Direct Probing of the Dielectric Scavenging-Layer Interface in Oxide

Publication Date (Web): March 7, 2017 ... the metal–insulator interface thus providing an unlimited availability of building blocks for the conducti...
0 downloads 0 Views 2MB Size
Research Article www.acsami.org

Direct Probing of the Dielectric Scavenging-Layer Interface in Oxide Filamentary-Based Valence Change Memory Umberto Celano,*,† Jonathan Op de Beeck,†,∥ Sergiu Clima,† Michael Luebben,‡ Paul M. Koenraad,∥ Ludovic Goux,‡ Ilia Valov,*,‡ and Wilfried Vandervorst†,§ †

IMEC, Kapeldreef 75, B-3001 Heverlee (Leuven), Belgium Peter Grünberg Institute and Jülich Aachen Research Alliance, Jülich 52425, Germany § KU Leuven, Department of Physics and Astronomy, Celestijnenlaan 200D, B-3001 Leuven, Belgium ∥ Department of Applied Physics, Eindhoven University of Technology, Eindhoven 5612AZ, The Netherlands ‡

S Supporting Information *

ABSTRACT: A great improvement in valence change memory performance has been recently achieved by adding another metallic layer to the simple metal−insulator−metal (MIM) structure. This metal layer is often referred to as oxygen exchange layer (OEL) and is introduced between one of the electrodes and the oxide. The OEL is believed to induce a distributed reservoir of defects at the metal−insulator interface thus providing an unlimited availability of building blocks for the conductive filament (CF). However, its role remains elusive and controversial owing to the difficulties to probe the interface between the OEL and the CF. Here, using Scalpel SPM we probe multiple functions of the OEL which have not yet been directly measured, for two popular VCMs material systems: Hf/HfO2 and Ta/Ta2O5. We locate and characterize in three-dimensions the volume containing the oxygen exchange layer and the CF with nanometer lateral resolution. We demonstrate that the OEL induces a thermodynamic barrier for the CF and estimate the minimum thickness of the OEL/oxide interface to guarantee the proper switching operations is ca. 3 nm. Our experimental observations are combined to first-principles thermodynamics and defect kinetics to elucidate the role of the OEL for device optimization. KEYWORDS: RRAM, scalpel SPM, filament, oxygen-exchange-layer, scavenging layer



the formation and disruption of nanosized conductive filaments thus altering the cell’s resistive state.2,3 A great improvement in the device performance has been recently achieved by adding another metallic layer to the simple metal−insulator−metal (MIM) structure as presented elsewhere.4−6 This metal layer is often referred to as oxygen exchange layer (OEL) or oxygen scavenger.5−7 Thus, the supporting oxygen exchange layer is introduced between the top-electrode and the oxide (Figure 1a). The OEL is believed to induce a distributed reservoir of defects at the metal−insulator interface, thus providing an unlimited availability of defects, i.e., which are the building blocks for the conductive filament (CF). In essence, the chemical reaction between the OEL (metal) and the oxide layer results in the transport of oxygen anions (O2−) from the dielectric into the metal and it is creating oxygen vacancies, VO•• or cation interstitials MeIx′ (in the Kröger−Vink notation) into the oxide. The oxygen ions (or equivalently VO••)

INTRODUCTION

Fueled by the improved understanding of the resistive switching (RS) mechanisms of recent years, the filamentary-based valence change memory (VCM) currently represents a fast, scalable, and low-power nonvolatile alternative for multiple memory applications. Although the introduction of 3D vertical NAND precludes the usage of VCMs as replacement for stand-alone memory in the short term, the latency/performance gap still existing between Flash and dynamic random access memory (DRAM) creates a clear opportunity for VCM to enter the applications offered by the Internet-of-Things (IoT) and storage class memory to name a few.1 However, additional understanding combined with material research and process development is required to overcome the reliability challenges (variability, endurance, and retention) that are still hampering the widespread application of VCMs. A basic VCM cell is constituted of two metal electrodes and one dielectric layer sandwiched between them, largely based on binary oxides and transition metal oxides (TMO). Inside the stack, the interplay between ion migration and atomic scale redox reactions induces © 2017 American Chemical Society

