Electrically Controlled Reversible Polarization Fatigue–Recovery in Ferroelectric Thin Film Capacitors

Abstract

Spontaneous polarization switchable by the externally applied electric field in ferroelectrics is a crucial feature for its technological applications; however, the reduction in switchable polarization due to bipolar electric cycling primarily hampers its potential application in nonvolatile memory. Among others, electric-field-controlled polarization recovery would be highly desirable from fundamental and technological viewpoints. In spite of recent progress in electrically controlled polarization recovery, polarization reversal behavior in fatigued and recovered states, associated device resistance change, and repeatability and reproducibility of fatigue–recovery cycles remain poorly understood. This study provides an in-depth insight into the recovery of fatigued polarization by an electric field for epitaxially grown Pb­(Zr0.3Ti0.7)­O3 thin film capacitor with oxide electrodes. Systematic analysis of time- and field-dependent polarization switching kinetics in pristine, fatigued, and recovered states using a suitable statistical model enabled the precise extraction of microscopic switching parameters such as mean characteristic switching times and effective activation fields. Our results demonstrate that fatigue and recovery are possibly mediated by domain pinning–depinning because of defect redistribution and trapping–detrapping of charges at nonelectroneutral domain walls. In addition, the consecutive fatigue–recovery cycles elucidate the repeatability and applicability of electric field controlled polarization recovery in ferroelectric capacitors.

Ferroelectric Size Effects on Statics and Dynamics of Domain Wall

“…[34] This has been ascribed to the weak pinning caused by lattice imperfections arising from defects and epitaxial strain fields, further enabling the dynamic response to subcritical fields in the presence of thermal activation with non-linear creep motion driven by the interplay between elasticity and pinning. [12,35,36] The average activation field E a and creep exponent 𝜇 extracted from fitting are 2.31 ± 0.36 and 0.54 ± 0.03 MV cm -1 , respectively, which are in good agreement with the reported values for epitaxial films. [29,37] Furthermore, it should be noted that the obtained E a and 𝜇 values are susceptible to intrinsic (e.g., local defect density) and extrinsic (e.g., tip abrasion effects and inhomogeneities in the applied field) factors, [26,32,38] thereby averaging over at least a few tens of independent realizations is imperative to achieve statistically more reliable results.…”

“…Notably, at any fixed 1/ E , the slope of the nonlinear fit lines reflecting activation field behavior decreased significantly for larger thicknesses and lower tip voltages (Figure S7, Supporting Information), which has been attributed to the larger depolarization fields in thinner films and the dominant role of thermal activations in relatively small field (or voltage) regions. [ 35,40,41 ]…”

“…The observed fluctuation in the DW profile could be due to weak pinning by inherent defects (e.g., oxygen vacancies and cation interstitials), interfacial defect states, or strain‐induced defects propagating from the substrate and bottom electrode. [ 35,36,49 ] To analyze the roughening of these DWs at a given length scale L , we employed the real space displacement autocorrelation function or roughness function B ( L ) $$$$\begin{equation}B\left( L \right)\ = \overline {{{\left\langle {{{\left[ {\Delta u\left( L \right)} \right]}}^2} \right\rangle }}_L} \ \end{equation}$$$$where Δ u ( L ) = u ( y + L ) − u ( y ) is the relative displacement for an elastically optimal configuration with respect to the flat configuration and u ( y ) is the transversal displacement at the longitudinal coordinate y along the wall (Figure 3c). [ 26 ] In Equation (), 〈⋅⋅⋅〉 L corresponds to an average over longitudinal coordinate y at fixed L , whereas $$\bar{ \cdots }$$ stands for an average over different disorder realizations.…”

“…Notably, at any fixed 1/E, the slope of the nonlinear fit lines reflecting activation field behavior decreased significantly for larger thicknesses and lower tip voltages (Figure S7, Supporting Information), which has been attributed to the larger depolarization fields in thinner films and the dominant role of thermal activations in relatively small field (or voltage) regions. [35,40,41] Figure 2e,f illustrates the thickness dependence of the fitting extracted 𝜇 n distribution mapping over the set of 16 independent data sets and averaged 𝜇, respectively. Remarkably, we observed the quasicontinuous transition of 𝜇 from ≈0.54 for thickness above 25 nm to ≈0.22 for thickness below 10 nm, which is highlighted by a shaded region.…”

“…The observed fluctuation in the DW profile could be due to weak pinning by inherent defects (e.g., oxygen vacancies and cation interstitials), interfacial defect states, or strain-induced defects propagating from the substrate and bottom electrode. [35,36,49] To analyze the roughening of these DWs at a given length scale L, we employed the real space displacement autocorrelation function or roughness function B(L)…”