† Corresponding author. E-mail:
Project supported by the National Natural Science Foundation of China (Grant Nos. 61006003 and 61674038), the Natural Science Foundation of Fujian Province, China (Grant Nos. 2015J01249 and 2010J05134), the Science Foundation of Fujian Education Department of China (Grant No. JAT160073), and the Science Foundation of Fujian Provincial Economic and Information Technology Commission of China (Grant No. 83016006).
As an industry accepted storage scheme, hafnium oxide (HfOx) based resistive random access memory (RRAM) should further improve its thermal stability and data retention for practical applications. We therefore fabricated RRAMs with HfOx/ZnO double-layer as the storage medium to study their thermal stability as well as data retention. The HfOx/ZnO double-layer is capable of reversible bipolar switching under ultralow switching current (< 3 μA) with a Schottky emission dominant conduction for the high resistance state and a Poole–Frenkel emission governed conduction for the low resistance state. Compared with a drastically increased switching current at 120 °C for the single HfOx layer RRAM, the HfOx/ZnO double-layer exhibits excellent thermal stability and maintains neglectful fluctuations in switching current at high temperatures (up to 180 °C), which might be attributed to the increased Schottky barrier height to suppress current at high temperatures. Additionally, the HfOx/ZnO double-layer exhibits 10-year data retention @85 °C that is helpful for the practical applications in RRAMs.
As one of the most promising nonvolatile storage technologies exhibiting fast write/read speed, good scalability, low power consumption, and good endurance, resistive random access memory (RRAM) has attracted much attention from industries and academic communities.[1–4] In recent years, various metal oxides have been studied as the storage medium of RRAMs, and conducting models have been adopted to interpret the oxide-based resistive switching.[4–10] Among them, hafnium oxide (HfOx) has been intensively investigated for its approbation from industries. Shrinking physical dimensions of a memory cell to meet the requirements of three-dimensional (3D) integration is an inevitable choice for further expanding the storage capability and meeting the demand of neuromorphic computing.[5,11–13] However, tiny space accommodating cells in a large number would face fluctuations in switching parameters due to the thermal disturbance triggered by joule-heating. Consequently, the significance of improving thermal stability and data retention is manifest.[14–17]
In order to achieve this goal, several schemes have been attempted.[18–22] Keeping the density of oxygen vacancy (Vo) high enough has been confirmed to effectively ensure thermal stability and data retention.[18,19] If a more thermally-stable thin film serves as a cap to suppress oxygen releasing from active switching layers, the thermal stability and data retention might also be enhanced due to the maintenance of Vo.[20] Additionally, metal electrodes with suitable thermodynamic ability are able to strengthen the thermal stability and data retention of an RRAM by forming conductive filaments with favorable geometry to alleviate oxygen stochastic movements.[21] Furthermore, a buffer layer could also be used to sustain the stability of a Vo assisted conductive filament (CF).[22] Therefore, sustaining the status of the Vos (or oxygen ions) is critical for improving thermal stability and data retention, and the thin films adjacent to the active switching layers may play an important role.
As wide band gap oxides exhibit high transmittance in visible regions, HfOx and zinc oxide (ZnO) with excellent switching properties have also been considered as the storage medium of a transparent RRAM.[14,23–27] Several schemes have been proposed to enhance switching properties. However, few efforts have been made to improve thermal stability and data retention. As we know, electron barrier at the interface might affect the switching properties of an RRAM, whose thermal stability and data retention might be affected as well. Recently, we have reported switching current reduction with an HfOx/ZnO double-layer,[28] confirming effects of interface on performance. However, thermal stability and data retention are still absent. In this paper, we further investigate the HfOx/ZnO double-layer (DL) to study the effects of the adjacent ZnO layer on the thermal stability and data retention of the RRAMs.
To fabricate an HfOx/ZnO double-layer RRAM as shown in the inset of Fig.
Figure
As shown in Fig.
