Increased ferroelastic domain mobility in ferroelectric thin films and its ...

INSTITUTE OF PHYSICS PUBLISHING

NANOTECHNOLOGY

Nanotechnology 17 (2006) 3154–3159

doi:10.1088/0957-4484/17/13/013

Increased ferroelastic domain mobility in ferroelectric thin films and its use in nano-patterned capacitors G Le Rhun, I Vrejoiu, L Pintilie, D Hesse, M Alexe and U G¨osele Max-Planck Institute of Microstructure Physics, Weinberg 2, D-06120 Halle, Germany E-mail: [email protected]

Received 15 March 2006 Published 5 June 2006 Online at stacks.iop.org/Nano/17/3154 Abstract Movements of ferroelastic 90◦ a -domains (twins) of sub-100 nm lateral size contribute greatly to the dielectric, piezoelectric, and elastic properties of ferroelectric materials. However, in thin films the mobility of twins is severely limited due to substrate clamping and pinning of domain walls. Here we demonstrate the possibility of moving ferroelastic domains by applying a non-uniform electric field. By means of piezoresponse force microscopy (PFM), it is shown that 90◦ domains in epitaxial Pb(Zr0.2 Ti0.8 )O3 (PZT) continuous films are highly mobile. Following this observation, capacitors with stripe-shaped top electrodes have been designed, which show a reversible increase in the piezoelectric signal, as well as an increase of the specific capacitance by a factor of 1.4. The present concept of non plane-parallel electrodes might be useful for redesigning devices such as multilayered capacitors or sensors and actuators based on ferroelectric thin films. (Some figures in this article are in colour only in the electronic version)

1. Introduction Ferroelastic 90◦ domain walls play an important role in the response of ferroelectric materials. Movements of 90◦ a domains are indeed responsible for the extrinsic contribution to the dielectric permittivity [1–4] and to the piezoelectric coefficient [4–6]. However, ferroelastic twin walls exhibit different mobilities depending on whether the material is in the bulk (ceramic or single crystal) or in thin film shape. The amount of switchable 90◦ a -domains in epitaxial PZT thin films was demonstrated to be relatively small compared with that reported for bulk PZT ceramics [7, 8]. In particular, it was found that the mobility of 90◦ a -domains is severely limited in sub-micron thin films mainly due to substrate clamping [4, 9, 10] and/or domain pinning [11, 12]. Nevertheless, switching of 90◦ a -domains in epitaxial films under an applied electric field was inferred from in situ synchrotron x-ray diffraction experiments [12]. The question of the mobility of ferroelastic domains in epitaxial ferroelectric thin films remains, however, controversial as it has also been 0957-4484/06/133154+06$30.00

reported that 90◦ domains do not move in continuous thin films, even upon application of high dc fields [13]. Recently, significant 90◦ a -domain displacements were observed in PZT micro-island capacitors, accompanied by a great enhancement of the piezoelectric coefficient [14, 15]. It was shown that only by patterning the ferroelectric film into micro-size islands is the clamping effect significantly reduced, enabling the movement of ferroelastic walls. Therefore, although the movement of 90◦ domain walls seems to be possible, this may depend on the sample itself (thickness, presence of defects, structure of the capacitor, etc) and, more generally, on the experimental conditions. In this paper we show by PFM investigations that 90◦ a -domains can move significantly in a continuous epitaxial PZT thin film on application of a non-uniform electric field. Thereafter we have exploited the effect of a non-uniform field on the mobility of the 90◦ domains to design metal–ferroelectric–metal patterned structures that involve a reversible and tunable increase of the piezoelectric coefficient, and a significant enhancement of the specific capacitance due to an increase of the extrinsic contribution of the moving 90◦ domain walls.

