Ferroelectric nanostructures Ionela Vrejoiu,a兲 Marin Alexe, Dietrich Hesse, and Ulrich Gösele Max Planck Institute of Microstructure Physics, Weinberg 2, 06120 Halle, Germany
共Received 16 July 2008; accepted 13 October 2008; published 9 February 2009兲 Device miniaturization poses not only technological and manufacturing challenges, it also requires the understanding of the phenomena occurring at nanoscale in condensed matter and at the involved interfaces. Herein, the authors present a summary of their results in fabricating and understanding the properties of ultrathin films and nanometer-size dots of ferroelectric perovskite and layered-perovskite oxides. © 2009 American Vacuum Society. 关DOI: 10.1116/1.3025907兴
I. INTRODUCTION Scaling sizes down to tens of nanometers or even below is appealing for device miniaturization, on one hand, and for exploring/exploiting novel phenomena arising from nanometer-size effects, on the other hand. Ferroelectric 共FE兲 materials, such as the renowned perovskites 共BaTiO3, PbZrxTi1−xO3, BiFeO3兲 and layered perovskites 关SrBi2Ta2O9, 共Bi, La兲4Ti3O12兴, have been proposed for nonvolatile ferroelectric random access memories since two decades.1 Ferroelectric perovskites can hold information in the form of the polarization direction of individual FE domains, because they usually have two thermodynamically equivalent groundstates of opposite ionic polarization. Additionally, the domain walls of FE thin films can be as thin as a few unit cells, which favors high-density data storage. Lateral dimensions of approximately 100 nm are required for FE dots in order to reach Gbit/ cm2 memory densities. Memory densities of more than 10 Gbit/ cm2 were achievable for BaTiO3 single crystals2 and 40 Gbit/ cm2 in an epitaxial PZT thin film.3 This is still much below the desirable memory density of 1 Tbit/ in.2 that one can readily get in ferromagnetic recordings. Hence, fundamental studies of the properties of FE nanostructures and ultrathin films are utterly necessary for overcoming the limiting factors that hinder the device optimization. Intense efforts have been devoted to the investigation of size effects in nanoscale ferroelectrics in the last decade. For the two-dimensional systems, namely, ultrathin FE films, the limit at which ferroelectric instability would disappear was established theoretically and experimentally to be only a few unit cells. Junquera and Ghosez calculated that BaTiO3 thin films between two metallic SrRuO3 electrodes in short circuit lose their ferroelectricity below a critical thickness of about 6 unit cells 共⬃2.4 nm兲: the reason behind was considered to be the depolarizing electrostatic field caused by dipoles at the ferroelectric-metal interface.4 Experimental investigations of complex oxide films that are only a few unit cells thick became possible due to recent advances in 共hetero-兲epitaxy, which facilitates the growth of single crystal films with very low density of extended structural defects. For real devices usually laterally structured films or patterned structures are a兲
Electronic mail: [email protected]
J. Vac. Sci. Technol. B 27„1…, Jan/Feb 2009
used in which the FE films might be only one 共active兲 part. This additional patterning process has major consequences for the ferroelectric properties. Patterning a complex material down to the nanoscale range is technically not a trivial task. Moreover, the volume of the patterned structure is now limited in all three dimensions and possible size effects are likely to be more severe than in the two-dimensional ultrathin film case. An additional degree of complexity arises from the measurement viewpoint. Assuming that it had been possible to fabricate successfully very small nanodots 共e.g., less than 10 nm in diameter兲, a detection method with resolution finer than 1 nm would be required to measure such dots with sufficient accuracy. Recently, the so-called scanning nonlinear dielectric microscopy 共SNDM兲 technique was developed and tested on a FE congruent lithium tantalate single crystal with the aim to investigate ultrahigh-density ferroelectric data storage.5
