- Email: [email protected]
Exploring the role of nitrogen incorporation in ZrO2 resistive switching film for enhancing the device performance
Journal of Alloys and Compounds 775 (2019) 1301e1306
Contents lists available at ScienceDirect
Journal of Alloys and Compounds journal homepage: http://www.elsevier.com/locate/jalcom
Exploring the role of nitrogen incorporation in ZrO2 resistive switching film for enhancing the device performance Xiaodi Wei a, Hong Huang a, Cong Ye a, b, *, Wei Wei a, b, Hao Zhou a, Yong Chen a, **, Rulin Zhang a, Li Zhang a, Qing Xia a a b
Faculty of Physics and Electronic Science, Hubei University, Wuhan 430062, China Key Laboratory of Microelectronics Devices & Integration Technology, Institute of Microelectronics of Chinese Academy of Sciences, Beijing, 100029, China
a r t i c l e i n f o
a b s t r a c t
Article history: Received 1 July 2018 Received in revised form 18 October 2018 Accepted 21 October 2018 Available online 22 October 2018
The role of nitrogen doping on the resistive switching (RS) performance in nitrogen doped ZrO2 memristive device is investigated. The Pt/N:ZrO2/TiN resistive random access memory (RRAM) shows smaller switching voltage, larger memory window as well as improved uniformity. Moreover, the multilevel storage capability can be successfully obtained by varying the compliance current in the SET process for the memory cell. It is considered that the connection and rupture of conducting oxygen vacancy filaments (CF) can be localized and the oxygen ions (O2) migration randomness is depressed due to nitrogen doping in ZrO2 film. Combining with the first-principle method, we theoretically calculate the formation energy (Evf), migration energy (Em) and density of states for oxygen vacancy (VO). Both Evf and Em values show noticeable decrease in N doped 2 2 2 ZrO2 supercell, which are related to the lower forming voltage and operating voltage. The density of states indicates that the oxygen vacancy midgap defect states can be eliminated as a result of N dopant, which neutralizes the excess defects in ZrO2 switching layer and may reduce the densities of the potential filaments. Herein the uniformity can be improved. All the theoretical results show reasonable agreement with the improved experimental RS performance for Pt/N:ZrO2/TiN device. © 2018 Elsevier B.V. All rights reserved.
Keywords: RRAM Nitrogen doping Multilevel storage Oxygen vacancy First-principle method
1. Introduction Resistive random access memory (RRAM) has been the most promising candidate for next generation nonvolatile memory in terms of its excellent scalability, simple structure, low operating power, rapid operation and high density integration [1e3]. Numerous transition metal oxides, such as Al2O3, Ta2O5, ZrO2, and HfO2 have been reported to show resistive switching (RS) characteristics [4e7]. It is well known that oxide-based RRAM switching behavior between high resistive states (HRS) and low resistance states (LRS) occurs by controlling the connection and rupture of the conducting oxygen vacancy (VO) filaments (CF) [8]. Herein, it is crucial to engineer the VO profile in the oxide film to achieve stable memory states and a large switching memory window [9]. In oxide
* Corresponding author. Key Laboratory of Microelectronic Devices & Integrated Technology，Institute of Microelectronics of ChineseAcademy of Sciences, Beijing100029, China. ** Corresponding author. E-mail addresses: [email protected] (C. Ye), [email protected] (Y. Chen). https://doi.org/10.1016/j.jallcom.2018.10.249 0925-8388/© 2018 Elsevier B.V. All rights reserved.
materials, ZrO2 has recently attracted significant attention for fabricating high quality RRAM devices [6,10,11]. However, one of the representative obstacles for ZrO2 RRAM device is the nonstability of RS parameters such as switching voltage and resistance states. Gao et al. found that Al doping ZrO2 can effectively reduce the VO formation energy and the forming voltage, and the conductive VO filament preferentially grows near the Al doping site by the first-principles calculations [12]. Recently, doping impurity elements have been reported to be effective for controlling the location of the CF and improving the switching parameters [13,14]. Also, it has been reported that N doping has often been used to enhance property of devices. For example, N doping Al2O3 RRAM significantly reduces the forming voltage and the operating current [15]; N incorporation into HfO2 reduces the leakage current by deactivating VO defect states in the band gap and elevating the VO level in the band gap to over the conduction band [16]. Therefore, it has been believed that the VO migration is the origin of resistive switching behaviors and N also plays an important role in oxide-based RRAM [17e19]. It is important to understand the origin of beneficial effects for N
1302
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
doping oxide-based RRAM on the VO properties. In this work, it is aimed to investigate the role of N incorporation into ZrO2-based RRAM on the VO properties by using the firstprinciples calculations. The calculated results show that the formation energy and migration energy of VO in ZrO2 are significantly decreased by N incorporation, which leads to the lower forming voltage and operating voltage in RRAM devices. Furthermore, we reveal that N atoms favorably couple with VO in ZrO2, which induces a more stable conductive filament by preventing VO diffusion. These results clearly explain how N incorporation into ZrO2based RRAM improves the performance of the memory device. The improved resistive switching behaviors and the multilevel storage capability in N-doped ZrO2 device are demonstrated in the experiment. 2. Calculation methods and experiments The first-principles calculations based on the density functional theory (DFT) is performed on monoclinic ZrO2 (m-ZrO2) with the generalized gradient approximation-Perdew Burke Ernzerh of (GGA-PBE) [20] for the exchange correlation potential, and Vanderbilt type ultrasoft pseudopotential [21] for core-valence interaction, as implemented in the CASTEP code [22]. The first-principle method is a standard theoretical solution to investigate the electronic behaviors of monoclinic ZrO2, such as the band structure, density of states and electron localization function. The inputs of the CASTEP code mainly include serials of parameters: the cutoff energy, the k-point, the force, the maximum displacement and the smearing width. After the self-consistent calculation, the material properties can be obtained, such as the band structure, density of states and electron localization function. The plane wave set basis cutoff energy is 450 eV, and the Brillouin zone is sampled at a grid of 2 2 2 Monkhorst-Pack k-point. The supercell is fully relaxed until the residual force of atoms is less than 0.001 eV/Å, the maximum ionic displacement is within 5.0 104 Å, and the smearing width is 0.2 eV. The valence electrons for zirconium are in the 4s24p64d25s2 configuration. The configuration 2s22p4 is used for the generation of the valence electrons of oxygen. The m-ZrO2 cell parameters are determined as a ¼ 5.192 Å, b ¼ 5.265 Å, c ¼ 5.358 Å, b ¼ 99.81, Zr(0.277, 0.044, 0.209), O(0.072, 0.338, 0.341) and O(0.447, 0.758, 0.479) [23]. The ZrO2 supercell containing 96 atoms is established by using the Materials Studio software. A m-ZrO2 (space group P21/c) 2 2 2 supercell of 96 atoms is chosen as a starting point to build these models. It is found that the threefold coordinated VO in ZrO2 networks has a small formation energy, low diffusion barrier, and forms a relatively stable configuration by binding with dopant [24], which is believed to be strongly related to the resistive switching behavior of the RRAM devices. Hence, this work focuses on investigating the three coordinated VO in ZrO2. Next, we construct four different models (model A, B, C, and D as shown in Fig. 1) by substituting two O atoms with N atoms and removing one O atom to investigate the VO interaction between N atoms. Distances from the VO to each N atom are 2.8 and 5.3 Å in Model B, 2.8 and 6.8 Å in Model C, and 6.8 and 7.5 Å in Model D, respectively. In the calculations, the formation energy (Evf) and the migration energy (Em) of VO can remarkably affect the performance of RRAM device [25], thus it is essential to clearly understand the N doping effect of Evf and Em for the VO in RRAM device. Based on the first-principles calculations, the VO formation energy can be calculated as [24].