© 2006 IOP Publishing Ltd Printed in the UK

3154

Ferroelastic domain mobility in ferroelectric thin films

2. Experiment The ferroelectric heterostructure investigated in this study consisted of a 150 nm thick Pb(Zr0.2 Ti0.8 )O3 film epitaxially grown onto a (100) single crystal SrTiO3 (STO) substrate with an intermediate SrRuO3 (SRO) layer as the bottom electrode. The PZT film was deposited by pulsed laser deposition at a substrate temperature T = 575 ◦ C under an oxygen partial pressure of 0.3 mbar. Details of the deposition process can be found elsewhere [16]. Transmission electron microscopy (TEM) was employed to investigate both plan-view and cross-sectional microstructures. TEM samples were prepared by standard mechanical and ion-beam thinning procedures, and TEM investigation was performed in a Philips CM20T microscope operated at 200 kV. One of the most powerful techniques for studying ferroelectric domain engineering is based on the use of scanning probe microscopy (SPM), namely piezoresponse force microscopy (PFM) [17]. In this work, piezoresponse investigations of the PZT film were performed using a scanning probe microscope (ThermoMicroscopes) equipped with PtIrcoated tips (Nanosensors, ATEC-EFM) with an elastic constant of about 2.5 N m−1 . An ac voltage (1 V rms) with a frequency f = 22.3 kHz used to induce a local piezoelectric vibration was applied to the AFM tip, while a dc bias voltage was applied to the bottom electrode in order to pole the ferroelectric sample. The mechanical oscillations of the sample surface were transmitted to the tip and extracted from the global deflection signal using a lock-in technique. Local piezoelectric hysteresis loops were acquired by positioning the tip at the desired site and then measuring the local piezoresponse signal as a function of dc bias voltages. For quantitative d33 measurements, the piezoresponse signal was previously calibrated using an X-cut quartz crystal, whose d33 is 2.17 pm V−1 . Top electrodes were fabricated by electronbeam (EB) lithography and a lift-off process. EB resist (PMMA/copolymer bilayer) was spin-coated on the surface of the film, and then exposed with an EB lithography system (ELPHY Plus) adapted to a JEOL JSM 6400 scanning electron microscope. Finally, metallization was provided by sputtering of 10 nm of chromium and 40 nm of platinum, followed by liftoff. Capacitance versus voltage curves were performed with dc biases ranging from −10 to +10 V with an oscillator amplitude of 300 mV and a 1 kHz frequency, using an HP4194A impedance/gain analyser.

3. Results and discussion Figure 1(a) shows a piezoresponse image of a 150 nm thick Pb(Zr0.2 Ti0.8 )O3 film in the as-grown state. In agreement with the prediction of Roytburd [18], the as-grown PZT film contains a uniform two-dimensional grid of 90◦ a -domains embedded into the ferroelectric matrix with the tetragonal c-axis being oriented perpendicular to the substrate. The uniform bright contrast of the image in between the twins (dark lines) suggests that the entire film is pre-poled in a preferential direction normal to the surface of the film [19]. The plan-view TEM image (figure 1(b)) shows a similar grid of 90◦ a -domains. Both figures 1(a) and (b) show a coarse mesh pattern of thick twins, the meshes of which contain

(a)

(b)

(c)

PZT

SRO STO

100 nm

Figure 1. Visualization of the ferroelastic 90◦ domains. (a) AFM piezoresponse, (b) plan-view TEM image and (c) cross-sectional TEM image of a 150 nm thick PbZr0.2 Ti0.8 O3 film in the as-grown state.

