Linearity improvement of HfOx-based memristor with multilayer structure
Introduction
Traditional computing system has suffered memory wall problems due to the physical separation of storage and computing [1]. Under such circumstances, memristor defined as the fourth basic element of the circuit is proposed to break through the limit of Von Neumann-architecture [2]. The resistance of memristor can switch between two stable resistance states or among less than two resistance states under the action of a specific external electrical signal. This resistance switching characteristic makes memristor attract attention as a new type of non-volatile memory used in the field of information storage which called resistance access memory (RRAM). Moreover, RRAM employed as memristor is generally considered to have a promising application in emulating synapses due to its high compatibility with CMOS technology and low energy consumption [[3], [4], [5], [6]]. The traditional RRAM comprised of a sandwich structure can store data during the formation and rupture of conductive filaments (CF) under the drive of an electric field [[7], [8], [9], [10]]. This characteristic of the two resistance states swtiching can be regarded as digital switching. However, for neuromorphic computing, memristor can also act as synapse to work in high-density neural networks [11]. Synaptic device can be optimized much continuous tunable resistance states equivalent to analog switching, which is widely used in kinds of training to mimic human brain, such as short-term plasticity (STP) to long-term plasticity (LTP), spike-rate-dependent plasticity (SRDP), and spike-timing-dependent plasticity (STDP) [[12], [13], [14]]. The linearity of incremental switching in memristor describes the uniformity of conductance change upon the adjusting of identical electric pulses [15]. When the memristor is applied with the same pulse, the linear increase in conductance (weight) highlights better predictability of synaptic modulation. It is worth noting that the resistance can be continuously decreased or increased between multiple intermediate states without going back to the original state [16,17].
According to the review of Wu et al. [18], the growth of the CF increases the local electric field or temperature, which in turn promotes VO (O vacancy) generation and migration and CF growth. This process forms a single strong CF, which hinders gradual control of SET. To avoid abrupt SET in filamentary RRAM, thermally enhanced layer [19], conductive filament lateral extension [20], channel composition modulation [21], interface engineering [22] are four dominant methods that have a relatively good effect. Among these four methods, interface engineering is considered to acquire more gradually switching. Hence, titanium serves as an inserting layer to recompose the interface state of hafnium oxide is a feasible way to achieve gradual change characteristic. H. Y. Lee et al. earlier put forward a thin Reactive Ti Buffer Layer in robust HfO2 based RRAM, the reaction between the Ti electrode and the HfO2 layer has been demonstrated [23]. Sk. ZiaurRahaman et al. further proposed an effortless Ti thickness modulation-based methodology to control the oxygen vacancy concentrations in the HfOx layer,thus indicates the drift of charged oxygen vacancies are crucial for filament formation and rupture [24]. So it is a effective way to improve linearity by inserting titanium layer.
In this work, based on TiN/HfOx/Ta device, titanium is inserted as a buffer layer to form a non-stoichiometric hafnium oxide at the interface, forming an oxygen vacancy concentration gradient from Ti/HfOx interface to the position away from the interface. The analog switching displays through Ti insert layer. Furthermore, a novel type HfOx-based multilayer structure device have been demonstrated to enhance the linearity by generating a roughly oxygen vacancy concentration gradient along the filament growth direction. The electrical characteristic more in line with the memristor are obtained, and linearity during pulse modulation has been raised to 98.4%, amplitude and symmetry have also been improved. Moreover, there exists potential to apply such devices to synaptic mimic measurement and neuromorphic computing.
Section snippets
Experimental section
To fabricate the HfOx-based memristor, 100 nm TiN layer is deposited on the Si substrate by DC magnetron sputtering as the bottom electrode (BE) under a mixed atmosphere with a total flow of nitrogen and argon of 33 sccm. Then 10 nm HfOx film is deposited by RF magnetron sputtering under a high vacuum of 10−4 Pa and a mixed atmosphere with a total flow of argon and oxygen of 36 sccm at room temperature. Finally, 80 nm Ta film is deposited by RF magnetron sputtering as the top electrode (TE)
Results and discussion
Fig. 1(a) exhibits schematic of synaptic devices structure and (b) shows the top view of the device. The current-voltage (I–V) curves of TiN/HfOx/Ta device are shown in Fig. 2(a). After an electroforming process (Vforming = 6 V, 400 μA) as is shown in Fig. 2(d), the applied voltage sweeps: 0 V→5 V→0 V→-5 V→0 V, Icc = 1 mA, the device performs bipolar resistance switching. Three curves of different colors distinguish three cycles of the same device and the arrows represent the direction of the
Conclusions
In summary, capitalizing on Ti buffer layer to construct a multilayer device to optimize linearity with configuration more excellent oxygen vacancy concentration gradient was accomplished. In addition, electrical characteristic was optimized through the multilayer device structure synaptic characteristic, such as pulse-modulated conductance linearity was improved, and mechanism explanations were also explained in detail, confirming that linearity is closely related to the oxygen vacancy
Author statement
Yutong Jiang: Methodology, Writing-Original Draft. Kai Hu: Writing-review & editing. Yujian Zhang: Software. Ange Liang: Software. Fang Wang: Data Curation, Investigation. Sannian Song: Writing-review & editing. Zhitang Song: Writing-review & editing. Kailiang Zhang: Supervision, Funding acquisition, Writing-review & editing.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by the National Key Research and Development Program of China (Grant No.2017YFB0405600), Natural Science Foundation of Tianjin City (Grant Nos.18JCZDJC30500, 17JCYBJC16100 and 17JCQNJC00900) and National Natural Science Foundation of China (Grant Nos.61404091, 61274113, 61505144, 51502203 and 51502204) and Open Project of State Key Laboratory of Functional Materials for Information(SKL202007) and Science and Technology Planning Project of Tianjin City (20ZYQCGX00070).
