Direct Nanoscale Analysis of TemperatureResolved Growth Behaviors of Ultrathin Perovskites on SrTiO3 Young Jun Chang† and Soo-hyon Phark*,‡,∥,§ †
Department of Physics, University of Seoul, Seoul 02504, Korea Center for Correlated Electron Systems, Institute for Basic Science, Seoul 08826, Korea ∥ Department of Physics and Astronomy, Seoul National University, Seoul 08826, Korea ‡
S Supporting Information *
ABSTRACT: Revealing growth mechanism of a thin film and properties of its film−substrate interface necessarily require microscopic investigations on the initial growth stages in temperature- and thickness-resolved manners. Here we applied in situ scanning tunneling microscopy and atomic force microscopy to investigate the growth dynamics in homo- (SrTiO3) and hetero- (SrRuO3) epitaxies on SrTiO3(001). A comparison of temperature-dependent surface structures of SrRuO3 and SrTiO3 films suggests that the peculiar growth mode switching from a “layer-by-layer” to “step-flow” type in a SrRuO3 films arises from a reduction of surface migration barrier, caused by the change in the chemical configuration of the interface between the topmost and underlying layers. Island densities in perovskite epitaxies exhibited a clear linear inverse-temperature dependence. A prototypical study on island nucleation stage of SrTiO3 homoepitaxy revealed that classical diffusion model is valid for the perovskite growths. KEYWORDS: perovskite, SrTiO3, scanning tunneling microscopy, growth mode, island nucleation
S
STO (001) is known to be maintained up to a thickness of 17 unit cells (uc).27−31 However, SrRuO3 shows a peculiar growth mode change from a “layer-by-layer” (LBL) type to a “stepflow” (SF) type on STO(001) at very initial growth regime, whereas the STO homoepitaxy shows only LBL growth over a wide temperature range.32 This suggests the possibility that surface termination switching during the initial growth of SrRuO3 on STO(001)16 contributes to the growth mode change. Moreover, to serve as a solid basis for investigating lateral complex oxide nanostructures, an extended study on surface diffusion dynamics in their growth stages is required. The design of nanoscale structures of metals and semiconductors33−35 has been successfully achieved based on the application of the microscopic model of surface diffusion dynamics, e.g., mean field nucleation (MFN) theory,36−38 which requires thermodynamically stable configurations of island nucleation. Recently, ‘diffusion-controlled growth’37,39 was utilized to reach the island nucleation stage in STO homoepitaxy.24 However, a quantitative survey of surface diffusion in the perovskite growth requires further effort,
tudies on surfaces and interfaces of oxide materials pave a path toward exciting and intriguing physical phenomena that originate from strongly correlated electrons under vertical or lateral confinements.1−5 Investigation of such phenomena and their technological applications requires an understanding of and control over film growth at the atomic scale. Low-dimensional (LD) complex oxide structures, perpendicular to the film plane, e.g., multilayers and superlattices, have been widely studied.6−14 For in situ monitoring of film growth, diffractive methods such as reflection high-energy electron diffraction (RHEED) and surface X-ray diffraction have been used extensively.15−18 However, LD lateral structures of complex oxides have been studied only recently.19−25 Due to the restricted lateral resolution of the diffraction methods and environmental sensitivity of the oxide surfaces, it is necessary to adopt in situ microscopic methods to study the individual lateral nanostructures. Recent scanning tunneling microscopy (STM) studies on perovskite growths on SrTiO3 (STO) (001) revealed that the initial growths are strongly influenced by substrate surface reconstructions (RCs),26 which usually arise not simply from structural changes but also, especially in oxides, accompanies distinct surface chemical configuration from the bulk counterpart. For instance, the misfit strain (−0.6%) in SrRuO3 film on © XXXX American Chemical Society
Received: March 4, 2016 Accepted: May 10, 2016
A
DOI: 10.1021/acsnano.6b01592 ACS Nano XXXX, XXX, XXX−XXX
Article
www.acsnano.org
Article
ACS Nano
Figure 1. Growth temperature TG dependences of island formation patterns in submonolayer SrTiO3 on SrTiO3(001) surfaces. (Left side) A schematic illustration of the experimental design. (a−c) STM images of 0.25-ML SrTiO3 on SrTiO3(001) for increasing three TGs. (d−f, g−i) STM images of SrTiO3(001) surfaces annealed at increasing three temperatures in oxygen partial pressure PO2 of 10−2 Torr (10−9 Torr). All STM images are measured at bias voltage VB = 2.5 V and tunneling current Iset = 50 pA. The TG is indicated at the bottom left corner of each image. All horizontal scale bars are 100 nm long. Color scales of all images refers the vertical scale bar at the left side of (g).
