A ferroelectric memristor based on the migration of oxygen vacancies

Abstract

Ferroelectric resistive switching memory is a non-destructive and easy to achieve multilevel storage, which is a breakthrough for further improving the density in the random access memory. However, the application of ferroelectric resistive switching memory is limited by the high operating voltage, the low switching ratio or slow write/read speed. Herein, we show a type of memristor with a thin ferroelectric film, the device can switch its resistance states by controlling the oxygen vacancy migration in the effect of an external electric field. The device still exhibits stable resistive switching phenomena after an endurance test of about 100 cycles and has high switch ratio about 10⁸% and good retention about 1.7 × 10⁵ s. Furthermore, the device has the potential of being a multi-states memory with a low write voltage below 2 V and a fast write/read speed of about 5 μs. Our results suggest new opportunities for the development of high storage density nonvolatile memory.

---

1. Introduction

Ferroelectric materials, having two stable spontaneous polarization states due to their low crystal symmetry, are a natural choice for binary code-based nonvolatile random access memory (RAM) applications.^1,2 However, the application of ferroelectric random access memory (FRAM) based on 1T1C is largely restricted by the destructive read. Ferroelectric tunnel junctions (FTJs) based on 1T can realize a non-destructive read, but is has a problem of retention.^3–9 In addition, high-quality FTJ demands epitaxial multilayer preparation, which causes high cost and limits mass production. Furthermore, the ON/OFF current of FTJs is too small, only special devices, such as conductive atomic force microscopy (CAFM) or piezoelectric atomic force microscope (PFM), can distinguish the different states. Furthermore, the retention of most FTJs is unsatisfactory, which also limits the application of the devices. The other concept realizing non-destructive read with ferroelectric materials is Schottky barrier variations between film interfaces. However, these devices are with high operate voltage and low switch ratio.^10–14 Therefore, it is difficult to achieve a high switch ratio, easy write/read and low operate voltage at the same time in ferroelectric devices based on the two abovementioned concepts. Therefore, it is necessary to develop a memory device that has high switching ratio, low write voltage and fast write speed with ferroelectric materials. Based on an investigation of the metal oxide memristor, it has been found that the migration of oxygen vacancies (or oxygen ions) plays an important role in realizing a high switch ratio and easy write/read.^15–19 In addition, the devices with thin film are much easier to achieve a low write voltage and high write speed. Is it possible to control the resistance states by controlling the migration of oxygen vacancies (or oxygen ions) in thin ferroelectric film? If it is possible, we can design a memristor with an easy write/read, high switch ratio, low operate voltage, fast write speed, and even more multi-states storage. All these will promote the application of ferroelectric storage devices.

Herein, we investigate a controllable filamentary memristor based on a PbZr_0.52Ti_0.48O₃ (PZT) thin film in a metal–ferroelectric–metal (MFM) structure. The resistive switching behaviors, which exhibit different resistive switching characteristics and memory performances, are investigated in detail. The possible transformation mechanisms are proposed systematically. The devices have high switching ratio, low write voltage and fast write speed. By controlling the preparation conditions and write processes, we are able to demonstrate multi resistance states in the ferroelectric memristor.

2. Experimental section

2.1 Sample preparation

30 nm thick PZT films were grown by laser molecular beam epitaxy (LMBE) on Pt-coated silicon. During deposition of the PZT layers, the substrate temperature was maintained at 625 °C with a chamber oxygen pressure maintained at 10 Pa, 20 Pa and 30 Pa and the samples were cooled to room temperature in same oxygen atmosphere. Pt and Ag dot electrodes were fabricated by d.c. sputtering through a shadow mask about 100 μm in diameter at room temperature.

2.2 Characterization

The film thicknesses were measured on Filmetrics model F20-UV and F50-UV. The current–voltage (I–V) measurement characteristics were measured using an Agilent B1500A semiconductor device analyzer. The XRD results have been performed using X-ray diffraction (D/MAX-2500/PC).

3. Results and discussion

Fig. 1(a) presents the structure of Pt/PZT/Pt and Ag/PZT/Pt. The low-magnification transmission electron microscopy (TEM) image for the cross section of the PZT/Pt shows that the thickness of PZT film is about 30 nm in Fig. 1(b). Structural characteristics of the films carried out using X-ray diffraction (XRD) in Fig. 2 present peaks at the (101), (111), (003) and (113) planes of PZT.