II. EXPERIMENTAL Epitaxial thin films and 3D nanodots of FE materials, such as PbZrxTi1−xO3 共PZT兲, SrBi2Ta2O9 共SBT兲, BaTiO3, and 共Bi, La兲4Ti3O12 共BLT兲, were fabricated by pulsed laser deposition6,7 共PLD兲 and chemical solution deposition.8 For PLD an excimer laser 共KrF, wavelength of 248 nm兲 was employed. Removable masks of self-assembled latex spheres,9 metal nanotube arrays,10,11 or nanoporous anodic aluminium oxide 共AAO兲 membranes12 were attached to the substrates for the PLD fabrication of ferroelectric nanodots and metalferroelectric-metal nanocapacitors. MgO, SrTiO3, and Nb: SrTiO3 single crystals with different orientations 关i.e., 共100兲, 共111兲, and 共011兲兴 were used as substrates for PLD and CSD growth of epitaxial FE thin films and nanodots, whereby epitaxial platinum or SrRuO3 films were used as bottom electrodes, when required. A schematic of the procedure how the free standing metal nanotube membranes are fabricated by using the nanoporous anodic aluminium oxide is given in Fig. 1, together with SEM images of various masks. The thin films and the nanodots 共NDs兲 were investigated by transmission electron microscopy 共TEM兲, high resolution TEM 共HRTEM兲, scanning electron microscopy 共SEM兲, atomic-force microscopy 共AFM兲, and piezoresponse scanning microscopy 共PFM兲, as well as by standard ferroelectric measurements 共TF Analyzer 2000 AixaCCT兲.
©2009 American Vacuum Society
Vrejoiu et al.: Ferroelectric nanostructures
FIG. 1. Schematic of the fabrication of metal nanotube membranes used as stencil masks for PLD of FE nanodots: 共1兲 Plasma sputtering of metal onto nanoporous Al2O3. 共2兲 Controlled electrochemical deposition 共ECD兲. 共3兲 Removal of nano porous Al2O3. The panel on the right shows representative SEM images of thus obtained gold nanotube membranes: 共a兲 Surface view, and 共b兲 a magnified view of 共a兲. 共c兲 Back surface view with a magnified image as inset and 共d兲 bent part of a membrane forming a ridge along the diagonal and allowing to view the 3D structure of the membrane.
FIG. 2. On the left a HRTEM micrograph of the interface between a PZT 52/ 48 nanostructure and the Nb:STO共100兲 substrate is shown, where an edge dislocation with a Burgers vector b = 具100典 is highlighted. The schematic of the “interface” between PZT and STO perovskite unit cells is displayed in the middle right. On the top right an SEM overview image of PZT nanoislands is given and the image on bottom right is the plan view TEM micrograph of such nanoislands displaying the network of misfit dislocations.
III. RESULTS AND DISCUSSION A. Structural defects in FE thin films and nanostructures
The effects of “going nano” become beneficial only when certain drawbacks are well understood and circumvented. The structural characterization is a key issue, because the role played by structural defects such as misfit 共MDs兲 and threading dislocations 共TDs兲 must be assessed carefully. The impact of the strain fields developed around MD cores can lead to FE polarization instabilities. Chu et al. performed a combined study by quantitative high-resolution electron microscopy and PFM on epitaxial PbZr0.52Ti0.48O3 nanoislands.8 Epitaxial PZT structures with different shapes and sizes were fabricated by chemical