Tingfeng
Song
,
Huan
Tan
,
Saúl
Estandía
,
Jaume
Gàzquez
,
Martí
Gich
,
Nico
Dix
,
Ignasi
Fina
* and
Florencio
Sánchez
*
Institut de Ciència de Materials de Barcelona (ICMAB-CSIC), Campus UAB, Bellaterra 08193, Barcelona, Spain. E-mail: ifina@icmab.es; fsanchez@icmab.es
First published on 11th January 2022
The metastable orthorhombic phase of Hf0.5Zr0.5O2 (HZO) can be stabilized in thin films on La0.67Sr0.33MnO3 (LSMO) buffered (001)-oriented SrTiO3 (STO) by intriguing epitaxy that results in (111)-HZO oriented growth and robust ferroelectric properties. Here, we show that the orthorhombic phase can also be epitaxially stabilized on LSMO/STO(110), presenting the same out-of-plane (111) orientation but a different distribution of the in-plane crystalline domains. The remanent polarization of HZO films with a thickness of less than 7 nm on LSMO/STO(110) is 33 μC cm−3, which corresponds to a 50% improvement over equivalent films on LSMO/STO(001). Furthermore, HZO on LSMO/STO(110) presents higher endurance, switchable polarization is still observed up to 4 × 1010 cycles, and retention of more than 10 years. These results demonstrate that tuning the epitaxial growth of ferroelectric HfO2, here using STO(110) substrates, allows the improvement of functional properties of relevance for memory applications.
Epitaxial growth can facilitate the strain control of functional properties. This is demonstrated for perovskite ferroelectrics on perovskite substrates,16,17 but the epitaxial growth of fluorite ferroelectrics on perovskite substrates is much less flexible. Probably because of the large structural difference between HfO2 and perovskite substrates, epitaxy occurs by the so-called domain matching epitaxy mechanism,18 which results in a non-significant change of the lattice parameters of HfO2 by changing the lattice mismatch via appropriate substrate selection.10,14 Furthermore, the orthorhombic phase grows epitaxially on LSMO and other La-doped manganites, but not on other popular oxide electrodes such as SrRuO3, LaNiO3 or La:BaSnO3.19 These facts limit the possibilities of controlling the microstructure and the ferroelectric properties of epitaxial HfO2 films. Aiming to open new ways of control, we have investigated the growth of Hf0.5Zr0.5O2 (HZO) films on STO(110) substrates buffered with LSMO electrodes. The films are epitaxial and (111) oriented like equivalent films on STO(001), but they present a different set of in-plane crystal variants. We observe that this different crystalline microstructure has an impact on functional properties, i.e. ferroelectric polarization is very high around 33 μC cm−2 in films of less than 7 nm thickness, representing an increase of 50% over equivalent films on STO(001). Films grown on STO(110) can be cycled with reduced electrical strength because of the higher polarization, which allows the enhancement of the endurance up to 4 × 1010 cycles.
Crystal phase identification and epitaxy characterization were performed by X-ray diffraction (XRD) with Cu Kα radiation using a Siemens D5000 diffractometer equipped with a point detector, and a Bruker D8-Advance diffractometer equipped with a 2D detector. Microstructural characterization was performed by scanning transmission electron microscopy using a JEOL ARM 200CF STEM with a cold field emission source, equipped with a CEOS aberration corrector and operated at 200 kV. Film topography was analyzed using atomic force microscopy (AFM) using a Keysight 5100 analyzer. Ferroelectric characterization was carried out using an AixACCT TFAnalyser2000 platform. Polarization loops were measured at 1 kHz by the dynamic leakage current compensation (DLCC)21 procedure at room temperature in top-bottom configuration, with the bottom LSMO electrode grounded. Residual leakage and series resistance contributions were subtracted using the reported methodology.22 Endurance was measured at room temperature by cycling the sample at a frequency of 100 kHz using bipolar square pulses of indicated amplitude and measuring polarization loops at 1 kHz. Retention was measured at 85 °C, poling the sample with a triangular pulse of 0.25 ms and determining the remanent polarization from the first polarization curve of the polarization loop measured at 1 kHz using the positive up-negative down protocol after a delay time. Piezoelectric force microscopy (PFM) measurements were performed with an MFP-3D microscope (Oxford Instrument Co.) using the BudgetSensors silicon (n-type) cantilevers with the Pt coating (Multi75E-G). To achieve better sensitivity, the dual AC resonance tracking (DART) method was employed.23 PFM voltage hysteresis loops were always performed at remanence at a dwell time of 100 ms. Due to the difficulties in quantifying PFM response in DART, piezoelectric coefficients have not been evaluated, but all the measurements in the characterized samples were performed under the same conditions, making them comparable.
