Dielectric relaxation and conduction in SrTiO3 thin films ... - Physics

3 downloads 194898 Views 62KB Size Report
veloping the frequency agile electronic devices, such as mi- crowave filters, resonators, and so on.1 .... Bi doped SrTiO3.12. Compared with the data, max(2280) ...
APPLIED PHYSICS LETTERS

VOLUME 79, NUMBER 6

6 AUGUST 2001

Dielectric relaxation and conduction in SrTiO3 thin films under dc bias Chen Ang,a) Zhi Yu, L. E. Cross, Ruyan Guo, and A. S. Bhalla Materials Research Laboratory, The Pennsylvania State University, University Park, Pennsylvania 16802

共Received 29 March 2001; accepted for publication 15 June 2001兲 The dielectric and conduction behavior of SrTiO3 thin films deposited on a SrTiO3 single-crystal substrate is studied. Without dc bias, an obvious dielectric ‘‘defect mode’’ in the dielectric loss is observed in the temperature range of ⬃100–200 K; however, no noticeable corresponding dielectric constant peak is observed. By applying a high dc bias 共⭓40 kV/cm兲, a dielectric constant peak with frequency dispersion appears in the same temperature range, the dielectric loss is increased, and simultaneously high dc conduction is observed. The induced dielectric constant peak is related to dc conduction and attributed to the coupling effect of the mobile carriers with the dielectric defect mode. © 2001 American Institute of Physics. 关DOI: 10.1063/1.1389771兴

High quality SrTiO3 thin films, with high field-tunable dielectric constant and low loss, are highly desirable for developing the frequency agile electronic devices, such as microwave filters, resonators, and so on.1,2 It is found that the dielectric loss of SrTiO3 thin films is increased by more than an order of magnitude from that of a bulk crystal, and that the dielectric constant is lower 共about 1000兲.3–7 Much effort has been devoted to reducing the dielectric loss. Recently with the progress in the preparation process of thin films, especially by pulsed laser deposition, high quality thin films have been prepared, with high ␧ max and low loss tan ␦.8 However, additional dielectric loss peaks with frequency dispersion were observed, for example, dielectric loss peaks in thin films deposited on SrTiO3 or LaAlO3 single-crystal substrates occurred around 150 or 70 K 共at 1 kHz兲, respectively.8,9 This increases the dielectric loss in the corresponding temperature range. In a previous letter,10 we reported the dielectric behavior of high quality SrTiO3 thin films with a high dielectric constant maximum ␧ max (⬃2280) and a low dielectric loss tan ␦ (⬃10⫺3 ) under high dc biases. However, we found that some samples having a lower dielectric constant maximum ␧ max (⬃1000) and higher dielectric loss tan ␦ (⬃10⫺2 ) display a dramatic increase in the loss under dc electric fields. In this letter, we discuss the dielectric behavior of these SrTiO3 thin film samples under dc biases. SrTiO3 thin film samples with thickness of 1 ␮m were prepared by the pulsed laser deposition technique on SrTiO3 single-crystal substrates. The samples adopt a parallel-plate capacitor structure, i.e., a Au/SrTiO3 thin film/ SrRuO3 /SrTiO3 single-crystal substrate. The complex dielectric permittivity was measured using an HP 4284A meter with an ac field of 100 mV/␮m. The temperature dependence of the dielectric properties was measured in a cryostat system 共ADP202 cryostat兲 in the temperature range 12–300 K. The dc voltage was applied to the sample and a blocking circuit was used to separate the high dc voltage from the HP4284A meter. The temperature dependence of the dielectric constant ␧ and dielectric loss tan ␦ for the SrTiO3 thin film is shown in a兲

Electronic mail: [email protected]

Figs. 1 and 2 without and with dc biases. At 0 kV/cm, as shown in Fig. 1共a兲, ␧ increases continuously with decreasing temperature from 300 to 12 K, attaining a value of 950 at 12 K. However, a set of tan ␦ peaks with frequency dispersive behavior occurred in the temperature range of 100–200 K 共hereafter denoted as mode I兲. With a further decrease in temperature, T⬍50 K, tan ␦ increases quickly, corresponding to the increase in ␧. The relaxation rate derived from the temperature dependence of the imaginary part of the permittivity for the sample is plotted in the inset of Fig. 1共a兲. The data were fitted to the Arrhenius law, v ⫽ v 0 exp共 E/k B T 兲 ,

共1兲

where v 0 is the relaxation rate at infinite temperature, E the activation energy for relaxation, and T is the temperature. The fitting parameters obtained are E⫽0.25 eV and v 0 ⫽ ⬃9⫻1011 Hz. It can be seen from Fig. 1 that at low dc biases (E ⭐20 kV/cm), the dielectric constant is suppressed by the application of dc bias. This behavior is similar to that observed in Ref. 10 except that ␧ is lower and tan ␦ is higher. In Ref. 10 the dielectric loss peak at ⬃150 K 共1 kHz兲 was assigned as the dielectric ‘‘defect mode,’’ which was attrib-

FIG. 1. Temperature dependence of ␧ and tan ␦ for the SrTiO3 thin film sample under different dc electric fields from 0 to 20 kV/cm at 0.1, 1, and 10 kHz 共␧ merges into one curve at different frequencies; tan ␦ peaks: from left to right兲.

