7756
J. Phys. Chem. B 2007, 111, 7756-7760
Study on Threshold Behavior of Operation Voltage in Metal Filament-Based Polymer Memory Won-Jae Joo,*,† Tae-Lim Choi,‡ Kwang-Hee Lee,† and Youngsu Chung§ Display DeVice and Material Laboratory, Samsung AdVanced Institute of Technology, P.O. Box 111, Suwon 440-600, Korea, Electronic Chemical Materials DiVision, Cheil Industries Inc. 332-2 Gocheon-dong, Uiwang-si, Gyeonggi-do, Korea, and AE Center, Samsung AdVanced Institute of Technology, P.O. Box 111, Suwon 440-600, Korea ReceiVed: December 11, 2006; In Final Form: May 7, 2007
In the metal filament formation-based organic memory, the positive voltage application over the threshold electric field strength (170 MV/m) is necessary for the filament formation in Cu/P3HT/Al device. By the positive voltage application, the copper ions are generated and drifted into polymer layer, which is clearly confirmed by the secondary ion mass spectroscopy. Also, the field strength (100 MV/m) required for the drift process could be independently determined with a new pulse operation method. We could conclude that the threshold field strength of 170 MV/m was determined by the ionization process of copper. Furthermore, the dependence of the positive field strength and the temperature on the memory behavior was studied.
1. Introduction During the past decade, organic electronics has received a great attraction from various fields, such as display, energy, and logic applications.1-3 In particular, the organic light-emitting device system has been successfully commercialized in mobile application and became a promising candidate for the next generation large scale display.4 Organic thin-film transistors also have shown the remarkable performance, resulting in carrier mobility higher than 1 cm2 V-1 s-1, which is applicable for driving transistor in liquid crystal display.3 Recently, nonvolatile memory application has emerged as a new field of organic electronics.5-10 With the rapid development of information technology, memory devices with various characteristics such as flexibility, low weight, and nonconventional shapes have been required,11 although their memory performances so far are poor compared to the silicon based devices represented by Flash memory. For this reason, organic memory has received lots of interests from scientists and engineers due to its potential to meet these requirements.5-11 However, development of organic memory is still in the early stage, suffering from premature performances, particularly in endurance, retention time, and device reliability.5-10,12,13 Therefore, investigating the memory mechanism in detail is necessary to find the key device that would lead to commercialization. Among the various organic memory devices reported previously, metal filament formation-based devices have unique potential in the aspects of prolonged retention time and good thermal stability, although it also has serious problems such as in large variation of switching delay time and high operation current.5,14-17 These devices have electrically controllable bistability arising from the metal filament formation (“0” state) and breakdown (“1” state) between two electrodes. Recently, * Corresponding author. Telephone: +82-31-280-6757. Fax: +82-31280-9349. E-mail:
[email protected]. † Display Device and Material Laboratory, Samsung Advanced Institute of Technology. ‡ Cheil Industries Inc. § AE Center, Samsung Advanced Institute of Technology.
we found out that the metal filament was reproducibly formed using heteroatom (S or N) containing conjugated polymers such as poly(3-hexyl thiophene(P3HT), polypyrrole, and polyaniline.13 In particular, the devices fabricated with P3HT showed the excellent endurance of 30 000 cycles without any switching failure.13 Herein, we investigate the kinetics of the metal filament formation in the P3HT device. The filament formation can be divided into three processes, ionization of metal, drift of the ions, and metallization of the ions. Among them, the first two processes are carried out with applying the positive voltage, and the last occurrs by the negative voltage.13 Therefore, applying both positive and negative voltages is required for the metal filament formation. Especially, since the role of the positive voltage is to meet the prerequisites in the device for making the filament, it should be investigated thoroughly to achieve reliable memory performance. We would examine the dependence of the positive voltage and the temperature on the memory behavior. 2. Experimental Methods The test devices were fabricated with the metal-polymermetal structure with aluminum and copper as bottom and top electrodes, respectively. Si wafer deposited with silicon oxide layer of 200 nm was used as substrate after sonication cleaning in acetone solution for 30 min. Aluminum (80 nm) was thermally evaporated onto the substrate as bottom electrode, followed by spin coating of the active organic layer with approximately 50 nm thickness. Regiorandom poly(3-hexylthiophene) (P3HT) obtained from Aldrich Inc. without further purification was dissolved in chlorobenzene with 1.3 wt % and was filtered through 0.2 µm membrane syringe filter. Because these fabrication processes were performed in air, the surface of aluminum bottom electrode was oxidized to alumina with 5-10 Å thickness. After spin casting, the residual solvent was removed by heating the device on hot plate at the temperature of 110 °C. Last, the copper was thermal evaporated as a top electrode (70 nm) with deposition rate of 1 Å/s, and the test
10.1021/jp0684933 CCC: $37.00 © 2007 American Chemical Society Published on Web 06/15/2007
Voltage for Metal Filament-Based Polymer Memory
Figure 1. Typical current-voltage curve of regiorandom P3HT device which shows the nonvolatile memory behavior.