quenching the sample temperature down to 300 °C within 1 min and forcing the surface structures of the sample to be frozen in the state they were in just before the time of power shutoff. The sample was then transferred into the SPM chamber. STM and AFM measurements were performed after waiting for 1 h. Strontium Titanate Homoepitaxy. The growth of STO in a sub-ML regime showed monotonic expansion of single uchigh islands (adatom islands) for increasing TG. Figures 1a−c show STM images of 0.25-ML STO films on STO(001), grown at three distinct TG values, spanning 100 °C. All of the islands displayed heights of ∼0.4 nm,24,26 approximately the same size as 1 uc of STO (0.3905 nm); this implies that the STO grows in the form of the LBL mode for all three TGs. As TG increased, the average island size became larger, and the island density decreased. Surface diffusion in oxide growth is strongly influenced by oxygen partial pressure PO2.32,42 To obtain microscopic insight into this issue, we studied the temperature-dependent evolution of the local surface structures of the STO(001) substrate. As the TiO2-STO surface was annealed, one uc-deep pits, so-called “pit islands”, formed. Figure 1d−f is STM images of STO(001) surfaces annealed at a PO2 of 10−2 Torr (O2-annealed), a PO2 for the STO film growth shown in Figure 1a−c, at three annealing temperatures (TAs). Pit islands were visible and became monotonically wider for increasing TA. Above 900 °C, the STO surface was pit-free. Figure 1g−i shows the STM images of STO(001) surfaces annealed at PO2 of 10−9 Torr (UHV-annealed) at three TAs. In comparison with the O2annealed samples, we note that the UHV-annealed samples reach a final state of similar pit size and density, in a TA range that was ∼60 °C higher; this supports the conjecture that either
based on more elaborate microscopic observations of temperature-resolved growth behaviors. In this work, we performed temperature-resolved STM and atomic force microscopy (AFM) studies to investigate the initial growth of STO and strontium ruthenate (SRO) films, within a few monolayers (MLs) on TiO2-terminated STO(001) (TiO2-STO) surfaces. Direct observations of surface termination conversion as well as of growth modes in the sub-ML regime of SRO growth on STO(001) were carried out; these phenomena are believed to drive the above-mentioned growth mode switching. We then introduced a quantitative approach to diffusion characteristics in complex perovskite oxide growth, based on microscopic data. Using an Arrhenius analysis of subML STO/STO(001) films combined with the MFN theory, we estimated the surface diffusion barrier in the STO homoepitaxy.
RESULTS AND DISCUSSION Experimental Design. To achieve simultaneous deposition of samples over a large growth temperature TG range, while holding all other growth parameters fixed, we employed the ‘temperature gradient method’40,41 in substrates of lateral size 2.5 × 10 mm2. Direct flow of current along the direction of the long edge of the crystal allowed for a temperature difference of over 200 °C (temperature gradient ∼20 °C/mm) between two of the crystal’s edges, due to different contact resistances (RC1,RC2) with the current electrodes, as shown in the temperature profile in Figure 1. Temperature measurement using an optical pyrometer of spot size ∼0.5 mm provided spatial temperature resolution of ∼10 °C. Following growth of the films, the samples were held at TG for more than 3 min, allowing the sample surface adatom islands to migrate enough that the surface structures reached an equilibrium configuration at the given TG. The heating power was then disconnected, B
DOI: 10.1021/acsnano.6b01592 ACS Nano XXXX, XXX, XXX−XXX
Article
ACS Nano
Figure 2. Growth temperature TG dependences of surface structures of SrRuO3 on SrTiO3(001) surfaces. (a−c, d−f) STM images of 0.25-ML (2.5 ML) SrRuO3 on SrTiO3(001) for increasing three TGs (VB = 2.5 V; Iset = 50 pA). The TG is indicated at the bottom left corner of each image. All horizontal scale bars are 100 nm long. (g) Schematic diagrams for the diffusion dynamics in the growths of (top) STO and (middle) SRO on bare STO(001), and (bottom) SRO on fully SRO-covered STO(001). Lattice model of chemistry at the interface between the diffusing species of the uppermost layer and compactly filled underlying layer is illustrated at the right side of each case. Color scales of all images refers the vertical scale bar at the right side of (f).
