Leakage Current-Voltage Characteristics of Ferroelectric Thin Film Capacitors

Transcription

1 Journal of the Korean Physical Society, Vol. 38, No. 6, June 2001, pp Leakage Current-Voltage Characteristics of Ferroelectric Thin Film Capacitors Kwangbae Lee and Byung Roh Rhee Department of Computer & Electronic Physics, Sangji University, Wonju Chanku Lee Optron-Tec Inc., Changwon (Received 28 December 2000, in final form 22 March 2001) The depolarization current-poling voltage (J D-V) characteristics of ferroelectric thin films, such as Pb(Zr 1 xti x)o 3 (x = 0.47, 0.70) (PZT) and SrBi 2Ta 2O 9 (SBT), have been investigated in order to interpret the true leakage current-voltage (J L-V) characteristics of ferroelectric thin films. We found that the contribution of the polarization current (J P ) in non-switching polarization states, as well as the additional one arising mainly from the partial switching polarization current, of PZT thin film capacitors was essentially included in the charging current-voltage (J C-V) curves measured by means of the conventional step-pulse method, as well as the staircase method. Hence, we propose a new method for measuring the leakage current, called the reversed step-pulse method, which can simultaneously determine both J C-V and J D-V at a given temperature. In addition, we proposed a method for finding the J L-V curves by subtracting the J P -V curves from the J C-V curves and discuss the true leakage conduction of PZT thin-film capacitors, which was determined to be due to a Poole-Frankel emission rather than a Schottky emission. I. INTRODUCTION The development of high-density nonvolatile ferroelectric memories (NvFRAM) has been one of the interesting subjects in the field of memory devices [1]. Ferroelectric Pb(Zr,Ti)O 3 (PZT) and SrBi 2 Ta 2 O 9 (SBT) thin films are promising NvFRAM materials, which have been tentatively studied in order to improve their properties including polarization-field (P-E) characteristics, degradations such as fatigue, retention and imprint, and leakage current. Extensive research has been devoted to understanding or improving/removing the P-E hysteresis characteristics and their degradations [2 4]. Meanwhile, although a variety of physical models describing the leakage current behaviors of ferroelectric thin films have been proposed and experimental data supporting these mechanisms have been presented in diverse publications, comprehensive understanding of them is not established yet available. The proposed mechanisms of leakage conduction for PZT are bulk ionic conduction [5], Schottky emission [6], space-charge-limited conduction (SCLC) [7], and mixed Schottky and Pool-Frankel emission [8]. Such a variety of experimental results and interpretations concerning the leakage conduction of ferroelectric thin film capacitors can be attributed to sev- kblee@mail.sangji.ac.kr eral factors, including the intrinsic effects due to different measuring techniques and the extrinsic effects due to different electrode/ferroelectrics interfaces, such interdiffusion layers, oxygen depletion layers and interface traps. Moreover, conclusions based only on analyzing the shapes of the J-V curves can be ambiguous because in many cases, different conduction mechanisms result in similar J-V curves. The main reason for this is that the true leakage current in ferroelectric films is hard to measure directly. The current measured for ferroelectric thin-film capacitors is the charging current (J C ) of the test capacitors, which includes the fully switching current (J S ), the polarization current (J P ), and the true leakage current (J L ). A current due to the aging effect (J A ) can also influence the measured charging current. Hence, we write J C as follows: J C = J S + J A + J P + J L (0) In Eq. (1), J S contributes during the few microseconds in the fully switching state while J A appears normally at high fields and/or at high temperature and can be discriminated from the J-t curves. Hence, for a given time interval, temperature, and measuring time, the measured J C is dominated by J P and J L : J C = JP + J L (0) However, the current response in ferroelectric films, in