Received: December 19, 2016 Accepted: March 7, 2017 Published: March 7, 2017 10820

DOI: 10.1021/acsami.6b16268 ACS Appl. Mater. Interfaces 2017, 9, 10820−10824

Research Article

ACS Applied Materials & Interfaces

formed, oxygen ions leave the oxide matrix under the effect of a high electric field. A certain amount of these oxygen ions are stored in the OEL to create the conductive path in the oxide constituting the low resistive state (LRS) of the device. On the contrary, some oxygen ions will be transferred from the OEL reservoir back into the conductive path during the reset to restore the high resistive state (HRS). The OEL can also exchange cations with the oxide matrix. Besides favoring the filament formation, it was shown experimentally that the OEL reduces the devices’ variability. The latter is achieved, by the uniform formation of the OEL/oxide interface in all the devices, which reduces the impact of the local defects and intrinsic film inhomogeneity, e.g., grain versus grain-boundaries, during the CF formation. In addition, the oxygen scavenged from the stoichiometric TMO leads to a reduced electrical thickness of the oxide and therefore lowers the forming voltage. Finally, an improved endurance and retention has been related to the presence of the OEL. Despite its central role, the direct analysis of the interface between the OEL and the CF, is still missing owing to the required complex analysis conditions. The most common methods used to study this region of the VCM cell are based on X-ray spectroscopy such as photoelectron spectroscopy (XPS) and absorption spectroscopy (XAS), or transmission electron microscopy (TEM).8,15,17−19 Despite their high chemical sensitivity these techniques have some disadvantages: (1) a relatively complex sample preparation, (2) complicated filament location, and (3) limited spatial resolution due to a strong background signal leading to an averaged measurement. Indeed, previous studies have shown that filamentary-based RS is generally triggered by the formation/dissolution process of localized conductive paths with dimensions in the range of ∼10 nm.18,20 In addition, there is the issue of sensitivity as only minimal compositional changes are induced by the generation and migration of the oxygen vacancies both in the CF and the OEL. Finally, for fully integrated devices, the OEL resides between the top-electrode and the oxide interface and is often buried beneath several conductive and insulating layers, thus complicating further the localization of the filament. Here we demonstrate that Scalpel SPM can overcome such problems, allowing us to probe multiple functions of the OEL

Figure 1. Structure of the VCM memory element and kinetic barriers at the metal/OEL/oxide interface. (a) Schematic of the conventional metal−insulator−metal VCM cell with the addition of the oxygen exchange layer. Two cases are reported here referred to HfO2 and Ta2O5. (b) Schematic energy landscape (not to scale) for the kinetic barriers of the metal/OEL/oxide interface based on Ru/Hf/HfO2 (inset TEM image scale bar =5 nm).

move through the energy potential wells with hopping distance a and energy barrier Ea, eventually assisted by an applied electric field E that lowers the energy barriers. By introducing the OEL, a thin buffer layer at the OEL/oxide interface is homogeneously induced in which the kinetic barrier for the oxygen ions, and metal cations motion is reduced.8,9 Such a layer with a shortrange structure and stoichiometry differing from this of main metal oxide is naturally formed at the Me/MeO interfaces and was suggested to play an essential role in the resistive switching by allowing easier transport of anions, cations, and electrons.10,11 The intermediate layer directly influences the kinetic barriers for the defect mobility in the VCM cell, thus defining among others the switching speed and dissipated power.5,6,10,12 Figure 1b shows a schematic example for the kinetic barriers at the metal/ OEL/oxide interface for the case of Ru/Hf/HfO2. Concurrently with an improved cell’s design13 and advanced processing,14 the presence of the OEL in the VCM cell results in positive effects (1) the filament formation, (2) the cell-to-cell variability, and (3) the endurance and CF’s stability.15,16 Indeed, when a CF is