solution deposition 共CSD兲.13 共Further principle details regarding the formation of the nanostructures are given in Sec. III C兲. It was noticed that the misfit between the substrate and the grown structures influences the final shape. A large misfit usually results in small structures, whereas a small misfit generates irregular shapes and a larger size distribution. TEM and HRTEM analysis of the obtained structures showed a high crystal quality, a good thickness uniformity over a large area, and a narrow size distribution of the nanostructures. All structures have a truncated pyramid shape with atomically flat surfaces and are free from extended defects, except for interfacial misfit dislocations, which relax the strain caused by the misfit between the substrate and the epitaxial structure. In the case of PZT 52/ 48 islands on STO the distance between the dislocations is from 9 to 13 nm, in agreement with the calculated spacing, which is about 12 nm. This network of dislocations 共see bottom right plan view TEM image in Fig. 2兲 was proven to be detrimental to the ferroelectric properties.8 A region of high strain is present around the dislocation core 共Fig. 2兲, which extends into the surrounding volume of the JVST B - Microelectronics and Nanometer Structures
ferroelectric nanostructure. The analysis of such HRTEM images as the one in Fig. 2 revealed a strain of about 3.5%, which is almost the same value as the tetragonality of the PZT unit cell and which extends only into the PZT layer.8 Consequently the ferroelectricity in the highly strained volume around the dislocation core is negatively affected. If this disturbed volume is sufficiently large compared to the entire volume of the nanostructure, e.g., more than about 60%, the ferroelectric properties of the entire nanostructure are severely affected even leading to the disappearance of the ferroelectric properties. Avoiding the formation of misfit dislocations either by substrate engineering or by choosing an appropriate composition of PZT mitigates this effect and in certain cases will lead to the reappearance of ferroelectricity. This extrinsic size effect based on extended lattice defects is not particular to nanostructures. Although it is intuitively accepted that MDs and TDs have an important impact in the case of epitaxial thin films, too, it is quite difficult to make a quantitative correlation between the presence of various types of dislocations and the effective ferroelectric and dielectric properties of the material. Nagarajan et al. reported a quantitative study of the thickness dependence of the polarization and piezoelectric properties of epitaxial PZT films with two compositions, PbZr0.52Ti0.48O3 共PZT 52/ 48兲 and PbZr0.2Ti0.8O3 共PZT 20/ 80兲, grown on SrRuO3-coated 共001兲 SrTiO3.14 MDs formed in a much higher density for the PZT 52/ 48 film, which has a larger in-plane lattice mismatch with the SrTiO3 substrate 共−3.9% for PZT 52/ 48, but only about −0.7% for PZT 20/ 80, at growth temperature兲. Cross-section TEM micrographs of 20 nm thin PZT films of both compositions confirmed that dislocations were present in these PZT films. Consequently, a drastic reduction of the switchable polarization ⌬P and the piezoelectric coefficient d33 was ob-
Vrejoiu et al.: Ferroelectric nanostructures
FIG. 3. FE thin films produced by PLD. Cross section TEM micrographs of 共a兲 a defective 250 nm PZT 20/ 80 film 共seen along 关110兴 STO direction兲 and 共b兲 of a 90 nm PZT 20/ 80 film coherently grown on SrRuO3 / SrTiO3共100兲. In 共c兲 and 共d兲, ferroelectric hysteresis loop acquired from a defective PZT 20/ 80 film and a defect-free PZT 20/ 80 film, respectively, are shown.