2 -724- Journal of the Korean Physical Society, Vol. 38, No. 6, June 2001 II. EXPERIMENT Fig. 1. P-V hysteresis loops of (a) PZT(53/47), (b) PZT(30/70) and (c) SBT capacitors for an applied bias starting at 1 V and increased to 7 V in 1-V steps. cluding PZT and SBT, as well as in high-permittivity films, such as SrTiO 3 (STO) [9] and Ba(Sr,Ti)O 3 (BST) [10], includes a time-dependent polarization current, J P (t), of dielectric relaxation. This time-dependent current depends strongly on the measuring technique and the prehistory of sample treatment, such as the applied bias and illumination. In addition, there is no clear evidence whether the true leakage current (J L ) is a steadystate one or not. Recently, Stolichnov and Tagantsev [11] asserted that the true leakage current of PZT thin films was time dependent because of the influence of injected charge entrapment during the measurement. However, no direct evidence has been given. In this paper, we show the importance of the choice of measuring method for extracting the true leakage current-voltage characteristics of ferroelectric thinfilm capacitors from their measured charging currentvoltage curves. We also show that J C can be influenced by the ferroelectric polarization states. Three different type of ferroelectric thin-film capacitors, Pb(Zr 0.53 Ti 0.47 )O 3 (PZT53/47) having high P r and low E c, Pb(Zr 0.3 Ti 0.7 )O 3 (PZT30/70) having high P r and high E c, and SBT having low P r and low E c, were used as test capacitors for this purpose. As a result, we propose a new method of current measurement, called the reversed step-pulse method, in order to estimate J P from J D qualitatively, which can determine both J C -V and J D -V simultaneously at each temperature. An attempt was made to find the true leakage current (J L ) from J C by separating J P for a PZT(53/47) capacitor and to discuss the mechanism for J L. PZT and SBT thin films were prepared on platinized silicon wafers by using the sol-gel method, which is described elsewhere [12,13]. The thickness of all films was about 250 nm. Pt-top electrodes with areas of cm 2 and thickness of 150-nm thickness were deposited by using the dc magnetron sputtering method with a shadow mask. The P-E hysteresis loop was measured using a RT66A standardized ferroelectric tester (Radiant Technologies). Charging and discharging currents were measured using an electrometer/source (Keithley 617). The temperature dependence of these currents was measured using a programmable temperature controller. The measurement control and data acquisitions were performed using self-made software. Three different methods were used for the study of the J-V relations: (i) the staircase method, (ii) the steppulse method, and (iii) the reversed step-pulse method. The first two methods have been used by many workers, and are described in detail elsewhere [11]. The last one is proposed to eliminate the contribution of the current arising from partially switching polarization states. The only difference between the second and the third methods is the method of applying the bias, i.e., increasing or decreasing from the start to the end voltages. In the reversed step-pulse method used in this study, the start voltage of 7.0 V was decreased to the end value of 0.0 V in steps of 0.2 V. Both in the second and the third methods, the voltage profile consists of successive steps separated by intervals of the same 10-s duration (t d ). During these intervals, the test capacitors were kept under a short-circuited condition during the same t d. The J-t characteristics of both the charging and the discharging currents were collected during t d with an interval of 1 s, and the J-V characteristics were collected at the end of t d. III. RESULTS AND DISCUSSION 1. J-V Characteristic of Ferroelectric Thin-Film Capacitors Measured by the Conventional Step-Pulse Method Figure 1 shows the P-V hysteresis loops of the ferroelectric test capacitors used in this study. The applied voltage dependence of the P-E hysteresis loops of all capacitors is shown for voltages from 1 V to 7 V in 1-V increments. Well-defined P-V loops were obtained for the test capacitors at applied voltages larger than 2 V. The values of P r and E c at an applied voltage of 7 V for PZT(53/47) [Fig. 1(a)], PZT(30/70) [Fig. 1(b)], and SBT [Fig. 1(c)] are 26 µc/cm 2 and 45 kv/cm, 47 µc/cm 2 and 115 kv/cm, and 8 µc/cm 2 and 51 kv/cm,

