Crystal Phase Distribution and Ferroelectricity in Ultrathin HfO₂–ZrO₂ Bilayers

Abstract

(Hf,Zr)O2 ultrathin films are used as ferroelectric layers in emerging digital logic and nonvolatile memory devices. The ferroelectric properties of (Hf,Zr)O2 can be improved by interface engineering, such as the formation of nanolaminates with distinct HfO2 and ZrO2 layers. Herein, the ferroelectric performance of HfO2–ZrO2 ultrathin bilayer devices is shown to depend on the stacking order of HfO2 and ZrO2, which affects the quantity of the noncentrosymmetric orthorhombic Pca21 crystal phase. By combining X‐ray diffraction with a novel extended X‐ray absorption fine structure (EXAFS) analysis procedure, the orthorhombic, tetragonal, and monoclinic phase fractions are quantified for bilayers composed of 3 nm HfO2 and 3 nm ZrO2. A significantly larger orthorhombic ZrO2 phase fraction is found when ZrO2 has an unconstrained surface during annealing, whereas the presence of a ZrO2 interface with the substrate results in a substantial tetragonal ZrO2 phase fraction and a 2.4× smaller remanent polarization. HfO2 is found to be less susceptible than ZrO2 to crystal phase templating. The methods presented herein enable mechanistic studies of ferroelectric wake‐up, fatigue, and processing effects in (Hf,Zr)O2 films, accelerating the development of electronic devices that rely on ferroelectric oxides.

Ferroelectric Phase Content in 7 nm Hf(1−x)ZrxO2 Thin Films Determined by X‐Ray‐Based Methods

“…Pca2 1 phase found in ultrathin films of hafnia‐based dielectrics are nonexistent in the bulk form to be used as reference foil for acquiring model or a calibration spectra. The theoretical spectra are simulated from the static average atomic positions and the thermal disorder in these structure models are included using the model proposed by McBriarty et al [ 34 ] A linear relation ( σ D 2 = a * R eff + b ) between the disorder coefficients σ D 2 and effective path length R eff was used to estimate the crystal phase distribution in HfO 2 –ZrO 2 nanolaminates. The constants a and b were appraised by analyzing a ZrO 2 ‐powdered sample and a ZrO 2 MIM stack and were used subsequently in the LCF analysis of the three HZO samples of interest here.…”

“…From the results of Table 2, it is shown that phase percentages estimates are within <20%, and there is no bias in estimating particular crystal phase fraction in these calculations. Two key factors that play a major role in LCF results of the data are binding energy offset and linear parametrization of disorder as explained by McBriarty et al [ 34 ] Simple disorder parametrization may lead to minor distortions in the amplitude of the LCF basis spectra and thereby differences in the fit compared with those obtained from true phase‐pure standard samples. However, the phase‐pure standards of Pca2 1 do not exist in the bulk form.…”

“…They also concluded that the powder samples were not appropriate references for the thin films deposited at high temperatures. McBriarty et al [ 34 ] estimated the phase fraction in HZO nanolaminates of 3 nm HfO 2 and 3 nm ZrO 2 with a combined XRD and EXAFS analysis. It was concluded that the phase distribution is dependent on the stacking order, clearly distinguishing the O‐fe and monoclinic phases.…”

“…A ZrO 2 powder sample and an MIM stack were used to confirm the previously parametrized disorder coefficients σ D 2 . [ 34 ] LCF is a powerful fingerprinting method used in both XANES and EXAFS analysis. For a sample with many constituents (elements, compounds, or crystal phases), its measured spectrum was fitted with a linear combination of spectra of known species to determine the proportion of each constituent in the sample.…”

A Brief Review on the Ferroelectric Fluorite-Structured Nanolaminate

“…Summary of ferroelectric properties on HfO 2 /ZrO 2 nanolaminate samples experimentally observed from ref [21][22][23][33][34][35][36]…”