CoChat
Sign Up
Cited by 127Advanced Memory and Neural ComputingNanowire Synthesis and ApplicationsAdvancements in Semiconductor Devices and Circuit Design

The DOI matches its Crossref record; 69 of the 70 references it cites were found in a registry or an index.

Nanowire FET Based Neural Element for Robotic Tactile Sensing Skin

Open access
Taube Navaraj, William, García Núñez, Carlos, Shakthivel, Dhayalan, Vinciguerra, Vincenzo, Labeau, Fabrice, Gregory, Duncan H., Dahiya, Ravinder·2017·Frontiers in Neuroscience
DOI 10.3389/fnins.2017.00501PMC PMC5611376
Open this paper in CoChat — free accountOpen PDF

Abstract

This paper presents novel Neural Nanowire Field Effect Transistors (υ­NWFETs) based hardware-implementable neural network (HNN) approach for tactile data processing in electronic skin (e-skin). The viability of Si nanowires (NWs) as the active material for υ­NWFETs in HNN is explored through modelling and demonstrated by fabricating the first device. Using υ­NWFETs to realize HNNs is an interesting approach as by printing NWs on large area flexible substrates it will be possible to develop a bendable tactile skin with distributed neural elements (for local data processing, as in biological skin) in the backplane. The modeling and simulation of υ­NWFET based devices show that the overlapping areas between individual gates and the floating gate determines the initial synaptic weights of the neural network - thus validating the working of υ­NWFETs as the building block for HNN. The simulation has been further extended to υ­NWFET based circuits and neuronal computation system and this has been validated by interfacing it with a transparent tactile skin prototype (comprising of 6x6 ITO based capacitive tactile sensors array) integrated on the palm of a 3D printed robotic hand. In this regard, a tactile data coding system is presented to detect touch gesture and the direction of touch. Following these simulation studies, a four-gated υ­NWFET is fabricated with Pt/Ti metal stack for gates, source and drain, Ni floating gate, and Al2O3 high-k dielectric layer. The current­voltage characteristics of fabricated υ­NWFET devices confirm the dependence of turn­off voltages on the (synaptic) weight of each gate. The presented υ­NWFET approach is promising for a neuro-robotic tactile sensory system with distributed computing as well as numerous futuristic applications such as prosthetics, and electroceuticals.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3389/fnins.2017.00501

