* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Ultrafast polarization switching in thin-film ferroelectrics
Survey
Document related concepts
Nuclear electromagnetic pulse wikipedia , lookup
Chirp spectrum wikipedia , lookup
Time-to-digital converter wikipedia , lookup
Opto-isolator wikipedia , lookup
Crossbar switch wikipedia , lookup
Electromagnetic compatibility wikipedia , lookup
Electrical substation wikipedia , lookup
Buck converter wikipedia , lookup
Chirp compression wikipedia , lookup
Pulse-width modulation wikipedia , lookup
Oscilloscope history wikipedia , lookup
Transcript
APPLIED PHYSICS LETTERS VOLUME 84, NUMBER 7 16 FEBRUARY 2004 Ultrafast polarization switching in thin-film ferroelectrics J. Li, B. Nagaraj, H. Liang, W. Cao, Chi. H. Lee, and R. Ramesha) Materials Research Science and Engineering Center, University of Maryland, College Park, Maryland 20742 共Received 3 July 2003; accepted 3 December 2003兲 We present an experimental approach to study the ultrafast polarization switching dynamics in thin-film ferroelectrics. A semiconductor photoconductive switch with femtosecond laser illumination is used as a ‘‘pulse generator’’ to produce jitter-free, sub-100 ps rise time step-function-like electrical pulses. Quantitative measurements yield a polarization switching time, t s , of ⬃220 ps when measured with a 5 V, 68 ps rise time input electrical pulse. Modeling of the switching transients using the Merz–Ishibashi model and Merz–Shur model of switching kinetics yields a quantitative estimate of the characteristic switching time constant, t 0 , of ⬃70–90 ps. © 2004 American Institute of Physics. 关DOI: 10.1063/1.1644917兴 The application of ferroelectric capacitors in nonvolatile ferroelectric random access memories 共FeRAMs兲 is due to their switchable remanent polarization, which can be reversed by application of a short voltage pulse.1,2 In such an application, the issue of how fast the polarization can be switched is of great interest. With this method, Larsen and co-workers have reported a polarization switching time of ⬃390 ps for PZT films.3,4 Nucleation and growth kinetics based theoretical models predict the switching time to be of the order of ⬃70 ps.5–7 Further, our earlier studies8 showed that the polarization switching time changed dramatically with the rise time of the input electrical pulse, suggesting that it is convoluted with the circuit parasitics. The resulting switching times were of the order of the input pulse rise time, showing that much faster input pulses were required. We report in this letter results of switching experiments in thin PNZT films using a femtosecond laser triggered experimental setup that has enabled us to probe the intrinsic switching dynamics. A semiconductor photoconductive switch is used as a fast ‘‘pulse generator’’ to produce step-like pulse with fast rise times. The resistance of the photoconductive switch in its dark state 共or high resistance state兲 without laser illumination is ⬃100 k⍀ to M⍀. With laser illumination, it is in its conductive state, with an ‘‘on-state’’ resistance of a few ⍀. The photoconductive switch produces fast electrical pulses with rise time of ⬃50–100 ps that depends on the laser power, bias level, etc. A Si photoconductive switch is chosen to generate the step-function-like pulse because the carrier lifetime in Si is of the order of microseconds. This enables us to systematically create pulses of up to several hundred nanoseconds. Si switches with interdigitated fingers were patterned into a coplanar transmission line structure with 50 ⍀ characteristic impedance in order to transmit wide band signals with little dispersion. Fully integrated Pt/共La0.5Sr0.5)CoO3 /Pb共Nb0.04Zr0.28 Ti0.68)O3 /(La0.5Sr0.5)CoO3 /Pt ferroelectric capacitors with thickness of 200 nm were used in this study. Details of the growth and long-term properties are reported elsewhere.8 Large bond pads were wire bonded and the device was inserted into a 50 ⍀ microwave transmission line setup. A a兲 Electronic mail: [email protected] HP54750A digitizing oscilloscope with time resolution of ⬃20 ps was used to monitor the polarization switching response. The