopy (DLTS) measurements. By separating out interface traps (IT) contribution from DLTS signals, we extracted the BT signals and calculated BT concentrations (N bt). BTs also exist in the SiO2/Si structure, and DLTS has been applied to evaluate N bt for Si MOSCAPs [1]. However, owing to the relatively low N bt in SiO2 and the large band-offsets at the SiO2/Si interface, BTs do not cause serious problems in Si MOSFETs. Therefore, an N bt evaluation has not been studied intensely. As for the GeO2/Ge structure, the BT problem becomes serious because the N bt in GeO2 is relatively high and the band offsets at the GeO2/Ge interface are lower than those of SiO2/Si. Actually, when we measured the IT density (D it) for a GeO2/Ge structure, we found that the DLTS signal intensity clearly increased with increasing intensity of the injection pulse [2], which must have resulted from the BTs in GeO2. We believe that DLTS is a promising method to evaluate BTs for Ge MOSCAPs. In this presentation, BT evaluation using the DLTS method for Ge MOSCAP is introduced. Both D it and N bt are evaluated for Ge MOSCAPs fabricated by post-passivation thermal oxidation (PTO) and electron cyclotron resonance (ECR) plasma oxidation. In addition, the effects of Al post metallization annealing (Al-PMA) are also investigated. 2) Experimental Both p- and n-type (100) Ge substrates with respective doping concentrations of 2.3x1016 and 9.3x1015 cm-3 were used. After substrate cleaning, SiO2/GeO2 bilayer passivation was performed [3]. Here, the thicknesses of SiO2 and GeO2 layers were ~1 nm each. Next, PTO was performed at 550°C for 15 min or 425°C for 9 h in O2 ambient. After 14-nm-thick SiO2 was deposited on the both samples, the annealing at 400°C for 30 min in N2 was performed. Then, the Al gate film was deposited on the SiO2 surface by thermal evaporation. Before the electrode patterning, an optional PMA was carried out at 300°C for 30 min in N2, which is the Al-PMA. The EOTs of the MOSCAPs with PTO at 550 and 425°C were 19.5 and 16.8 nm, corresponding to GeO2 thicknesses of 6.0 and 3.3 nm, respectively. DLTS measurements were performed using a lock-in integrator. The MOSCAP was reverse-biased at V R and was applied the pulse bias of V P from V R. V AP (=|V P-V FB|, V FB: flat band voltage) is the accumulation pulse voltage, which is an important parameter for the BT analysis, because the injection pulse intensity (E AP) at the GeO2/Ge interface is given by E AP =V AP/EOT. Frequency (f ) and pulse width (t w) are also important parameters for selecting the observed BT position (z 0) and for filling with carriers into BTs, respectively. 3) Summary of the r esults By using p-type MOSCAPs, BTs at the position of 0.4 nm from the GeO2/Ge interface were measured. The energy of these BTs was centralized at the position near to the valence band edge of Ge, and their N bt was in the range of 1017–1018 cm-3. By using n-type MOSCAPs, BTs at the position range of 2.8–3.4 nm from the GeO2/Ge interface were measured, of which Nbt varied little in the depth direction. The energy of these BTs was distributed in a relatively wide range near to the conduction band edge of Ge, and their N bt was approximately one order of magnitude higher than those measured by p-MOSCAPs. This high N bt value might originate from the states of the valence alternation pair with energy close to 1 eV above the conduction band edge of Ge. We also found that Al-PMA can passivate both ITs and BTs near to the valence band edge of Ge but not those near to the conduction band edge. Acknowledgement This work was supported by (JSPS) KAKENHI (grant No. 17H03237).