Figure 2. Direct observation of the OEL/oxide interface by scalpel SPM. (a) 3 × 3 μm2 VCM cell is shown after being studied by scalpel SPM. A cell area ca. 2 × 2 μm2 is exposed. Note, scalpel SPM is a destructive technique, owing to the pressure (∼few GPa) needed for the tip-induced material removal of metals and dielectrics, a cantilever spring constant >10 N/m is generally required. (b) Evolution of the 2D conductive profiles as a function of depth in the last 10 nm prior to the bottom-electrode exposure. (c) Schematic representation of the role of the tip−sample system while approaching (still in the Ta) the interface of interest. (d) 2D current profile in the location of the CF for a Ta/Ta2O5 device showing the strong reduction of current as observed in proximity of the OEL/CF interface. 10821

DOI: 10.1021/acsami.6b16268 ACS Appl. Mater. Interfaces 2017, 9, 10820−10824

Research Article

ACS Applied Materials & Interfaces

Figure 3. Scalpel SPM 3D reconstruction of the OEL/oxide interface. (a) The material in correspondence of the CF location is interpolated in 3D (5 nm before the oxide layer is reached). Sections at different depths (the distance from the bottom electrode is indicated on each images) are shown in a1−4, note that the 3D linear interpolation between each slice is suppressed to highlight the presence of the internal conductive variations of the OEL while approaching the oxide interface. (b) Similar observations are repeated on Hf/HfO2 and Ta/Ta2O5 for multiple cells and the spots dimension compared at the top of the CF. First-principles thermodynamics and defect kinetics are used in Figure S3 to elucidate these findings. In the inset, the atomic models of am-HfO2 (Hf- blue, O-red) and am-Ta2O5 that are used to compute O vacancies formation energy in Figure S3.

Ta2O5-based cell, made of 5 nm Ta2O5 and 15 nm Ta sandwiched between two Pt electrodes (30 nm thick). Compared to other methods, Scalpel SPM provides a relatively simple way for the CF localization.22 Indeed after the electrical programming of the cell, the top electrode is removed with the AFM tip, to progressively approach the CF region. A bias between tip and sample is applied during the entire procedure with the tip grounded. As the measurement is done in air, we bias our tip with 500 mV during the entire procedure, for the rest all the standard conditions of scalpel SPM are applied.22−24 Different 2D current maps can be observed measured during this phase. Their evolution as a function of the cell depth is shown in Figure 2b for the Pt/Ta/Ta2O5/Pt cell. The Pt top-electrode shields the observation of the layer underneath thus making the scanned area appear as a fully conductive area. Through the removal, when the Ta layer is reached the location of the CF becomes already visible as a localized conductive spot with lateral dimensions of still several few tens of nm (Figure 2b). Note, this happens while the tip is still several nanometers away from the OEL/oxide interface. This is the combined result of the partial oxidation of the OEL metal during its progressive removal, and the formation of resistive network which is established between the tip and the bottom electrode and is function of the relative distance between tip and CF (Figure 2c). While approaching the oxide interface the most relevant current profiles related to the OEL layer are acquired. Indeed, by continuing the material removal, a remarkable drop in the conductivity of the CF is visible when in proximity of the OEL/ oxide interface. This observation coincides well and extends the results reported on the OEL layer properties in Wedig et al.10 providing a direct experimental argument. From this point, the 2D profiles of the filament containing shape and size of the constriction can be collected. The oxide thickness is further reduced until becomes too small indicated by the appearance of direct tunnel leakage currents in proximity of the (Pt) bottom electrode (Figure 2b). It is worth noting that the reduction in current flowing through the CF when the OEL/oxide interface is reached is similar for both Hf/HfO2 and Ta/Ta2O5 (Figure S2). Figure 2b shows the 2D profile of the current flowing in the Ta/Ta2O5 stack as the material is progressively removed. Observing the evolution of