Fig. 1 (a) Structure of Pt/PZT/Pt and Ag/PZT/Pt. (b) TEM image for the cross section of PZT/Pt. The PZT layer was prepared at an oxygen pressure of 20 Pa.

Fig. 2 XRD patterns of PZT/Pt heterostructures and Pt substrate at various 2θ. The PZT layers were prepared at oxygen pressure of 10, 20 and 30 Pa.

The measured I–V curve of Pt/PZT/Pt memory cell (with a PZT layer prepared at 20 Pa oxygen pressure) with Pt dot electrodes is shown in Fig. 3. The corresponding current is measured at a d.c. voltage V sweeping in a sequence of 0 V → −2 V → 0 V → +2 V → 0 V. The devices show rectifying and hysteretic I–V characteristics indicative of resistive switching. It has two pronounced changes in resistance from the high resistance state to the low resistance state at 0.8 V and −1.3 V. The I–V characteristics of the Ag/PZT/Pt memory cell with a PZT layer prepared at a 20 Pa oxygen pressure were studied by DC voltage sweep measurements to evaluate the memory effects of the obtained devices and the results are illustrated in the semi logarithmic (Fig. 3) scales. During the measurements, neither a forming process nor current compliance was necessary for activating the memory effect. The voltage was swept in a sequence of 0 V → −2 V → 0 V → +2 V → 0 V. By the steady increase in the negative voltages imposed on the device, a pronounced change in resistance from the high resistance state (HRS or OFF) to the low resistance state (LRS or ON) was observed at about −1.3 V, which is called the “SET” process. Subsequently, an opposite “RESET” process could also be observed when the voltage is swept reversely to positive values about 1.6 V. The OFF/ON resistance ratio is high to about 10⁸%, which is relative high comparing with other ferroelectric memory.^3,6–13 It is an ultrahigh memory margin, making the periphery circuit very easy to distinguish the storage information (‘1’ or ‘0’).

Fig. 3 Current of Pt/PZT/Pt and Ag/PZT/Pt films at various applied voltages. The PZT layer was prepared at an oxygen pressure of 20 Pa.

Because the filamentary rupture controlled by Joule heating shows no polarity dependence in phase-change memory, thermal effects are not expected to play a dominant role in the SET process of the present device, even if they exist.¹⁶ The resistive switching effect in our case is very likely to be mediated by another way. Because the current of Ag/PZT/Pt is much larger than that of Pt/PZT/Pt at the same voltage, which cannot be explained by the ferroelectric resistance switching discussed in the former reports.^10–14 When the negative electric field is applied on the devices, the LRS of Ag/PZT/Pt stays stable, but the LRS of Pt/PZT/Pt disappears. All these might serve as a clue to the filamentary switching in Ag/PZT/Pt. The active medium is in fact much smaller than the cell size, providing a potential of scaling. As reported previously, the migration of oxygen vacancies (or oxygen ions) under an applied electrical field plays an important role in metal oxide materials.^10–20 Oxygen ions can be stable with Ag as AgO_x, and it may demonstrate the special resistance switching on Pt/PZT/Pt and the filamentary resistive switching on Ag/PZT/Pt. On the other hand, it may explain that V_RESET is rather larger than V_SET of Ag/PZT/Pt because the reduction reaction of AgO_x will expend much more energy in the RESET process. This is different to what occurs in the Ag/ZnO:Mn(ZMO)/Pt device, which exhibits resistance switching due to the formation and rupture of nanoscale Ag bridges in the ZMO films.²¹ Because the formation process is determined by the competition among different filamentary paths²² and the ions migration potential is affected by lattice constant. Considering the atomic radius, oxygen ions and oxygen vacancies might be migrating much more positive in the PZT films. Even more in the device with Ag filamentary paths, the V_SET is always larger than V_RESET. In our case, it may not be the Ag bridges determining the resistance switching like Ag/ZMO/Pt and it is much like oxygen vacancies filament.