a cross-hatch pattern of fine twins. Figure 1(c) is a cross section TEM micrograph of the investigated PZT/SRO/STO heterostructure showing broad and narrow 90◦ a -domains wedges and confirming the above description. The domain structure was investigated after poling the sample. This was performed by scanning the surface of the PZT film with a biased tip. Immediately after poling, the same area was rescanned with a zero dc voltage. On the application of a negative dc voltage, a 1 µm × 1 µm area of the sample was switched from its original positive state (bright contrast) to the reversed state (dark contrast); see figures 2(a) and (b). In addition to the 180◦ c-domain switching, a significant change in the 90◦ domain structure was observed. This clearly shows that the twins are mobile in a continuous film despite the substrate clamping effect, as previously reported for an epitaxial PbTiO3 thin film [20]. The same area was further poled with positive bias and this resulted again in a displacement of several twins (figures 2(c) and (d)). We made similar observations of 90◦ a -domain movement on PZT films with a thickness ranging from 100 to 240 nm. It is worth mentioning that the evolution of the ferroelastic domain pattern under an applied electric field, shown in figure 2, is very similar to that observed by Nagarajan et al for a PZT film of the same composition [14]. However, in their case, this was obtained after patterning the film into 1 µm2 islands in order to reduce the constraint and diminish internal stresses [21]. Detailed analysis of figure 1(c) revealed two distinct twin types: i.e. long and wide twins, and short and thin twins. Therefore, it may be assumed that the mobility of the 90◦ a domains under the application of an electric field or a mechanic stress differs according to their size. The smaller ones most probably move easily and are able to disappear or merge. 3155

G Le Rhun et al

Figure 3. (a) Piezoresponse image of a 150 nm thick PZT 20/80 film before local poling. (b) Corresponding PFM image after local poling at the two sites marked ‘+’ in figure (a), showing the distortion or bending of the 90◦ a -domain.

Figure 2. Piezoresponse images of the same 1 µm2 area showing the evolution of the 90◦ domain structure of a 150 nm thick PbZr0.2 Ti0.8 O3 film as a function of applied dc bias voltages: (a) as-grown state, (b) after −6 V dc, (c) after +3 V dc, and (d) after +6 V dc.

Twins with medium size may move towards neighbouring twins and merge. Finally, twins that cross along the whole PZT film are most probably immobile. More precisely, they end either at the bottom electrode or at a defect in the bulk of the layer where they might be pinned (see figure 1(c)). However, as is shown in figure 3, large 90◦ a -domains may simply bend, and not necessarily entirely move, in response to a non-uniform applied electric field. In this experiment, the AFM tip was placed close to a twin at the positions marked ‘+’ (figure 3(a)), and a dc pulse voltage was applied (five times the coercive voltage). A subsequent piezoresponse scan of the same region of the film (figure 3(b)) shows that the 90◦ domain line became curved as indicated by the arrows, which is a direct proof that the 90◦ domain walls (DWs), even large ones, can bend. The bending of 90◦ a -domains in the vicinity of the surface was demonstrated by Ishibashi et al [22] in a model treating the polarization reversals in the presence of 90◦ DWs. Additionally, the extent of bending appears to depend upon the anisotropy of the local potential. Moreover, along with the motion of the 90◦ a -domains, new twins may also be formed, as has already been observed [20, 23]. Figure 4(a) shows piezoelectric loops measured on a c-domain and on an a -domain. The piezoelectric loop obtained on the c-domain shows a marked asymmetry indicating a built-in internal field, thus confirming the preferred domain orientation deduced from the uniform contrast of the piezoelectric image. However, the asymmetric configuration of the electrodes (i.e., needle-like top electrode and a large flat bottom electrode) in PFM may also contribute to this asymmetry. A d33 value of ∼50 pm V−1 was determined, which is in good agreement with previously reported values [21, 24]. The second loop was obtained on a 90◦ a -domain after positioning the tip on the corresponding 3156