Evf ¼ EðVO Þ EðbulkÞ þ mO where E(VO) and E(bulk) denote the total energy of each model with or without a VO in the undoped or N-doped ZrO2 bulk crystal,
respectively. mo is the chemical potential of an oxygen atom, which is half of the oxygen molecule energy in this calculation. To simulate the VO diffusion process, one oxygen ion is moved from the lattice site to the neighboring vacant sites in ZrO2. The migration energy (Em) is defined as the total energy difference between saddle point energy of the diffusion process and the VO energy at the initial site [26]. The migration barrier is a chemical reaction from initial state to the final state. The most important process is to find the middle of the transition state which is the most difficult between the initial position and the final position. Thus, transition state search is used to search the transition state and calculate the Em of VO in ZrO2. Linear synchronous transit (LST), followed by repeated conjugate gradient minimizations and quadric synchronous transit (QST) maximizations calculations are performed to search the saddle point in the VO migration process [27]. In the experiment, undoped and N-doped ZrO2 memory devices were deposited on TiN/SiO2/Si substrates by radio frequency magnetron sputtering in pure Ar atmosphere and N2/Ar mixed atmosphere. Undoped ZrO2 films were deposited by a ZrO2 target (99.99%) with the power of 60 W, in a pure 45 sccm Ar gas at room temperature by sputtering. N-doped ZrO2 films were deposited in a gas mixture of 30 sccm Ar and 15 sccm N2. The switching layer was same thickness of 25 nm. Then, the top electrode Pt with 200 nm thickness was deposited by direct current (DC) magnetron sputtering by a Pt target (99.99%) with the power of 100 W in a pure argon atmosphere. The working pressure was kept at 0.5 Pa. Lastly, in order to shape and complete an individual TiN/ZrOx/Pt RRAM, photolithography and the lift-off technique were employed. All the electrical properties were measured by using an Agilent B1500A semiconductor parameter analyzer. 3. Results and discussion Fig. 1 presents the VO formation energy in ZrO2 for all models and the current-voltage (IeV) curves of the formation process for the Pt/ZrO2/TiN and Pt/N:ZrO2/TiN devices. In order to calculate the VO formation energy, we establish oxygen molecule model and ZrO2 supercell models with or without the oxygen vacancy by using the Materials Studio software. We carry out geometry optimization for the established models by the CASTEP code. We choose energy option to calculate the Kohn-Sham energy for the optimized models by the CASTEP code. After analyzing energy evolution of the optimized models, we can obtain mo, E(bulk) and E(VO) of ZrO2 supercell models with or without an oxygen vacancy. According to the equation, we can calculate the VO formation energies (Evf), which are 5.82 eV, 3.15 eV, 3.25 eV and 3.45 eV for Models A, B, C and D in Fig. 1(a), respectively. As shown in Fig. 1(a), it is found that the calculated Evf in ZrO2 is significantly reduced by introducing N atoms, which indicates that VO easily forms when N atoms are incorporated into ZrO2-based RRAM. The Evf of Model B is the lowest, and the Evf values of Model C and Model D are calculated to be 3.25 and 3.45 eV, respectively. These results indicate that the most stable configuration is Model B, and N atoms are inclined to couple with the VO in ZrO2 with closer distance site. It can be explained by dipole formation and the Coulomb interaction energy [28]. In the Pt/ZrO2/TiN device and the Pt/N:ZrO2/TiN device, the voltage was biased on the TiN bottom electrode while the Pt top electrode was grounded. Fig. 1(b) presents IeV characteristics of forming process for both devices. As the positive bias voltage was applied on the TiN bottom electrode, a soft breakdown occurred at about 7.2 V under a compliance current of 500 mA for the Pt/ZrO2/ TiN memory cell. For the Pt/N:ZrO2/TiN device, the forming voltage decreased to 3.6 V. Normally, the forming process is interpreted to be equivalent to the dielectric soft breakdown accompanying with creation of the VO, which plays the crucial role in the memory
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
1303
Fig. 1. (a) The formation energy of oxygen vacancy in undoped and N-doped ZrO2, the upper inset shows ZrO2 networks. The gray, blue, pink, red, and yellow balls denote zirconium, nitrogen, 3-fold oxygen, 4-fold oxygen and the VO site, respectively. (b) The forming I-V characteristics for ZrO2 device and N:ZrO2 device. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
operation for oxide-based RRAM [18]. Here, it is assumed that N doping stimulates the VO formation as a result of lower Evf and influences the formation of the conductive filament, so the forming voltage decreased for Pt/N:ZrO2/TiN device. In fact, IeV curves for RRAM devices can be calculated and show excellent consistency with the experimental results in our previous work [29]. Here, we aim to obtain the experimental results from the IeV measurement of forming process, and explain it by calculating the VO formation energy of undoped and N-doped ZrO2 supercells. The conductive VO filament formation depends not only on the VO formation energy, but also on the VO migration energy in the system. The calculated values of Em are presented with different distance of migration paths in Fig. 2(a). In terms of the distance from VO to N atom, it can be considered as VO near to N atom and VO away from N atom. Thus, the migration paths of VO can be classified as migration close to N atom and outward from N atom. For migration of VO, the calculated Em decreases in N doped ZrO2. In contrast, the calculated Em increases with migration distance in ZrO2 with or without N atoms. These results indicate that the VO easily migrates towards N atoms and the conductive VO filament preferentially grows near N doping sites. Therefore, N incorporation would improve the reliability of the conductive VO filament in RRAM by suppression of VO diffusion. Similar results have been proved in N-doped Al2O3 film [15]. On microscopic aspect, the
degree of difficulty in the VO migration process determines the macroscopic properties of devices such as the SET voltage and the RESET voltage. After the electroforming process, both devices can be reversibly switched from the HRS to the LRS and from the LRS to the HRS. Fig. 2(b) presents the IeV curves for the Pt/ZrO2/TiN and Pt/N:ZrO2/TiN device. During the electrical test, the voltage was applied on the TiN electrode when the Pt electrode was grounded, and a compliance current of 1 mA was used. The result clearly shows that the switching voltages (VSET and VRESET) became smaller for the Pt/N:ZrO2/TiN device. It should be noted that the SET voltage reduces from 0.61 to 0.52 V in the N-doped device as compared with the device without doping. All the theoretical Em results show reasonable agreement with the experimental SET voltage for Pt/ N:ZrO2/TiN device. To analyze the electronic properties of ZrO2, the calculated energy band structure and density of states (DOS) of undoped ZrO2 with VO (Model A) and N-doped ZrO2 with VO (Model B) are shown in Fig. 3. The zero-point energy is taken as the Fermi level. The calculated band gap of perfect ZrO2 supercell is about 3.73 eV, which is consistent with other DFT calculations [30,31]. It should be noted that the calculated data is smaller than the experimental data of 5.4 eV [32] due to the well-known band gap underestimations of DFT calculations [23]. As shown in Fig. 3(a), the single VO in ZrO2 network will introduce an additional band localized at the Fermi
Fig. 2. (a) The VO migration energy as a function of migration distance, the upper inset shows the possible VO site (number between 1 and 7) in ZrO2 networks. The gray, blue, pink and red balls denote zirconium, nitrogen, 3-fold oxygen, 4-fold oxygen, respectively. (b) Typical I-V curves for Pt/N:ZrO2/TiN and Pt/N:ZrO2/TiN devices. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
1304
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
Fig. 3. The band structure and density of states for (a) and (b) undoped ZrO2 with VO (Model A), (c) and (d) N-doped ZrO2 with VO (Model B), respectively.