Epitaxy has been studied by XRD using a 2D detector. Fig. 2a shows the reciprocal space map (RSM) around the asymmetric STO{111} reflections for the HZO/LSMO/STO(001) sample. The map was obtained from ϕ-scans with a 2D detector, recorded at ϕ angles ranging from 0° to 360°, with a step Δϕ = 1°. In addition to the STO{111} spot, the integrated frame shows the presence of a spot that corresponds to o-HZO{11−1} reflections (Qx = 0.316 Å−1, Qz = 0.109 Å−1) and a lower intensity spot of o-HZO{200} reflections (Qx = 0.316 Å−1, Qz = 0.224 Å−1). The presence of both spots at the same ϕ is due to 180° twinning. The diffraction spots appear at specific ϕ angles, as can be seen in Fig. 2b. Fig. 2b shows a Qz − ϕ map obtained by slicing the RSM around Qx = 0.316 Å−1 (ΔQ = ±0.01 Å−1). The integrated region corresponds to the region enclosed by a dashed rectangle in Fig. 2a. There are 12 spots, 30° apart, corresponding to both o-HZO{11−1} and o-HZO{200} reflections. The pole figure of o-HZO{11−1} reflections shown in Fig. 2c allows us to visualize more clearly the presence of four families of in-plane crystal variant consequences of the epitaxy of o-HZO(111) (3-fold symmetry surface) on LSMO/STO(001) (4-fold symmetry surface).18 The four families are indicated by colored triangles. Similar measurements for the HZO/LSMO/STO(110) sample are presented in Fig. 2d–f. Fig. 2d shows the RSM around the asymmetric STO{011} reflections. In addition to the substrate spot, there are o-HZO{11−1} (Qx = 0.313 Å−1, Qz = 0.115 Å−1) and o-HZO{200} reflections (Qx = 0.313 Å−1, Qz = 0.225 Å−1). The Qz − ϕ map (Fig. 2e) shows 12 o-HZO{11−1} and 12 o-HZO{200} spots. The 12 spots are distributed in 6 sets of 2 spots, each set is about 60° apart and the two spots in each set are 8.5° apart. Other heterostructures with complex domain structures have shown similar sets of peaks in ϕ-scans.25,26 The pole figure shown in Fig. 2f proves the angular distribution of o-HZO{11−1} reflections. The pole figure indicates that there are four HZO crystal domains (Fig. S3, ESI†), corresponding to the following epitaxial relationships: [−211]HZO(111)//[−112]STO(110), [−211]HZO(111)//[1−12]STO(110), [2−1−1]HZO(111)//[−112]STO(110), and [2−1−1]HZO(111)//[1−12]STO(110).
To better understand the epitaxy, we have conducted cross-sectional STEM characterization along the [001] zone axis of STO. Fig. 3a shows an inverse-intensity annular bright field (ABF) image where the HZO and LSMO layers and the STO substrate are visualized (horizontal yellow arrows mark the position of the interfaces). The original ABF image and simultaneously acquired high-angle annular dark field images are shown in Fig. S4, ESI.†Fig. 3a shows that HZO is orthorhombic, and different crystal variants were observed (vertical yellow arrows mark the position of boundaries between variants). There is absence of the monoclinic phase or other HZO polymorphs in Fig. 3a, although monoclinic grains were occasionally observed in other regions of the STEM specimen. Thus, the HZO film on LSMO/STO(110) is almost purely orthorhombic. This is remarkable since equivalent films on LSMO/STO(001) show the coexistence of orthorhombic and monoclinic phases.14 The rectangular area marked with a dashed yellow line is shown in an enlarged view in Fig. 3b. This image confirms the high crystalline quality of HZO and shows a well-defined (semi) coherent HZO/LSMO interface. The lattice mismatch, around −11%, is too high for conventional epitaxy. Indeed, the comparison of LSMO(110) and HZO(111) unit cells does not reveal direct matching (Fig. S5, ESI†). Domain matching epitaxy recently observed in the high-mismatch growth of HZO on LSMO(001),18 is also expected to occur in films grown on LSMO(110). To confirm this, we obtained reconstructed STEM images by filtering in the Fourier space (Fig. S6, ESI†). The analysis confirms the presence of extra planes at the HZO/LSMO interface with the periodicity anticipated in domain matching epitaxy. Finally, we note that HZO is (111) oriented on both LSMO(001) and LSMO(110) surfaces. This suggests the relevance of the surface energy contribution in addition to the interface energy. On the other hand, atomic force microscopy measurements confirm that the film on STO(110) is very flat, with a root mean square roughness of 0.25 nm (Fig. S7, ESI†), which is comparable to the films on STO(001).10
The measurements of polarization loops confirm that the HZO films on LSMO/STO(001) and LSMO/STO(110) are ferroelectric (Fig. 4). The current–voltage curve (black line) of the film on STO(001) shows two ferroelectric switching peaks at −2.7 and +1.8 V. The average coercive field (EC) is EC = 3.4 MV cm−1 and there is an internal field of 1.3 MV cm−1, pointing from the upper Pt contact towards the lower LSMO electrode, as shown by PFM characterization. The polarization loop (Fig. 3b, black line) is well saturated and the remanent polarization (Pr) is 22 μC cm−2, similar to the reported values for films of comparable thickness.10Fig. 3 also shows the current–voltage curve and the polarization loop of the equivalent film on LSMO/STO(110) (red lines). In this sample, the coercive voltages are −2.3 and +1.9 V, corresponding to an average EC of 3.4 MV cm−1 and an internal field of 0.6 MV cm−1. The remanent polarization, Pr = 33 μC cm−2, is much higher than the usual values reported for HZO.2,3 Remarkably, Pr is 50% higher than the equivalent epitaxial film grown simultaneously on STO(001). The larger polarization is likely the consequence of the higher amount of the orthorhombic phase in the film on LSMO/STO(110). STEM characterization showed that it is almost purely orthorhombic (Fig. 3), while films on LSMO/STO(001) present an important amount of the parasitic monoclinic phase.14 Theoretical calculations predicted a spontaneous polarization of 52–55 μC cm−2 for orthorhombic HfO2.27,28 The projected polarization along the [111] direction of a pure orthorhombic phase film would be around 31 μC cm−2, which is similar to the measured remanent polarization. On the other hand, the HZO film on STO(110) presents a leakage current of about 5 × 10−7 A cm−2 at 1 MV cm−1, less than about 1 × 10−6 A cm−2 of the film on STO(001) (Fig. S8, ESI†).