which have not yet been measured directly before. In this study, we investigate the two popular VCMs material systems: Hf/ HfO2 and Ta/Ta2O5. Using the electronic current probed by an AFM tip as information carrier used to generate an image, scalpel SPM is able to locate and characterize in three-dimensions the volume containing the oxygen exchange layer and the CF with nanometer lateral resolution.21 The experimental observations are combined to first-principles thermodynamics simulations and defect kinetics to elucidate the role of the OEL in device optimization. Our measurements suggest that a thermodynamic barrier is formed directly on top of the CF, limiting the interaction with undesired sources of oxygen ions (e.g., etch damages), thereby improving the filament stability and cell’s performance. We demonstrate that the local reservoir of defects extends mostly in the OEL implying that an optimal thickness exists for this layer in functional devices. Finally, comparing Hf/ HfO2 and Ta/Ta2O5 we can interpret the superior electrical behavior of Ta2O5-based VCMs in the light of the processes happening at the OEL/oxide interface such as an improved CF stability and wider defects reservoir formed at this interface. The generated fundamental understanding of the role of the scavenger layer allows the interpretation of the electrical results shown for the two material systems and offers fundamental insights for device engineering thus enabling a further optimization of the VCM cells.



RESULTS AND DISCUSSION We make use of cross-point cells (3 × 3 μm2), based on two material systems Hf/HfO2 and Ta/Ta2O5. Our devices have comparable thickness for the oxide layers (5 nm), and were programmed under similar electrical conditions. The electrical programming (not shown) is carried out by an initial electroforming process leading to a bipolar resistive-switching behavior with two stable states, i.e., a high-resistance (HRS) and a low-resistance (LRS) one. The change in resistance can be triggered by using a voltage variation (±1 V for VCM) whereby we limit the current through the device at 100 μA to avoid the breakdown of the oxide and to limit the degradation (Figure S1 of the Supporting Information). After the cells are programmed, we use scalpel SPM for the direct three-dimensional (3D) observation of the CF region. Figure 2a shows the case of a Ta/ 10822

DOI: 10.1021/acsami.6b16268 ACS Appl. Mater. Interfaces 2017, 9, 10820−10824

Research Article

ACS Applied Materials & Interfaces

for the vacancies formation and the broad range of values for the diffusion barrier of oxygen in pure metals, can account for major differences at the OEL/oxide interface.11,25 For example, the reduced ΔG Ta2O5 (vacancies formation in the Ta/Ta2O5) material system accounts for more defects in the reservoir available to the CF resulting in a wider lateral extent of the scavenged area. For a fully integrated cell, this results in a higher oxygen chemical potential and optimized kinetic barrier for the O ion diffusion (Figure S3a,b). Particularly for the retention of the VCM cell, the latter reflects in longer CF lifetime and stability. In essence, the CF formation process is triggered by the high electric field inside the oxide during the electroforming (∼109 V· m−1), with oxygen ions that leave the initial oxide matrix and drift toward the anode leaving behind an oxygen vacancy. The newly formed conductive path bridges a reservoir of oxygen ions facilitated by the OEL, whose properties are mostly depending on the thermodynamics of the metal−oxide interface. Our results clearly indicate that the lateral dimensions of the reservoir is 2 to 5 times larger than the CF depending on the material combination. In addition, as shown by Scalpel SPM, the OEL acts a thermodynamic barrier shielding the CF from the interaction with undesired oxygen atoms not belonging to the CF’s reservoir. These interactions are detrimental as they can induce undesired fluctuation in the filament resistance during the cell lifetime. The uncontrolled interaction of the CF with the environmental oxygen induced by the tip-based material removal consistently results in the CFs’ passivation. The latter appears as the spontaneous reoxidation of the CF when the tip reaches ca. 2 nm from the oxide. This is due to the nature of the LRS state in VCM as an unstable state of the system that has a thermodynamic driving force to be stabilized toward a fully oxidized CF. In summary, our observations show that the improved retention and endurance of Ta2O5-based VCM cells is a result of the wide lateral growth of the thermodynamic barrier protecting the CF at the OEL/oxide interface. Depending on the ΔG for oxide formation, the material will tend to oxidize and the number of VO•• will have a tendency to decrease, thus degrading the retention properties of the cell. A less negative ΔG as in the case of Ta/Ta2O5 can furthermore contribute to broaden the scavenged area hence improving the filament stability, i.e., retention. On each subsequent memory cycle, a number of VO•• will be added/removed at the constriction of the CF inducing the resistance change, i.e., flipping the bit-state. Therefore, a tailored VCM cell requires an isolated CF solely interacting with its oxygen reservoir. By monitoring the drop in conductivity of the CF as a function of the progressive size reduction of the thermodynamic barrier, we estimate that a minimum thickness of OEL/oxide interface to guarantee the proper switching operations is ca. 3 nm.