STO and SRO molecular species may experience similar diffusion barrier landscapes, which could lead to similar growth modes. On the other hand, SRO film, which is thicker than 1 ML, is expected to serve a different chemical configuration to the adatom diffusion, as compared with the sub-ML SRO film. Note that the SRO film on STO(001) undergoes conversion of the termination layer from RuO2 to SrO at the initial growth stage.16,43 Thus, in view of the surface diffusion, the adatom migration in a SRO film thicker than 1 ML will experience a RuO2-SrO chemical configuration with the underlying layer, leading to a change in the diffusion barrier, compared with that of the sub-ML case. A comparison of the growth modes in the 0.25 ML (Figure 2a) and 2.5 ML (Figure 2d) film surfaces at TG = 700 °C leads us to conclude that the surface diffusion barrier is lowered due to termination switching (i.e., migration interface change). In addition, this may explain why the SRO film completely changes its growth mode only at the initial growth stage.16 The interface chemistry between the adatoms and underlying layers, therefore, plays a governing role in the observed thickness-dependent crossover of the growth mode as well as in the TG-dependent dynamic changes in the SRO epitaxy on STO(001). Intra-Island Structures of a Submonolayer Strontium Ruthenate. To obtain a deeper understanding of the conversion process of surface termination from RuO2 to SrO during the initial stage of the SRO growth, the microscopic structures of the individual islands in the sub-ML regime were examined. On oxide surfaces, STM contrast accounts not only for topographic information but also for the local electronic density of states, at a given bias voltage VB. However, AFM, which measures the force between the instrument tip and sample surface, is mostly sensitive to surface topography. Ru oxide is known to be so volatile that the SRO, besides SrRuO3, easily appears to be stoichiometric variants such as Sr2RuO4, Sr3Ru2O7, and so on, the so-called ‘Ruddlesden−Popper series’, depending sensitively on the growth conditions.47 To avoid such complications and keep the focus on the simple cubic perovskite phase SrRuO3, we introduced the growth conditions
atomic or molecular oxygen species assist in surface-layer diffusion.43−45 Strontium Ruthenate Films. Similar measurements were performed for the SRO adatom islands. Figure 2a−c shows STM images of 0.25-ML SRO, grown on STO(001) at three TGs over a range of 700−800 °C. Interestingly, similar island growth patterns of a height of 1 uc (∼0.4 nm; Figure 3) appear at a TG interval of ∼100 °C, shifted lower by ∼100 °C from the STO case. SRO also grows in the LBL mode at this coverage. On the other hand, SRO films of thickness >1 ML exhibited completely different surface structures. Figure 2d−f shows STM images of 2.5-ML SRO, grown on STO(001) at the same TGs as those in Figure 2a−c. The presence of flat terraces in the film at TG = 700 °C (Figure 2d) indicates growth in the SF mode, wherein the adatoms completely diffuse and subsequently merge to form step edges, which is desirable for the preparation of atomically flat SRO films.9,17,46 A further increase in TG accommodates surface structures with more complex step boundaries. At TG = 750 °C, 1 uc-high islands appeared in the topmost layer (labeled as ‘1’); only a small amount of 2 uc-high layers were observed (labeled as ‘2’). At TG = 800 °C, the surface structure exhibited step bunching and islands that were multiple layers thick (i.e., three-dimensional island growth); the thickness of a specific region is labeled as ‘1’, ‘2’, and ‘3’ in uc. Theoretical surveys on the growth dynamics of SRO films predicted temperature- and thicknessdependent transitions among such growth modes depending on various parameters, such as misfit strain, terrace width, growth speed, and temperature;20,46 however, only for film thicknesses larger than a few tens of nm. Our observations in the ultrathin limit, including sub-MLs, require analysis on the atomic scale. To this end, we surveyed the interface chemistry between adatom islands and their underlying layer, which plays a key role in determining microscopic motion in surface diffusion. We schematically summarize the cases in this study in Figure 2g. SRO shares the same A-site cation (Sr) with STO, but its Bsite cation (Ru) is different. In the sub-ML growth regime, the TiO2-STO surface provides the same chemical configuration, i.e., SrO-TiO2, at the interfaces between the adatom islands and the substrates for both STO and SRO on STO(001). Both C
DOI: 10.1021/acsnano.6b01592 ACS Nano XXXX, XXX, XXX−XXX
Article
ACS Nano
Figure 3. Intra-island structures of a submonolayer SrRuO3 on SrTiO3(001). (a−c) AFM (a) and STM (b and c) images of 0.5-ML SRO film on STO(001). Note that the STM images in (b) and (c) are measured with VBs of 3.5 and 2.7 V, respectively (Iset = 50 pA). (d−f) Height profiles along the lines in the images a (for d), b (for e), and c (for f). (g) Hard sphere model of a submonolayer SRO on STO(001) substrate. α−δ indicate the four subsequent half-uc unit terminations from the TiO2-terminated STO(001) surface (α).