served for PZT 52/ 48 films thinner than 100 nm. The better lattice-matched PZT 20/ 80 films, with low density of dislocations, showed no scaling in ⌬P or d33 down to 15 nm. The severe drop in ⌬P and d33 values in the PZT 52/ 48 system occurring for much thicker layers indicated the significant role of dislocations in the size effects in ferroelectric thin films and nanostructures. Therefore, a lot of effort was dedicated to the optimization of the heteroepitaxy of PZT films on closely matched single crystalline substrates, such as SrRuO3-coated 共001兲 SrTiO3. B. Single crystalline PbZr0.2Ti0.8O3 thin films
Pursuing the growth of FE epitaxial thin films free of extended structural defects, Vrejoiu et al. reported on the PLD fabrication of single crystalline PZT 20/ 80 thin films grown on SrRuO3-coated vicinal SrTiO3 共STO兲 共100兲 substrates.6 The quality of the vicinal STO substrates 共CrysTec, Berlin兲 and their prior treatment by etching in buffered hydrofluoric acid and annealing at 1000– 1100 ° C is an extremely critical step in the fabrication, which results in TiO2-terminated single unit cell-stepped STO共100兲 surfaces. This is a prerequisite for achieving the step flow growth regime for the oxide electrode, SrRuO3 共SRO兲.15 Step flow growth renders the SRO layer pseudomorphic and in-plane coherent to the STO共100兲 substrate. Furthermore, step flow grown SRO serves as a template for the subsequent layer-bylayer growth of the PbZr0.2Ti0.8O3 epitaxial films, which occurs under carefully optimized PLD conditions. The particular composition PZT 20/ 80 was chosen because it is nominally tetragonal at room temperature, and c-axis grown films would also have relatively low in-plane lattice mismatch on SRO/STO 共100兲. Figure 3 summarizes the results of structural and ferroelectric characterization of our PZT 20/ 80 films. Figures 3共a兲 and 3共b兲 show TEM cross section micrographs taken on PZT 20共80兲/SRO/STO 共100兲 heterostructures grown under nonopJ. Vac. Sci. Technol. B, Vol. 27, No. 1, Jan/Feb 2009
timized and under optimized PLD conditions, respectively. In the defective 250 nm thick PZT 20/ 80 film imaged in Fig. 3共a兲, a high density of extended structural defects, such as TDs and stacking faults 共SFs兲 along with 90° domains were found by TEM and HRTEM. The abundant TD dipoles included a SF in between and HRTEM of the SFs indicated that they are most probably Pb rich.16 These defects were correlated with the microscopic studies of the backswitching of the polarization observed by PFM.16 Further evidence of the detrimental role played by these defects on the ferroelectric properties of the PZT films is given by the macroscopic polarization measurements performed through large top Pt/ SRO electrodes 共electrode area 艋0.09 mm2兲. Figures 3共c兲 and 3共d兲 show the polarization hysteresis and the switching current loops measured on a PZT 20/ 80 film with high density of such defects and on a PZT 20/ 80 film free from TDs and SFs, respectively. The remnant polarization is considerably higher for the high structural quality film, Pr = 95 C / cm2, whereas the defective film has only around Pr = 65 C / cm2. Ferroelectric 180° and 90° domains play important roles in tetragonal ferroelectric thin films and nanostructures, especially in the polarization switching.17–21 Remarkable progress has been made in understanding the nature of the 180° domain walls in our ultrathin PZT 20/ 80 films by employing the negative spherical-aberration imaging technique in an aberration corrected HRTEM.17 Quantitative analysis of the domain micrographs revealed that two types of 180° domains can exist in a 10 nm thin PZT 20/ 80 film, with head-to-head and head-to-tail arrangement of the local dipoles, and precise values of the lattice parameters, and the spontaneous polarization inside the domains could be calculated. The domain wall thickness of the head-to-tail coupled 180° boundaries was estimated to ⬃4 nm 共⬃10 unit cells兲. The ferroelastic twins that are often present in tetragonal FE thin films and crystals, the so-called 90° a-c domains, were studied in the case of PZT 20/ 80 epitaxial films as well and pinpointed as nucleating centers for the polarization switching.18 Moreover, their mechanical mobility under external electric field was found responsible for extrinsic enhancement of the piezoelectric coefficients and dielectric permittivity.19,20 C. Ferroelectric nanostructures: From self assembly to geometrically ordered arrays