To verify that the migration of oxygen ions (oxygen vacancies) is the main reason in our device, further tests were performed. In the tests, current compliance was necessary to control the amount of transfer ions (vacancies). After a RESET process, the device was set to the HRS and the voltage was swept from 0 V to 0.3 V without current compliance. The voltage was then swept from 0 V to 0.3 V without current compliance 3 times to read the resistance states after the resistance states were written with voltage sweepings from 0 V to −2 V with current compliances of 10⁻⁷ A, 10⁻⁴ A and 10⁻² A. As shown in Fig. 4(a), we achieved four resistance states at last. In the Ag bridge system like Ag/ZMO/Pt,²¹ the Ag bridge determines the resistance states. If the bridge is good, it will be the LRS, when the bridge is broken, it will be HRS. Furthermore, if the intermediate states may be stable between HRS and LRS, the current compliances of 10⁻⁴ A and 10⁻² A are higher than the current at which the Ag bridge may form (about 10⁻⁹ to 10⁻⁶ A in the sweeping from 0 V to −2 V in our case if it exists), so the resistive switching is not mainly caused by the Ag bridge. In addition, the Ag bridge device may need a forming process, but our devices do not need this. In a PZT film, oxygen vacancies are the main factor for the leakage current. If the density of oxygen vacancies increases, the leakage current would be much larger at a constant voltage. The stable intermediate states prove the migration of oxygen ions (vacancies) in our device and prove the device is not only a resistive switching memory, but also a memristor.

Fig. 4 (a) Four states of the device after different write process. ‘0’ was achieved after a RESET process. ‘1’, ‘2’ and ‘3’ were achieved after processes from 0 V to −2 V with a current compliance of 10⁻⁷ A, 10⁻⁴ A and 10⁻² A, respectively. The PZT layer was prepared at an oxygen pressure of 20 Pa. (b) Six states of the device after different processes. ‘0’ was achieved after a RESET process. ‘1’, ‘2’, ‘3’, ‘4’and ‘5’ were achieved after the write processes from 0 V to −2 V with current compliance of 10⁻⁶ A, 10⁻⁵ A, 10⁻⁴ A, 10⁻³ A and 10⁻² A, respectively. The PZT layer was prepared at an oxygen pressure of 10 Pa.

One important thing is that the preparation conditions largely affect the stable resistance states that the devices can achieve. If the oxygen pressure is much lower in the preparation of PZT, more stable resistance states will be achieved. We show six states of a device with a PZT film prepared at 10 Pa oxygen in Fig. 4(b), but its ratio of the highest resistance and the lowest resistance is lower than the former, which was with PZT film prepared at 20 Pa oxygen. The oxygen pressure affects the original number of oxygen vacancies in the PZT layer and oxygen vacancies affect the electrical conductivity. The PZT films prepared at a lower oxygen pressure have much more oxygen vacancies causing a lower ratio of the highest resistance and the lowest resistance, but increasing the possibility of oxygen ion migration. Fig. 5(a) shows the current of the device with the 30 nm PZT layer prepared at an oxygen pressure of 30 Pa at various applied voltages. This is like the Pt/PZT/Pt device, which means that the initial oxygen vacancy concentration is important for realizing the resistance switching. If the Ag/PZT/Pt is caused mainly by Ag migration, the initial oxygen vacancy concentration may affect the resistance switching slightly. Both the initial oxygen vacancies in the PZT film and the oxygen vacancies formed in the electric field are important in resistance switching. In the write processes of the multi-level device, the current compliances are necessary to control the migration of oxygen ions (vacancies), but the step widths of the write voltages are unnecessary. Fig. 5(b) shows the write processes of the device prepared at 10 Pa oxygen with different step width. No matter in SET or RESET process, 1 step, 4 steps, 20 steps or 50 steps are all the same to achieve the HRS or LRS, which is beneficial for realizing ultrafast write. Especially, 1 step write operation only needs about 5 μs. Otherwise, the operating voltage is below 2 V, which is lower than many ferroelectric memory devices.^6–13

Fig. 5 (a) Current of the device with the PZT layer prepared at an oxygen pressure of 30 Pa at various applied voltages. (b) Write processes with different step numbers of device with the PZT layer prepared at an oxygen pressure of 10 Pa.