region, where the piezoelectric signal is initially around zero. This loop appeared strongly distorted and can be seen as consisting of two minor loops: one complete open hysteresis loop on the negative branch (0 V to −5 V and back to 0 V) and a closed loop at positive applied voltages (0 V to +5 V and back to 0 V). We interpret this as a 90◦ a -domain movement under the tip (figure 4(b)). In the specific configuration of the PFM experiments, a highly non-uniform electric field develops under the AFM tip [25–28]. This non-uniform field has a certain in-plane component, which represents a driving force to move or bend the 90◦ domains. We know that the actual volume that is probed under the tip depends on the tip–surface contact area. According to the calculations of Harnagea et al [27], 90% of the voltage applied between the AFM tip and the bottom electrode is confined into 30% of the film thickness. In our case, the estimated probed depth is 30–50 nm. In the negative branch of the hysteresis loop the 90◦ a -domain moves totally outside the probing region, whereas in the positive branch it moves in the opposite direction, the probing volume being a mixture of a - and c-domains. At large field, the 90◦ a -domain moves out from the probed region under the tip, there remaining only a c-domain (note the equivalent d33 signal amplitude). After removing the bias voltage, the piezoelectric signal returns slowly to zero, suggesting that the twin domain moves reversibly. A plane-parallel capacitor is currently employed in most ferroelectric devices including random access memories (FeRAMs). In this configuration the electric field is perpendicular to the surface electrodes and uniform inside the capacitor. In order to use the above observed mobility of the 90◦ a -domains, we have designed the top electrode of a metal–ferroelectric–metal (MFM) capacitor as a patterned stripe-like structure. The width of the stripes was chosen to be in the same range as the thickness of the film in order to induce a non-uniform electric field within the ferroelectric layer. Moreover, the width of the stripe electrode was also close to the periodicity of the 90◦ domains, knowing that at the edges of the stripe electrode the fringing electric field has a certain in-plane component. Figure 5 shows a piezoelectric hysteresis loop measured on a 0.4 µm wide, 2 µm long single stripe electrode. At remanence, the piezoelectric coefficient is about 20 pm V−1 , which is well below the value measured on c-domains. This is mainly because the volume probed under the electrode consists of both a - and c-domains.

Ferroelastic domain mobility in ferroelectric thin films

Figure 4. The effect of the motion of a twin on the piezoelectric hysteresis loop. (a) Comparison of the piezoelectric loops measured on a 90◦ a -domain and on a 180◦ c-domain. During the hysteresis measurement, the dc voltage was swept as follows: 0 → +5 V → −5 V → +5 V. (b) A schematic representation of the motion of the twin under the tip during the hysteresis measurement. Note that the twins, as well as their motion, are intentionally represented in a simplified way.

Figure 5. Piezoelectric loop measured on a single stripe electrode and on a square electrode on the same sample. The stripe electrode was 2 µm long with a width of 0.4 µm, while the area of the square electrode was 4 µm2 .

By applying a sufficient dc voltage, the d33 value increases up to 40 pm V−1 . This is due to the motion of 90◦ a -domains out of the region underneath the top electrode. We emphasize that it is a reversible process, as indicated by the hysteresis behaviour of the loop. The acquired loop consists indeed of three loops. The main loop corresponds to the switching of the 180◦ c-domains, and the two minor loops at higher voltages originate from the motion of the 90◦ a -domains. It is worth noting that this was not observed in planar capacitors patterned on the same sample, near the stripe electrode, and in which the electric field is uniform (see figure 5). The high mobility of 90◦ a -domains should also be reflected in the capacitance of the MFM structure. For this purpose, an array of stripe electrodes connected all together with a 2 µm wide line was fabricated (figure 6(a)). Each