This section addresses fabrication and patterning of nanoscale ferroelectric structures, respectively structures with all dimensions smaller than 100 nm. A classical top-down patterning processing starts with a thin film, fabricating structures of desired geometry out of it by photolithography and etching. There are obvious advantages of this process, one being the starting “material” itself, which can be a wellcharacterized thin film of high structural quality, along with the possibility of large area processing and controlled size and positioning. Nevertheless, etching a complex material such as PZT or other perovskite oxides is a complex task. The etching process usually renders a damaged region at the
Vrejoiu et al.: Ferroelectric nanostructures
rim of the patterned structure that, depending on the particular process, can be as large as the final structures.22,23 The only way to avoid these drawbacks and to replace the expensive top-down processes is to develop self-assembly processes, which would allow to pattern large area arrays of high-quality, possibly epitaxial nanostructures with lateral dimensions in the range well below 100 nm. Epitaxial structures are particularly useful, because they may enable us to pinpoint those factors that influence the behavior of the ferroelectric materials at nanoscale. The epitaxial character of the patterned nanostructures would result in a single orientation of all structures, which consequently will give the possibility of detailed structural analysis using standard macroscopic methods such as x-ray diffraction, and will ease the assessment of the role played by structural defects for the electrical and optical properties. One possibility to grow epitaxial FE nanostructures is based on the analysis of growth of an ultrathin film by chemical solution deposition 共CSD兲.24 Briefly, an ultrathin amorphous film is deposited on a single-crystal substrate by spin coating an appropriate precursor solution, followed by a simple pyrolyse process at temperatures of about 300 ° C. The nanosized structures are formed by a high temperature annealing, e.g., at 900 ° C for one hour. During the final annealing the amorphous film breaks up and due to enhanced surface mobility forms islands that eventually crystallize. The ultimate shape and dimensions of the structures are a function of the involved interfacial and surface energies and are related to the initial thickness of the amorphous layer, the final annealing temperature, and the annealing time. The method is versatile and can principally be used for any substrate. Such self-assembled PZT epitaxial nanostructures 共Fig. 2兲 enabled us to correlate the occurrence of “size effects” to their structural quality, as it was shown in the discussion in Sec. III A. There we pointed out that interface and surface related defects, as well as extended defects, may play a detrimental role for the FE properties of the nanostructures. Additionally, the work by Chu et al.8 showed that analysis of the size effects without a complex structural, electrical and optical analysis might be erroneous. However, two features were not achieved by the selfassembly approach: 共i兲 geometrical large scale order of the nanostructures and 共ii兲 patterning of metal-ferroelectricmetal 共MFM兲 nanocapacitors. The fabrication of wellordered arrays of nanoscale FE structures without using the classic top-down process is technically very demanding. Arrays of ferroelectric structures of few hundred nanometre sizes were prepared by electron-beam direct writing25 and imprint lithography.26 Both methods, however, cannot produce nanostructures smaller than about 100 nm lateral size. An alternative method is to employ a stencil mask attached to the substrate and to deposit the ferroelectric material through this mask. The stencil mask may in turn be fabricated using self-assembly methods. Ma and Hesse used monolayers of latex spheres as stencil mask and PLD to deposit—through the interstices—BaTiO3 islands on Nb: JVST B - Microelectronics and Nanometer Structures
FIG. 4. FE nanodos 共NDs兲 produced by PLD: SEM micrographs of 共a兲 a self-assembled monolayer of latex spheres used as mask for 共b兲 NDs of SrBi2Ta2O9, 共c兲 gold nanotube membrane used as stencil mask for 共d兲 共Bi, La兲4Ti3O12 NDs 共inset: cross section TEM of one of the dots兲, 共e兲 wafer-scale SiO2 membrane used for 共f兲 PZT NDs.