3 Leakage Current-Voltage Characteristics of Ferroelectric Thin Film Capacitors Kwangbae Lee et al Fig. 3. Typical P-E hysteresis loops for elevated bias voltages in ferroelectric thin-film capacitors. In the figure, the point O denotes the initial zero polarization state, and the A-B curve denotes the non-switching polarization states after applying V(A). Fig. 2. J C-V and J D-V curves measured by means of the conventional step-pulse method at elevated temperature of 298 to 448 K in increments of 25 K in (a) PZT(53/47), (b) PZT(30/70), and (c) SBT capacitors. In the figure, the arrows denote the shoulders attributed to the partially switching polarization current. respectively. Such P-V hysteresis characteristics are expected, as mentioned in Section I. Hence, we confirmed that those capacitors were appropriate for the study of the ferroelectric polarization effect on the leakage current behavior of ferroelectric thin-film capacitors. Figure 2 shows the typical J C -V curves measured by means of the conventional step-pulse method at elevated temperatures ranging from 298 K to 448 K in increments of 25 K. In Fig. 2, the negative value of J D was also plotted as a function of the poling voltage (V). Shoulders (denoted by arrows) can be seen in both the J C -V and the J D -V curves for PZT(53/47) [Fig. 2(a)], PZT(30/70) [Fig. 2(b)], and SBT [Fig. 2(c)]. These shoulders decreased slightly with increasing temperature. These shoulders are referred as crossover voltages associated with the crossover from the ohmic to Schottky region [6]. However, we suggest that these shoulders in both the J C -V and the J D -V curves should be closely related to the polarization states of the ferroelectric thin-film capacitors because no shoulders were found in either the J C -V or the J D -V curves for non-ferroelectric SrTiO 3 capacitors (omitted). The polarization states in ferroelectric thin-film capacitors may be influenced by the bias applied during the J-V measurement. As an illustration, the polarization state corresponding to the conventional step-pulse bias used for the J-V measurements in Fig. 2 has to follow a typical O-A curve, as shown in Fig. 3. If that happens, the polarization current (J P ) is determined by the poling rate of polarization along such an O-A curve whereas the depolarization current (J D ) is determined by the rate of the polarization along an O-B line. Here, one can expect J P J D, but can hardly find their exact relations. Actually, however, the initial polarization states may not be at the point O, but possibly at any point along the O-B line, for example, O. Hence, the partial switching current can be attributed to the measured charging current. 2. J-V Characteristics of Ferroelectric Thin- Film Capacitors Measured by Reversed Step-Pulse Method The above results of Fig. 2 and Fig. 3 imply that neither the step-pulse method nor the staircase one is available for the study of the J-V characteristics of ferroelectric thin-film capacitors and that the current measurement has to be performed for well-defined polarization states. For this purpose, we devised a new method for measuring the charging current, as well as the discharging current, in ferroelectric thin-film capacitors, called the reversed step-pulse method. The basic principle of this method is that the initial polarization state of the test capacitor starts at point A (fully switching state) in Fig. 3 with the start pulse of V(A); then, the polarization states during the current measurement are confirmed to be in non-switching polarization states along the A-B curve in Fig. 3. Figure 4 shows the typical J C -V curves of PZT(53/47) [Fig. 4(a)], PZT(30/70) [Fig. 4(b)], and SBT [Fig. 4(c)] capacitors measured by using the reversed step-pulse method at elevated temperatures in increments of 25 K. The negative of J D was also plotted as a function of the poling voltage. Figure 4 shows that no shoulder exists in either the J C -V or the J D -V curves of any test capacitor. Moreover, J D was almost constant in the poling voltage of the step pulse.