cell size is 0.25 mm2. The current voltage curve was measured in air with the sourcemeter (model: 2400, Keithley Inc.). Measurement in nitrogen atmosphere did not give any significant difference in memory phenomenon. To obtain the switching probability of the device with various environmental conditions, continuous voltage sweep was applied to the device with the programmable power supply of Yokogawa instrument (model 7631), and the current was measured with digital oscilloscope (model TDS3502B). The copper top electrode was connected to signal line, and the bottom electrode was used as ground. Cu density within P3HT layer was investigated using time-of-flight secondary ions mass spectroscopy (TOF-SIMS, model ION-TOF IV). Cu depth profiles were acquired in the dual beam interlaced mode using 500 eV Cs primary ion and 25 keV Bi analysis gun. 3. Results and Discussion The regiorandom P3HT was chosen as an active material because the device showed highly reproducible metal filament formation based memory behavior in our previous work.13 Typical switching behavior of the P3HT device is represented in Figure 1. A sudden decrease in the device resistance was
J. Phys. Chem. B, Vol. 111, No. 27, 2007 7757 observed at -2 V as a result of the metal filament formation between two metal electrodes. The higher conductive state (set state) showed the typical metal characteristics confirmed by lowtemperature experiment where the resistance linearly increased with the temperature from 160 to 300 K.13 The set state could be retained for several months and the application of 4 V switches back to the initial low conductive state (reset state) due to the cleavage of filament by joule heating. In this memory behavior, it was notable that high positive voltage application over 7 V was necessary for reproducible formation of the metal filament. The high positive voltage is believed to play important roles of ionizing the copper electrode and to inject the ions into polymer layer. Assisted by coordination to the heteroatom (S) of the P3HT, copper ions are distributed uniformly throughout the polymer layer. Then, copper ions are metallized to form the filament by the injected electrons under negative voltage bias. This concept of memory behavior is schematically represented in Figure 2. According to the concept mentioned above, the density and the distribution of copper ions throughout P3HT layer are important factors controlled by the positive electric field. For this reason, the memory behavior was investigated at various positive voltage strengths. In voltage sweep mode with frequency of 0.6 Hz, the maximum positive voltage was varied from 5 to 12 V with negative voltage set to -3 V. The switching probability was defined as the probability of occurrence of switching to the set state during a negative voltage sweep. For accuracy, the probability was determined from 60 cycles obtained after waiting for 180 s under the continuous voltage sweep. For 50 nm thick devices, the switching probability was near zero when the positive voltage below 7 V was applied. However, sudden increase up to unity was observed over 8 V as shown in Figure 3a. As the thickness of polymer layer in the devices increased, the threshold voltage also increased from 8 to 11 V. Interestingly, when represented in terms of electric field, Figure 3b indicates that all devices with different thickness had identical threshold electric field of about 170 MV/m. It was surprising that the switching probability increased sharply around the threshold electric field. To study the density of copper ions within polymer layer before and after the positive voltage application over the threshold value, we used the TOF-SIMS analysis to measure the depth profile of copper ions. To avoid the broadening of
Figure 2. Schematic concept for the formation of metal filament within polymer layer: (a) device structure, (b) ionization and drift processes of copper caused by positive voltage, (c) metal filament formation by the reduction of copper ions, and (d) the breakdown of copper filament by joule heating. Practically, the regiorandom P3HT is not aligned and has a random coil conformation in the organic layer.
