Abhijeet L.
Sangle
*a,
Oon Jew
Lee
b,
Ahmed
Kursumovic
a,
Wenrui
Zhang
c,
Aiping
Chen
d,
Haiyan
Wang
e and
Judith L.
MacManus-Driscoll
*a
aDepartment of Materials Science and Metallurgy, University of Cambridge, UK. E-mail: sangle.abhijeet@gmail.com; jld35@cam.ac.uk
bSchool of Fundamental Science, Universiti Malaysia Terengganu, 21300 Kuala Terengganu, Malaysia
cCenter for Functional Nanomaterials, Brookhaven National Laboratory, Bldg. 735 – P.O. Box 5000, Upton, NY 11973-5000, USA
dCenter for Integrated Nanotechnologies (CINT), Los Alamos National Laboratory, Los Alamos, NM 87545, USA
eSchool of Materials Engineering, Purdue University, West Lafayette, IN 47907, USA
First published on 5th February 2018
We report on nanoengineered SrTiO3–Sm2O3 nanocomposite thin films with the highest reported values of commutation quality factor (CQF or K-factor) of >2800 in SrTiO3 at room temperature. The films also had a large tunability of dielectric constant (49%), low tangent loss (tanδ = 0.01) and a Curie temperature for SrTiO3 > 300 °C, making them very attractive for tunable RF applications. The enhanced properties originate from the unique nanostructure in the films, with <20 nm diameter strain-controlling Sm2O3 nanocolumns embedded in a SrTiO3 matrix. Very large out-of-plane strains (up to 2.6%) and high tetragonality (c/a) (up to 1.013) were induced in the SrTiO3. The K-factor was further enhanced by adding 1 at% Sc3+ (acceptor) dopant in SrTiO3 to a value of 3300 with the tangent loss being ≤0.01 up to 1000 kV cm−1.
(1) |
Ferroelectrics can offer an excellent alternative to the conventional semiconductor-based tunable microwave devices.2,6 They show many potential benefits such as high tunability, low loss, high tuning speed, small leakage currents, low power consumption, radiation hardness and high breakdown electric field, making them very attractive for use in communications.2,7,8 Over the last five decades or so, ferroelectrics have been studied in different forms such as pure or doped,6,9 composites or single-phase systems10 and bulk or films.11 Sometimes, these different forms have been combined, e.g., doped ferroelectric containing composite films.12 Out of these different forms, bulk materials face major drawbacks such as the need of high tuning voltages and large sizes, which are not ideal for the increasingly smaller communication devices.11 These disadvantages can be removed by growing ferroelectric thin films, wherein the required high electric field is generated at a fraction of the voltage needed for the bulk devices. Their small size also gives the possibility of integrating them into conventional microwave technology.11 Ferroelectric perovskites such as PbZrxTi1−xO3, BaTiO3 and BaxSr1−xTiO3 (BSTO) are widely used for tunable microwave devices. However, they face several challenges. Lead-based ferroelectrics such as PbZrxTi1−xO3, PbTiO3, PbxLa1−xZryTi1−yO3 are increasingly discouraged by the environmental agencies owing to the toxicity of Pb.13,14 BaxSr1−xTiO3 (BSTO) is beneficial over BTO because it has a relatively low loss while maintaining a high tunability,15 but it does not exhibit both these optimised properties at the same time. Indeed, BSTO has a broad ferroelectric-paraelectric phase transition over hundreds of degrees Celsius, deteriorating its tunable microwave performance.16 It has been suggested that the heterogeneity associated with the chemical substitution of Sr2+ in BaTiO3 lattice is responsible.16 Moreover, BSTO usually possesses a very high dielectric constant.17,18 This is not suitable for impedance matching,6,19–21 high power phase shifters, or high power accelerating structures with dielectric loading,22 where low dielectric permittivity is needed in addition to high tunability and low loss. It has been estimated that the dielectric constant (relative permittivity (εr)) should be <500 for impedance matching purposes.20,21
On the other hand, the incipient ferroelectric SrTiO3 (STO) can offer a viable solution to the problem. STO has low permittivity (εr = 300 for bulk STO at room temperature23). However, owing to its bulk ferroelectric Tc of <100 K, STO is useful only for low temperature applications.24,25 At room temperature, STO has low tunability24 because here it is a cubic non-ferroelectric, rather than a highly tetragonal ferroelectric which is needed to obtain high tunability.16,24,26 If high tetragonality can be induced in STO at room temperature its ferroelectric properties can be maintained. Indeed, Haeni et al. achieved room temperature ferroelectricity in STO by epitaxial straining of 70–80 nm thin film by clamping it to a substrate.16 They achieved a very high tunability of 70% at room temperature, and at 10 GHz.27 However, the loss was high at 0.067, which is relatively high, yielding an overall CQF of ∼360.
