Equivalent oxide thickness ͑EOT͒ scaling still remains one of the main facilitators to increase transistor performance. The use of rare-earth ͑RE͒ elements has been shown to effectively increase permittivity of the gate-stack interfacial layer between substrate and high-k dielectric, reducing its EOT contribution. In this paper, we have studied the optimal RE-to-SiO 2 ratio using Dy 2 O 3 as a test case. Capacitance-voltage and leakage current-voltage measurements were performed on Pt-gated capacitors with a SiO 2 /Dy 2 O 3 /Sc-doped HfO 2 gate stack and varying ratios of Dy 2 O 3 to SiO 2 after a 1000°C anneal. Optimal EOT-leakage performance was found for a Dy 2 O 3 -to-SiO 2 ratio of ϳ0.5 to 0.6. Improvement of complementary metal oxide semiconductor ͑CMOS͒ transistor performance is largely driven by the scaling of the equivalent oxide thickness ͑EOT͒ of the gate dielectric. 1 To allow continued scaling, beyond the limitations of SiON gate dielectrics due to excessive gate leakage, Hf-based dielectric layers have been introduced in the gate stack. 2 Because of their higher permittivity ͑k value͒ compared to SiON, the dielectric layer can be made thicker while maintaining high capacitance values, i.e., without having to compromise with respect to the EOT. 3,4 In practice, there is typically a SiO 2 -like interfacial layer ͑IL͒ present between the high-k layer and the silicon substrate. The presence of a SiO 2 IL is often desired because it helps to alleviate the reduction in carrier mobility in the transistor channel due to interaction with the high-k dielectric. 5 However, the total EOT of the gate stack will therefore always consist of contributions from both the high-k layer and the IL. For scaled high-k gate stacks, the EOT contribution of the IL can become even more significant than the one from the high-k layer on top because of its lower k value ͑as a first approximation by a factor of k high-k /3.9͒. This implies that EOT scaling of thin stacks is more efficient by reducing the thickness ͑or more generally the EOT contribution͒ of the IL than that of the high-k layer.The IL that is present at the end of the CMOS device process can have different origins. First, a SiO 2 layer is often grown deliberately prior to the high-k deposition. This can be done to reduce the influence of the high-k dielectric on channel mobility or to accommodate the high-k deposition itself, as it has been shown that, for example, atomic layer deposition ͑ALD͒ HfO 2 by HfCl 4 /H 2 O needs a proper starting surface for optimal growth. 6,7 Next, depending on thickness and composition of the starting surface, the starting layer can further oxidize ͑thicken͒ during high-k growth. Especially for metallorganic chemical vapor deposition type depositions, this typically occurs because of the higher deposition temperatures as compared to, for example, ALD or physical vapor deposition. [8][9][10] Finally, even when the IL thickness can be controlled before and during high-k deposition, there is still the possibility of IL regrowth during h...