level between the conduction band and the valence band, and the additional band is mainly composed of the Zr 4d and O 2p orbitals. Fig. 3(b) shows that the VO gives rise to a defect state in the mid bandgap region, with the defect state orbital localized on the nearby Zr atom and O atom. Fermi level is at midgap state, indicating that the defect level of the neutrally charged VO is occupied. It means that the oxygen vacancy acts as an electron trap and traps electrons from surrounding atoms [33]. The effect of N doping has been studied by substituting two O atoms with N atoms nearest the VO site (Model B). The additional band and the original VO defect state in the bandgap of undoped ZrO2 with VO is removed due to N doping in Fig. 3(c) and (d), leaving some N non-bonding states near the valence band of ZrO2 [34,35]. The original VO as a neutral charge state introduces two excess electrons localized on the nearby Zr atom. During the N atoms doping process, each N atom substitutes one O atom in the oxide network and creates a valence band hole. The electrons transfer from the higher VO level to the lower 2p orbitals of N atoms at the valence band top. As a result, the two excess electrons fill the valence states and form a closed shell electronic configuration. It will make the system extremely stable. Hence, the VO becomes a V2þ O center, involving a certain lattice distortion, and pushes the original defect state up into the conduction band [35]. The energy gain of this process comes from the two electrons falling from mid gap states into the valence band holes, which compensates the energy required for local distortion. In this way, the VO in the ZrO2 network is nullified by the presence of two N atoms. The neutralization of the defect states association with the VO eliminates the excess conductive paths and reduces the numbers of the potential filaments. Electron localization function (ELF) was initially proposed to measure electron pairing by defining [36].
ELF ¼
1 1 þ ½DðrÞ=Dh ðrÞ2
where the term D(r)/Dh(r) normalizes the same-spin probability by the uniform-density electron gas as reference, and thus ELF is a dimensionless localization index restricted to the range of [0,1]. The ELF value at 0e0.5 and at 0.5e1 denote low electron density area and high electron density area, respectively. A high ELF value stands for a low probability of finding a second electron with the same spin in the neighboring region of the reference electron, and the reference electron is highly localized. Based on its definition, one can simply interpret high ELF values as covalent bonds, lone pairs, or inert cores. Here, we focus on the [001] surface which has proven to be the most stable surface over a wide range of electron state. Fig. 4 indicates that N atoms tend to capture electrons from VO and thus neutralize the defect states of oxygen vacancies, which may eliminate the excess conductive paths. Fig. 5(a) shows the multilevel IeV curves for the Pt/N:ZrO2/TiN device with different compliance currents of 500 mA, 1 mA, and 5 mA. From Fig. 5(b), controlling varying the compliance current in the SET process, intermediate resistances of LRS occur and we can observe three resistances of LRS (level 2, level 3 and level 4). It is well acknowledged that with the increase of compliance current for the SET process, the conductive filament connecting both electrodes becomes strong and the resistance of LRS becomes small [37]. Here, for the forming compliance current of 5 mA, the conductive filament is the strongest, whereas the conductive filament is the weakest for the forming compliance current of 500 mA. Therefore, by varying the compliance currents of 500 mA, 1 mA, and 5 mA in the SET process, three resistances of LRS can be obtained, which is called multilevel storage. ZrO2-based RRAM with
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
1305
Fig. 4. ELF for (a) undoped ZrO2 and (b) N doped ZrO2. The ELF value at 0e0.5 and at 0.5e1 indicate low electron density area and high electron density areas, respectively.
Fig. 5. (a) Typical I-V curves for the Pt/N:ZrO2/TiN RRAM device under different compliance currents, (b) The HRS and LRS of the Pt/N:ZrO2/TiN RRAM device with different compliance currents.
multilevel storage can greatly achieve large integration and highdensity storage. In addition, the multilevel memory device opens application areas in logic functions, neural networks and passive crossbar memory selectors [38]. Lastly, for multiple application of ZrO2-based RRAM in the future work, it is noted that more detailed experimental and calculation methods are performed by varying temperature, different doping concentration and stoichiometric tools [39e42].
Acknowledgements This study was supported by the National Nature Science Foundation of China (No.61474039; No.61774057); The Opening Project of Key Laboratory of Microelectronic Devices & Integrated Technology, Institute of Microelectronics of the Chinese Academy of Sciences (Y7YS013001); National Key Research and Development Program under Grant 2017YFB0405602. References
4. Conclusions In summary, the role of N incorporation into ZrO2-based RRAM has been investigated by using first-principles calculations. The calculated results showed that the VO formation energy and migration energy in ZrO2 is significantly reduced by N incorporation, resulting in a lower forming voltage and SET voltage in RRAM device. Herein, N atoms favorably couple with VO and stabilize the conductive filament by preventing VO diffusion in N-doped ZrO2. Furthermore, the density of states for N-doped ZrO2 supercell indicates that the oxygen vacancy midgap defect states can be eliminated. Hence, nitrogen incorporation can neutralize the excess defects in N-doped ZrO2 switching layer and may reduce the densities of the potential filaments. Finally, multilevel storage was achieved with different compliance currents in the SET process of the Pt/N:ZrO2/TiN device. Thus, the results are conductive to designing ZrO2-based RRAM devices for optimized performance.
[1] B. Ku, Y. Abbas, A.S. Sokolov, C. Choi, Interface engineering of ALD HfO2-based RRAM with Ar plasma treatment for reliable and uniform switching behaviors, J. Alloys Compd. 735 (2018) 1181e1188. [2] X.L. Zhao, S. Liu, J.B. Niu, L. Liao, Q. Liu, X.H. Xiao, H.B. Lv, S.B. Long, W. Banerjee, W.Q. Li, S.Y. Si, M. Liu, Confining cation injection to enhance CBRAM performance by nanopore graphene layer, Small 13 (2017), 1603948. [3] S. Lee, T. Kim, B. Jang, W.Y. Lee, K.C. Song, H.S. Kim, G.Y. Do, S.B. Hwang, S. Chung, J. Jang, Impact of device area and film thickness on performance of sol-gel processed ZrO2 RRAM, IEEE Electron. Device Lett. 39 (2018) 668e671. [4] H.D. Kim, S. Kim, M.J. Yun, Self-rectifying resistive switching behavior observed in Al2O3-based resistive switching memory devices with p-AlGaN semiconductor bottom electrode, J. Alloys Compd. 742 (2018) 822e827. [5] J.X. Xiao, T.S. Herng, J. Ding, K.Y. Zeng, Resistive switching behavior in copper doped zinc oxide (ZnO:Cu) thin films studied by using scanning probe microscopy techniques, J. Alloys Compd. 709 (2017) 535e541. [6] L.P. Fu, Y.T. Li, G.L. Han, X.P. Gao, C.B. Chen, P. Yuan, Stable resistive switching characteristics of ZrO2-based memory device with low-cost, Microelectron. Eng. 172 (2017) 26e29. [7] C. Sun, S.M. Lu, F. Jin, W.Q. Mo, J.L. Song, K.F. Dong, Control the switching mode of Pt/HfO2/TiN RRAM devices by tuning the crystalline state of TiN electrode, J. Alloys Compd. 749 (2018), 487-486.