4 -726- Journal of the Korean Physical Society, Vol. 38, No. 6, June 2001 Fig. 5. Plot of J D vs 1000/T for both PZT(53/47) and PZT(30/70) capacitors. Fig. 4. J C-V and J D-V curves measured by means of the reversed step-pulse method at elevated temperature of 298 to 448 K in increments of 25 K in (a) PZT(53/47), (b) PZT(30/70) and (c) SBT capacitors. J C -V and J D -V curves similar to those shown in Fig. 4 can be obtained by using a non-switching pulse applied before using the conventional step-pulse method to measure J C and J D [14]. However, the optimum conditions for measuring the current were hard to determine in this case because the resultant current measurement depended on the duration and the voltage of the nonswitching pulse and high voltage, and a long duration time, even at low temperature, could cause degradations in the leakage current characteristics of the test capacitors. Actually, in this method, the partial switching polarization current can also be found, even though such contributions were smaller than those observed in J C - V curves measured by using the conventional step-pulse method without non-switching pulse. This result implies that the polarization state during the current measurement may not be followed by perfect non-switching polarization states, such as the A-B curve in Fig. 3, but may be followed by unidentified non-switching polarization states, such as the A-B curve, as typically shown in Fig. 3. We assumed that such a partial switching current was dominated by induced dipoles, which might cause a space-charge effect in ferroelectric films. 3. Leakage Current Mechanism of Ferroelectric Thin-Film Capacitors As mentioned above, the reversed step-pulse method, which can measure J C as well as J D simultaneously, confirms the current measurement held in non-switching po- larization states, as shown in Fig. 4. Hence, we can regard J C as the sum of J P and J L, as written in Eq. (2), under the condition the aging effect in the measured current is excluded. Hereafter, we refer to J P as the ferroelectric polarization current in non-switching polarization states. In order to interpret the leakage current mechanism of PZT capacitors, we tried to separate J L from J C as follows: Since the contribution of J P to J C was hard to find, we used the depolarization current (J D ) as a measure of J P. That is, the only possibility of finding the relation between J P -V and J D -V was believed to be the empirical determination of the J P -V curves from the shapes of the J D -V curves. However, as seen in Fig. 4, the contribution of J P to J C might only be dominant at low temperatures of less than about 348 K. This was compatible with the result that the activation energy (E a ) of J D for PZT(53/47) and PZT(30/70) is about 0.05 ev, as shown in Fig. 5, which is smaller than that of J C by about one order. The most probable mechanisms for J L in this study are Schottky emission and Poole-Frankel emission. The distinction between these two mechanisms is based on whether the effect is interface- or bulk-limited and on the respective activation energies, which can be related to differences in the work function between the electrode material and the dielectric or possible traps in the dielectric. If the leakage is assumed to be controlled by the reverse-biased interface, the current for a Schottky diode is given by [15] J L = A T 2 exp ( α S E 1/2 W b ) ; αs = ( q 3 ) 1/2 (3) k B T 4πɛ r ɛ 0 where A is the effective Richardson constant, which incorporates the carrier mobility, k B is the Boltzmann constant, T is the temperature, W b is the zero-field interfacial barrier height, q is the charge of the carriers,