7758 J. Phys. Chem. B, Vol. 111, No. 27, 2007
Joo et al.
Figure 4. Copper profiles in P3HT devices applied by the higher and lower voltages (5 and 13 V, respectively) than the threshold using TOFSIMS analysis. Dashed line indicates the carbon profile. The devices were fabricated with the structure of Al/P3HT/Cu/wafer for TOF-SIMS analysis.
Figure 3. (a) Switching probability as a function of the magnitude of positive voltage in voltage sweep mode in which negative voltage set to -3 V. The devices were used with the active layer thickness of 50, 55, 60, and 65 nm. (b) Changes in the switching probability with the electric field, which have an identical threshold value (170 MV/m) at different thickness of active layer.
Cu profile caused by the ion beam mixing effect during the SIMS measurement, new devices were fabricated with inverse structure of Al/P3HT/Cu on wafer substrate.18 The switching behavior in the Cu/P3HT/Al structure was described in Supporting Information. Three samples were prepared with different voltage bias conditions; the first device was virgin and the second and the third ones were biased for 100 s with lower and higher voltages than the threshold value (5 and 13 V, respectively). There was no notable difference in the copper depth profiles between the virgin and the device applied by the lower voltage (not shown in Figure 4). On the other hand, the density of copper ions was much higher near the aluminum electrode for the device applied with the high voltage. The SIMS analysis strongly implies that the copper ions are injected only when higher voltage than the threshold voltage is applied. Although the threshold behavior in the positive voltage was clearly observed in the measurement of the switching probability, it is still necessary to evaluate the numerical validity of the threshold electric field of 170 MV/m. The studies on the electromigration of copper had been extensively carried out to improve the reliability in the copper-insulator-silicon (MIS) capacitor.19-22 The bias temperature stress (BTS) was usually adopted to accelerate the electromigration of metal into the insulating layer in the performance characteristics of MIS capacitor.19-22 For the copper-polyimide capacitor, the electric field between 50 and 150 MV/m at 150 °C had been generally used as the BTS condition.20-22 Although the test condition and
the materials used in our memory devices are different than those of the MIS capacitors, it would be possible that the electric field over 170 MV/m could ionize the copper and drift the copper ions within polymer layer. In the case of MIS capacitors, it has been reported that electromigration of copper shows the Arrhenius behavior under BTS condition, where the logarithm of the lifetime to breakdown has a linear relation with the inverse temperature (1/T).19 Similarly, our memory phenomenon is expected to follow the Arrhenius behavior. However, since the operation method for our device is different from MIS capacitors, a new parameter, Ts, was defined in the memory device as the time when the switching probability approached to 80% in pulse voltage mode, instead of breakdown lifetime for MIS capacitors. Experimentally, after applying a positive voltage (6.5 V) to the P3HT device for a relatively long time, one tested whether filament formation occurred by short negative pulse (-2 V) or not. (The detailed experimental method was described in Supporting Information.) The positive and negative voltage pulses were continuously applied to measure the switching probability. The metal filament based switching being occasionally occurred at first became more reproducible as the positive voltage application time increased. The Ts was measured at different temperature as shown in Figure 6a. The logarithm of Ts decreased linearly with increasing the temperature which fits the Arrhenius model. It was fitted with the Arrhenius equation, Ts ) To exp(Ea/kT), where Ea is activation energy, k Boltzmann constant, and T the temperature, and the resulting activation energy was found to be 0.81 eV. With similar experimental method, the dependence of the time to failure (Tf) on the external electric field was investigated at room temperature. The Tf was defined as the time when the set state could not be switched back to the reset state any more by the voltage application. This failure would be due to the excess copper ions existing within polymer layer which could lead to the formation of thick metal filament or multiple current paths between two electrodes. From this point of view, Tf would be a useful parameter to characterize the endurance performance of this kind of memory devices. As the magnitude of positive voltage was changed from 7 to 9 V, Tf decreases from 1560 to 36 s. (Figure 6b) The relation between the external field and
Voltage for Metal Filament-Based Polymer Memory
Figure 5. The temperature dependence of the parameter Ts defined as the time when the switching probability approached to 80% in pulse voltage mode. Solid line is a result fitted with Arrhenius equation, Ts ) To exp(Ea/kT), and the activation energy is obtained to be 0.81 eV.