current in the last 7 nm before reaching the bottom-electrode (Figure 2d), a three-order of magnitude current drop is visible between 4.5 and 3 nm. We attribute this effect to two main reasons: (1) the progressive thinning of the Ta acting as a physical barrier for the oxygen inward diffusion inside the CF and (2) the metastable state of a generic VCM system with tendency to relax to the HRS.11 Therefore, the high reactivity of the Ta with the environmental oxygen leads to a partial reoxidation (reducing the VO•• concentration) of the top part of the filament such that in the top-layer the conductivity is strongly reduced. We estimate the physical extend of the OEL acting as a barrier on top of the filament as ∼2 nm. This value is similar for Hf/HfO2 and Ta/Ta2O5. These results suggest that the same physical principles are governing the two material systems concerning the OEL\oxide interface. However, we noticed a stronger decay of the conductivity in case of the CFs which are induced in HfO2. In essence, when the last 2 nm of OEL are removed the CF inside the HfO2 recovers to a relatively higher resistance values compared to Ta2O5 even if the two cells showed comparable lowresistive-states. The latter is ascribed to the lower (more negative) Gibbs energy of formation for HfO2 compared to the case of Ta2O5 reported also in Figure S3a.7,25 Furthermore, for both material systems, our observation demonstrates that few substoichiometric nanometers (ca. 2 nm) play the role of oxygen exchange layer limiting the inward diffusion of oxygen. Focusing on the electrical properties of the barrier formed directly on top of the CF, we collect a set of 2D conductive profiles from the 5 nm of material before the oxide interface. Once aligned and interpolated, a 3D tomogram of this area is obtained (Figure 3). Note, this is not the observation of the CF but rather the volume of material immediately on top of it. Figure 3a shows the interpolated volume (iso-surface at 100 nA, as a blue region) while Figure 3a1−4 presents the tomogram before the interpolation to highlight the internal structure of the spot with its local variations in 3D. Although the size of the conductive spot ca. 60 nm at its wide side, the 3D cross sections (Figure 3a1−4) show a fine internal structure of the spot with local conductance variations in the range of tens of nanometers. In essence, the observed variations in the localized conductivity inside the OEL in proximity of the CF are the first indication of the smaller dimension of the CF once in the oxide. In addition, this is consistent with first-principle simulations that reveal how the presence of diffused oxygen at the OEL/oxide interface induces variations in the density of states available near the Fermi level in the oxide.5,26 The latter is reflected in the evolution of the conductive paths in the CF region within the last 5 nm before the Ta/Ta2O5 interface (Figure 3a1−4). As introduced in Figure 2d, together with changes in conductivity, we also observe the shrinkage of the spot sizes when approaching the OEL/oxide interface from 950 to 190 nm2 spot area. One can also consider the inner core part of the spot as the only relevant part for its reduced resistance, in this case the values are reduced, and the shrinkage is from 570 to 190 nm2. Similar observation relative to the case of the Hf/HfO2 are available elsewhere.20 Finally, by comparing the size of multiple CF spots measured at few nanometers from the OEL/oxide interface it is possible to interpret the improved electrical performance of Ta/Ta2O5 compared to Hf/HfO2.6,15 Particularly, in light of the different lateral dimensions for the thermodynamic barrier which is formed on top of the CF at the OEL/oxide interface. Figure 3b shows that in Ta/Ta2O5, the lateral extent of the scavenged area is wider compared to Hf/ HfO2. Indeed, differences in the ΔG (Gibbs formation energy)



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsami.6b16268. Electrical programming of the devices presented. We report another direct observation of the barrier for HfO2, removed by means of SPM tomography with a comparison between the two materials. Finally the ab initio calculations for the formation energy in HfO2 and Ta2O5 are presented, together with the statistics for the kinetic barriers (i.e., Nudged Elastic Band) (PDF) 10823

DOI: 10.1021/acsami.6b16268 ACS Appl. Mater. Interfaces 2017, 9, 10820−10824

Research Article

ACS Applied Materials & Interfaces



(12) Chen, C. Y.; Goux, L.; Fantini, A.; Clima, S.; Degraeve, R.; Redolfi, A.; Chen, Y. Y.; Groeseneken, G.; Jurczak, M. Endurance Degradation Mechanisms in TiN\Ta2O5\Ta Resistive Random-Access Memory Cells. Appl. Phys. Lett. 2015, 106 (5), 53501. (13) Hayakawa, Y.; Himeno, A.; Yasuhara, R.; Boullart, W.; Vecchio, E.; Vandeweyer, T.; Witters, T.; Crotti, D.; Jurczak, M.; Fujii, S.; Ito, S.; Kawashima, Y.; Ikeda, Y.; Kawahara, A.; Kawai, K.; Wei, Z.; Muraoka, S.; Shimakawa, K.; Mikawa, T.; Yoneda, S. Highly Reliable TaOx ReRAM with Centralized Filament for 28-nm Embedded Application. In Symposium on VLSI Technology (VLSIT); IEEE, 2015; Vol. 724, pp T14−T15. (14) Park, T. H.; Song, S. J.; Kim, H. J.; Kim, S. G.; Chung, S.; Kim, B. Y.; Lee, K. J.; Kim, K. M.; Choi, B. J.; Hwang, C. S. Thickness Effect of Ultra-Thin Ta2O5 Resistance Switching Layer in 28 Nm-Diameter Memory Cell. Sci. Rep. 2015, 5, 15965. (15) Lee, M.-J.; Lee, C. B.; Lee, D.; Lee, S. R.; Chang, M.; Hur, J. H.; Kim, Y.-B.; Kim, C.-J.; Seo, D. H.; Seo, S.; Chung, U.-I.; Yoo, I.-K.; Kim, K. A Fast, High-Endurance and Scalable Non-Volatile Memory Device Made from Asymmetric Ta2O(5-x)/TaO(2-X) Bilayer Structures. Nat. Mater. 2011, 10 (8), 625−630. (16) Govoreanu, B.; Kar, G. S.; Chen, Y.; Paraschiv, V.; Kubicek, S.; Fantini, A.; Radu, I. P.; Goux, L.; Clima, S.; Degraeve, R.; Jossart, N.; Richard, O.; Vandeweyer, T.; Seo, K.; Hendrickx, P.; Pourtois, G.; Bender, H.; Altimime, L.; Wouters, D. J.; Kittl, J. A.; Jurczak, M. 10 × 10 nm2 Hf/HfO X Crossbar Resistive RAM with Excellent Performance, Reliability and Low-Energy Operation. In IEDM Tech. Dig.; Washington, DC, 2011; p 31.6.1-31.6.4. (17) Dittmann, R.; Muenstermann, R.; Krug, I.; Park, D.; Menke, T.; Mayer, J.; Besmehn, A.; Kronast, F.; Schneider, C. M.; Waser, R. Scaling Potential of Local Redox Processes in Memristive SrTiO3 Thin-Film Devices. Proc. IEEE 2012, 100 (6), 1979−1990. (18) Miao, F.; Strachan, J. P.; Yang, J. J.; Zhang, M.-X.; Goldfarb, I.; Torrezan, A. C.; Eschbach, P.; Kelley, R. D.; Medeiros-Ribeiro, G.; Williams, R. S. Anatomy of a Nanoscale Conduction Channel Reveals the Mechanism of a High-Performance Memristor. Adv. Mater. 