5 Leakage Current-Voltage Characteristics of Ferroelectric Thin Film Capacitors Kwangbae Lee et al Table 1. Physical parameters for PZT(53/47) and PZT (30/70) capacitors, such as barrier height (W b ), trap ionization energy (W I), and high-frequency relative permittivity (ɛ r), obtained by fitting the true leakage current (J L) to the Schottky emission model and to the Poole-Frankel (PF) emission model. Parameter Schottky model PF model Specimen W b (ev) ɛ r W I (ev) ɛ r PZT(53/47) PZT(30/70) Fig. 6. Typical fitting of J P and J L to J C in order to find the true leakage current, J L, at temperatures of 298 K and 323 K for a PZT(53/47) capacitor. Here, J L was fitted using Eq. (5), which has the same form for Schottky emission and Poole-Frankel emission. ɛ r is the high-frequency permittivity of the ferroelectric material, and ɛ 0 is the permittivity of free space. Poole- Frankel (PF) emission is due to the field-assisted thermal ionization of trapped carriers in the conduction band of an insulator, and the PF current is given by [16] J L = c exp ( α P F E 1/2 W I ) ; αp F = ( q 3 ) 1/2 (4) k B T πɛ r ɛ 0 where W I is the trap ionization energy. Both in Eq. (3) and Eq. (4), αe 1/2 is the barrier lowering due to the applied field. Using Eq. (3) and Eq. (4), we fitted a curve of the measured current J C by using two contributions, J P and J L, as follows: J C = J P + J L = λj D + s exp(βe 1/2 ) (5) where J D is the measured value of depolarization current, as shown in Fig. 4, and λ, s, and β are fitting parameters, where s = A T 2 exp( W b /k B T ) and β S = α S /k B T for Schottky emission (6a) Fig. 7. Plot of (a) ln(s/a T 2 ) vs. 1000/T and (b) ln s vs. 1000/T for PZT(53/47) and PZT(30/70). s = c exp( W I /k B T ) and β P F = α P F /k B T for PF emission. (6b) Figure 6 shows typical figures for the contributions of J P and J L to J C of a PZT(53/47) capacitor at relatively low temperatures of 298 K and 323 K, where J L was fitted using Eq. (5). For a proper curve fit using Eq. (5), the data of J C for a bias of less than 1 V were excluded in order to avoid confusion about the conduction mechanism in the low-field region. The rapid drop of J C for biases lower than 1 V was explained by means of ohmic conduction [6,17,18], but was presumed to be due to the negative contribution of J P to J C. Actually, negative charging currents could be observed for bias voltages of less than 0.5 V even at high temperature, and were larger than the noise level of about A for the current measurement. Because the form of J L is the same for both the Schottky model and the PF model, as seen in Eq. (5), the obtained values of fitting parameters, such as λ, s, and β (= β S and β P F ), can be used for both models. Hence, the value of ɛ r for the PF model is four times that for Schottky model, i.e., ɛ r (PF model) = 4ɛ r (Schottky model). This functional relationship means that the barrier lowering, αe 1/2, in the case of PF emission is twice the Schottky-barrier lowering, because the distance between the escape electron and the trap site is half that between the escape electron and the image charge in the case of Schottky emission. Physical parameters, such as W b and W I, were determined from plots of ln(s/a T 2 ) vs. 1000/T [Fig. 7(a)] and plots of ln s vs. 1000/T [Fig. 7(b)], respectively, for both PZT(53/47) and PZT(30/70). The obtained values of W b and W I, including the values of ɛ r, are written in Table 1. In the case of PZT(30/70), the contribution of J P to J C seems to be smaller than that in PZT(53/47), which may be attributed to the larger coercive field of PZT(30/70). Hence, the λ in Eq. (3) for PZT(30/70) is fixed as 1 while the λ for PZT(53/47) ranges from 0.3 to 1.0. At this point, we will try to determine the proper conduction mechanism for PZT capacitors. The refractive index (n) of a PZT(30/70) film at a wavelength of 630 nm is 2.35 [19] ; in addition, the high-frequency permittivity (n 2 ) of (Pb,La)TiO 3 and BaTiO 3 films at a