internal 50 ⍀ impedance of the oscilloscope is matched to the transmission line in order to ensure that no signals reflects back to the ferroelectric device. A schematic layout of the experimental setup is shown in Fig. 1. A femtosecond Ti:sapphire laser 共frequency⫽10 kHz兲 illuminates both the photodiode 共PD兲 trigger and the Si photoconductive switch 共PCS兲. The PD generates a trigger signal to synchronize the oscilloscope and two pulse generators. Conventional pulse generators 1 and 2 produce pulses with rise time of ⬃7 ns. Generator 1 creates a write pulse train which sets the ferroelectric device to the stable positive or negative polarization state by applying a positive or negative square pulse. In our experiment, pulses with 8 V amplitude and 1 ms duration were used as the write pulses. Pulse generator 2 generates a biasing pulse train for the Si switch. The delay of this pulse train is tuned so that it arrives at the Si switch coincident with the laser pulse. The Si switch generates a read pulse train, the timing of which is controlled by the relative delay between generators 1 and 2. The delay is tuned to ensure that the writing pulses occur in between the reading pulses. The resulting output pulse train to the ferroelectric device is a combination of the write and read pulse trains. The new pulse train can be a sequence of positive– positive pulses or negative–positive pulses by varying the FIG. 1. Experimental setup with a Si photoconductive switch as the fast pulse generator. 0003-6951/2004/84(7)/1174/3/$22.00 1174 © 2004 American Institute of Physics Downloaded 15 Mar 2004 to 128.32.120.53. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp Appl. Phys. Lett., Vol. 84, No. 7, 16 February 2004 FIG. 2. Input electrical pulse generated by a Si photoconductive switch and corresponding remanent (⌬ P) responses of a 4.5 m⫻5.4 m LSCO/ PNZT/LSCO capacitor. The amplitude of the electric field applied is 250 kV/cm 共5 V兲. polarity of the writing pulse generator. The oscilloscope is used to monitor the displacement current flowing through the ferroelectric capacitor. Due to the small dispersion from transmission lines and the high bandwidth of the monitoring oscilloscope, the time resolution of this setup depends on the rise time of the pulse generated by the Si switch, which can be less than 100 ps. A negative–positive pulse sequence applied to the ferroelectric device produces a switched response P * whereas a positive–positive sequence gives the nonswitched response P∧ . ⌬ P⫽ P * ⫺ P ∧ is the remanent polarization, which is directly connected to domain switching dynamics. For each capacitor size, a minimum of five capacitors was measured; each capacitor was tested a minimum of five times to obtain data at each voltage. Figure 2 shows the early stages of the 5 V input electrical pulse with a rise time 共10%–90%兲 of 68 ps and the corresponding output voltage response of the P * ⫺ P ∧ for a 4.5 m⫻5.4 m capacitor; the inset shows the full input pulse. The switching time, t s , which is defined as the time from the onset of switching to a point 90% from the maximum, is ⬃220 ps. We find that this switching transient is quite sensitive to the capacitor size, the rise time of the input pulse, and the pulse shape. An important fundamental question is the limit of the switching speed in thin film ferroelectrics. If we assume that the domain walls propagate with field depen- Li et al. 1175 dent speed v ⫽ E and mobility of ⬃2.4⫻10⫺4 m2 /V s, 9 then the switching time in a 200 nm PNZT film is ⬃80 ps, corresponding to a 5 V pulse. The switching process can be described in terms of a classical solid-state phase transformation that involves nucleation and growth of reverse domains that causes polarization reversal.5,10,11 The volume fraction of the domains switched is given by f (switched) ⫽1⫺exp关⫺⌽(E)⌿(t) 兴 , where ⌽(E) represents the field dependence and ⌿(t) represents the time dependence of the switching process. Combining the Merz model with that of Ishibashi’s, yields f (switched)⫽1⫺exp兵⫺关exp(⫺␣/E) ⫻(t/t0)n兴其, where t 0 is the characteristic switching time and n is the dimensionality of the nucleating