1306
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
[8] T. Park, Y.J. Kwon, H.J. Kim, H.C. Woo, G.S. Kim, C.H. An, Y. Kim, D.E. Kwon, C.S. Hwang, Balancing the source and sink of oxygen vacancies for the resistive switching memory, ACS Appl. Mater. Interfaces 6 (2018) 9731e9738. [9] W. Kim, S. Menzel, D.J. Wouters, Y. Guo, J. Robertson, B. Roesgen, R. Waser, V. Rana, Impact of oxygen exchange reaction at the ohmic interface in Ta2O5based ReRAM devices, Nanoscale 8 (2016) 17774e17781. [10] C. Wang, B. Song, Z.M. Zeng, Excellent selector performance in engineered Ag/ ZrO2:Ag/Pt structure for high-density bipolar RRAM applications, AIP Adv. 7 (2017), 125209. [11] J. Jang, V. Subramanian, Effect of electrode material on resistive switching memory behavior of solution-processed resistive switches: realization of robust multi-level cells, Thin Solid Films 625 (2017) 87e92. [12] B. Gao, H.W. Zhang, S. Yu, B. Sun, L.F. Liu, X.Y. Liu, Y. Wang, R.Q. Han, J.F. Kang, B. Yu, Y.Y. Wang, Oxide-based RRAM: uniformity improvement using a new material-oriented methodology, in: Symp. on VLSI Technol., 2009, p. 978. [13] T.T. Tan, Y.H. Du, A. Cao, Y.L. Sun, G.q. Zha, H. Lei, X.S. Zheng, The resistive switching characteristics of Ni-doped HfOx film and its application as a synapse, J. Alloys Compd. 766 (2018) 918e924. [14] H. Jiang, D.A. Stewart, Using dopants to tune oxygen vacancy formation in transition metal oxide resistive memory, ACS Appl. Mater. Interfaces 9 (2017) 16296e16304. €pe, Y. Nishi, K. Shiraishi, [15] M.Y. Yang, K. Kamiya, H. Shirakawa, B. Magyari-Ko Role of nitrogen incorporation into Al2O3-based resistive random-access memory, Appl. Phys. Express 7 (2014), 074202. [16] N. Umezawa, K. Shiraishi, T. Ohno, H. Watanabe, T. Chikyow, K. Torii, K. Yamabe, K. Yamada, H. Kitajima, T. Arikado, First-principles studies of the intrinsic effect of nitrogen atoms on reduction in gate leakage current through Hf-based high-k dielectrics, Appl. Phys. Lett. 86 (2005), 143507. [17] D.H. Kwon, K.M. Kim, J.H. Jang, J.M. Jeon, M.H. Lee, G.H. Kim, X.S. Li, G.S. Park, B. Lee, S. Han, M. Kim, C.S. Hwang, Atomic structure of conducting nanofilaments in TiO2 resistive switching memory, Nat. Nanotechnol. 5 (2010) 148e153. €pe, Y. Nishi, Impact of oxygen vacancy ordering on [18] S.G. Park, B. Magyari-Ko The formation of a conductive filament in TiO2 for resistive switching memory, IEEE Electron. Device Lett. 32 (2011) 197e199. €pe, Y. Nishi, M. Niwa, [19] K. Kamiya, M.Y. Yang, S.G. Park, B. Magyari-Ko K. Shiraishi, ON-OFF switching mechanism of resistive-random-accessmemories based on the formation and disruption of oxygen vacancy conducting channels, Appl. Phys. Lett. 100 (2012), 073502. [20] W. Mahmood, Q. Zhao, First principles calculations on structural, elastic, acoustic and optical properties of fluorite phase TiO2 under pressure, Int. J. Mod. Phys. C 25 (2014), 1450020. [21] T.T. Tan, T.T. Guo, Z.T. Liu, Au doping effects in HfO2-based resistive switching memory, J. Alloys Compd. 610 (2014) 388e391. [22] P. Maleeswaran, D. Nagulapally, R.P. Joshi, A.K. Pradhan, Leakage current in high dielectric oxides: role of defect-induced energies, J. Appl. Phys. 113 (2013), 184504. [23] A.S. Foster, V.B. Sulimov, F. Lopez Gejo, A.L. Shluger, R.M. Nieminen, Structure and electrical levels of point defects in monoclinic zirconia, Phys. Rev. B 64 (2001), 224108. [24] N. Umezawa, M. Sato, K. Shiraishi, Reduction in charged defects associated with oxygen vacancies in hafnia by magnesium incorporation: first-principles
study, Appl. Phys. Lett. 93 (2008), 223104. [25] Y.H. Dai, Z.Y. Pan, F.F. Wang, X.F. Li, Oxygen vacancy effects in HfO2-based resistive switching memory: first principle study, AIP Adv. 6 (2016), 085209. [26] L.J. Zhang, Z.Y. Chen, Q.M. Hu, R. Yang, On the abnormal fast diffusion of solute atoms in a-Ti: a first-principles investigation, J. Alloys Compd. 740 (2018) 156e166. [27] N. Govind, M. Petersen, G. Fitzgerald, D. King-Smith, J. Andzelm, A generalized synchronous transit method for transition state location, Comput. Mater. Sci. 28 (2003) 250e258. [28] H.W. Zhang, B. Gao, S.M. Yu, L. Lai, L. Zeng, B. Sun, L.F. Liu, X.Y. Liu, J. Lu, R.Q. Han, J.F. Kang, Effects of ionic doping on the behaviors of oxygen vacancies in HfO2 and ZrO2: a first principles study, in: SISPAD, Int. Conf. Simul. Semicond. Processes and Devices IEEE, 2009, pp. 1e4. [29] W. Wei, X.C. Chuai, N.D. Lu, Y. Wang, L. Li, C. Ye, M. Liu, Simulation of doping effect for HfO2-based RRAM based on first-principles calculations, in: SISPAD, Int. Conf. Simul. Semicond. Processes Devices IEEE, 2017, pp. 21e24. [30] H. Jiang, R.I. Gomez-Abal, P. Rinke, M. Scheffler, Electronic band structure of zirconia and hafnia polymorphs from the GW perspective, Phys. Rev. B 81 (2010), 085119. [31] D. Ceresoli, D. Vanderbilt, Structural and dielectric properties of amorphous ZrO2 and HfO2, Phys. Rev. B 74 (2006), 125108. [32] J.H. Hur, S. Park, U.I. Chung, First principles study of oxygen vacancy states in monoclinic ZrO2: interpretation of conduction characteristics, J. Appl. Phys. 112 (2012), 113719. [33] P. Broqvist, A. Pasquarello, Oxygen vacancy in monoclinic HfO2: a consistent interpretation of trap assisted conduction, direct electron injection, and optical absorption experiments, Appl. Phys. Lett. 89 (2006), 262904. [34] G. Shang, P.W. Peacock, J. Robertson, Stability and band offsets of nitrogenated high-dielectric-constant gate oxides, Appl. Phys. Lett. 84 (2004) 106. [35] H. Li, Y. Guo, J. Robertson, Dopant compensation in HfO2 and other high K oxides, Appl. Phys. Lett. 104 (2014), 192904. [36] B. Li, H. Sun, C.F. Chen, First-principles calculation of the indentation strength of FeB4, Phys. Rev. B 90 (2014), 014106. [37] C. Rohde, B.J. Choi, D.S. Jeong, S. Choi, J.S. Zhao, C.S. Hwang, Identification of a determining parameter for resistive switching of TiO2 thin films, Appl. Phys. Lett. 86 (2005), 262907. [38] Y.T. Li, P. Yuan, L.P. Fu, R.R. Li, X.P. Gao, C.L. Tao, Coexistence of diode-like volatile and multilevel nonvolatile resistive switching in a ZrO2/TiO2 stack structure, Nanotechnology 26 (2015), 391001. ~ oz-García, F. Colo, G. Meligrana, A. Lamberti, M. Destro, [39] F. Bella, A.B. Mun M. Pavone, C. Gerbaldi, Combined structural, chemometric, and electrochemical investigation of vertically aligned TiO2 nanotubes for Na-ion batteries, ACS Omega 3 (2018) 8440e8450. [40] B. Miccoli, V. Cauda, A. Bonanno, A. Sanginario, K. Bejtka, F. Bella, M. Fontana, D. Demarchi, One-dimensional ZnO/gold junction for simultaneous and versatile multisensing measurements, Sci. Rep. 6 (2016) 29763. €kmena, Behaviour of Trolox with [41] E.E. Çelika, J.M. Rubiob, V. Go macromolecule-bound antioxidants in aqueous medium: inhibition of autoregeneration mechanism, Food Chem. 243 (2018) 428e434. [42] S. Gallianoa, F. Bella, G. Piana, G. Giacona, G. Viscardia, C. Gerbaldi, M. Gr€ atzel, C. Barolo, Finely tuning electrolytes and photoanodes in aqueous solar cells by experimental design, Sol. Energy 163 (2018) 251e255.