Fig. 4 (a) Current–voltage curves of HZO/LSMO/STO(001) (black line) and HZO/LSMO/STO(110) (red line) samples, measured in the pristine state. (b) Corresponding polarization loops. |
Fig. 5 shows the endurance measurements of the films. The film on STO(001) (Fig. 5a) was cycled with sub-coercive field voltage pulses of 3.5 V (electric field E = 5.4 MV cm−1) having 2Pr = 24.2 μC cm−2 in the pristine state (solid blue circles). Wake-up effect is not observed. Polarization does not change significantly after 105 cycles, but additional cycles cause fatigue. The measurements were stopped when the polarization was as low as 2Pr = 3.3 μC cm−2 after 4 × 109 cycles. The current leakage (open blue circles in Fig. 5a) was constant up to 107 cycles and increased dramatically with additional cycles (see current–voltage curves in Fig. S8, ESI†). The test of endurance at 3 V (solid green diamonds) shows a similar dependence. The polarization in the pristine state is lower, but the fatigue after 105 cycles is less pronounced and the capacitor was cycled up to 1010 cycles (the test was stopped due to the low 2Pr at 3 μC cm−2, see polarization–voltage loops shown in Fig. S9, ESI†). The current leakage (open green diamonds in Fig. 5a) was constant during the test. The film on STO(110), switched at the same voltage of 3.5 V (Fig. 5b, solid black squares), has a larger initial polarization, 2Pr = 47 μC cm−2. It presents slight fatigue from the first cycles, being more pronounced after 105 cycles, and there was a breakdown after 2 × 109 cycles. The breakdown is likely due to the increased leakage after cycling (open black squares), which was larger than in the sample grown on STO(110) after 109 cycles. Reducing the electric field is essential to avoid the breakdown. The remanent polarization of the film on STO(110) is very high, and by applying a lower voltage of 3 V (E = 4.6 MV cm−1), the initial polarization, 2Pr = 30 μC cm−2, is greater than in the film on STO(001) cycled with 3.5 V pulses. The evolution of polarization with cycling (solid red triangles in Fig. 5b) is similar to that at a higher field, but in this case, there is no electrical breakdown up to 4 × 1010 cycles when 2Pr decreased to 4 μC cm−2 and the measurement was stopped. Polarization fatigue is the main factor that limits endurance in epitaxial HZO films. The robustness against breakdown is probably because of the fact that the current leakage (open red triangles in Fig. 5b), contrary to what was observed for the 3.5 V cycling voltage, does not increase significantly during cycling.
Fig. 6 shows the retention measurements of the HZO/LSMO/STO(110) sample, after poling a capacitor with pulses of amplitudes +3 V (solid triangles) and −3 V (open triangles). The measurements were performed at a temperature of 85 °C. Dashed red lines are fits to Pr = P0·τdn equation, where τd is the delay time and n a fitting parameter.29 The vertical black dashed line indicates a time of 10 years. The extrapolated polarization is high after 10 years for both positive and negative poling at 3 V, the poling amplitude that allows an endurance of 4 × 1010 cycles. Therefore, the film shows high polarization (2Pr = 30 μC cm−2), endurance (4 × 1010 cycles) and retention (more than 10 years) under the same poling voltage.
Fig. 6 Polarization retention measurements at 85 °C of HZO/LSMO/STO(110) for the positive and negative poling of 3 V. Lines are fits to Pr = P0·τdn equation. |
Footnote |
† Electronic supplementary information (ESI) available: Simulation of Laue oscillations. Piezoresponse force microscopy amplitude images. Sketches of the epitaxial relationships between crystal variants and substrates. STEM: simultaneous ABF and HAADF images. Top view of HZO(111) on LSMO(110). STEM characterization (reconstructed image from reflections in the Fourier space). Atomic force microscopy image of the HZO/LSMO/STO(110) sample. Leakage curves during endurance tests. Polarization loops during endurance tests. See DOI: 10.1039/d1nr06983g |
This journal is © The Royal Society of Chemistry 2022 |