Film thickness non-uniformity and crucial role of water dosage for atomic layer deposition of yttrium oxide from yttrium cyclopentadienyl precursor

Yttrium oxide (Y2O3) is a promising dielectric material for various applications in micro- and optoelectronics due to its high permittivity (k = 10 ~ 15), wide band gap (ΔE ~ 5.5eV), high refractive index (n ~ 1.9), and high thermal stability [1]. Among the existing techniques to grow Y2O3, atomic layer deposition (ALD) gained an increasing attention due to the ease to form layers of excellent conformality and desired stoichiometry with an accurate thickness control [1]. In the past, yttrium β-diketonates, such as Y(thd)3, have been widely investigated as precursors for ALD of Y2O3, but they required highly reactive co-reactants such as ozone (O3) or oxygen (O2) plasma [2]. In contrast, the cyclopentadienyl-type yttrium precursors, namely Y(RCp)3 (R = H, Me, Et, or iPr), were found to be more reactive and they enabled thermal ALD with H2O [3] as co-reactant even without applying plasma. Besides, ALD of Y2O3 from Y(RCp)3 is more attractive because of the higher growth per cycle (GPC) rate, as compared to that of the β-diketonates, considerably improving the throughput. For example, ALD from Y(MeCp)3 and H2O shows a GPC of ~ 0.12 nm/cycle, being nearly 5 times higher than that for ALD from Y(thd)3 and O3 [3]. The Y(MeCp)3-type precursors are also available from commercial suppliers, in opposite to the precursor family recently developed by introducing O or N-coordination into Y(MeCp)x (x = 1 or 2).

ALD of Y2O3 from Y(MeCp)3 and H2O was reported to occur between 200 and 400 °C with a near-constant GPC of approx. 0.12 nm/cycle [3]. The need for an accurate control of the water vapor was mentioned elsewhere [4]. These conclusions were obtained solely via ex-situ characterization of the film growth kinetics. Studying the kinetics of Y2O3 ALD from Y(MeCp)3 and H2O in real time, specifying the influence of water dosage, and finally demonstrating a high level of film thickness uniformity on a wafer scale, naturally expected for ALD, is still missing.

This work explores in detail the temperature window and the crucial role of introducing well-controlled dosage of H2O for ALD of Y2O3 from Y(MeCp)3, by utilizing real-time in-situ characterization of the process kinetics by means of spectroscopic ellipsometry (SE) (Fig. 1 (a)). We verify the reported ALD window by combining the in-situ SE measurements with ex-situ wafer-scale mapping of the film thickness. This work demonstrates that growing a Y2O3 film at conditions corresponding to standard saturation of GPC curves versus pulse- and purge-duration (Figs. 1 (b), (d) and (e)) does not necessarily guarantee a high level of the film thickness uniformity (Fig. 1 (c)), as it must be for an ALD process. Our results suggest a significant narrowing of the reported ALD window from 200 ~ 400 oC to just ~ 200 oC, in order to keep the film thickness non-uniformity within 5% over a 4” Si wafer and a decent GPC of ~ 0.08 nm/cycle (Fig. 1 (f)). Increasing the temperature from 200 to 225 oC already provides a GPC of 0.11 nm/cycle, i.e., close to the value reported in the literature. However, the corresponding thickness non-uniformity increases up to 16%. Further raise of the temperature to 250 oC provides a slightly higher GPC of 0.13 nm/cycle but increases the thickness non-uniformity to 22% (Fig. 1 (f)). This implies a significant non-self-limiting (CVD) component of the ALD process, being rapidly enhanced by the temperature.