Contents lists available at ScienceDirect
Journal of Alloys and Compounds journal homepage: http://www.elsevier.com/locate/jalcom
Exploring the role of nitrogen incorporation in ZrO2 resistive switching film for enhancing the device performance Xiaodi Wei a, Hong Huang a, Cong Ye a, b, *, Wei Wei a, b, Hao Zhou a, Yong Chen a, **, Rulin Zhang a, Li Zhang a, Qing Xia a a b
Faculty of Physics and Electronic Science, Hubei University, Wuhan 430062, China Key Laboratory of Microelectronics Devices & Integration Technology, Institute of Microelectronics of Chinese Academy of Sciences, Beijing, 100029, China
a r t i c l e i n f o
a b s t r a c t
Article history: Received 1 July 2018 Received in revised form 18 October 2018 Accepted 21 October 2018 Available online 22 October 2018
The role of nitrogen doping on the resistive switching (RS) performance in nitrogen doped ZrO2 memristive device is investigated. The Pt/N:ZrO2/TiN resistive random access memory (RRAM) shows smaller switching voltage, larger memory window as well as improved uniformity. Moreover, the multilevel storage capability can be successfully obtained by varying the compliance current in the SET process for the memory cell. It is considered that the connection and rupture of conducting oxygen vacancy filaments (CF) can be localized and the oxygen ions (O2) migration randomness is depressed due to nitrogen doping in ZrO2 film. Combining with the first-principle method, we theoretically calculate the formation energy (Evf), migration energy (Em) and density of states for oxygen vacancy (VO). Both Evf and Em values show noticeable decrease in N doped 2 2 2 ZrO2 supercell, which are related to the lower forming voltage and operating voltage. The density of states indicates that the oxygen vacancy midgap defect states can be eliminated as a result of N dopant, which neutralizes the excess defects in ZrO2 switching layer and may reduce the densities of the potential filaments. Herein the uniformity can be improved. All the theoretical results show reasonable agreement with the improved experimental RS performance for Pt/N:ZrO2/TiN device. © 2018 Elsevier B.V. All rights reserved.
Keywords: RRAM Nitrogen doping Multilevel storage Oxygen vacancy First-principle method
1. Introduction Resistive random access memory (RRAM) has been the most promising candidate for next generation nonvolatile memory in terms of its excellent scalability, simple structure, low operating power, rapid operation and high density integration [1e3]. Numerous transition metal oxides, such as Al2O3, Ta2O5, ZrO2, and HfO2 have been reported to show resistive switching (RS) characteristics [4e7]. It is well known that oxide-based RRAM switching behavior between high resistive states (HRS) and low resistance states (LRS) occurs by controlling the connection and rupture of the conducting oxygen vacancy (VO) filaments (CF) [8]. Herein, it is crucial to engineer the VO profile in the oxide film to achieve stable memory states and a large switching memory window [9]. In oxide
* Corresponding author. Key Laboratory of Microelectronic Devices & Integrated Technology，Institute of Microelectronics of ChineseAcademy of Sciences, Beijing100029, China. ** Corresponding author. E-mail addresses: [email protected] (C. Ye), [email protected] (Y. Chen). https://doi.org/10.1016/j.jallcom.2018.10.249 0925-8388/© 2018 Elsevier B.V. All rights reserved.
materials, ZrO2 has recently attracted significant attention for fabricating high quality RRAM devices [6,10,11]. However, one of the representative obstacles for ZrO2 RRAM device is the nonstability of RS parameters such as switching voltage and resistance states. Gao et al. found that Al doping ZrO2 can effectively reduce the VO formation energy and the forming voltage, and the conductive VO filament preferentially grows near the Al doping site by the first-principles calculations [12]. Recently, doping impurity elements have been reported to be effective for controlling the location of the CF and improving the switching parameters [13,14]. Also, it has been reported that N doping has often been used to enhance property of devices. For example, N doping Al2O3 RRAM significantly reduces the forming voltage and the operating current [15]; N incorporation into HfO2 reduces the leakage current by deactivating VO defect states in the band gap and elevating the VO level in the band gap to over the conduction band [16]. Therefore, it has been believed that the VO migration is the origin of resistive switching behaviors and N also plays an important role in oxide-based RRAM [17e19]. It is important to understand the origin of beneficial effects for N
1302
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
doping oxide-based RRAM on the VO properties. In this work, it is aimed to investigate the role of N incorporation into ZrO2-based RRAM on the VO properties by using the firstprinciples calculations. The calculated results show that the formation energy and migration energy of VO in ZrO2 are significantly decreased by N incorporation, which leads to the lower forming voltage and operating voltage in RRAM devices. Furthermore, we reveal that N atoms favorably couple with VO in ZrO2, which induces a more stable conductive filament by preventing VO diffusion. These results clearly explain how N incorporation into ZrO2based RRAM improves the performance of the memory device. The improved resistive switching behaviors and the multilevel storage capability in N-doped ZrO2 device are demonstrated in the experiment. 2. Calculation methods and experiments The first-principles calculations based on the density functional theory (DFT) is performed on monoclinic ZrO2 (m-ZrO2) with the generalized gradient approximation-Perdew Burke Ernzerh of (GGA-PBE) [20] for the exchange correlation potential, and Vanderbilt type ultrasoft pseudopotential [21] for core-valence interaction, as implemented in the CASTEP code [22]. The first-principle method is a standard theoretical solution to investigate the electronic behaviors of monoclinic ZrO2, such as the band structure, density of states and electron localization function. The inputs of the CASTEP code mainly include serials of parameters: the cutoff energy, the k-point, the force, the maximum displacement and the smearing width. After the self-consistent calculation, the material properties can be obtained, such as the band structure, density of states and electron localization function. The plane wave set basis cutoff energy is 450 eV, and the Brillouin zone is sampled at a grid of 2 2 2 Monkhorst-Pack k-point. The supercell is fully relaxed until the residual force of atoms is less than 0.001 eV/Å, the maximum ionic displacement is within 5.0 104 Å, and the smearing width is 0.2 eV. The valence electrons for zirconium are in the 4s24p64d25s2 configuration. The configuration 2s22p4 is used for the generation of the valence electrons of oxygen. The m-ZrO2 cell parameters are determined as a ¼ 5.192 Å, b ¼ 5.265 Å, c ¼ 5.358 Å, b ¼ 99.81, Zr(0.277, 0.044, 0.209), O(0.072, 0.338, 0.341) and O(0.447, 0.758, 0.479) [23]. The ZrO2 supercell containing 96 atoms is established by using the Materials Studio software. A m-ZrO2 (space group P21/c) 2 2 2 supercell of 96 atoms is chosen as a starting point to build these models. It is found that the threefold coordinated VO in ZrO2 networks has a small formation energy, low diffusion barrier, and forms a relatively stable configuration by binding with dopant [24], which is believed to be strongly related to the resistive switching behavior of the RRAM devices. Hence, this work focuses on investigating the three coordinated VO in ZrO2. Next, we construct four different models (model A, B, C, and D as shown in Fig. 1) by substituting two O atoms with N atoms and removing one O atom to investigate the VO interaction between N atoms. Distances from the VO to each N atom are 2.8 and 5.3 Å in Model B, 2.8 and 6.8 Å in Model C, and 6.8 and 7.5 Å in Model D, respectively. In the calculations, the formation energy (Evf) and the migration energy (Em) of VO can remarkably affect the performance of RRAM device [25], thus it is essential to clearly understand the N doping effect of Evf and Em for the VO in RRAM device. Based on the first-principles calculations, the VO formation energy can be calculated as [24].