The possible transformation mechanisms are shown in Fig. 6. Two steps are needed in the realization of the high on/off ratio. In the first step, the oxygen ions by the interfaces between Ag electrode and PZT layer are absorbed by the Ag ions in the effect of applied electrical field and many more oxygen vacancies are formed in the PZT layer (the formation of the oxygen vacancies is caused by oxidation reaction, Ag + O → AgO_x). In the second step, the oxygen vacancies will migrate in the PZT film and change the ferroelectric barrier with the initial oxygen vacancies together with the effect of the electric field. Both the width and height of the ferroelectric barrier are affected by these oxygen vacancies causing switching from the HRS to LRS, as shown in Fig. 6. Therefore, a high on/off ratio is reasonable and it is very different with the ferroelectric resistance switching devices in former reports, which are based only on the initial oxygen vacancies in the preparation of ferroelectric films. In our case, the resistance of the HRS, R_HRS = R_TE1 + R_BE + R_PZT (R_TE1 is the resistance of Ag top electrode, R_BE is the resistance of Pt below electrode and R_PZT is the resistance of PZT without oxygen vacancies filament), and the resistance of the LRS, R_LRS = R_TE2 + R_BE + R_PZTF (R_TE2 is the resistance of Ag top electrode with AgO_x, R_PZTF is the resistance of PZT with oxygen vacancies or filament). While R_TE2 > R_TE1 and R_PZT ≫ R_PZTF, it may be difficult to compare R_HRS and R_LRS. However, one important thing is that AgO_x is a material that can be used as conductor. Therefore, R_PZT ≫ R_TE2 and R_HRS ≫ R_LRS, which demonstrates the result in our case. The migration distance of oxygen vacancies (as shown in Fig. 6) may be different in the effect of different voltage sweepings, so there are 6 stable resistance states in the device with the PZT layer prepared at an oxygen pressure of 10 Pa. When an opposite electrical field was applied, oxygen ions were set free and merged with the oxygen vacancies, causing switching from the LRS to HRS. As shown in the former articles, the Ti/Fe:STO₃/Nb:STO₃ device also shows a filamentary resistance switching phenomenon, induced by the redox-reaction at the Ti/SrTiO₃ interface.²³ Such resistance switching phenomenon can also been observed in the NiO thin films device with special metal electrodes.²⁴ However, the switching ratio and operating voltage of the device are all better than these devices. In the device of Au/Co/BaTiO₃/La_2/3Sr_1/3MnO₃/NdGaO₃,²⁵ the resistance switching phenomenon is quite different from the case in our study. Because of the CoO_x formed in the redox-reaction enhances the ferroelectric barrier, however, the AgO_x in our case decreased the ferroelectric barrier. It is interesting that the device with a PZT film prepared at 10 Pa oxygen is at the LRS after the preparation. Because the PZT film prepared at 10 Pa oxygen pressure has many more oxygen vacancies, the device can easily be at the LRS after the Ag ions absorbed the oxygen ions in preparation.

Fig. 6 Possible mechanism of the resistive switching phenomenon.

As shown in Fig. 7, after an endurance test of about 100 cycles, the I–V curve of the device with the Ag electrode and PZT layer prepared at an oxygen pressure of 20 Pa has little difference compared to the I–V curve shown in Fig. 3, especially, the V_SET and V_RESET are almost the same. Retention is an important issue for the nonvolatile memory cells. For convenience, the HRS and LRS were set at 2 V and −2 V by voltage pulses, respectively, and read at 0.15 V. After 24 hours, there is not much difference from the beginning, as shown in Fig. 8. Therefore, the Ag/PZT/Pt thin film memristor has good retention properties.

Fig. 7 Current of Ag/PZT/Pt films with the PZT layer prepared at an oxygen pressure of 20 Pa at various applied voltage after endurance test of ∼100 cycles.

Fig. 8 Retention of the Ag/PZT/Pt thin film memristor; the PZT layer was prepared at an oxygen pressure of 20 Pa. Current as a function of time. The HRS was set by a 2 V voltage pulse and read at 0.15 V. The LRS is set by a −2 V voltage pulse and read at 0.15 V.

4. Conclusion

In summary, we fabricated PZT thin film memristors. The forming or rupture of oxygen vacancies filament causes the resistive switching in the effect of the external electric field. In addition, we find the switching ratio of PZT thin film memristor can be high as about 10⁸%, the operating voltage is as low as 2 V, and the write speed is very fast at about 5 μs. Considering the retention characteristics, these excellent switching performances and the potential of being multi-states memory, the Ag/PZT/Pt memory device has potential for future nonvolatile ferroelectric resistive switching memory applications.

Acknowledgements

This study was supported by the National Natural Science Foundation of China (No. 51572233, 61574121, 11372266 and 11272274), the Hunan Provincial Natural Science Foundation of China (No. 12JJ1007), the Specialized Research Fund for the Doctoral Program of Higher Education (No. 20114301110004) and the Foundation for the Author of National Excellent Doctoral Dissertation of PR China (201143).