The aim of this work is to produce a superior microwave tunable material based on STO. With the very strong benefit of much lower dielectric constant than analogue perovskite ferroelectrics, SrTiO3 has the potential to be an ideal microwave tunable material. Two challenges need to be overcome. First, since in ferroelectric materials the highest tunability is close to the Curie temperature, for operation at room temperature or higher, Tc needs to be enhanced to well above room temperature.16,28,29 Second, the loss must be kept low. An enhanced Tc in STO films of practical thickness has never been shown before and neither has reduced loss been shown in such films.
The novelty of the work lies in the engineering of STO thin films using a combined nanocomposite and doping approach, to give properties which cannot be achieved by any other method. As we show later, our films have the highest values of commutation quality factor ever reported at room temperature.
We use nanocomposite films to achieve high vertical tensile strain in the films independent of the substrate. Using the nanocomposite approach, there is no intrinsic limitation on strain control of thickness (with micron thick films possible). This is in strong contrast to standard, plain epitaxial films where strain is controlled by the substrate, at least for the first few 10's of nm, with gradual relaxation occurring above this up to ∼100–200 nm. As we have shown before for BTO/Sm2O3 films31 and BSTO/Sm2O3 films,26 the nanocomposite approach allows us to maintain ferroelectricity to a temperature well above room temperature in micron thick films. Strain is controlled in a direction perpendicular to the substrate using stiff Sm2O3 nanocolumns which grow by self-assembly in the STO matrix, with Sm2O3 substituting only minimally in the titanate ferroelectrics.30 Sm2O3 is a low-loss, low permittivity, passive dielectric material which does not show any tunability with electric field. The tunability in the STO is expected to be controlled solely by the strain induced in it by the presence of the stiff Sm2O3.26
Achieving low loss in STO could be very challenging, because of possible formation of oxygen vacancies while straining, and yet to achieve a high K-factor, it is very important to pay attention both to increasing η and lowering loss. To address the potential loss problem from oxygen deficiency, we explored acceptor-doping of the lower valent ion (Sc3+) onto the Ti4+ site. Leakage has been shown to be lower in such SrTiO3 and BaxSr(1−x)TiO3 films than in standard plain films.9,32
The phases in the film are both aligned along (001), with STO being aligned in-plane, cube-on-cube, and the Sm2O3 being rotated by 45° in-plane, both as expected.26,31 The X-ray information on the films is shown in Fig. S1a and S1b in the ESI.†
Fig. 1c shows an X-ray reciprocal space map around the STO (113) peak. Here, the SrTiO3 film peak is clearly separated from and lower than the STO substrate peak along Qz (i.e., in the out-of-plane direction). On the other hand, the film peak is nearly fully in-plane strained to the substrate, as seen by the very close overlap of the Qx peak positions. Thus, the strong vertical strain-controlling effect of the stiff Sm2O3 scaffold nanopillars on the SrTiO3 film matrix is evident. The nanocolumns of Sm2O3 dispersed in the SrTiO3 matrix can be identified from the atomic force microscope (AFM) image shown in Fig. 1d. The films are very smooth, the root-mean-square (RMS) roughness being around only 4.5 nm.
The structural and dielectric properties of the films with varying Sm2O3 content (from 0 to 100 wt%) in them are shown in Fig. 2.