Evf ¼ EðVO Þ EðbulkÞ þ mO where E(VO) and E(bulk) denote the total energy of each model with or without a VO in the undoped or N-doped ZrO2 bulk crystal,
respectively. mo is the chemical potential of an oxygen atom, which is half of the oxygen molecule energy in this calculation. To simulate the VO diffusion process, one oxygen ion is moved from the lattice site to the neighboring vacant sites in ZrO2. The migration energy (Em) is defined as the total energy difference between saddle point energy of the diffusion process and the VO energy at the initial site [26]. The migration barrier is a chemical reaction from initial state to the final state. The most important process is to find the middle of the transition state which is the most difficult between the initial position and the final position. Thus, transition state search is used to search the transition state and calculate the Em of VO in ZrO2. Linear synchronous transit (LST), followed by repeated conjugate gradient minimizations and quadric synchronous transit (QST) maximizations calculations are performed to search the saddle point in the VO migration process [27]. In the experiment, undoped and N-doped ZrO2 memory devices were deposited on TiN/SiO2/Si substrates by radio frequency magnetron sputtering in pure Ar atmosphere and N2/Ar mixed atmosphere. Undoped ZrO2 films were deposited by a ZrO2 target (99.99%) with the power of 60 W, in a pure 45 sccm Ar gas at room temperature by sputtering. N-doped ZrO2 films were deposited in a gas mixture of 30 sccm Ar and 15 sccm N2. The switching layer was same thickness of 25 nm. Then, the top electrode Pt with 200 nm thickness was deposited by direct current (DC) magnetron sputtering by a Pt target (99.99%) with the power of 100 W in a pure argon atmosphere. The working pressure was kept at 0.5 Pa. Lastly, in order to shape and complete an individual TiN/ZrOx/Pt RRAM, photolithography and the lift-off technique were employed. All the electrical properties were measured by using an Agilent B1500A semiconductor parameter analyzer. 3. Results and discussion Fig. 1 presents the VO formation energy in ZrO2 for all models and the current-voltage (IeV) curves of the formation process for the Pt/ZrO2/TiN and Pt/N:ZrO2/TiN devices. In order to calculate the VO formation energy, we establish oxygen molecule model and ZrO2 supercell models with or without the oxygen vacancy by using the Materials Studio software. We carry out geometry optimization for the established models by the CASTEP code. We choose energy option to calculate the Kohn-Sham energy for the optimized models by the CASTEP code. After analyzing energy evolution of the optimized models, we can obtain mo, E(bulk) and E(VO) of ZrO2 supercell models with or without an oxygen vacancy. According to the equation, we can calculate the VO formation energies (Evf), which are 5.82 eV, 3.15 eV, 3.25 eV and 3.45 eV for Models A, B, C and D in Fig. 1(a), respectively. As shown in Fig. 1(a), it is found that the calculated Evf in ZrO2 is significantly reduced by introducing N atoms, which indicates that VO easily forms when N atoms are incorporated into ZrO2-based RRAM. The Evf of Model B is the lowest, and the Evf values of Model C and Model D are calculated to be 3.25 and 3.45 eV, respectively. These results indicate that the most stable configuration is Model B, and N atoms are inclined to couple with the VO in ZrO2 with closer distance site. It can be explained by dipole formation and the Coulomb interaction energy [28]. In the Pt/ZrO2/TiN device and the Pt/N:ZrO2/TiN device, the voltage was biased on the TiN bottom electrode while the Pt top electrode was grounded. Fig. 1(b) presents IeV characteristics of forming process for both devices. As the positive bias voltage was applied on the TiN bottom electrode, a soft breakdown occurred at about 7.2 V under a compliance current of 500 mA for the Pt/ZrO2/ TiN memory cell. For the Pt/N:ZrO2/TiN device, the forming voltage decreased to 3.6 V. Normally, the forming process is interpreted to be equivalent to the dielectric soft breakdown accompanying with creation of the VO, which plays the crucial role in the memory
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
1303
Fig. 1. (a) The formation energy of oxygen vacancy in undoped and N-doped ZrO2, the upper inset shows ZrO2 networks. The gray, blue, pink, red, and yellow balls denote zirconium, nitrogen, 3-fold oxygen, 4-fold oxygen and the VO site, respectively. (b) The forming I-V characteristics for ZrO2 device and N:ZrO2 device. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
operation for oxide-based RRAM [18]. Here, it is assumed that N doping stimulates the VO formation as a result of lower Evf and influences the formation of the conductive filament, so the forming voltage decreased for Pt/N:ZrO2/TiN device. In fact, IeV curves for RRAM devices can be calculated and show excellent consistency with the experimental results in our previous work [29]. Here, we aim to obtain the experimental results from the IeV measurement of forming process, and explain it by calculating the VO formation energy of undoped and N-doped ZrO2 supercells. The conductive VO filament formation depends not only on the VO formation energy, but also on the VO migration energy in the system. The calculated values of Em are presented with different distance of migration paths in Fig. 2(a). In terms of the distance from VO to N atom, it can be considered as VO near to N atom and VO away from N atom. Thus, the migration paths of VO can be classified as migration close to N atom and outward from N atom. For migration of VO, the calculated Em decreases in N doped ZrO2. In contrast, the calculated Em increases with migration distance in ZrO2 with or without N atoms. These results indicate that the VO easily migrates towards N atoms and the conductive VO filament preferentially grows near N doping sites. Therefore, N incorporation would improve the reliability of the conductive VO filament in RRAM by suppression of VO diffusion. Similar results have been proved in N-doped Al2O3 film [15]. On microscopic aspect, the
degree of difficulty in the VO migration process determines the macroscopic properties of devices such as the SET voltage and the RESET voltage. After the electroforming process, both devices can be reversibly switched from the HRS to the LRS and from the LRS to the HRS. Fig. 2(b) presents the IeV curves for the Pt/ZrO2/TiN and Pt/N:ZrO2/TiN device. During the electrical test, the voltage was applied on the TiN electrode when the Pt electrode was grounded, and a compliance current of 1 mA was used. The result clearly shows that the switching voltages (VSET and VRESET) became smaller for the Pt/N:ZrO2/TiN device. It should be noted that the SET voltage reduces from 0.61 to 0.52 V in the N-doped device as compared with the device without doping. All the theoretical Em results show reasonable agreement with the experimental SET voltage for Pt/ N:ZrO2/TiN device. To analyze the electronic properties of ZrO2, the calculated energy band structure and density of states (DOS) of undoped ZrO2 with VO (Model A) and N-doped ZrO2 with VO (Model B) are shown in Fig. 3. The zero-point energy is taken as the Fermi level. The calculated band gap of perfect ZrO2 supercell is about 3.73 eV, which is consistent with other DFT calculations [30,31]. It should be noted that the calculated data is smaller than the experimental data of 5.4 eV [32] due to the well-known band gap underestimations of DFT calculations [23]. As shown in Fig. 3(a), the single VO in ZrO2 network will introduce an additional band localized at the Fermi
Fig. 2. (a) The VO migration energy as a function of migration distance, the upper inset shows the possible VO site (number between 1 and 7) in ZrO2 networks. The gray, blue, pink and red balls denote zirconium, nitrogen, 3-fold oxygen, 4-fold oxygen, respectively. (b) Typical I-V curves for Pt/N:ZrO2/TiN and Pt/N:ZrO2/TiN devices. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
1304
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
Fig. 3. The band structure and density of states for (a) and (b) undoped ZrO2 with VO (Model A), (c) and (d) N-doped ZrO2 with VO (Model B), respectively.
level between the conduction band and the valence band, and the additional band is mainly composed of the Zr 4d and O 2p orbitals. Fig. 3(b) shows that the VO gives rise to a defect state in the mid bandgap region, with the defect state orbital localized on the nearby Zr atom and O atom. Fermi level is at midgap state, indicating that the defect level of the neutrally charged VO is occupied. It means that the oxygen vacancy acts as an electron trap and traps electrons from surrounding atoms [33]. The effect of N doping has been studied by substituting two O atoms with N atoms nearest the VO site (Model B). The additional band and the original VO defect state in the bandgap of undoped ZrO2 with VO is removed due to N doping in Fig. 3(c) and (d), leaving some N non-bonding states near the valence band of ZrO2 [34,35]. The original VO as a neutral charge state introduces two excess electrons localized on the nearby Zr atom. During the N atoms doping process, each N atom substitutes one O atom in the oxide network and creates a valence band hole. The electrons transfer from the higher VO level to the lower 2p orbitals of N atoms at the valence band top. As a result, the two excess electrons fill the valence states and form a closed shell electronic configuration. It will make the system extremely stable. Hence, the VO becomes a V2þ O center, involving a certain lattice distortion, and pushes the original defect state up into the conduction band [35]. The energy gain of this process comes from the two electrons falling from mid gap states into the valence band holes, which compensates the energy required for local distortion. In this way, the VO in the ZrO2 network is nullified by the presence of two N atoms. The neutralization of the defect states association with the VO eliminates the excess conductive paths and reduces the numbers of the potential filaments. Electron localization function (ELF) was initially proposed to measure electron pairing by defining [36].