that have successfully resulted in high quality SrRuO3 films on STO(001) in the present growth chamber.9 Figure 3a shows an in situ AFM image of 0.5-ML SRO on STO(001); Figure 3b,c shows STM images of the same sample measured at VB = 3.5 and 2.7 V, respectively. All three images show similar island and terrace structures, with slight differences in island perimeters. Figure 3d−f displays the profiles along the lines in Figure 3a−c, respectively. The AFM image (Figure 3d) of the SRO islands shows three distinct height levels (β−δ) distinguished by 0.2 nm-high steps, increasing from the substrate level (α); whereas the STM image at VB = 3.5 V (Figure 3e) shows only the islands of 0.4 nm in height. Remarkably, the STM image taken at the same sample area, but at VB = 2.7 V (Figure 3c), shows depressions of 1−2 Å inside and around the island edges (δ and β in Figure 3f). Because both SrO and RuO2 sublayers of SRO have thickness ∼0.2 nm, each 0.2 nm-high step in the AFM contrast can be assigned as either an SrO or RuO2 sublayer. Considering the much lower local electrical conductivity of SrO compared with RuO2,48−51 the regions of the depressions appearing in the STM image of the lower VB (Figure 3c) arise from a reduced electronic density due to the SrO capping layer. At the higher VB (Figure 3b), an increased electronic contribution of the SrO layer to the STM tunneling current results in a higher STM contrast in β and δ regions, compared with those shown in Figure 3c. Hence, we assigned the regions of α−δ to TiO2, SrO, RuO2, and SrO, respectively, as schematically indicated in Figure 3g. Ru oxides are known to be much more volatile than the other sublayer components of this heterosystem at the TG range of the SRO film.52,53 This could result in shrinkage of the RuO2 layer with respect to the underlying SrO layer. The distinct line profiles and schematic model shown in Figure 3 clearly reveal that termination switching from RuO2 to SrO occurs, even from sub-ML coverage of SRO, and, furthermore, that the SrO capping layer may stabilize the RuO2 layer. Additionally, alongside SrO-RuO2-SrO layered regions, the SrO perimeters and islands show lateral growth of a
single layer in height (∼0.2 nm), i.e., LBL growth. This implies that the similarity in growth modes between the sub-MLs of SRO and STO may arise from the chemical configuration of the SrO−TiO2 interface, independent of any internal island structure. Further studies using different terminations (SrOterminated STO(001) or other perovskite substrates)15−17,21 with full span of various growth parameters20 would provide more comprehensive microscopic understanding of the SRO growths. Arrhenius Analysis of TG Dependences of Strontium Titanate Homoepitaxy. Arrhenius plots of the island density n with respect to TG have been successfully applied to extract energy barriers governing the surface diffusion of metal film growth.37,39 Figure 4a shows the TG dependences of n, on a logarithmic scale, for 0.25-ML STO (black), 0.25-ML SRO (gray), and O2-annealed (red) and UHV-annealed (blue) STO(001) surfaces, as extracted from the STM images shown in Figures 1 and 2. Note the clear linear dependence of ln(n) on the reciprocal of TG, in each case. This indicates that the surface diffusion in each case can be characterized by a single energy barrier ΔE and implies a simple exponential relationship between n and 1/TG as n ∼ exp(ΔE/TG). The linear fits (solid lines) of the data deduce ΔE values of 1.7 ± 0.03 eV for the STO adatom islands (black) and 2.8 ± 0.03 (3.5 ± 0.15) eV for the O2-annealed (UHV-annealed) STO pit islands. The classical MFN theory36 predicts (eq 1) for stable twodimensional islands with complete condensation: n ∼ exp[(iEm + E i)/(i + 2)kBT ]
(1)
where Em, kB, and T represent the surface migration barrier, the Boltzmann constant, and temperature, respectively. The i and Ei in eq 1 represent the critical nucleus, specifically, the size i of the island that can be stabilized by adding just one more monomer and its internal bonding energy Ei. We apply eq 1 to the data in Figure 4a. Assuming that the critical nucleus can be represented by a single uc of STO (Figure 4d; Supporting Information, note 1), eq 1 simplifies to n ∼ exp(Em/3TG) (i = D
DOI: 10.1021/acsnano.6b01592 ACS Nano XXXX, XXX, XXX−XXX
Article
ACS Nano
Figure 4. Arrhenius analysis of TG dependences of SrTiO3 homoepitaxy. (a) Island densities in log scale as a function of the inverse TG, 1/TG, extracted from STM images in Figures 1 and 2 (black: 0.25-ML STO, red: STO(001) annealed at PO2 = 10−2 Torr, blue: STO(001) annealed at PO2 = 10−9 Torr, gray: 0.25-ML SRO). The solid lines are the linear fits. (b, c) A high-resolution STM image (b) of the smallest stable STO nanoisland (ref 21) and height profile (c) taken along the line A. (d) An illustration of the square-shaped 1.6 × 1.6 nm2 STO island shown in (b), sectionalized by the STO unit cell size with the depiction of the uc structure. (e, f) High-resolution STM images of the various size STO nanostructures formed in the surface of the 2 pulses sample (ref 21), showing the bidirectional growth of 4 uc width. (g) Number distribution vs island size in uc for the 1 pulse (blue) and 2 pulse (red) samples. The island number was counted for an area of 100 × 100 nm2. All STM images are measured with VB = 2.5 V and Iset = 50 pA.