The STO out-of-plane lattice parameter (c) increased from 3.905 Å for pure STO up to a maximum of 4.005 Å for 60 wt% Sm2O3 films (Fig. 2a). This behaviour can be attributed to greater interfacial area with increasing Sm2O3 in the film, hence availing more fraction of material at the interface for vertical straining.26 For high volume fractions of Sm2O3, the nanopillars will begin to be connected and the interfacial area with the SrTiO3 will reduce, thus explaining the reduction in out-of-plane lattice parameter above 60 wt% Sm2O3. Similar results have been obtained before.33 The in-plane lattice parameter (a) also increased from the pure film to the 60 wt% Sm2O3 film, albeit to a much lesser extent, from 3.905 Å for pure SrTiO3 to 3.916 Å for the 60 wt% Sm2O3 film (Fig. 2a). The increase in in-plane lattice parameter can be attributed to the restricted thermal contraction, upon cooling of the SrTiO3 matrix (because it is epitaxially pinned by the stiff Sm2O3 nanopillars) post-growth, leading to ‘auxetic-like’ behaviour as modelled previously for the Ba(1−x)SrxTiO3/Sm2O3 system.34 A maximum tetragonality of 1.023 was achieved here which is very high compared to 1.014 obtained by Haeni et al.16 by bi-axial in-plane straining of their films to substrates.
The tunability of the dielectric constant was computed from the capacitance vs. electric field curves. A small AC disturbance of 50 mV amplitude and electric field of 400 kV cm−1 were used. The dielectric tunability peaked with Sm2O3 content (Fig. 2b) coincident with the peak in tetragonality (Fig. 2a), underscoring the importance of tetragonality in determining the tunability of a ferroelectric material. The increased tunability with increasing tetragonality can be explained by the increased c lattice parameter. Hence, there is increased space along c (out-of-plane direction) for the vibration of Ti4+ ions in the oxygen tetrahedron of a SrTiO3 unit cell. The relative permittivity (εr) is directly related to the electrical susceptibility (χe = εr − 1), i.e. it reflects the ability of the material to polarise upon application of an external electric field. This tendency of the material to be polarised (i.e. electrical susceptibility) is highest at zero external dc bias field, as there is plenty of room for Ti4+ to be pushed from its central position in the direction of external electric field. With increasing c/a ratio, the electrical susceptibility of the material, at zero external dc bias field, increases, because of increasingly larger room available for Ti4+ to be pushed in the direction of applied external electric field. Similar behaviour has been observed by Hyun and Char35 and showed computationally by Antons et al.36
The inverse correlation of tunability and loss is a very beneficial aspect of the composite films, since normally these two parameters directly correlate.26 However, for the x = 0.6 sample, the loss values are rather high at 0.15. Fig. 2c shows capacitance vs. electric field data for the optimum composition film (60 wt% Sm2O3). The presence of butterfly-shaped loops confirms the presence of ferroelectricity.37 As expected, the relative permittivity of the columnar nanocomposite films decreased with increasing Sm2O3 content due to the lower permittivity of Sm2O3 (Fig. S3 in ESI†).
The K-factor follows the same dependence on Sm2O3 fraction (Fig. 2d), similar to the tunability relation (Fig. 2b). However, for a 400 kV cm−1 applied field, the highest K-factor achieved was 20. The K-factor and tunability further increased to 53 and 57%, respectively, upon increasing the maximum dc electric field to ∼1000 kV cm−1. While the tunability value is one of the best for SrTiO3, and also comparable to literature reports for other ferroelectric thin films,8,38 the K-factor is insufficiently high for practical applications, where >900 is a target value.6 Therefore, subsequent experiments were focussed on reducing the loss (and hence on enhancing K).
It has been reported that the interface between the electrode and the film plays a crucial role in deciding the loss, with smooth, epitaxial and highly crystalline interfaces being necessary for minimising the loss.9 The presence of disorder at the interface, even at the smallest scale, can cause significant loss39 as atomistic defects at the interface scatter the electric field.39 Even though the SrRuO3 electrode was smooth (RMS roughness ∼0.30 nm) and highly conductive (280 μΩ cm), and the nanocomposite film grown on it was highly epitaxial, the SrRuO3 film is granular (grains and grain boundaries) which means there will be more defects at the interface with the composite film.