STO共100兲 共Ref. 27兲 and SrBi2Ta2O9 islands on Nb:STO共111兲.28 After lifting off the mask, the dots were crystallized by postannealing, resulting in regular hexagonally arranged structures 关Figs. 4共a兲 and 4共b兲兴 that in case of BaTiO3 were epitaxial. Size and pitch between the structures were determined chiefly by the diameter of the latex spheres. For example, for a diameter of the latex spheres of one micron, the pitch and lateral size of the obtained nanostructures were 500 and 300 nm, respectively 关see Fig. 4共b兲兴. Again, the substrate played an important role in establishing the ultimate crystal quality of the patterned structures. On SROcoated STO substrates high-quality epitaxial structures were grown and had good ferroelectric properties. Although this method is quite facile and elegant, it has a few drawbacks, such as the lack of long-range order and the relatively large size of the individual nanostructures. Latex beads self-organize in hexagonal symmetry, but in rather small domains, as shown in Fig. 4共a兲. Nevertheless, this work proved the capability of PLD in combination with liftoff masks to fabricate epitaxial nanostructures. The next steps we undertook aimed at improving the long-range order and decreasing the lateral size of the structures. Lee et al. have developed an alternative mask, based on a self-assembly process.10 They used porous anodic alumina 共AAO兲 as template and electrodeposition to fabricate gold nanotube membranes 关Fig. 4共c兲兴 which later on were
Vrejoiu et al.: Ferroelectric nanostructures
FIG. 5. A typical piezoelectric hysteresis loop obtained by piezoresponsescanning force mocroscopy 共PFM兲 on a single PbZr0.4Ti0.6O3 nanodot.
used as stencil mask for the deposition of ferroelectric nanostructures. Perfectly ordered lanthanum-doped bismuth titanate 共BLT兲 关Fig. 4共d兲兴, strontium bismuth tanatalate 共SBT兲, and PZT nanostructures were fabricated on large area substrates 共typically 1 cm2兲 by deposition of ferroelectric material through the gold nanotube membrane.10,11 PLD deposition was performed at room temperature to avoid the thermal damage of the membrane. The large area periodic arrangement of the gold nanotube membranes is determined by the AAO template. Perfectly ordered, large-area AAO templates with hexagonal, quadratic, or complex Moiré-type symmetry were achieved by controlling the anodization process of aluminium using imprint methods.29 After mechanical lift-off of the membrane and high temperature crystallization, epitaxial BLT nanostructures with a height of about 100 nm and a lateral size of about 150 nm were obtained. Using substrates with different crystallographic orientation, epitaxial BLT nanostructures with different orientations could be achieved.30 Investigations of the local ferroelectric switching properties of individual FE nanodots were performed by means of PFM: an example of a piezoelectric hysteresis loop acquired on a PZT 40/ 60 nanostructure is shown in Fig. 5 共see Refs. 10 and 30 for details on PFM of the FE nanodots兲. Moreover, the relatively large area covered by nanostructures and their periodic arrangement enabled detailed structural investigations based on x-ray diffraction analysis, demonstrating the advantage of these structurally uniform nanostructures to determine structural properties using macroscopicsize samples.30 Although the patterning method based on gold nanotube membranes