0003-6951/2001/79(6)/818/3/$18.00 818 © 2001 American Institute of Physics Downloaded 15 Nov 2001 to 146.186.113.137. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp

Chen et al.

Appl. Phys. Lett., Vol. 79, No. 6, 6 August 2001

FIG. 2. Temperature dependence of ␧ and tan ␦ for the SrTiO3 thin film sample under different dc electric fields from 40 to 80 kV/cm at 1, 10, and 100 kHz 共␧ peaks: from top to bottom; tan ␦ peaks: from left to right for 40 kV/cm; for 60 and 80 kV/cm, only the tan ␦ data at 10 kHz are shown兲.

uted to the coupling effect between the dipoles and the soft mode of the host lattice of SrTiO3, This loss peak is suppressed and finally disappears with increasing dc bias. However, an interesting characteristic of the loss peak is that the temperature of the dielectric constant maximum is independent of dc biases. In the present work, in the absence of a dc bias and at low dc biases (E⭐20 kV/cm), the dielectric loss peak observed around 150 K has the same physical nature as that reported in Ref. 10. However, at higher electric fields (E⭓40 kV/cm), interestingly, an obvious dielectric constant peak with frequency dispersion appears in the temperature range where mode I occurs, as shown in Fig. 2. This is quite different from the results reported in Ref. 10, in which no noticeable dielectric anomalies occurred in the dielectric constant up to 80 kV/ cm. It is seen that the induced dielectric constant peak is strongly frequency dispersive. At 105 Hz, the ␧ peak almost disappears. This indicates that the induced dielectric constant peak is probably related to a low frequency relaxation process that originated from dc conduction. In order to study the physical mechanism of the induced dielectric constant peak, the frequency dependence of the

819

FIG. 4. Frequency dependence of the ␧, ␧⬙, and ac conductivity ( ␴ ac) as a function of temperature 共120, 140, 160, and 180 K兲 at 60 kV/cm.

dielectric constant ␧, the imaginary part 共␧⬙兲 of permittivity, and ac conductivity ( ␴ ac) under different dc biases of 10, 60, and 80 kV/cm is plotted in Figs. 3–5, respectively. As shown in Fig. 3, at 10 kV/cm, ␧ is almost independent of frequency in the frequency range of 102 – 105 Hz, similar to the observation at 0 kV/cm, the imaginary part of the permittivity ␧⬙ is also independent of frequency in the range of 102 – 104 Hz, and ␴ ac increases monotonically with increasing frequency. This is consistent with that observed in most insulating dielectric materials. At 60 and 80 kV/cm, shown in Figs. 4 and 5, ␧ exhibits a dramatic drop with increasing frequency in the low frequency range of 102 – 104 Hz. The slope of ␧⬙ versus frequency 共f 兲 is found to be near ⫺1 in the low frequency range. According to dielectric physics, the dielectric loss tan ␦ mainly consists of two contributions, one from the dielectric polarization relaxation process tan ␦relax , and the other from dc conduction tan ␦dc . So tan ␦dc can be expressed as tan ␦ dc⫽ ␴ dc / ␻ ␧,

共2兲

⬙ where ␻ ⫽2 ␲ f . Equation 共2兲 can be rewritten as ␧ dc ⬙ vs log ␻ 共or ⫽ ␴ dc / ␻ . This means that the slope of log ␧dc

FIG. 3. Frequency dependence of the ␧, ␧⬙, and ac conductivity ( ␴ ac) as a FIG. 5. Frequency dependence of the ␧, ␧⬙, and ac conductivity ( ␴ ac) as a function of temperature 共120, 140, 160, and 180 K兲 at 10 kV/cm. function of temperature 共120, 140, 160, and 180 K兲 at 80 kV/cm. Downloaded 15 Nov 2001 to 146.186.113.137. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp

820

Chen et al.

Appl. Phys. Lett., Vol. 79, No. 6, 6 August 2001

⬙ – log f plot with a log f 兲 is ⫺1. In Figs. 4 and 5, the log ␧dc slope value near ⫺1 at lower frequencies indicates that the dc conduction contribution is predominant. On the other hand, as shown in Figs. 4 and 5, ␴ ac is almost independent of frequency in the range of 102 – 104 Hz and only as frequency exceeds 104 Hz, does ␴ ac increase with increasing frequency. This also indicates an obvious dc conduction contribution at low frequencies. The ac conductivity can be described as ␴ ac⫽ ␴ dc⫹ ␻ s