Hence, the nanocomposite films were grown on conducting (001) Nb-doped single crystal STO (Nb-STO) substrates which also served as bottom electrodes, while at the same time giving a more perfect homoepitaxial interface. Hence, a 250 nm thick film of the optimum 60 wt% Sm2O3 composition was grown on Nb-STO. Fig. 3a shows polarisation-electric field (PE) loops. The lack of frequency dispersion in the PE loops rules out a relaxor character of ferroelectricity.40 The narrow shape of the PE loop can be attributed to finer lateral dimensions leading to formation of smaller domains.41,42
The variation of capacitance with electric field for the films is shown in Fig. 3b. From these curves, the calculated tunability was 33% and the K-factor was 1500 (at 1 MHz, 500 kV cm−1 and at room temperature). The tunability and K-factor further increased to 49% and 2800, respectively, for 1000 kV cm−1 (see Fig. S2b in ESI†). To the best of the authors’ knowledge, this is the highest ever reported K-factor at 1 MHz and at room temperature for SrTiO3 in any form – pure or doped, single phase or composite and bulk or thin/thick film. We see from Fig. 3b (and Fig. S2b†) that the loss tangent is roughly an order of magnitude less than what we have observed with the SrRuO3 bottom electrode, indicating the strong influence of the quality of the nanocomposite/electrode interface.
The polarisation vs. electric field (PE) loops (shown in Fig. S2a in ESI†) showed the ferroelectric nature of the films to at least 270 °C, after which the loops became too lossy to conclude whether the films were ferroelectric or paraelectric. The variation in relative permittivity (εr) and loss tangent (tanδ) (Fig. 3c and d, respectively) with temperature show peaks at 305 °C with no frequency dispersion of the peak positions with temperature, confirming no relaxor type ferroelectricity in these films.40 The results corroborate the high ferroelectric temperature determined from the PE loops. This is the highest ever reported Curie temperatures for SrTiO3.
Finally, to further reduce the loss in the ferroelectric films, Sc3+ doping of the SrTiO3 films was undertaken at the doping level of 1 at% on B-site. Sc3+ in place of Ti4+ is expected to produce acceptor doping to compensate electronic doping arising from the formation of oxygen vacancies. Indeed, Fe2+, Fe3+, Sc3+, Mg2+, Mn3+, etc. have previously been studied as dopants for loss reduction with varying degrees of success.43–45 The preparation of the Sc-doped SrTiO3 ceramic powder for the target is described in the methods section.
As shown in Fig. 4a, the Sc-doped SrTiO3 films were found to be ferroelectric with identical PE loops obtained for different frequencies. A tunability of 45% was obtained at 1000 kV cm−1 electric field (as determined from capacitance versus electric field plots, Fig. 4b and c). A loss tangent of ≤0.01 (Fig. 4b and c) at all electric fields. This gives a K-factor of 3300. The tunability is similar as for the undoped STO nanocomposite films. The lower loss of ≤0.01 across the whole field range, compared to the undoped film led to an increase in the K-factor to 3300. Table 1 compares the high frequency properties of our columnar nanocomposite films measured at room temperature with some of the best results reported in the literature.
Device | Tunable high frequency performance (η in %) | Frequency | Ref. |
---|---|---|---|
SrTiO3–Sm2O3 columnar composite films | η = 49, tanδ ≤ 0.02 | Measured at 1 MHz and room temperature | This work |
K = 2.8 × 103, εr = 135 | |||
(1 at%) Sc-doped SrTiO3–Sm2O3 columnar composite films | η = 45, tanδ ≤ 0.01 | Measured at 1 MHz and room temperature | This work |
K = 3.3 × 103, εr = 180 | |||
SrTiO3 thin film, molecular beam epitaxy (MBE) grown | η = 70, tanδ = 0.067 | Measured at 10 GHz and room temperature | 27 |
K(estimated) ∼ 3.6 × 102, εr ∼ 3000 | |||
Ba(1−x)SrxTiO3 thick film | η = 60, tanδ = 0.01 | Measured at 1 MHz and room temperature | 46 |
K ∼ 1.8 × 104, εr ∼ 270 | |||
Ba(1−x)SrxTiO3 thin film, PLD grown | η = 45, tanδ = 0.0057 | Measured at 1 MHz and room temperature | 47 |