domain. Shur et al. refined this model to account for finite-size corrections of the KA theory through the insertion of an impingement time, given by the following equation: f 共 switched兲 ⫽1⫺exp兵 ⫺ 关 exp共 ⫺ ␣ /E 兲共 t/t 0 兲 n 共 1⫺t/t m 兲兴 其 , 共1兲 where both t 0 and n have the same definition as in the Ishibashi model and t m accounts for impingement of the growing domains on the boundary of the media. We have used these two related models to fit the experimental current transients for the 4.5 m⫻5.4 m capacitor as a function of the field applied. The f (switched) vs t data are obtained by integrating the current transients and normalizing with respect to the maximum polarization. This normalized switched domain fraction is plotted as a function of the transient time. A 500 kV/cm activation field8 was used to fit the experimental data. Figures 3共a兲 and 3共b兲 show plots of normalized, switched polarization versus time, and the corresponding fit using both Ishibashi and Shur et al. models. The general approach for fitting the switching current transients using the Ishibashi model is by varying both t 0 and n. As a result, a nonintegral value of dimensionality n is typically obtained. In our preliminary fitting of the experimental data, we varied both t 0 and n, which yielded a n value of 3.1. However we recognized that interpretation of the switching data is more meaningful when n is an integer, thereby enabling us to identify the actual nucleation and growth mechanism. Therefore in our fit, we fixed n to an integer value of 3.0 and extracted the t 0 value from experimental current transient data. The excellent fit of the experi- FIG. 3. 共a兲, 共b兲 Plots of measured 共dotted line兲 and fitted 共solid line兲 f (t) curves of a 4.5 m⫻5.4 m LSCO/PNZT/LSCO capacitor. The fit was carried out with both 共a兲 Ishibashi and 共b兲 Shur et al. models. 共c兲 Characteristic switching time t 0 as a function of the voltage applied. Downloaded 15 Mar 2004 to 128.32.120.53. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp 1176 Li et al. Appl. Phys. Lett., Vol. 84, No. 7, 16 February 2004 FIG. 4. 共a兲 Remanent (⌬ P) response of a 4.5 m⫻5.4 m LSCO/PNZT/LSCO capacitor with different rise times of input electric pulse. 共b兲 Characteristic switching time t 0 as a function of the rise time of the input electric pulse. The amplitude of the electric field applied is 250 kV/cm. mental data with the model, Fig. 3共a兲, is noteworthy. Furthermore, we did not observe any significant changes in the regression coefficient of the fit as a consequence of changing the n value from 3.1 to 3.0, thus supporting that validity of fixing n⫽3. t 0 of 68 ps is obtained for this capacitor with n⫽3 and the field dependence is shown in Fig. 3共c兲. We have also attempted to fit the switching transients to the Shur et al. model, Fig. 3共b兲, in which the polarization switching process is separated into two or three stages with different dimensionalities.12 All stages of the kinetic process can be described by the same equation, Eq. 共1兲, except that the dimensionality changes by integral values for each stage. However, our data could not be fitted well with such a multistage switching process. What is the meaning of the observed value of n⫽3? Two limiting situations of the reversal process are usually considered in both Ishibashi11 and Shur et al.12 models, called ␣ and  models. In the ␣ model, reverse nuclei arise throughout the whole process and in the  model reversal is caused by nuclei that pre-exist at the beginning of the process. The fitting equations for ␣ and  models are the same except that dimensionality n is different. Therefore, n⫽3 corresponds to two-dimensional 共2D兲 growth of reverse domains in the ␣ model and three-dimensional 共3D兲 growth in the  model. Differentiating between these two possibilities is complicated by the polycrystalline nature of the films used in this study; however, we suspect that 3D reversal is unlikely 共as surmised by Ishibashi and Takagi11兲. We note that the value of n determined from fitting the switching transients is extremely sensitive to the size of the capacitor; although not presented in this letter, increasing the capacitor