Figure 6. Electric field strength dependence of the parameter Tf defined as the time when the set state cannot be switched back to the reset state any more by the voltage application. Solid line indicates a result fitted with the equation, Tf ) To exp(-βE), with the field acceleration factor β of 0.09 (m/MV).
the Tf could be explained with the equation: Tf ) To exp(-βE), where β and E were a field acceleration factor and an electric field strength, respectively.19 The result fitted with the equation above agrees well with the experimental data and the parameter β was determined to be 0.09 (m/MV). This value is five times larger than the β of 0.018 (m/MV) for the coppersilicon oxide interface which could be explained with its material dependence.19 The parameter β is expected to be a key factor to represent the memory performance of the device. The foregoing investigation on the threshold electric field and the factors (Ts and Tf) dealt with the prerequisites of the device condition for reproducibly producing the metal filament. The prerequisite is consisted of the two processes, the ionization of copper and the drift of the ions which have different activation
J. Phys. Chem. B, Vol. 111, No. 27, 2007 7759
Figure 7. Qualitative determination of the activation energy for the drift of copper ions within P3HT layer using three voltage pulses. Inset shows the three voltage pulses and the measured current through the device indicating that the metal filament is formed only by the third pulse. The second pulse was varied from -3.5 to -6 V.
energies. Since both processes occur by the positive voltage application, it is difficult to control the each process independently. In other words, the distribution and the density of the copper ions existing in polymer layer were determined with the levels of the activation energies. If the activation energy for the drift is larger than that for the ionization, excessive copper ions will be injected in organic layer. For the contrary case, a large portion of metal ions will be deeply drifted toward the opposite electrode leading to asymmetric distribution. Hence, it is important to optimize the activation energy levels of the two processes to enhance memory performance. To qualitatively estimate the active energy of the drift for the P3HT device, three pulses were subsequently applied to the device. The first pulse of 9 V is responsible for the ionization, the drift, and the breakdown of the filament formed previously and the third pulse of -2 V is for the metal filament formation. These two pulses have been typically used for the pulse operation of these memory devices. The second pulse with negative voltage was newly added in order to drift the ions in reverse direction. If the distribution of copper ions changes due to the backward drift caused by the second pulses, it will affect the metal filament formation by the third negative pulse, leading to the change in the switching probability. The second voltage pulse was varied from -3.5 to -6 V, and fortunately, the metal filament was not formed by the second pulses within this range. Figure 7 shows the probability obtained in the pulse operation mode with varying the magnitude of the second pulse. The three voltage pulses and the measured current are shown in the inset where the metal filament is formed not by the second pulse but by the third one. The filament was almost perfectly formed when the magnitude of the second pulse voltage was below 4.5 V. However, the probability substantially decreased in half at -5 V, and it even became close to zero with the second pulse over -6 V. Compared to the positive threshold electric field (ca. 170 MV/m) of Figure 3, the negative threshold electric strength (100 MV/m) corresponding to the drift of metal ions is smaller. Therefore, the threshold of the positive voltage would be governed by the ionization process of copper.