K = 1.5 × 104, εr = unknown | |||
Mn-doped Pb(1−x)SrxTiO3 thin film, sol–gel grown | η = 70, tanδ = 0.03 | Measured at 100 kHz at room temperature | 48 |
K = 2.1 × 103, εr ∼ 1000 | |||
Ba(1−x)SrxTiO3–Bi1.5ZnNb1.5O7 (80–20) (probably) granular composite thin film, PLD grown | η > 90, tanδ = 0.007 | Measured at 1 MHz and room temperature | 49 |
K = 1.16 × 106, εr ∼ 200 | |||
Ba(1−x)SrxTiO3–Sm2O3 (25–75 wt%) columnar composite film, PLD grown | η = 75, tanδ < 0.01 | Measured at 1 MHz and room temperature | 26 |
K = 8–9 × 105, εr ∼ 4000 | |||
Bi1.5ZnNb1.5O7 – Mn-doped Ba(1−x)SrxTiO3 laminated composite film, PLD grown | η = 60, tanδ < 0.005 | Measured at 100 kHz and room temperature | 12 |
K = 9.0 × 104, εr ∼ 400 |
The columnar STO nanocomposite films of this work have the highest reported K-factor for STO. While better K-factors have been reported for BaxSr(1−x)TiO3 (BSTO)-containing films (Table 2), the SrTiO3 composite films are more suited for applications which require lower dielectric permittivity e.g. high power phase shifters or high power accelerating structures with dielectric loading or impedance matching in complex circuits.
Parameter | Value |
---|---|
Deposition temperature | 750–800 °C |
Oxygen gas pressure during the deposition | 0.2 mbar |
Laser fluence | ∼2.05 mJ cm−2 |
Laser pulse frequency | 1 Hz |
Deposition time | 25 minutes for (∼250 nm thick films) |
Post-annealing conditions | 650 °C, 400 mbar O2 partial pressure and annealing time of 60 minutes |
Finally, we note that the relative permittivity, εr, was around 30% higher for the Sc-doped samples. This 30% increase in εr is not significant when experimental error is taken into account (from run to run, film thickness, variation in Pt pad area, and difference in amount of remnant Ag paint on the substrate bottom, Ag being used to glue the substrate to the heater during growth).
The targets used for the SrTiO3 nanocomposite films were made by standard ceramic procedures. Pure SrTiO3 and Sm2O3 powders were mixed in the desired proportion and finely ground to create a homogenous mixture. This powder mixture was then uniaxially pressed under a >100 kN force and then sintered at ∼1100 °C for 6 hours. Sc-doped SrTiO3 powder was prepared by mixing SrCO3, TiO2 and Sc2O3 powders in the desired stoichiometric proportion followed by grinding and pressing, as before. These pellets were then calcined at ∼900 °C for ∼6 hours and reground again into fine powders and pressed and calcined again. The Sc-doped SrTiO3–Sm2O3 target was then pressed and sintered just as for the undoped SrTiO3 composite target above.
X-ray diffraction characterisation of the films was performed using a four-circle diffractometer and the data was fitted and analysed using X'pert HighScore Plus software suite. An atomic force microscope in tapping mode was used for the AFM studies. The microstructures were then characterized by (scanning) transmission electron microscopy (STEM) (FEI Tecnai G2 F20 operated at 200 kV) to collect the STEM figures used in this study. The polarisation hysteresis loops were acquired by employing Radiant Precision Premier which used a Sawyer-Tower circuit and an in-house made probe station. The impedance analyser Agilent 4294A was used to perform the capacitance, tangent loss vs. electric field and capacitance, tangent loss vs. frequency studies. An in-house-made heater coupled with the probe station was used to perform the high temperature measurements. An in-house written LabVIEW program was used to collect the data. The K-factor was usually computed from the dielectric constant and loss values for 0 kV cm−1 and negative maximum values for electric field (i.e., −400 kV cm−1, −500 kV cm−1, −1000 kV cm−1, etc.), as the loss was usually the lowest for negative electric fields.
To enable dielectric measurements, Pt sputter-coated top electrode pads of diameter ∼100 μm were deposited on films grown on SrRuO3/SrTiO3 or Nb-doped SrTiO3. A parallel-plate configuration was used for these measurements.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7nr06991j |
This journal is © The Royal Society of Chemistry 2018 |