size systematically lowers the n value, presumably due to significant convolution from capacitive RC effects. We have observed that the capacitor size and therefore the RC-time constant directly influence the shape of the ⌬ P response 共i.e., it produces a long ‘‘tail’’兲, and consequently the fitting parameters. Therefore, it is clearly important to reduce the RC-time constant of the test structure to a value smaller than the expected t 0 value 共i.e., smaller than ⬃50 ps兲. For the 4.5 m⫻5.4 m size capacitor, we estimate an RC-time constant value of about 45 ps. Finally, we show that the rise time of the input pulse significantly impacts the switching dynamics. Figure 4共a兲 shows the ⌬ P response with different rise times of the input electrical pulse. Figure 4共b兲 shows the rise time dependence of characteristic switching time t 0 for this 4.5 m⫻5.4 m capacitor and another 12.5 m⫻14 m capacitor, obtained by fitting the transients to the Ishibashi model. We have not been able to observe saturation in the t 0 value as a function of the rise time, since the smallest rise time we have been able to obtain is of the order of 50 ps. In conclusion, we demonstrated direct measurement of the ferroelectric polarization switching time, t s of ⬃220 ps using a femtosecond optically generated electrical pulse with a rise time of 68 ps. We estimated the characteristic switching time t 0 to be ⬃70–90 ps with fixed dimensionality n⫽3. The switching dynamics are impacted most by the rise time of the input electric pulses to the FE device and the size of the capacitor, i.e., the RC-time constant of the capacitor. In order to obtain the real intrinsic polarization switching time, the rise time of the input electric pulses has to be smaller than or at least of the same order as the intrinsic switching time, while the capacitor size has to be small enough so that the RC-time constant is equal to or less than the intrinsic switching time. The polarization switching time measurement is limited by the larger of the two parameters, namely, the RC-time constant and the rise time of the input electric pulse. In other words, the rise time becomes important only when the capacitor size is small enough. For larger capacitors, a fast rise time will still result in a large switching time since it is now limited by the RC-time constant of the capacitor. For smaller capacitors, a longer rise time will still result in a larger switching time. Only when the above two parameters work together can one obtain a switching time that is close to the intrinsic switching time. This work was supported by the NSF-MRSEC under Contract No. DMR-00-80008. J. F. Scott and C. A. Araujo, Science 246, 1400 共1989兲. O. Lohse, S. Tiedke, M. Grossman, and R. Waser, Integr. Ferroelectr. 22, 123 共1998兲; T. A. Bartic, D. J. Wouters, and H. Maes, ibid. 22, 181 共1998兲. 3 P. K. Larsen, G. L. M. Kampschoer, M. B. van der Mark, and M. Klee, Proceedings of the 8th IEEE International Symposium on Applications of Ferroelectrics, 1992, 共ISAF ’92兲, pp. 217–224. 4 P. K. Larsen, R. Cuppens, and G. J. M. Dormans, in Science and Technology of Electronic Thin Films, edited by O. Auciello and R. Waser 共Kluwer Academic, Dodrecht, The Netherlands, 1995兲, Vol. 284, pp. 201–221. 5 E. Fatuzzo and W. J. Merz, Ferroelectricity 共Wiley, New York, 1967兲; W. J. Merz, J. Appl. Phys. 27, 938 共1956兲. 6 H. L. Stadler and P. J. Zachmanidis, J. Appl. Phys. 34, 3255 共1963兲. 7 J. F. Scott, IEICE Trans. Electron. E81–C, 477 共1998兲; Ferroelectr. Rev. 1, 1 共1998兲. 8 T. K. Song, S. Aggarwal, Y. Gallais, B. Nagaraj, and R. Ramesh, Appl. Phys. Lett. 73, 3366 共1998兲. 9 P. K. Larsen, G. L. M. Kampschoer, M. J. E. Ulenaers, and G. A. C. M. Spierings, Appl. Phys. Lett. 59, 611 共1991兲. 10 W. J. Merz, Phys. Rev. 95, 690 共1954兲. 11 Y. Ishibashi and Y. Takagi, J. Phys. Soc. Jpn. 31, 506 共1971兲; Y. Ishibashi, ibid. 59, 4148 共1990兲; V. Y. Shur, E. L. Rumyantsev, S. D. Makarov, A. L. Subbotin, and V. V. Volegov, Integr. Ferroelectr. 10, 223 共1995兲. 12 V. Shur, E. Rumyantsev, and S. Makarov, J. Appl. Phys. 84, 445 共1998兲. Downloaded 15 Mar 2004 to 128.32.120.53. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp 1 2