Abstract:3D monolithic integration of logic and memory devices has emerged as one of the key enablers for next-generation electronic devices to address the ever-increasing demand for higher integration density, better performance and energy efficiency. BEOL-compatible oxide semiconductors show great potential to revolutionize the field thanks to their unique properties [1]. We present our recent advancement related to BEOL-compatible oxide semiconductors for logic and memory applications. The digital-etch-enabled nanowire transistors and α -IGZO-based eDRAM will first be discussed, followed by the ferroelectric (FE) memories with high performance and novel structures utilizing α -IGZO or ALD-deposited ZnO as the channel. Ultra-scaled amorphous IGZO ( α -IGZO) nanowire field-effect transistors (NW-FETs) hold great potential for applications demanding high performance and integration density. To realize the fabrication of high-quality and aggressively-scaled nanowire structures, a novel digital etching method for amorphous α -IGZO materials has been proposed and demonstrated [2]. Confirmed by the SEM images of an α -IGZO nanowire before and after digital etching, a notable nanowire width W NW reduction can be observed. We have the thinnest α -IGZO nanowire achieved by the digital etching with a W NW of approximately 20 nm. By further developing the transistor upon the nanowire, the α -IGZO NW-FET attains a good subthreshold swing (SS) of 80 mV/decade and a high peak extrinsic transconductance ( G m, ext ) of 612 μS/μm at V DS of 2 V (456 μS/μm at V DS = 1 V). Compared to the previous studies, our IGZO NW-FET achieves one of the highest peak G m values among all IGZO-based FETs. Besides the BEOL compatibility and high on-current, ultra-low subthreshold leakage makes α -IGZO FETs extremely promising for eDRAM. We further investigated the potential of α -IGZO eDRAM by developing the charge-domain compute-in-memory (CiM) [3]. Experiments have demonstrated small SS, large on-state current, and long-time charge retention with our dedicated 4T1C memory cell. With differential cell structure, an even higher tolerance for charge loss can be attained. With experiment-calibrated benchmarking in the VGG-8 network for CIFAR-10 image classification tasks, 2092 TOPS/W power efficiency for the CiM core can be expected, outperforming the prior TFT and CMOS-based CiM approaches, exhibiting significant advantages for ultra-low-power applications. The integration of doped-HfO 2 FE material and oxide semiconductor, both BEOL-compatible with large-scale and cost-effective deposition, presents a promising avenue for advancing data storage in the future. We have developed the high-performance α -IGZO Fe-FET with a metal-ferroelectric-metal-oxide-semiconductor (MFMIS) structure as well as the Fe TCAM [4]. The α -IGZO Fe-FET achieves a large memory window of ~3 V with high reliability, while the Fe TCAM reduces the transistor number from 16 to 2 compared to the traditional SRAM-based one. Besides α -IGZO, we believe that ALD-based oxide semiconductors featuring high controllability of film thickness and conformal coverage of the 3D structures can create new opportunities for novel device structure and integration. The fin-gate ZnO Fe-FET stands for a great example while not only outstanding device performance but also suppressed device-to-device variation due to the unique structure have been demonstrated, holding tremendous promise for high-density 3D integration [5]. Acknowledgments: This work is supported by Singapore Ministry of Education (Tier 2: MOE2018-T2-2-154, Tier 1: R-263-000-D65-114). References : [1] S. Dutta et al. , IEDM, p. 36.4., 2020. [2] K. Han et al. , VLSI , 2021, p. T10-1. [3] J. Liu et al. , IEDM , 2021, p. 46.2. [4] C. Sun et al. , VLSI , 2021, p. T7-4. [5] Q. Kong et al. , IEDM , 2022, p. 12.3.

Low-Power and Scalable BEOL-Compatible IGZO TFT eDRAM-Based Charge-Domain Computing

Low-Power and Scalable BEOL-Compatible IGZO TFT eDRAM-Based Charge-Domain Computing

Low-Power and Scalable Retention-Enhanced IGZO TFT Edram-Based Charge-Domain Computing

High-density Energy-Efficient Charge-Domain Computing Based on CAA-IGZO TFT with BEOL-compatible 3D Integration

An 8-Bit in Resistive Memory Computing Core with Regulated Passive Neuron and Bitline Weight Mapping

Computing-in-memory Circuits and Cross-Layer Integrated Design and Optimization: from SRAM to FeFET

Counteractive Coupling IGZO/CNT Hybrid 2T0C DRAM Accelerating RRAM-based Computing-In-Memory Via Monolithic 3D Integration for Edge AI

Monolithic 3-D Integration of Counteractive Coupling IGZO/CNT Hybrid 2T0C DRAM and Analog RRAM-Based Computing-In-Memory

30 Mb/mm<sup>2</sup>/layer 3D eDRAM Computing-in-Memory with Embedded BEOL Peripherals and Local Layer-wise Calibration based on First-Demonstrated Vertically-stacked CAA-IGZO 4F<sup>2</sup> 2T0C Cell

A Hybrid Computing-In-Memory Architecture by Monolithic 3D Integration of BEOL CNT/IGZO-based CFET Logic and Analog RRAM

Almost-Nonvolatile IGZO-TFT-Based Near-Sensor In-Memory Computing.

Low Power and Temperature-Resilient Compute-In-Memory Based on Subthreshold-FeFET

Area-and Power-Efficient Compute-in-Memory Macro Based on Oxide Semiconductor 2T0C Edram with Cell-System Co-Optimization

TFT-Based Near-Sensor In-Memory Computing: Circuits and Architecture Perspectives of Large-Area Edram and ROM CiM Chips

Computing-in-memory with thin-filmtransistors: challenges and opportunities

(Invited) BEOL-Compatible Oxide Semiconductor Logic and Memory Devices

Capacitor-less, Long-Retention (>400s) DRAM Cell Paving the Way towards Low-Power and High-Density Monolithic 3D DRAM

Cryogenic quasi-static embedded DRAM for energy-efficient compute-in-memory applications

Computing-in-Memory with Thin-Film-Transistors: Challenges and Opportunities

A Dual Core Oxide 8T SRAM Cell with Low Vccmin and Dual Voltage Supplies in 45nm Triple Gate Oxide and Multi Vt CMOS for Very High Performance Yet Low Leakage Mobile SoC Applications

A 28nm 8928Kb/mm 2 -Weight-Density Hybrid SRAM/ROM Compute-in-Memory Architecture Reducing >95% Weight Loading from DRAM.