ELF ¼
1 1 þ ½DðrÞ=Dh ðrÞ2
where the term D(r)/Dh(r) normalizes the same-spin probability by the uniform-density electron gas as reference, and thus ELF is a dimensionless localization index restricted to the range of [0,1]. The ELF value at 0e0.5 and at 0.5e1 denote low electron density area and high electron density area, respectively. A high ELF value stands for a low probability of finding a second electron with the same spin in the neighboring region of the reference electron, and the reference electron is highly localized. Based on its definition, one can simply interpret high ELF values as covalent bonds, lone pairs, or inert cores. Here, we focus on the [001] surface which has proven to be the most stable surface over a wide range of electron state. Fig. 4 indicates that N atoms tend to capture electrons from VO and thus neutralize the defect states of oxygen vacancies, which may eliminate the excess conductive paths. Fig. 5(a) shows the multilevel IeV curves for the Pt/N:ZrO2/TiN device with different compliance currents of 500 mA, 1 mA, and 5 mA. From Fig. 5(b), controlling varying the compliance current in the SET process, intermediate resistances of LRS occur and we can observe three resistances of LRS (level 2, level 3 and level 4). It is well acknowledged that with the increase of compliance current for the SET process, the conductive filament connecting both electrodes becomes strong and the resistance of LRS becomes small [37]. Here, for the forming compliance current of 5 mA, the conductive filament is the strongest, whereas the conductive filament is the weakest for the forming compliance current of 500 mA. Therefore, by varying the compliance currents of 500 mA, 1 mA, and 5 mA in the SET process, three resistances of LRS can be obtained, which is called multilevel storage. ZrO2-based RRAM with
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
1305
Fig. 4. ELF for (a) undoped ZrO2 and (b) N doped ZrO2. The ELF value at 0e0.5 and at 0.5e1 indicate low electron density area and high electron density areas, respectively.
Fig. 5. (a) Typical I-V curves for the Pt/N:ZrO2/TiN RRAM device under different compliance currents, (b) The HRS and LRS of the Pt/N:ZrO2/TiN RRAM device with different compliance currents.
multilevel storage can greatly achieve large integration and highdensity storage. In addition, the multilevel memory device opens application areas in logic functions, neural networks and passive crossbar memory selectors [38]. Lastly, for multiple application of ZrO2-based RRAM in the future work, it is noted that more detailed experimental and calculation methods are performed by varying temperature, different doping concentration and stoichiometric tools [39e42].
Acknowledgements This study was supported by the National Nature Science Foundation of China (No.61474039; No.61774057); The Opening Project of Key Laboratory of Microelectronic Devices & Integrated Technology, Institute of Microelectronics of the Chinese Academy of Sciences (Y7YS013001); National Key Research and Development Program under Grant 2017YFB0405602. References
4. Conclusions In summary, the role of N incorporation into ZrO2-based RRAM has been investigated by using first-principles calculations. The calculated results showed that the VO formation energy and migration energy in ZrO2 is significantly reduced by N incorporation, resulting in a lower forming voltage and SET voltage in RRAM device. Herein, N atoms favorably couple with VO and stabilize the conductive filament by preventing VO diffusion in N-doped ZrO2. Furthermore, the density of states for N-doped ZrO2 supercell indicates that the oxygen vacancy midgap defect states can be eliminated. Hence, nitrogen incorporation can neutralize the excess defects in N-doped ZrO2 switching layer and may reduce the densities of the potential filaments. Finally, multilevel storage was achieved with different compliance currents in the SET process of the Pt/N:ZrO2/TiN device. Thus, the results are conductive to designing ZrO2-based RRAM devices for optimized performance.
[1] B. Ku, Y. Abbas, A.S. Sokolov, C. Choi, Interface engineering of ALD HfO2-based RRAM with Ar plasma treatment for reliable and uniform switching behaviors, J. Alloys Compd. 735 (2018) 1181e1188. [2] X.L. Zhao, S. Liu, J.B. Niu, L. Liao, Q. Liu, X.H. Xiao, H.B. Lv, S.B. Long, W. Banerjee, W.Q. Li, S.Y. Si, M. Liu, Confining cation injection to enhance CBRAM performance by nanopore graphene layer, Small 13 (2017), 1603948. [3] S. Lee, T. Kim, B. Jang, W.Y. Lee, K.C. Song, H.S. Kim, G.Y. Do, S.B. Hwang, S. Chung, J. Jang, Impact of device area and film thickness on performance of sol-gel processed ZrO2 RRAM, IEEE Electron. Device Lett. 39 (2018) 668e671. [4] H.D. Kim, S. Kim, M.J. Yun, Self-rectifying resistive switching behavior observed in Al2O3-based resistive switching memory devices with p-AlGaN semiconductor bottom electrode, J. Alloys Compd. 742 (2018) 822e827. [5] J.X. Xiao, T.S. Herng, J. Ding, K.Y. Zeng, Resistive switching behavior in copper doped zinc oxide (ZnO:Cu) thin films studied by using scanning probe microscopy techniques, J. Alloys Compd. 709 (2017) 535e541. [6] L.P. Fu, Y.T. Li, G.L. Han, X.P. Gao, C.B. Chen, P. Yuan, Stable resistive switching characteristics of ZrO2-based memory device with low-cost, Microelectron. Eng. 172 (2017) 26e29. [7] C. Sun, S.M. Lu, F. Jin, W.Q. Mo, J.L. Song, K.F. Dong, Control the switching mode of Pt/HfO2/TiN RRAM devices by tuning the crystalline state of TiN electrode, J. Alloys Compd. 749 (2018), 487-486.
