Plan view STEM analysis of the domain structure in anisotropically strained epitaxial K 0.95 Na 0.05 NbO 3 ferroelectric films with giant piezoelectricity

Abstract

In the past years lead free K x Na 1‐x NbO 3 ceramics with huge d 33 ‐values, comparable to that of PbZr 1‐x Ti x O 3 , have been demonstrated [1]. The high piezoelectric coefficients are generally attributed to the presence of a monoclinic phase at the morphotropic phase boundary and the associated symmetry reduction. A recent work by Schwarzkopf et al. [2] reports that monoclinic phases can also be stabilized in ferroelectric perovskite thin films by utilizing anisotropic epitaxial strain. Another advantage of using epitaxial strain is to engineer by an appropriate choice of the substrate and film composition the arrangement and size of domains, which in turn influences on the piezo‐ and ferroelectric material properties. Ferroelectric thin films and their domain structure are commonly characterized by combining X‐ray diffraction (XRD) and piezoresponse force microscopy (PFM). While XRD gives useful crystallographic information, its local resolution is limited and a detailed study of the domains is impossible. On the other hand PFM has sufficient spatial resolution but does not provide structural information. Plan‐view transmission electron microscopy (TEM) with its ability to study structural data with high accuracy at atomic resolution is in principle ideally suited to gain detailed information on the domain structure of the film. There is, however, one major challenge. In order to preserve the as‐grown strain state of the film and thus its domain structure plan‐view TEM samples with a thick substrate layer are necessary. Since TEM is a projective method it images typically the projected potential averaged along the beam direction including both film and substrate for a plan‐view observation geometry. In this paper we will show that this challenge can be overcome by appropriate imaging conditions using high resolution scanning TEM (STEM) annular dark field (ADF) imaging. Due to the channeling effect and the reduced depth of focus in STEM mode we are able to obtain information mainly from the monoclinic distorted film, while the substrate gives rise to a uniform background intensity. As an example we study a 26 nm thin film of K 0.95 Na 0.05 NbO 3 , which was epitaxially grown on top of a (110) NdScO 3 substrate and shows giant piezoelectricity. PFM reveals a complex periodic domain structure with the coexistence of two different monoclinic domains (yellow and purple stripes in Fig. 2 a). According to XRD these domains differ from each other by the orientation of the in‐plane monoclinic distortion. Based on the structural data provided by XRD and PFM we created a supercell consisting of a K 0.95 Na 0.05 NbO 3 film, which contains two domains with a difference of the in‐plane monoclinic distortion of 2β=8 mrad, on top of a 150 nm thick rigid NdScO 3 substrate without any lattice distortion (Fig. 1 a). The chosen thickness of the substrate is a minimum value necessary to preserve the as‐grown strain state of the film. Frozen phonon simulations reveal that if the incident electron probe is transmitted first through the film, then the final high resolution STEM ADF image pattern directly reflects the domain structure of the K 0.95 Na 0.05 NbO 3 film with the evaluated shear angle 2β as in the supercell (Fig. 1 b). However, if the electron beam direction is inverted, i.e. the beam is transmitted first through the substrate, then no lattice distortion corresponding to the domain structure of the film is found in the obtained high resolution STEM ADF image. An analysis of the evolution of the electron probe inside the supercell (Fig. 1 c) shows that this 3‐dimensional sensitivity is caused by the electron channeling effect. In the present case the high resolution pattern in the final image is produced basically in the first 20‐30 nm of transmitted material. After this thickness the wave function of the electron beam is widely dispersed, irrespectively of the initial position of the electron probe. Thus the subsequent part of the specimen mainly contributes to a uniform background intensity in the final image. Consequently, the frozen phonon simulations demonstrate the feasibility of our approach to study domains in epitaxially strained ferrolectic films by plan‐view high resolution STEM ADF imaging. An experimental example for the STEM analysis of the domain structure for the above mentioned K 0.95 Na 0.05 NbO 3 film is shown in Fig 2 c. A model of the complex domain pattern including the sub‐domain structure will be discussed in the paper.