2011, 23 (47), 5633−5640. (19) Moors, M.; Adepalli, K. K.; Lu, Q.; Wedig, A.; Bäumer, C.; Skaja, K.; Arndt, B.; Tuller, H. L.; Dittmann, R.; Waser, R.; Yildiz, B.; Valov, I. Resistive Switching Mechanisms on TaO X and SrRuO 3 Thin-Film Surfaces Probed by Scanning Tunneling Microscopy. ACS Nano 2016, 10 (1), 1481−1492. (20) Celano, U.; Goux, L.; Degraeve, R.; Fantini, A.; Richard, O.; Bender, H.; Jurczak, M.; Vandervorst, W. Imaging the ThreeDimensional Conductive Channel in Filamentary-Based Oxide Resistive Switching Memory. Nano Lett. 2015, 15 (12), 7970−7975. (21) Celano, U.; Goux, L.; Belmonte, A.; Schulze, A.; Opsomer, K.; Detavernier, C.; Richard, O.; Bender, H.; Jurczak, M.; Vandervorst, W. Conductive-AFM Tomography for 3D Filament Observation in Resistive Switching Devices. IEDM Technol. Dig. 2013, 21.6.1−21.6.4. (22) Celano, U. Metrology and Physical Mechanisms in New Generation Ionic Devices; Springer Theses; Springer International Publishing: Cham (ZG), Switzerland, 2016. (23) Celano, U.; Goux, L.; Opsomer, K.; Iapichino, M.; Belmonte, A.; Franquet, A.; Hoflijk, I.; Detavernier, C.; Jurczak, M.; Vandervorst, W. Scanning Probe Microscopy as a Scalpel to Probe Filament Formation in Conductive Bridging Memory Devices. Microelectron. Eng. 2014, 120, 67−70. (24) Celano, U.; Hantschel, T.; Giammaria, G.; Chintala, R. C.; Conard, T.; Bender, H.; Vandervorst, W. Evaluation of the Electrical Contact Area in Contact-Mode Scanning Probe Microscopy. J. Appl. Phys. 2015, 117 (21), 214305. (25) Clima, S.; Sankaran, K.; Chen, Y. Y.; Fantini, A.; Celano, U.; Belmonte, A.; Zhang, L.; Goux, L.; Govoreanu, B.; Degraeve, R.; Wouters, D. J.; Jurczak, M.; Vandervorst, W.; De Gendt, S.; Pourtois, G. RRAMs Based on Anionic and Cationic Switching: A Short Overview. Phys. Status Solidi RRL 2014, 8, 501−511. (26) Guo, Y.; Robertson, J. Oxygen Vacancy Defects in Ta2O5 Showing Long-Range Atomic Re-Arrangements. Appl. Phys. Lett. 2014, 104 (11), 112906.

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected] (U.C.). *E-mail: [email protected] (I.V.). ORCID

Umberto Celano: 0000-0002-2856-3847 Ilia Valov: 0000-0002-0728-7214 Author Contributions

U.C., I.V., L.G., and W.V. designed this work and U.C. prepared the manuscript. The experiments were carried out by U.C., J.O.D.B., S.C., and M.L, while P.M.K., I.V., and W.V. have contributed in results analysis and discussion of the manuscript during all phases of the preparation. All authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We acknowledge the partial funding by IMEC’s Industrial Affiliation program on RRAM.