共3兲

where s is a constant ranging from 0 to 1. The frequencyindependent part of ␴ ac gives rise to ␴ dc⫽4⫻10⫺5 and 15 ⫻10⫺5 (⍀ cm兲⫺1 at 60 and 80 kV/cm, respectively. The values indicate that the dc current increases with increasing electric field. The above results indicate that the dielectric constant peak induced by the dc bias is closely related to dc conduction. According to dielectric physics, the existence of dielectric loss peak-mode I implies a corresponding ␧ peak. However, at low electric fields (E⭐20 kV/cm) in this work, no visible dielectric anomaly in ␧ appears. This is due to the intensity of such a ␧ peak that is too small to be observed below 20 kV/cm. A possible explanation for the high dc bias induced ␧ peak is that the concentration of electrons in the sample is greatly enhanced under high dc biases, and that the electrons strongly couple with the existing dielectric peakmode I, and thus lead to the obvious dielectric peak in ␧. A similar explanation of the ‘‘electron-defect-mode’’ coupling model was also reported in La doped BaTiO3 共Ref. 11兲 and Bi doped SrTiO3. 12 Compared with the data, ␧ max(⬃2280) and tan ␦ (⬃10⫺3 ), of the SrTiO3 thin films without dc bias in Ref. 10, the sample in the present work has lower ␧ max(⬃950) and higher tan ␦ (⬃10⫺2 ). A question arises then as to whether the reason is the same for the lower ␧ max and higher tan ␦ under zero bias, and also for the occurrence of the dielectric constant peak under high dc bias. It is known that the dielectric constant is lower and the dielectric loss is higher in SrTiO3 thin films than that observed in bulk single crystals.6 – 8 It is commonly recognized that several possible factors in thin films, including defects, oxygen vacancies, high strain, interfacial layers, and nonstoichiometry, are the main reasons.13,14 In this work, it is reasonable to assume that there is a high density of oxygen vacancies, as well as

poor interfacial layers. These factors result in lower ␧ max . Furthermore, the ionization of oxygen vacancies produces weakly bonded electrons, which leads to high leakage current and increases the loss tan ␦. Under high dc biases, similar to the description in the electrical breakdown theory for dielectric materials, the weakly bonded electrons become more mobile, and the concentration of the electrons is increased, and thus contributes to the high dc conductivity. Consequently the coupling effect of electrons with the dielectric defect mode leads to the occurrence of the dielectric peak in ␧. In conclusion, we have observed dc bias induced dielectric relaxation behavior and high dc conduction in SrTiO3 thin films under high dc bias. The induced dielectric constant peak is tentatively attributed to the coupling effect of mobile carriers with the dielectric defect mode. The authors thank Dr. X. X. Xi and Dr. Jian-hua Hao for providing the samples. This work was supported by a grant from DARPA under Contract No. DABT63-98-1-002.

O. G. Vendik, Ferroelectrics 12, 85 共1976兲. O. G. Vendik, I. G. Mironenko, and L. T. Ter-Martirosyan, Microwaves RF 33, 67 共1994兲; O. G. Vendik, E. Kollberg, S. S. Gevorgian, A. B. Kozyrev, and O. I. Soldatenkov, Electron. Lett. 31, 654b 共1995兲. 3 A. B. Kozyrev, T. B. Samoilova, A. A. Golovkov, E. K. Hollmann, D. A. Kalinikos, V. E. Loginov, A. M. Prudan, O. I. Soldatenkov, D. Galt, C. H. Mueller, T. V. Rivkin, and G. A. Koepf, J. Appl. Phys. 84, 3326 共1998兲. 4 F. W. Van Keuls, R. R. Romanofsky, D. Y. Bohman, M. D. Winters, F. A. Miranda, C. H. Mueller, R. E. Treece, T. V. Rivkin, and D. Galt, Appl. Phys. Lett. 71, 3075 共1997兲. 5 R. E. Treece, J. B. Thompson, C. H. Mueller, T. Rivkin, and M. W. Cromar, IEEE Trans. Appl. Supercond. 7, 2363 共1997兲. 6 D. Galt, J. Price, J. A. Beall, and R. H. Ono, Appl. Phys. Lett. 63, 3078 共1993兲; D. Galt, J. Price, J. A. Beall, and T. E. Harvey, IEEE Trans. Appl. Supercond. 5, 2575 共1995兲. 7 D. Fuchs, C. W. Schneider, R. Schneider, and H. Rietschel, J. Appl. Phys. 85, 7362 共1999兲. 8 H.-C. Li, W. Si, A. D. West, and X. X. Xi, Appl. Phys. Lett. 73, 190 共1998兲; 73, 464 共1998兲. 9 X. X. Xi, A. M. Clark, J. H. Hao, and Weidong Si, Integr. Ferroelectr. 24, 239 共1999兲. 10 A. Chen, L. E. Cross, Y. Zhi, A. S. Bhalla, R. Guo, and J.-h. Hao, Appl. Phys. Lett. 78, 2754 共2001兲. 11 O. Bidault, P. Goux, M. Kchikech, M. Belkaoumi, and M. Maglione, Phys. Rev. B 49, 7868 共1994兲. 12 A. Chen and Y. Zhi, Phys. Rev. B 61, 11363 共2000兲. 13 C. Zhou and D. M. Newns, J. Appl. Phys. 82, 3081 共1997兲. 14 C. Basceri, S. K. Streiffr, A. I. Kingon, and R. Waser, J. Appl. Phys. 82, 2497 共1997兲. 1 2

Downloaded 15 Nov 2001 to 146.186.113.137. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp