Long-term stability and oxidation of ferroelectric AlScN devices: An operando HAXPES study

Aluminum scandium nitride (Al${1-x}$Sc$x$N) is a promising material for ferroelectric devices due to its large remanent polarization, scalability, and compatibility with semiconductor technology. By doping AlN with Sc, the bonds in the polar AlN structure are weakened, which enables ferroelectric switching below the dielectric breakdown field. However, one disadvantage of Sc doping is that it increases the material’s tendency towards oxidation. In the present study, the oxidation process of tungsten-capped and uncapped Al${0.83}$Sc${0.17}$N thin films is investigated by hard X-ray photoelectron spectroscopy (HAXPES). The samples had been exposed to air for either two weeks or 6 months. HAXPES spectra indicate the replacement of nitrogen by oxygen, and the tendency of oxygen to favor oxidation with Sc rather than Al. The appearance of an N$2$ spectral feature thus can be directly related to the oxidation process. We present an oxidation model that mimics these spectroscopic results of the element-specific oxidation processes within Al${1-x}$Sc$_x$N. Finally, in operando HAXPES data of uncapped and capped AlScN-capacitor stacks are interpreted using the proposed model.

💡 Research Summary

This paper investigates the long‑term oxidation behavior of scandium‑doped aluminum nitride (Al₀.₈₃Sc₀.₁₇N), a material of great interest for ferroelectric (FE) devices because of its large remanent polarization, CMOS compatibility, and scalability. While Sc incorporation weakens the Al‑N bonds and enables FE switching below the dielectric breakdown field, it also makes the material more prone to oxidation, which can degrade device performance. The authors therefore set out to quantify how oxidation proceeds under realistic processing conditions, i.e., exposure to ambient air for extended periods, and to develop a simple yet predictive model of the elemental oxidation processes.

Three sample sets were prepared: (i) an in‑situ tungsten‑capped Al₀.₈₃Sc₀.₁₇N film (3 nm W), (ii) an uncapped film exposed to air for two weeks, and (iii) an uncapped film exposed for six months. Hard X‑ray photoelectron spectroscopy (HAXPES) was employed using photon energies of 6 keV and 2.8 keV, with emission angles of 5° and 30°, providing information depths of roughly 9–18 nm. Core‑level spectra of Sc 2p, Al 2s, and N 1s were recorded and deconvoluted by peak‑fitting.

Key spectroscopic observations are:

1. Sc‑O versus Al‑O formation – The Sc 2p region shows a pronounced Sc‑O doublet whose intensity exceeds that of the Al‑O component in the Al 2s region. This reflects the larger thermodynamic driving force for Sc‑O bond formation (≈207 kJ mol⁻¹) compared with Al‑O (≈134 kJ mol⁻¹). In the two‑week sample, the Sc‑O fraction reaches ~58 % of the Sc 2p intensity, while in the six‑month sample it drops to ~33 %, still remaining larger than the Al‑O fraction.

2. Emergence of an N‑N feature at ~404 eV – In the six‑month, uncapped sample a distinct peak appears in the N 1s region at a binding energy ~404 eV. The authors assign this to nitrogen‑nitrogen bonding, i.e., molecular N₂ released during oxidation. The same feature is absent in the W‑capped reference, confirming its link to oxidation rather than intrinsic nitride chemistry.

3. Depth dependence – By comparing spectra taken at 2.8 keV (≈9 nm) and 6 keV (≈18 nm), the authors find that the relative intensities of Sc‑O and Al‑O increase for the more surface‑sensitive measurements, indicating that oxidation is not confined to the extreme surface but penetrates at least 10 nm into the film.

To rationalize these findings, the authors construct a two‑dimensional lattice model of the AlScN alloy. Sc atoms are randomly substituted for Al, with the constraint that no two Sc atoms occupy adjacent sites (an approximation that fails at higher Sc concentrations). Oxygen is assumed to replace only nitrogen atoms that are bonded to Sc, because the energy gain for forming Sc‑O is larger than for Al‑O. Under this assumption the following relationships are derived:

• Al‑O fraction = 3 · c_Sc · c_ScO / (1 − c_Sc)
• Released N fraction (forming N₂) = 4 · c_Sc · c_ScO

where c_Sc is the overall Sc concentration (0.17) and c_ScO is the fraction of Sc that has been oxidized. Using the experimentally determined Sc‑O intensity as input, the model predicts Al‑O and N‑N intensities that match the six‑month data reasonably well. For the two‑week sample the model overestimates Al‑O and underestimates N‑N, which the authors attribute to early‑stage formation of Sc‑O‑Sc bridges that consume oxygen without generating as many Al‑O bonds.

The model also allows estimation of the oxide layer thickness. If oxidation were self‑limiting, the same oxide thickness should be inferred from both photon energies. Instead, the calculated thicknesses differ, supporting the conclusion that oxidation proceeds continuously rather than stopping after forming a thin passivation layer.

Finally, operando HAXPES measurements were performed on full capacitor stacks (AlScN sandwiched between Ti bottom electrode and Pt top electrode) both with and without the W capping layer while applying voltage pulses. The spectra show negligible changes in the Sc‑O, Al‑O, and N‑N components during switching, indicating that the electric field does not accelerate oxidation under the tested conditions. The W‑capped stack exhibits virtually no oxidation signatures, confirming the effectiveness of a thin metal cap in protecting the ferroelectric layer.

In summary, the paper provides a comprehensive picture of AlScN oxidation: Sc atoms act as preferential oxidation sites, leading to a higher Sc‑O/Al‑O ratio; nitrogen is released as N₂, giving rise to a characteristic N‑N peak; oxidation penetrates several nanometers and is not self‑limiting; and a simple bond‑counting model can quantitatively reproduce the observed elemental ratios. These insights are directly relevant for the design of reliable FeRAM, FeFET, and other ferroelectric nitride devices, where controlling oxidation—through capping layers, processing atmosphere, or alloy composition—will be crucial for long‑term device stability.