REFERENCES

(1) Wong, H.-S. P.; Salahuddin, S. Memory Leads the Way to Better Computing. Nat. Nanotechnol. 2015, 10 (3), 191−194. (2) Waser, R.; Aono, M. Nanoionics-Based Resistive Switching Memories. Nat. Mater. 2007, 6 (11), 833−840. (3) Wong, H.-S. P.; Lee, H.-Y.; Yu, S.; Chen, Y.-S.; Wu, Y.; Chen, P.-S.; Lee, B.; Chen, F. T.; Tsai, M.-J. Metal−Oxide RRAM. Proc. IEEE 2012, 100 (6), 1951−1970. (4) Zhong, X.; Rungger, I.; Zapol, P.; Nakamura, H.; Asai, Y.; Heinonen, O. The Effect of Ta Oxygen Scavenger Layer on HfO2-Based Resistive Switching Behavior: Thermodynamic Stability, Electronic Structure, and Low-Bias Transport. Phys. Chem. Chem. Phys. 2016, 18, 7502−7510. (5) Kim, W.; Menzel, S.; Wouters, D. J.; Guo, Y.; Robertson, J.; Roesgen, B.; Waser, R.; Rana, V. Impact of Oxygen Exchange Reaction at the Ohmic Interface in Ta2O5-Based ReRAM Devices. Nanoscale 2016, 8 (41), 17774−17781. (6) Goux, L.; Fantini, A.; Redolfi, A.; Chen, C. Y.; Shi, F. F.; Degraeve, R.; Chen, Y. Y.; Witters, T.; Groeseneken, G.; Jurczak, M. Role of the Ta Scavenger Electrode in the Excellent Switching Control and Reliability of a Scalable Low-Current Operated TiN\Ta2O5\Ta RRAM Device. In 2014 Symposium on VLSI Technology (VLSI-Technology): Digest of Technical Papers; IEEE, 2014; pp 1−2. (7) Zhuo, V. Y.-Q.; Li, M.; Guo, Y.; Wang, W.; Yang, Y.; Jiang, Y.; Robertson, J. CMOS Compatible Electrode Materials Selection in Oxide-Based Memory Devices. J. Appl. Phys. 2016, 120 (2), 24504. (8) Park, G.-S.; Kim, Y. B.; Park, S. Y.; Li, X. S.; Heo, S.; Lee, M.-J.; Chang, M.; Kwon, J. H.; Kim, M.; Chung, U.-I.; Dittmann, R.; Waser, R.; Kim, K. In Situ Observation of Filamentary Conducting Channels in an Asymmetric TaO5-x/TaO2-X Bilayer Structure. Nat. Commun. 2013, 4, 2382. (9) Hayakawa, Y.; Himeno, A.; Yasuhara, R.; Boullart, W.; Vecchio, E.; Vandeweyer, T.; Witters, T.; Crotti, D.; Jurczak, M.; Fujii, S. Highly Reliable TaO X ReRAM with Centralized Filament for 28-Nm Embedded Application T14 T15 2015, 724 (2011), 14−15. (10) Wedig, A.; Luebben, M.; Cho, D.-Y.; Moors, M.; Skaja, K.; Rana, V.; Hasegawa, T.; Adepalli, K. K.; Yildiz, B.; Waser, R.; Valov, I. Nanoscale Cation Motion in TaOx, HfOx and TiOx Memristive Systems. Nat. Nanotechnol. 2015, 11 (1), 67−74. (11) Clima, S.; Chen, Y. Y.; Chen, C. Y.; Goux, L.; Govoreanu, B.; Degraeve, R.; Fantini, A.; Jurczak, M.; Pourtois, G. First-Principles Thermodynamics and Defect Kinetics Guidelines for Engineering a Tailored RRAM Device. J. Appl. Phys. 2016, 119 (22), 225107. 10824

DOI: 10.1021/acsami.6b16268 ACS Appl. Mater. Interfaces 2017, 9, 10820−10824