is a versatile method, providing large-area arrays of very well geometrically ordered nanostructures and a wide range of structure dimensions and pitches, there are two main disadvantages: 共i兲 the impossibility to grow the structures directly in epitaxial form, and consequently, 共ii兲 the impossibility to deposit dissimilar crystalline materials using the same mask. As a result, nanoscale ferroelectric capacitors, or any other complex multilayered structures, cannot be achieved due to the high temperature anneal needed to crysJ. Vac. Sci. Technol. B, Vol. 27, No. 1, Jan/Feb 2009
tallize the ferroelectric. In order to overcome this deficiency, Lee et al. have developed a fabrication method based on ultrathin AAO membranes.31 Instead of using gold nanotube membranes as lift-off masks, ultrathin AAO membranes are used. Due to the high temperature stability of the Al2O3 mask, PLD deposition can be performed at relatively high temperatures.12 Thus epitaxial ferroelectric nanostructures can be directly obtained. Moreover, since the crystallization by postannealing is not necessary, multiple subsequent depositions of dissimilar materials can be performed. This was demonstrated by Lee et al., who fabricated well-ordered nanoscale Pt-PZT-Pt capacitor arrays with sizes smaller than 100 nm and thus closely approaching the desirable memory density of the order of Tbit/ in.2.12 The epitaxially grown capacitors can be addressed individually and possess good switching properties, as probed by PFM investigations. Another approach employed to produce masks for depositing ordered FE nanodots was laser interference lithography.32,33 Figure 4共e兲 shows an SEM image taken on the patterned SiO2 mask obtained by this method and Fig. 4共f兲 is an SEM image of the corresponding well-ordered epitaxial PZT nanodots 共diameters down to ⬃250 nm were achieved兲 fabricated on SrRuO3-coated SrTiO3共100兲 by PLD through such a mask. IV. CONCLUSIONS A broad spectrum of ultrathin ferroelectric films and nanodots were fabricated by various methods and thoroughly characterized, with emphasis on reducing the size below 100 nm while maintaining good ferroelectric properties. Structural defects and ferroelectric domain patterns were investigated in order to have a good understanding of their influence on nanosized ferroelectrics and to figure out how to avoid their harmful effects. Epitaxial PZT 20/ 80 films were grown under optimized PLD conditions in order to achieve single crystallike FE thin films with very low density of extended structural defects. These high quality PZT films enabled fundamental studies of the electrical and ferroelectric properties, with minimized contribution from structural defects. Epitaxial FE nanodots of various materials were successfully fabricated and geometrical ordering was achieved, allowing to approach the desirable memory density of Tbit/ in.2 for potential application in nonvolatile ferroelectric data storage. ACKNOWLEDGMENTS We are grateful to M.-W. Chu, I. Szafraniak, W. Ma, C. Harnagea, S. K. Lee, W. Lee, K. Nielsch, H. Han, and R. Ji for their valuable contribution to herein presented work. J. F. Scott and C. A. P. D. Araujo, Science 246, 1400 共1989兲; A. Rüdiger, T. Schneller, A. Roelofs, S. Tiedke, T. Schmitz, and R. Waser, Appl. Phys. A: Mater. Sci. Process. 80, 1247 共2005兲. 2 L. M. Eng, M. Bammerlin, Ch. Loppacher, M. Guggisberg, R. Bennewitz, R. Lüthi, E. Meyer, Th. Huser, H. Heinzelmann, and H.-J. Güntherodt, Ferroelectrics 222, 153 共1999兲. 3 P. Paruch, T. Tybell, and J.-M. Triscone, Appl. Phys. Lett. 79, 530 共2001兲. 1