7760 J. Phys. Chem. B, Vol. 111, No. 27, 2007 4. Conclusions We observed that the metal filament was reproducibly formed when the positive voltage over the threshold (170 MV/m) was applied prior to the negative voltage application. The roles of the positive voltage were to ionize copper electrode and to drift the ions within a P3HT layer, which are the prerequisites for the metal filament formation. The TOF-SIMS analysis showed these two processes indeed occurred by applying the positive voltage over the threshold. In P3HT devices, we experimentally determined the threshold electric field for the drift of the copper ions to be 100 MV/m which was smaller than the total threshold voltage (ca. 170 MV/ m). Therefore, the activation energy of the ionization process is larger than that of the drift process and it determined the total threshold voltage. We could conjecture that the combination of the activation levels for the two processes would be dominant factor to determine the performance of the metal filament based organic memory. In future, dependence of active material on the memory performance would be investigated by determining the activation energies for the ionization and the drift. Supporting Information Available: TOF-SIMS results for the virgin device and the devices biased with the lower and higher voltage than the threshold, detailed measurement methods for the parameters, Ts and Tf, and the switching behavior in Cu/P3HT/Al structure. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Forsythe, E. W.; Abkowitz, M. A.; Gao, Y. J. Phys. Chem. B 2000, 104, 3948.
Joo et al. (2) O’Regan, B.; Gratzel, M. Nature 1991, 353, 737. (3) Kuo, C. C.; Payne, M. M.; Anthony, J. E. Jackson, T. N. IEDM Tech. Digest 2004, 373-376. (4) Joo, W.-J.; Choi, T.-L.; Lee, S. K.; Chung, Y.; Jung, M.-S.; Kim, J. M. Org. Electron. 2006, 7, 600. (5) Ma, L.; Xu, Q.; Yang, Y. Appl. Phys. Lett. 2004, 84, 4908. (6) Bandhopadhyay, A.; Pal, A. J. J. Phys. Chem. B 2003, 107, 2531. (7) Co¨lle, M.; Bu¨chel, M.; de Leeuw, D. M. Org. Electron. 2006, 7, 305. (8) Chen, J.; Ma, D. Appl. Phys. Lett. 2005, 87, 023505. (9) Tondelier, D.; Lmimouni, K.; Vuillaume, D.; Fery, C.; Haas, G. Appl. Phys. Lett. 2004, 85, 5763. (10) Pradhan, B.; Batabyal, S. K.; Pal, A. J. J. Phys. Chem. B 2006, 110, 8274. (11) Lauters, M.; McCarthy, B.; Sarid, D.; Jabbour, G. E. Appl. Phys. Lett. 2005, 87, 231105. (12) Ouyang, J.; Chu, C.; Szmanda, C. R.; Ma, L.; Yang, Y. Nat. Mater. 2004, 3, 918. (13) Joo, W.-J.; Choi, T. L.; Lee, J.; Lee, S. K.; Jung, M.-S.; Kim, N.; Kim, J. M. J. Phys. Chem. B 2006, 110, 23812. (14) Segui, Y.; Ai, B.; Carchano, H. J. Appl. Phys. 1976, 47, 140. (15) Pender, L. F.; Fleming, R. J. J. Appl. Phys. 1975, 46, 3426. (16) Carchano, H.; Lacoste, R.; Segui, Y. Appl. Phys. Lett. 1971, 19, 414. (17) Sliva, P. O.; Dir, G.; Griffiths, C. J. Non-Cryst. Solids 1970, 2, 316. (18) Cheng, J.; Winograd, N. Anal. Chem. 2005, 77, 3651. (19) Takeda, K.; Ryuzaki, D.; Mine, T.; Hinode, K.; Yoneyama, R. J. Appl. Phys. 2003, 94, 2572. (20) Mallikarjunan, A.; Murarka, S. P.; Lu, T.-M. Appl. Phys. Lett. 2001, 79, 1855. (21) Yoshino, T.; Tata, N.; Kikkawa, T. Jpn. J. Appl. Phys. 2004, 43, 8026. (22) Loke, A. L. S.; Wetzel, J. T.; Townsend, P. H.; Tanabe, T.; Vrtis, R. N.; Zussman, M. P.; Kumer, D.; Ryu, C.; Wong, S. S. IEEE Trans. Electron. DeVices 1999, 46, 2178.