single stripe electrode had a length of 45 µm and a width of (250 ± 10) nm. A detail of the stripe-electrode pattern is shown in the inset of figure 6(a). Figure 6(b) is a crosssection normal to the film plane showing a modelling of the equipotential lines inside the previously described MFM structure. This was obtained with finite element analysis software, FlexPDE® , from PDE Solutions Inc. For modelling, the capacitor pattern was designed so that it was as similar as possible to the real one. The top electrodes, with a width 1.6 times as large as the thickness of the film, were biased to the same potential, while the bottom electrode was grounded. As can be inferred from figure 6(b), the electric field is highly non-homogeneous inside the MFM structure. It is uniform in the middle part below each stripe, and becomes strongly nonuniform at the edges of the stripes, having a strong in-plane component. We expect that the 90◦ a -domains located at the edges of the stripes will show a higher mobility in response to the inhomogeneous field, similarly to what we observed throughout PFM investigations. In order to compare the results with a conventional plane-parallel capacitor, a squareshaped electrode was deposited simultaneously on the same sample, using the same lithography process. Capacitance versus voltage curves measured on both the plane capacitor and stripe capacitor are shown in figure 6(c). An increase of the specific capacitance with a factor of 1.4 was obtained for the non-uniform capacitor compared with that measured for the planar capacitor1. We attribute this enhancement to a high mobility of the 90◦ a -domains made now possible by the application of a non-uniform electric field. In 1 Due to the particular geometry of the striped capacitor, it is not possible to apply the formula C = ε0 εr S/d used for a plane-parallel capacitor for calculating the permittivity of the MFM structure. That is why we present the results as a capacitance per surface unit. Moreover, we notice that a thin parasitic layer (copolymer) with a very small permittivity (ε < 10) might be present between the PZT layer and the top electrode because of the lithography process, thereby lowering the capacitance of all fabricated capacitors.

3157

G Le Rhun et al

a non-uniform electric field in the ferroelectric capacitor and make use of the extrinsic contribution of the twins. This results in a reversible and tunable increase of the piezoelectric signal, as well as a significant increase by 40% of the specific capacitance. Finally, we point out, as a general conclusion, that this concept of non-plane-parallel electrodes might be useful for (re)designing devices based on ferroelectric thin films.

(a) 20 µm

Acknowledgments The work has been partly funded by the Volkswagen Foundation through the ‘Nanosized ferroelectric hybrids’ project no I/80897. The authors acknowledge Dr W Erfurth and U Doß for performing e-beam exposures, and S Swatek and N Schammelt for the TEM sample preparation and assistance in PLD-system maintenance, respectively.

250 nm

(b)

References 0.0

0.5

1.0

C (nF/mm2)

(c)

d.c. Voltage

Figure 6. (a) SEM image of the stripe-electrode structure (width of stripe = 240–260 nm) with a detail of the electrode in inset. (b) Cross-section normal to the film plane showing the non-parallel equipotential lines in the stripe capacitor. Below is the intensity scale bar normalized to 1. Note that the electric field vector is perpendicular to the equipotential lines. (c) Specific capacitance as a function of dc voltage for the uniform capacitor (square electrode) and the non-uniform capacitor (stripe electrode).

this enhancement, the flexoelectric effect might also play a role [29]. It is worth noting that our capacitor can be further optimized by modifying the spacing between the stripes, the width of the stripes, or by using a top electrode of a different shape. For instance, the use of an electrode with a grid shape similar to the 90◦ domain pattern observed by PFM and TEM would allow us to obtain an effect of the non-uniform electric field on both ‘vertical’ and ‘horizontal’ twins. We believe that the above results will have an impact in designing the next generation of embedded thin film capacitors based on tetragonal BaTiO3 by maximizing the extrinsic contribution of the 90◦ a -domains to the dielectric permittivity.

4. Conclusion In summary, we have shown that 90◦ a -domains can be moved in a PbZr0.2 Ti0.8 03 continuous film provided that the applied electric field is non-uniform. Following this observation, stripe-shaped top electrodes were designed in order to induce 3158