To highlight the switching mechanism in the DL-RRAM, oxygen migration should be taken into account. Once the HfOx layer is deposited on the ZnO layer, the oxygen ions in the ZnO thin film would migrate into the HfOx thin film due to a difference in Gibbs free energy of oxide formation (
If further increased electric field is applied to the memory cell to complete the set process, some Vos in the HfOx thin film would be generated to serve as trapping centers for electron transportation.[28] Unlike the Vos in the ZnO with a high concentration to form conductive channels, Vos in the HfOx thin film are still insufficient for electron direct transport between them. Considering the increased electric field to reduce Coulomb potential energies of the electrons, the probability of an electron getting into the conduction band of the HfOx thin film by thermal excitation out of the trapping centers (Vos in the HfOx thin film) would be increased.[32] Therefore, we can observe the Poole–Frenkel emission conduction in the LRS, whose current density (
Thermal stability of the switching current is very important for maintaining normal working status at high temperatures. The Schottky emission dominant conduction in the HRS suggests a close relationship between the current and the Schottky barrier height
Increased temperature may facilitate oxygen ions getting out of lattice matrix by leaving Vos in the oxides. For the SL-RRAM under an electric field, more Vos would be generated at high temperatures to favor the formation of Vo assisted conductive filament. Consequently, we observe a drastically increased current at high temperatures as shown in Fig.
Data retention is also crucial for the practical applications of the RRAMs. To investigate the retention reliability of the RRAMs, the thermal accelerating measurement for resistance switching under various temperatures was carried out. Figure
Retention failure should be associated with redistribution of Vos to affect electron transport and device resistance. As previously discussed, the Vos assisted CFs would be responsible for the LRS conduction in an HfOx-based SL-RRAM, instead of the Vos sparsely distributing to assist electron transport governed by a Poole–Frenkel emission conduction in a DL-RRAM. Driven by the concentration gradient, oxygen ions easily move to CFs and combine with Vos to narrow or even cut off CFs. Therefore, the redistribution of the Vos in an SL-RRAM usually exhibits obvious resistance changes. As for the DL-RRAMs, redistribution of Vos might also be affected by temperature rise. However, since the Vos in the HfOx are sparsely and topologically distributed, slight stochastic movement of the Vos is able to secure electron transport with neglectful variations in resistance. Consequently, resistance variations of the LRS SL-RRAM would be more sensitive to temperature rise than that for an LRS DL-RRAM as shown in Fig.
When it comes to the HRS data retention, the formation of Vos is obviously necessary to be responsible for the retention failure. A rise in temperature produces Vos in both the DL-RRAMs and the SL-RRAMs, which favors the formation of a conductive filament in the SL-RRAMs and results in an HRS retention failure. However, once Vos generate in the HfOx thin films of an HRS DL-RRAM, it might be filled with some oxygen ions migrating from the ZnO layer, as ZnO has a greater Gibbs free energy of oxide formation than HfO
The sputter-deposited HfOx/ZnO double-layer capable of reversible bipolar switching has been prepared with a Schottky emission dominant conduction in the HRS and a Poole–Frenkel emission governed conduction in the LRS. Switching current of the DL-RRAMs can be sustained at high temperatures (up to 180 °C) with neglectful fluctuations, which is ascribed to the increase of Schottky barrier height at the HfOx/ZnO interface to suppress the increase of current at high temperatures. In addition, the HfOx/ZnO double-layer shows no less than 10 years of data retention for the HRS and the LRS.
[1] | |
[2] | |
[3] | |
[4] | |
[5] | |
[6] | |
[7] | |
[8] | |
[9] | |
[10] | |
[11] | |
[12] | |
[13] | |
[14] | |
[15] | |
[16] | |
[17] | |
[18] | |
[19] | |
[20] | |
[21] | |
[22] | |
[23] | |
[24] | |
[25] | |
[26] | |
[27] | |
[28] | |
[29] | |
[30] | |
[31] | |
[32] | |
[33] | |
[34] | |
[35] | |
[36] | |
[37] | |
[38] | |
[39] |