Vrejoiu et al.: Ferroelectric nanostructures
J. Junquera and Ph. Ghosez, Nature 共London兲 422, 506 共2003兲. K. Tanaka, Y. Kurihashi, T. Uda, Y. Daimon, N. Odagawa, R. Hirose, Y. Hiranaga, and Y. Cho, Jpn. J. Appl. Phys. 47, 3311 共2008兲. 6 I. Vrejoiu, G. Le Rhun, L. Pintilie, M. Alexe, D. Hesse, and U. Gösele, Adv. Mater. 共Weinheim, Ger.兲 18, 1657 共2006兲. 7 H. N. Lee, D. Hesse, N. Zakharov, and U. Gösele, Science 296, 2006 共2002兲. 8 M.-W. Chu, I. Szafraniak, R. Scholz, C. Harnagea, D. Hesse, M. Alexe, and U. Gösele, Nature Mater. 3, 87 共2004兲. 9 W. Ma, D. Hesse, and U. Gösele, Nanotechnology 17, 2536 共2006兲. 10 W. Lee, M. Alexe, K. Nielsch, and U. Gösele, Chem. Mater. 17, 3325 共2005兲. 11 S. K. Lee, W. Lee, M. Alexe, K. Nielsch, D. Hesse, and U. Gösele, Appl. Phys. Lett. 86, 152906 共2005兲. 12 W. Lee, H. Han, A. Lotnyk, S. Senz, M. Alexe, D. Hesse, S. Baik, and U. Gösele, Nat. Nanotechnol. 3, 402 共2008兲. 13 I. Szafraniak, C. Harnagea, R. Scholz, S. Bhattacharyya, D. Hesse, and M. Alexe, Appl. Phys. Lett. 83, 2211 共2003兲. 14 V. Nagarajan, C. L. Jia, H. Kohlstedt, R. Waser, I. B. Misirlioglu, S. P. Alpay, and R. Ramesh, Appl. Phys. Lett. 86, 192910 共2005兲. 15 W. Hong, H. N. Lee, M. Yoon, H. M. Christen, D. H. Lowndes, Z. Suo, and Z. Zhang, Phys. Rev. Lett. 95, 095501 共2005兲. 16 I. Vrejoiu, G. Le Rhun, N. D. Zakharov, L. Pintilie, M. Alexe, and D. Hesse, Philos. Mag. 86, 4477 共2006兲. 17 C.-L. Jia, S.-B. Mi, K. Urban, I. Vrejoiu, M. Alexe, and D. Hesse, Nature Mater. 7, 57 共2008兲. 18 S. Jesse, B. J. Rodriguez, S. Choudhury, A. P. Baddorf, I. Vrejoiu, D. Hesse, M. Alexe, E. A. Eliseev, A. N. Morozovska, J. Zhang, L. Q. Chen, 4 5
JVST B - Microelectronics and Nanometer Structures
503 and S. V. Kalinin, Nature Mater. 7, 209 共2008兲. G. Le Rhun, I. Vrejoiu, L. Pintilie, D. Hesse, M. Alexe, and U. Gösele, Nanotechnology 17, 3154 共2006兲. 20 G. Le Rhun, I. Vrejoiu, and M. Alexe, Appl. Phys. Lett. 90, 012908 共2007兲. 21 W. Li and M. Alexe, Appl. Phys. Lett. 91, 262903 共2007兲. 22 A. Stanishevsky, B. Nagaraj, J. Melngailis, R. Ramesh, L. Khriachtchev, and E. McDaniel, J. Appl. Phys. 92, 3275 共2002兲. 23 A. Stanishevky, S. Aggarwal, A. S. Prakash, J. Melngailis, and R. Ramesh, J. Vac. Sci. Technol. B 16, 3899 共1998兲. 24 A. Seifert, A. Vojta, J. S. Speck, and F. F. Lange, J. Mater. Res. 11, 1470 共1996兲. 25 M. Alexe, C. Harnagea, D. Hesse, and U. Gösele, Appl. Phys. Lett. 75, 1793 共1999兲. 26 C. Hamagea, M. Alexe, J. Schilling, J. Choi, R. B. Wehrspohn, D. Hesse, and U. Gösele, Appl. Phys. Lett. 83, 1827 共2003兲. 27 W. Ma and D. Hesse, Appl. Phys. Lett. 84, 2871 共2004兲. 28 W. Ma and D. Hesse, Appl. Phys. Lett. 85, 3214 共2004兲. 29 J. Choi, K. Nielsch, M. Reiche, R. B. Wehrspohn, and U. Gösele, J. Vac. Sci. Technol. B 21, 763 共2003兲. 30 S. K. Lee, D. Hesse, M. Alexe, Woo Lee, K. Nielsch, and U. Gösele, J. Appl. Phys. 98, 124302 共2005兲. 31 W. Lee, R. Ji, U. Gösele, and K. Nielsch, Nature Mater. 5, 741 共2006兲. 32 D. S. Kim, R. Ji, H. J. Fan, F. Bertram, R. Scholz, A. Dadgar, K. Nielsch, A. Krost, J. Christen, U. Gösele, and M. Zacharias, Small 3, 76 共2007兲. 33 H. Han, Y. Park, S. Lee, R. Ji, G. Le Rhun, M. Alexe, K. Nielsch, D. Hesse, U. Gösele, and S. Baik, Nanotechnology 共in press兲. 19