1, Ei = 0). The ΔEs obtained from the linear fits of our data result in Em values of 5.1 ± 0.1 eV for the STO adatom islands and 8.3 ± 0.1 eV (10.5 ± 0.5 eV) for the O2-annealed (UHVannealed) STO pit islands. These values deviate greatly from the previously reported Em range of 0.3−3.8 eV for the STO homoepitaxy.43,54,55 At TG < T*, where T* represents the characteristic temperature below which the island density n no longer depends on TG,37,56−58 the governing growth mechanism falls into the so-called “kinetic limit”,35 due to suppressed diffusion. We estimate a T* value of 1920 ± 175 °C for Em = 5.1 eV (Supporting Information, note 2). Note that the TG values in this study lie well below the above-estimated T*, contradicting our observations of LBL growth and the clearly observed linear dependence of ln(n) on 1/TG. Hence, our fundamental assumption on the nature and configuration of the critical nucleus i in the perovskite oxide epitaxies must be questioned. To obtain insight into the critical nuclei i, we utilized ‘diffusion-controlled growth’37,39 for the STO homoepitaxy.24 We prepared two samples distinguished by the number of laser pulses in the pulsed laser deposition (PLD) growth: one with one laser pulse (1-p) and the other with two laser pulses (2p)24 (Supporting Information, figures). Various types of nanostructures were observed, with heights of STO 1 uc (∼0.4 nm) and of different shapes and sizes (Figure 4b,e−h). Figure 4b shows a high-resolution STM image of the smallest observed nanostructure type: a square with edges aligned in the two primary directions of the STO(001) surface. The lateral size of the structure was measured to be ∼1.6 nm (Figure 4c),
corresponding to four times the size of 1 uc of STO, identifying its lattice configuration to be 4 × 4 uc2 of the STO (Figure 4d). We also statistically determined the smallest island in the nucleation stage of STO homoepitaxy by counting the numbers of islands observed in the STM images of the 1-p and 2-p samples, with respect to their lateral size. We did not find any island of a size smaller than 4 × 4 uc2, as shown in the histogram of Figure 4i. Assuming the 4 × 4 uc2 species to be the smallest stable island leads to i = 15, which results in Ei = 22Eb, where Eb is the single uc−uc bonding energy between a pair of STO molecules. Utilizing the MFN model, we deduced the TG dependence of the island density to be n ∼ exp[(37/17) Em/kBTG] with the assumption that Em is equal to Eb in homoepitaxial cases (Supporting Information, note 3). The slope (ΔE) of the linear fit shown in Figure 4a gives an Em value of 0.87 ± 0.02 eV for the STO adatom islands, which lies in the previously reported Em range of 0.3−3.8 eV43,54,55 and is much larger than those of the metallic systems.37,56−58 This is expected, due to the fact that the LBL growth occurs at TG values that are much higher than those of the metallic systems. Two features in the Em values of our studies elicit further discussion: (1) The excessively wide distribution of Em values (0.3−3.8 eV)43,54,55 and the relatively small Em observed in this study, compared with previous reports. One possible explanation is that a considerable contribution of the kinetic energy Ek of the particles in the PLD process, which effectively lowers migration barrier. A variety of growth factors such as PO2 level, target-substrate distance, and laser fluence determine the Ek contribution to the effective (measured) Em of the diffusing E
DOI: 10.1021/acsnano.6b01592 ACS Nano XXXX, XXX, XXX−XXX
Article
ACS Nano
248 nm with energy density of 3 (2.5) J/cm2 at the target surface.9 Coverage of a film was determined from the RHEED oscillations (for film thickness >1 ML) and STM image (for film thickness