References (39)
- et al.
A Pt/ITO/CeO2/Pt memristor with an analog, linear, symmetric, and long-term stable synaptic weight modulation
Apl. Mater.
(2019) - et al.
The resistive switching characteristics of Ni-doped HfOx film and its application as a synapse
J. Alloys Compd.
(2018) - et al.
Improved switching reliability achieved in HfOx based RRAM with mountain-like surface-graphited carbon layer
Appl. Surf. Sci.
(2018) - et al.
Analog-type resistive switching devices for neuromorphic computing
Phys. Status Solidi Rapid Res. Lett.
(2019) - et al.
Improved resistive switching and synaptic characteristics using Ar plasma irradiation on the Ti/HfO2 interface
J. Alloys Compd.
(2019) - et al.
Switching mechanism of double forming process phenomenon in ZrOx/HfOy bilayer resistive switching memory structure with large endurance
Appl. Phys. Lett.
(2014) - et al.
XPS and IPE analysis of HfO2 band alignment with high-mobility semiconductors
Mater. Sci. Semicond. Process.
(2008) - et al.
Flexible neuromorphic electronics for computing, soft robotics, and neuroprosthetics
Adv. Mater.
(2020) Resistance switching memories are memristors
Appl. Phys. Mater. Sci. Process
(2011)- et al.
Nano resistive memory (Re-ram) devices and their applications
Rev. Adv. Mater. Sci.
(2019)
Nonvolatile memory materials for neuromorphic intelligent machines
Adv. Mater.
Interfacial interactions and their impact on redox-based resistive switching memories (ReRAMs)
Semicond. Sci. Technol.
RRAM device models: a comparative analysis with experimental validation
IEEE Access
Methodology for the characterization and observation of filamentary spots in HfOx-based memristor devices
Microelectron. Eng.
Self-doping memristors with equivalently synaptic ion dynamics for neuromorphic computing
ACS Appl. Mater. Interfaces
Artificial synapses with short- and long-term memory for spiking neural networks based on renewable materials
ACS Nano
Analog synaptic behavior of a silicon nitride memristor
ACS Appl. Mater. Interfaces
Performance enhancement of TaOx resistive switching memory using graded oxygen content
Appl. Phys. Lett.
Engineering incremental resistive switching in TaOx based memristors for brain-inspired computing
Nanoscale
Cited by (19)
Graded oxide layer for high-performing nanosized synaptic emulator
2023, Applied Surface ScienceALD and PEALD deposition of HfO<inf>2</inf> and its effects on the nature of oxygen vacancies
2022, Materials Science and Engineering: BCitation Excerpt :An alternative route thus would require optimization of the HfO2 deposition technique to obtain a material of better quality, which, together with transistor structure improvement, have the capacity for further enhancement of the resulting device performance. In addition to transistors, oxide layers are of considerable interest for memristor applications, which require materials with conductivity controllable by applied electric field [5]. The pronounced variation of conductivity is achieved with charged oxygen vacancies that can move inside the material, forming paths for preferential carrier transport (high conductivity state) or rupturing them (low conductivity state) [6,7].
Electronic synaptic plasticity and analog switching characteristics in Pt/TiO<inf>x</inf>/AlO<inf>x</inf>/AlTaON/TaN multilayer RRAM for artificial synapses
2022, Applied Surface ScienceCitation Excerpt :In the Pt/Ti/Al2O3/TaN system, high-resolution TEM creates a TiOx buffer layer at the top interface and the AlTaON interlayer at the bottom interface, as illustrated in Fig. 2(a). Developing a TiOx buffer layer and the AlTaON bottom interface layer is predicted to improve RS properties significantly [26]. A close view of the oxide AlOx sandwich layer is shown in Fig. 1(b), which corresponds to a high-resolution TEM image.
Reliable resistive switching and synaptic plasticity in Ar<sup>+</sup>-irradiated single-crystalline LiNbO<inf>3</inf> memristor
2022, Applied Surface ScienceCitation Excerpt :The resistance in realistic memristors tends to change rapidly at the pulse-programming beginning but saturates at the end of the process [18-20]. It is noted that those imperfect device characteristics originate from their inherent lattice imperfections, hence they are difficult to be eliminated by simply device structural optimization [21,22]. Therefore, innovations in materials and devices are highly desired to yield more reliable resistive switching characteristics and further construct memristive hardware computing systems [23].