[1] Fouskova A 1965 J. Phys. Soc. Japan 20 1625 [2] Pertsev N A, Arlt G and Zembilgotov A G 1995 Micro. Eng. 29 135 [3] Topolov V Y, Bolten D, Bottger U and Waser R 2001 J. Phys. D: Appl. Phys. 34 711 [4] Xu F, Trolier-McKinstry S, Ren W, Xu B, Xie Z-L and Hemker K J 2001 J. Appl. Phys. 89 1336 [5] Pertsev N A and Emelyanov A Y 1997 Appl. Phys. Lett. 71 3646 [6] Bondarenko E I, Topolov Y Y and Turik A V 1991 Ferroelectr. Lett. Sect. 13 13 [7] Kholkin A L, Wutchrich C, Taylor D V and Setter N 1996 Rev. Sci. Instrum. 67 1935 [8] Xu F, Chu F and Trolier-McKinstry S 1999 J. Appl. Phys. 86 588 [9] Kholkin A L, Akdogan E K, Safari A, Chauvy P-F and Setter N 1999 J. Appl. Phys. 89 8066 [10] Lefki K and Dormans G J 1994 J. Appl. Phys. 76 1764 [11] Kohli M, Muralt P and Setter N 1998 Appl. Phys. Lett. 72 3217 [12] Lee K, Kim Y K, Baik S, Kim J and Jung S 2001 Appl. Phys. Lett. 79 2444 [13] Nagarajan V, Roytburd A and Ramesh R 2004 Nanoscale Characterization of Ferroelectric Materials ed M Alexe and A Gruverman (Berlin: Springer) pp 163–91 Ganpule C S, Nagarajan V, Li H, Ogale A S, Martinez A D, Ogale S B, Aggarwal S, Williams E, De Wolf P and Ramesh R 2001 Integr. Ferroelectr. 32 199 [14] Nagarajan V, Roytburd A, Stanishevsky A, Prasertchoung S, Zhao T, Chen L, Melngailis J, Auciello O and Ramesh R 2003 Nat. Mater. 2 43 [15] Ma Z, Zavaliche F, Chen L, Ouyang J, Melngailis J, Roytburd A L, Vaithyanathan V, Schlom D G, Zhao T and Ramesh R 2005 Appl. Phys. Lett. 87 072107 [16] Vrejoiu I, Le Rhun G, Pintilie L, Hesse D, Alexe M and G¨osele U 2006 Adv. Mater. at press [17] Gruverman A, Auciello O and Tokumoto H 1998 Annu. Rev. Mater. Sci. 28 101 [18] Roytburd A L 1976 Phys. Status Solidi a 37 329 [19] Ganpule C S, Roytburd A L, Nagarajan V, Hill B K, Ogale S B, Williams E D, Ramesh R and Scott J F 2001 Phys. Rev. B 65 014101 [20] Kim Y K, Kin S S, Shin H and Baik S 2004 Appl. Phys. Lett. 84 5085 [21] Roytburd A L, Alpay S P, Nagarajan V, Ganpule C S, Aggarwal S, Williams E D and Ramesh R 2000 Phys. Rev. Lett. 85 190 [22] Ishibashi Y, Iwata M and Salje E 2005 Japan. J. Appl. Phys. 44 7512 [23] Chen L, Ouyang J, Ganpule C S, Nagarajan V, Ramesh R and Roytburd A L 2004 Appl. Phys. Lett. 84 254

Ferroelastic domain mobility in ferroelectric thin films

[24] Kuffer O, Maggio-Aprile I, Triscone J M, Fischer O and Renner Ch 2000 Appl. Phys. Lett. 77 1701 [25] Hong S, Colla E L, Kim E, Taylor D V, Tagantsev A K, Muralt P, No K and Setter N 1999 J. Appl. Phys. 86 607 [26] Durkan C, Welland M E, Chu D P and Migliorato P 1999 Phys. Rev. B 60 16198

[27] Harnagea C 2001 PhD Dissertation Halle, Germany Harnagea C and Pignolet A 2004 Nanoscale Characterization of Ferroelectric Materials ed M Alexe and A Gruverman (Berlin: Springer) pp 45–85 [28] Kalinin S, Karapetian E and Kachanov M 2004 Phys. Rev. B 70 184101 [29] Ma W and Cross L E 2003 Appl. Phys. Lett. 82 3293

3159