Functionalization of transparent oxide resistive random-access memories for visible light photosensitivity

• Main Author: Kalaga, Pranav Sairam
• Other Authors: Ang Diing Shenp
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering::Microelectronics; Engineering::Electrical and electronic engineering::Optics, optoelectronics, photonics
• Online Access: https://hdl.handle.net/10356/160030
• Institution: Nanyang Technological University

Resistive Random-Access Memory (RRAM) by virtue of its simple metal-insulator-metal structure, high scalability, excellent device density and the potential to become a future universal memory format of an industry fast approaching a limit to Moore’s law scaling has seen tremendous research over the past decade. Its high-speed operation, excellent reliability, low silicon foot-print and a vast array of CMOS compatible material systems have provided a compelling argument in this regard. As research in the field of resistive switching devices grew, various attempts to further enhance the functionality of these devices have gained traction. These features allow for more closely integrated circuitry, lesser routing constraints and an overall smaller system footprint with vastly extended real-world applications. Several studies have made efforts to incorporate the various advantages conferred by RRAM devices, least-of-all not being its intrinsic memory feature. This is driven by the dual need of requiring closely integrated systems with multiple functionality enabled into it as well as the scaling limitations of conventional flash memory which is the non-volatile mainstay of modern electronics. Other features include, but are not limited to, epidermal electronics, transparent electronics, optoelectronics, green electronics, environmental robustness, flexible devices, stretchable devices, chemical sensing, neuromorphic computing, synaptic behaviour etc. Of these various functionalities, the field of invisible electronics integrated with RRAM devices has been of particular interest. While proposed as early as 1997, the lack of advanced transparent conductive materials and substrates has limited the proliferation of these circuits in real-world applications. With significant advances in transparent conductive oxides (TCOs), as well as using transparent substrates for processing including but not limited to glass and polymers, the field of invisible electronics has seen a leap forward over the last decade. These have primarily been confined to display technologies but there has been a push to integrate logic and memory circuitry into these circuits. Of these elements, the flash memory is a poor-platform owing to its reliance on crystalline silicon to achieve the required memory density and performance. RRAM devices show potential to surmount the scaling, memory density and material limitations that flash memory currently suffers from. Several devices have made use of conventional RRAM resistive switching materials in conjunction with TCOs to form fully transparent RRAM (T-RRAM) that can be deployed onto smart glasses and other display and industrial applications. Several of these devices are deployed as standalone transparent memory elements. Research into optical functionality integration on the invisible electronics T-RRAM platform however faces significant challenges owing to several reasons. The prime reason being conventional wide-bandgap oxides used in RRAM devices do not exhibit a response to illumination outside the UV spectrum. While these have been demonstrated in visible-blind, UV photodetecting RRAM devices, there is a dearth of demonstrations in the important visible-spectrum of light. Efforts have been made to address this area of critical importance through the incorporation of photoresponsive material layers, such as semiconductors, 2D materials, perovskites, quantum dots etc. to be able to demonstrate a photocurrent. Significant attention has been given particularly to semi-transparent 2D material systems in this regard and regardless of the material system incorporated, efforts have been made to minimize the impact on the overall transparency of the devices. However, this approach of utilizing photocarrier generation based sensing involves significant trade-offs. These materials naturally have a bandgap smaller than the photon energy of light in the visible regime and passively absorb energy regardless of the device being in use or not. This reduces the overall transparency of the total material stack. Furthermore, for efficient light absorption, the photoresponsive layers must be thick enough to generate enough photocarriers to be collected. The introduction of semiconductor layers such as Si into the device further exacerbates the transparency issue. Photo-reponsive semiconductors are conventionally opaque to visible light, while invisible electronics applications require >80% transparency to be viable. Another issue arising from the use of exotic materials is CMOS compatibility, if they wish to make use of ubiquitous manufacturing technology for the roll-out of invisible electronics applications as well as ease of integration on other non-transparent platforms, particularly the Si platform. This has led to investigation of photoresponsive devices that do not rely on the photocarrier generation effect, or positive photoconductivity (PPC). Efforts have been made in the field of negative photoconductivity (NPC) devices but most research is still limited by the use of exotic material systems. Single-filament studies on uncapped wide-bandgap oxide stacks have demonstrated the potential for using conventional resistive switching oxides for photosensing. These have however been constrained to ultra-high vacuum phenomenological studies without device demonstration. In this thesis, we report the demonstration of all-oxide, CMOS compatible, transparent RRAM devices that utilize the NPC phenomenon for visible light photosensing that effectively eliminates several, if not all the issues, that have limited visible-light photodetectors on the invisible electronics platform. The defect generation from the single filament studies for visible light sensing has been made use of in demonstrating ITO/HfO2/ITO devices on a silica substrate for realizing on-demand visible light sensing on a completely transparent platform. The sensing mechanism does not negatively impact the transparency and furthermore, the use of wide bandgap oxides to demonstrate this strengthens the case for several other conventional resistive switching oxide layers to demonstrate a similar effect. These devices present an optical response in the form of a sudden current quenching post illumination that is recorded as a resistive state change in the oxide, allowing for memory integrated photodetector devices. A dual use case has also been presented as a continuous-time sensor mitigating any trade-off compared to conventional photodetectors that inherently are continuous-time only. The devices have been investigated for endurance, high-temperature retention, reliability as well as optical switching cycles. An investigation into the response speed shows tremendous potential for scaling the devices into high-speed photodetection circuitry and a mechanism has been theorized for the behaviour. Based on this study and the proposed mechanism of defect generation, a pathway to enhance the performance of the device through defect engineering of the oxide layer has been demonstrated using ITO/HfO2/MgO/ITO devices. These devices present best-in-class performance in terms of ON/OFF ratio, endurance as well as high temperature retention as well as validating the use of conventional wide-bandgap oxides for visible light sensing. By virtue of the material systems used, these devices have been demonstrated to be able to incorporate multiple functionality into a single RRAM device, namely transparency, visible-light sensing as well as all-oxide CMOS compatible systems presenting a compelling case for integration across various platforms and as a potential high-density image sensor due to the RRAM device’s intrinsic scaling potential. An alternative crossbar architecture has also been proposed and a method to mitigate the line-resistance component arising from the ITO electrode lines has been presented to inspire further studies towards large scale integration of these components as image sensors on various platforms.