6 -728- Journal of the Korean Physical Society, Vol. 38, No. 6, June 2001 wavelength of 400 nm is about 6.5 [20]. Hence, if the ferroelectric films are assumed to have a similar highfrequency permittivity (ɛ r = n 2 ) of around 6, the obtained results of ɛ r for Poole-Frankel emission, as seen in Table 1, indicate that Poole-Frankel emission is more probable than Schottky emission for the leakage conduction of PZT capacitors. However, the ɛ r of PZT(53/47) is still smaller at 2.4. In addition, the ionization energies of the traps (W I ) for PZT(53/47) and PZT(30/70) are as 1.29 ev and 0.80 ev, respectively. Hence, if the conduction of PZT films is dominated by a unique mechanism, such as Poole-Frankel emission, despite the compositional difference of the Zr/Ti ratio, such differences in ɛ r and W I imply that the origins of the bulk defects for PZT(53/47) and PZT(30/70) are quite different from each other. Such different bulk defects may cause the different P-V hysteresis characteristics shown in Fig. 1. IV. SUMMARY AND CONCLUSION In the present study, using our new proposed method called the reversed step-pulse method, we found that the effect of a partial switching polarization current could be successfully excluded, so the polarization current could be confirmed as only following non-switching polarization states during the measurement. We found that the charging current-voltage curves of PZT(53/47) and PZT(30/70) measured by the reversed step-pulse method consisted of two main contributions, i.e., the polarization current and the pure leakage current. From a parametric analysis of the high-frequency permittivity, we showed that the Poole-Frankel emission is more probable than the Schottky emission for pure leakage conduction. The values of the trap ionization energies for PZT(53/47) and PZT(30/70) capacitors having Pt electrodes were estimated as 1.29 ev and 0.80 ev, respectively. In summary, the measured charging current in ferroelectric thin film capacitors is dominated by the bulk effect, which includes the polarization current under non-switching polarization states and the Poole-Frankel emission. While the former is dominant for low fields and low temperatures, the latter is dominant for high fields and/or high temperatures. Conclusively, our reversed step-pulse method is available as a standard method for studying the leakage current characteristics of ferroelectric thin-film capacitors for NvFRAM applications. REFERENCES [1] J. F. Scott and C. A. Paz de Araujo, Science 246, 1400 (1989). [2] I. K. Yoo, S. B. Desu and J. Xing, Ferroelectric Thin Films III (Material Research Society Symposium Proceedings 310, Pittsburgh, 1993), p [3] W. L. Warren, D. Dimos and R. Waser, MRS Bulletin 21, 40 (1996). [4] W. M. Warren, D. Dimos, B. A. Tuttle, G. E. Pike, R. W. Schwartz, P. J. Clews and D. C. McIntrye, J. Appl. Phys. 77, 6695 (1997). [5] R. Moazzami, C. Hu and W. Shepherd, in Proceedings International Reliability Physics Symposium, (1990), p [6] C. Sudhama, A. C. Campbell, P. D. Manier, R. E. Jones, R. Moazzami, C. J. Mogab and J. C. Lee, J. Appl. Phys. 75, 1014 (1994). [7] J. F. Scott, B. M. Melnick, J. D. Cuchiaro, R. Juleeg, C. A. Paz de Araujo, A. D. McMillan and M. C. Scott, Intergr. Ferroelectr. 4, 85 (1994). [8] T. Mihara and H. Watanabe, Jpn. J. Appl. Phys. 34, 5664 (1995). [9] G. W. Dietz, W. Antpohler, M. Klee and R. Waser, J. Appl. Phys. 78, 6113 (1995). [10] G. W. Dietz, M. Schmacher, R. Waser, S. K. Streiffer, C. Basceri and A. I. Kingon, J. Appl. Phys. 82, 2359 (1997). [11] I. Stolichnov and A. Tagantsev, J. Appl. Phys. 84, 3216 (1998). [12] K. B. Lee, S. Tirumala and S. B. Desu, Appl. Phys. Lett. 74, 1484 (1999). [13] K. B. Lee and B. K. Ju, Thin Solid Films 334, 65 (1998). [14] Y. S. Yang, S. A. Lee, S. H. Kim, B. G. Chae and M. S. Jang, J. Appl. Phys. 84, 5005 (1998). [15] A. M. Cowly and S. M. Sze, J. Appl. Phys. 36, 3212 (1965). [16] J. J. O Dwyer, The Theory of Electrical Conduction and Breakdown in Solid Dielectrics (Clarendon, Oxford, 1973), p [17] B. Nagaraj, S. Aggawal, T. K. Song, T. Sawhney and R. Ramesh, Phys. Rev. B 59, 3212 (1965). [18] C. J. Peng, H. Hu and S. B. Kurpanidhi, Appl. Phys. Lett. 63, 1038 (1993). [19] C. H. Peng, J. F. Chang and S. B. Desu, Ferroelectric Thin Films II (Material Research Society Symposium Proceedings 243, Boston, MA, 1991), p. 21. [20] K. H. Hellwege and O. Madelung, Ferroelectrics and Related Substance. Non-Oxides (Springer-Verlag, Berin, 1991), p. 232.