1306
X. Wei et al. / Journal of Alloys and Compounds 775 (2019) 1301e1306
[8] T. Park, Y.J. Kwon, H.J. Kim, H.C. Woo, G.S. Kim, C.H. An, Y. Kim, D.E. Kwon, C.S. Hwang, Balancing the source and sink of oxygen vacancies for the resistive switching memory, ACS Appl. Mater. Interfaces 6 (2018) 9731e9738. [9] W. Kim, S. Menzel, D.J. Wouters, Y. Guo, J. Robertson, B. Roesgen, R. Waser, V. Rana, Impact of oxygen exchange reaction at the ohmic interface in Ta2O5based ReRAM devices, Nanoscale 8 (2016) 17774e17781. [10] C. Wang, B. Song, Z.M. Zeng, Excellent selector performance in engineered Ag/ ZrO2:Ag/Pt structure for high-density bipolar RRAM applications, AIP Adv. 7 (2017), 125209. [11] J. Jang, V. Subramanian, Effect of electrode material on resistive switching memory behavior of solution-processed resistive switches: realization of robust multi-level cells, Thin Solid Films 625 (2017) 87e92. [12] B. Gao, H.W. Zhang, S. Yu, B. Sun, L.F. Liu, X.Y. Liu, Y. Wang, R.Q. Han, J.F. Kang, B. Yu, Y.Y. Wang, Oxide-based RRAM: uniformity improvement using a new material-oriented methodology, in: Symp. on VLSI Technol., 2009, p. 978. [13] T.T. Tan, Y.H. Du, A. Cao, Y.L. Sun, G.q. Zha, H. Lei, X.S. Zheng, The resistive switching characteristics of Ni-doped HfOx film and its application as a synapse, J. Alloys Compd. 766 (2018) 918e924. [14] H. Jiang, D.A. Stewart, Using dopants to tune oxygen vacancy formation in transition metal oxide resistive memory, ACS Appl. Mater. Interfaces 9 (2017) 16296e16304. €pe, Y. Nishi, K. Shiraishi, [15] M.Y. Yang, K. Kamiya, H. Shirakawa, B. Magyari-Ko Role of nitrogen incorporation into Al2O3-based resistive random-access memory, Appl. Phys. Express 7 (2014), 074202. [16] N. Umezawa, K. Shiraishi, T. Ohno, H. Watanabe, T. Chikyow, K. Torii, K. Yamabe, K. Yamada, H. Kitajima, T. Arikado, First-principles studies of the intrinsic effect of nitrogen atoms on reduction in gate leakage current through Hf-based high-k dielectrics, Appl. Phys. Lett. 86 (2005), 143507. [17] D.H. Kwon, K.M. Kim, J.H. Jang, J.M. Jeon, M.H. Lee, G.H. Kim, X.S. Li, G.S. Park, B. Lee, S. Han, M. Kim, C.S. Hwang, Atomic structure of conducting nanofilaments in TiO2 resistive switching memory, Nat. Nanotechnol. 5 (2010) 148e153. €pe, Y. Nishi, Impact of oxygen vacancy ordering on [18] S.G. Park, B. Magyari-Ko The formation of a conductive filament in TiO2 for resistive switching memory, IEEE Electron. Device Lett. 32 (2011) 197e199. €pe, Y. Nishi, M. Niwa, [19] K. Kamiya, M.Y. Yang, S.G. Park, B. Magyari-Ko K. Shiraishi, ON-OFF switching mechanism of resistive-random-accessmemories based on the formation and disruption of oxygen vacancy conducting channels, Appl. Phys. Lett. 100 (2012), 073502. [20] W. Mahmood, Q. Zhao, First principles calculations on structural, elastic, acoustic and optical properties of fluorite phase TiO2 under pressure, Int. J. Mod. Phys. C 25 (2014), 1450020. [21] T.T. Tan, T.T. Guo, Z.T. Liu, Au doping effects in HfO2-based resistive switching memory, J. Alloys Compd. 610 (2014) 388e391. [22] P. Maleeswaran, D. Nagulapally, R.P. Joshi, A.K. Pradhan, Leakage current in high dielectric oxides: role of defect-induced energies, J. Appl. Phys. 113 (2013), 184504. [23] A.S. Foster, V.B. Sulimov, F. Lopez Gejo, A.L. Shluger, R.M. Nieminen, Structure and electrical levels of point defects in monoclinic zirconia, Phys. Rev. B 64 (2001), 224108. [24] N. Umezawa, M. Sato, K. Shiraishi, Reduction in charged defects associated with oxygen vacancies in hafnia by magnesium incorporation: first-principles
study, Appl. Phys. Lett. 93 (2008), 223104. [25] Y.H. Dai, Z.Y. Pan, F.F. Wang, X.F. Li, Oxygen vacancy effects in HfO2-based resistive switching memory: first principle study, AIP Adv. 6 (2016), 085209. [26] L.J. Zhang, Z.Y. Chen, Q.M. Hu, R. Yang, On the abnormal fast diffusion of solute atoms in a-Ti: a first-principles investigation, J. Alloys Compd. 740 (2018) 156e166. [27] N. Govind, M. Petersen, G. Fitzgerald, D. King-Smith, J. Andzelm, A generalized synchronous transit method for transition state location, Comput. Mater. Sci. 28 (2003) 250e258. [28] H.W. Zhang, B. Gao, S.M. Yu, L. Lai, L. Zeng, B. Sun, L.F. Liu, X.Y. Liu, J. Lu, R.Q. Han, J.F. Kang, Effects of ionic doping on the behaviors of oxygen vacancies in HfO2 and ZrO2: a first principles study, in: SISPAD, Int. Conf. Simul. Semicond. Processes and Devices IEEE, 2009, pp. 1e4. [29] W. Wei, X.C. Chuai, N.D. Lu, Y. Wang, L. Li, C. Ye, M. Liu, Simulation of doping effect for HfO2-based RRAM based on first-principles calculations, in: SISPAD, Int. Conf. Simul. Semicond. Processes Devices IEEE, 2017, pp. 21e24. [30] H. Jiang, R.I. Gomez-Abal, P. Rinke, M. Scheffler, Electronic band structure of zirconia and hafnia polymorphs from the GW perspective, Phys. Rev. B 81 (2010), 085119. [31] D. Ceresoli, D. Vanderbilt, Structural and dielectric properties of amorphous ZrO2 and HfO2, Phys. Rev. B 74 (2006), 125108. [32] J.H. Hur, S. Park, U.I. Chung, First principles study of oxygen vacancy states in monoclinic ZrO2: interpretation of conduction characteristics, J. Appl. Phys. 112 (2012), 113719. [33] P. Broqvist, A. Pasquarello, Oxygen vacancy in monoclinic HfO2: a consistent interpretation of trap assisted conduction, direct electron injection, and optical absorption experiments, Appl. Phys. Lett. 89 (2006), 262904. [34] G. Shang, P.W. Peacock, J. Robertson, Stability and band offsets of nitrogenated high-dielectric-constant gate oxides, Appl. Phys. Lett. 84 (2004) 106. [35] H. Li, Y. Guo, J. Robertson, Dopant compensation in HfO2 and other high K oxides, Appl. Phys. Lett. 104 (2014), 192904. [36] B. Li, H. Sun, C.F. Chen, First-principles calculation of the indentation strength of FeB4, Phys. Rev. B 90 (2014), 014106. [37] C. Rohde, B.J. Choi, D.S. Jeong, S. Choi, J.S. Zhao, C.S. Hwang, Identification of a determining parameter for resistive switching of TiO2 thin films, Appl. Phys. Lett. 86 (2005), 262907. [38] Y.T. Li, P. Yuan, L.P. Fu, R.R. Li, X.P. Gao, C.L. Tao, Coexistence of diode-like volatile and multilevel nonvolatile resistive switching in a ZrO2/TiO2 stack structure, Nanotechnology 26 (2015), 391001. ~ oz-García, F. Colo, G. Meligrana, A. Lamberti, M. Destro, [39] F. Bella, A.B. Mun M. Pavone, C. Gerbaldi, Combined structural, chemometric, and electrochemical investigation of vertically aligned TiO2 nanotubes for Na-ion batteries, ACS Omega 3 (2018) 8440e8450. [40] B. Miccoli, V. Cauda, A. Bonanno, A. Sanginario, K. Bejtka, F. Bella, M. Fontana, D. Demarchi, One-dimensional ZnO/gold junction for simultaneous and versatile multisensing measurements, Sci. Rep. 6 (2016) 29763. €kmena, Behaviour of Trolox with [41] E.E. Çelika, J.M. Rubiob, V. Go macromolecule-bound antioxidants in aqueous medium: inhibition of autoregeneration mechanism, Food Chem. 243 (2018) 428e434. [42] S. Gallianoa, F. Bella, G. Piana, G. Giacona, G. Viscardia, C. Gerbaldi, M. Gr€ atzel, C. Barolo, Finely tuning electrolytes and photoanodes in aqueous solar cells by experimental design, Sol. Energy 163 (2018) 251e255.