Thickness‐Programmable Opto‐Memristive Dynamics in 6‐In Wafer‐Scale Atomic‐Layer‐Controlled WS 2 for Reservoir Computing
ABSTRACT Neuromorphic optoelectronic hardware requires 2D channels whose thickness, and therefore band structure, is reproducible across an entire wafer, yet atomic‐layer control of transition metal dichalcogenides (TMDs) must reconcile precision, uniformity, and scalability simultaneously. Here, 6‐inch wafer‐scale atomic‐layer control of WS 2 is reported by integrating low‐temperature plasma‐assisted synthesis with a three‐component mixed‐gas (Ar+O 2 +CF 4 ) reactive ion etching process. WS 2 synthesized from electron‐beam‐evaporated W seed layers preserves structural integrity and achieves uniform atomic‐layer control across the wafer, whereas sputter‐derived WS 2 etches non‐uniformly with rougher surfaces and partial lattice disruption, as confirmed by ellipsometry, atomic force microscopy, Raman mapping, and cross‐sectional transmission electron microscopy. Density functional theory (DFT) calculations qualitatively support the experimentally observed optical band gap according to the number of layers and provide insight into the layer‐dependent electronic properties of WS 2 . The layer‐controlled WS 2 was subsequently employed to fabricate opto‐memristive devices, which exhibited pulse‐history‐dependent photocurrent modulation at zero read bias. The 4L‐WS 2 device balances optical sensitivity against inter‐pulse memory, resolving 16 four‐pulse optical histories into distinguishable analog states. A dual‐state recursive model calibrated from the measured device dynamics achieved a mean waveform‐classification accuracy of 95.8 ± 1.69%. Atomic‐layer control thus provides a manufacturable route from wafer‐scale TMD processing to band‐structure‐programmed neuromorphic optoelectronics.
Authors
- Byeong‐Seon An (ORCID: https://orcid.org/0000-0002-3994-555X)
- Hyeong‐U Kim (ORCID: https://orcid.org/0000-0002-0018-9603)
- Muyoung Kim (ORCID: https://orcid.org/0000-0002-1102-8194)
- Hyunwoo Shim
- Min Sup Choi (ORCID: https://orcid.org/0000-0002-8448-4043)
- Taehyun Park (ORCID: https://orcid.org/0000-0001-9119-9518)
- Seungwoo Song (ORCID: https://orcid.org/0000-0002-8524-3444)
- Hocheon Yoo (ORCID: https://orcid.org/0000-0003-0772-3327)
- Taesung Kim (ORCID: https://orcid.org/0000-0001-6280-7668)
- Ki Seok Kim (ORCID: https://orcid.org/0000-0002-7958-4058)
- Jin-Hyoung Lee (ORCID: https://orcid.org/0009-0003-0679-629X)
- Eunseo Go
- Junhyung Cho (ORCID: https://orcid.org/0009-0007-0043-3294)
- Seongho Kim
- Sooji Byun
- Yeonho Choi (ORCID: https://orcid.org/0009-0003-0936-904X)
- Minchul Roh (ORCID: https://orcid.org/0009-0004-6979-3480)
- Gunhyoung Kim
Institutions
- Chungnam National University (KR)
- Korea Research Institute of Standards and Science (KR)
- Korea Institute of Machinery & Materials (KR)
- Korea Institute of Energy Research (KR)
- Hanyang University (KR)
- Imperial College London (GB)
- Sungkyunkwan University (KR)
- Korea University of Science and Technology (KR)
Publication Details
- Journal
- Advanced Functional Materials
- Published
- 2026-10-29
- DOI
- https://doi.org/10.1002/adfm.79354
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00