References

1. J. Valasek, Phys. Rev., 1921, 17, 475.
2. M. H. R. Lankhorst, B. W. Ketelaars and R. A. M. Wolters, Nat. Mater., 2005, 4, 347.
3. A. Chanthbouala, A. Crassous, V. Garcia, K. Bouzehouane, S. Fusil, X. Moya, J. Allibe, B. Dlubak, J. Grollier, S. Xavier, C. Deranlot, A. Moshar, R. Proksch, N. D. Mathur, M. Bibes and A. Barthélémy, Nat. Nanotechnol., 2012, 7, 101.
4. E. Y. Tsymbal and H. Kohlstedt, Science, 2006, 313, 181–183.
5. R. Ramesh, Thin Film Ferroelectric Material and Devices, Kluwer Academic Publishers, Norwell, MA, 1997.
6. J. Seidel, L. W. Martin, Q. He, Q. Zhan, Y. H. Chu, A. Rother, M. E. Hawkridge, P. Maksymovych, P. Yu, M. Gajek, N. Balke, S. V. Kalinin, S. Gemming, F. Wang, G. Catalan, J. F. Scott, N. A. Spaldin, J. Orenstein and R. Rames, Nat. Mater., 2009, 8, 229.
7. D. Lee, S. M. Yang, T. H. Kim, B. C. Jeon, Y. S. Kim, J. G. Yoon, H. N. Lee, S. H. Baek, C. B. Eom and T. W. Noh, Adv. Mater., 2012, 24, 402.
8. E. A. Eliseev, A. N. Morozovska, G. S. Svechnikov, V. Gopalan and V. Ya Shur, Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 83, 235313.
9. P. Maksymovych, A. N. Morozovska, R. K. Vasudevan, A. N. Morozovska, E. A. Eliseev, J. Britson, J. C. Yang, Y. H. Chu, P. Maksymovych, L. Q. Chen, V. Nagarajan and S. V. Kalinin, Nano Lett., 2012, 12, 209.
10. C. Wang, K.-J. Jin, Z.-T. Xu, L. Wang, C. Ge, H.-B. Lu, H.-Z. Guo, M. He and G.-Z. Yang, Appl. Phys. Lett., 2011, 98, 192901.
11. L. Liu, A. Tsurumaki-Fukuchi, H. Yamada and A. Sawa, J. Appl. Phys., 2015, 118, 204104.
12. D. Pantel, S. Goetze, D. Hesse and M. Alexe, ACS Nano, 2011, 5, 6032.
13. D. Jiménez, E. Miranda, A. Tsurumaki-Fukuchi, H. Yamada, J. Suñé and A. Sawa, Appl. Phys. Lett., 2013, 103, 263502.
14. F. Yan, G. Z. Xing and L. Li, Appl. Phys. Lett., 2014, 104, 132904.
15. C. Ederer and N. A. Spaldin, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 71, 224103.
16. R. Waser and M. Aono, Nat. Mater., 2007, 6, 833–840.
17. A. Sawa, Mater. Today, 2008, 11, 28–36.
18. R. Waser, R. Dittmann, G. Staikov and K. Szot, Adv. Mater., 2009, 21, 2632.
19. C. H. Huang, J. S Huang, S. M. Lin, W. Y. Chang, J. H. He and Y. L. Chueh, ACS Nano, 2012, 6, 8407.
20. L. Pintilie, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 75, 104103.
21. Y. C. Yang, F. Pan, Q. Liu, M. Liu and F. Zeng, Nano Lett., 2009, 9, 1636.
22. Z. Wang, P. B. Griffin, J. McVittie, S. ong, P. C. cIntyre and Y. Nishi, IEEE Electron Device Lett., 2007, 28, 14.
23. S. Stille, C. Lenser, R. Dittmann, A. Koehl, I. Krug, R. Muenstermann, J. Perlich, C. M. Schneider, U. Klemradt and R. Waser, Appl. Phys. Lett., 2012, 100, 223503.
24. C. B. Lee, B. S. Kang, A. Benayad, M. J. Lee, S.-E. Ahn, K. H. Kim, G. Stefanovich, Y. Park and I. K. Yoo, Appl. Phys. Lett., 2008, 93, 042115.
25. D. J. Kim, H. Lu, S. Ryu, C.-W. Bark, C.-B. Eom, E. Y. Tsymbal and A. Gruverman, Nano Lett., 2012, 12, 5697–5702.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra08257b

---