Cites

69 of 69 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 The distribution of mechanoreceptors in various parts of human body. (Inset: 100% glabrous area of hand, which corresponds to 1.2% of the total skin area). The number of mechanoreceptors indicate the typical number of sensors required in various parts to achieve a full body touch/pressure sensing to mimic human body.
Figure 1 The distribution of mechanoreceptors in various parts of human body. (Inset: 100% glabrous area of hand, which corresponds to 1.2% of the total skin area). The number of mechanoreceptors indicate the typical number of sensors required in various parts to achieve a full body touch/pressure sensing to mimic human body.CC BY · PubMed Central
Figure 2 E-skin neural element and proposed tactile data processing scheme: (A) illustration of a biological neuron, (B) block diagram of an artificial neuron corresponding to the implemented weights. Complementary υ-NWFETs followed by an inverter are needed to realize this function, (C) symbolic representation of υ-NWFET, and (D) illustration of υ-NWFET based tactile e-skin covering an artificial hand. The simulated tactile signal is shown as stimuli to the input layer of υ-NWFET based network for coding of tactile information.
Figure 2 E-skin neural element and proposed tactile data processing scheme: (A) illustration of a biological neuron, (B) block diagram of an artificial neuron corresponding to the implemented weights. Complementary υ-NWFETs followed by an inverter are needed to realize this function, (C) symbolic representation of υ-NWFET, and (D) illustration of υ-NWFET based tactile e-skin covering an artificial hand. The simulated tactile signal is shown as stimuli to the input layer of υ-NWFET based network for coding of tactile information.CC BY · PubMed Central
Figure 3 The work flow of the research presented in this paper. (A) Three levels of modeling. (B) e-Skin. (C) Device fabrication.
Figure 3 The work flow of the research presented in this paper. (A) Three levels of modeling. (B) e-Skin. (C) Device fabrication.CC BY · PubMed Central
Figure 4 Modeling of υ-NWFET device. (A) 3D Structure of a simulated floating gate υ-NWFET. (a1–a3) Cross-sectional view of the υ-NWFET at drain/source region, floating gate and gate region. (B) Equivalent circuit model of a υ-NWFET.
Figure 4 Modeling of υ-NWFET device. (A) 3D Structure of a simulated floating gate υ-NWFET. (a1–a3) Cross-sectional view of the υ-NWFET at drain/source region, floating gate and gate region. (B) Equivalent circuit model of a υ-NWFET.CC BY · PubMed Central
Figure 5 Schematic illustration of fabrication steps for υ-NWFET, using a SOI wafer as substrate (A), and comprising (B) PMMA spin-coating, (C) e-beam lithographic patterning of lines on PMMA, (D) deposition of NiCr on the pre-patterned substrate, (E) PMMA lift-off, (F) Si sloped etching resulting in a trapezoidal shape Si NW, (G) ALD deposition of Al2O3 on top of the NW, (H) doping of drain and source, (I) Deposition of floating gate, (J) ALD Al2O3 deposition, (K) Definition of Source/Drain windows, and (L) Lift-off based patterning of Drain, Source and Multi-gates.
Figure 5 Schematic illustration of fabrication steps for υ-NWFET, using a SOI wafer as substrate (A), and comprising (B) PMMA spin-coating, (C) e-beam lithographic patterning of lines on PMMA, (D) deposition of NiCr on the pre-patterned substrate, (E) PMMA lift-off, (F) Si sloped etching resulting in a trapezoidal shape Si NW, (G) ALD deposition of Al2O3 on top of the NW, (H) doping of drain and source, (I) Deposition of floating gate, (J) ALD Al2O3 deposition, (K) Definition of Source/Drain windows, and (L) Lift-off based patterning of Drain, Source and Multi-gates.CC BY · PubMed Central
Figure 6 (A) Transfer characteristics from simulation of υ-NWFET device (dotted) and circuit (continuous) based on them. (B) Drain current (ID) vs. D-S voltage (VDS) as the gates are turned on one by one.
Figure 6 (A) Transfer characteristics from simulation of υ-NWFET device (dotted) and circuit (continuous) based on them. (B) Drain current (ID) vs. D-S voltage (VDS) as the gates are turned on one by one.CC BY · PubMed Central
Figure 7 Equivalent circuit implementation of (A) equally and (B) non-equally weighted υ-NWFET based neurons and (C,D) their respective input-output characteristics.
Figure 7 Equivalent circuit implementation of (A) equally and (B) non-equally weighted υ-NWFET based neurons and (C,D) their respective input-output characteristics.CC BY · PubMed Central
Figure 8 (A) Schematic and (B) output of a floating gate programmable synapse simulation.
Figure 8 (A) Schematic and (B) output of a floating gate programmable synapse simulation.CC BY · PubMed Central
Figure 9 (A) Schematic of a neuron with sensory memory and adaptable toward STM and LTM. (B) Post synaptic current vs. time graph demonstrating sensory memory. (C) Post synaptic current vs. time graph demonstrating short term memory (STM).
Figure 9 (A) Schematic of a neuron with sensory memory and adaptable toward STM and LTM. (B) Post synaptic current vs. time graph demonstrating sensory memory. (C) Post synaptic current vs. time graph demonstrating short term memory (STM).CC BY · PubMed Central
Figure 10 (A) Schematic illustration of sparse coding of tactile data from e-skin neural information. (B) Goal of the sparse coding. (C) Typical implementation of one Level 1 sub neural network cell on SimBrain; the output of the decay sensory neuron is passed to the input of the sparse coder which gives the sparse coded output. (D) Mean squared error vs. epochs during training of the network.
Figure 10 (A) Schematic illustration of sparse coding of tactile data from e-skin neural information. (B) Goal of the sparse coding. (C) Typical implementation of one Level 1 sub neural network cell on SimBrain; the output of the decay sensory neuron is passed to the input of the sparse coder which gives the sparse coded output. (D) Mean squared error vs. epochs during training of the network.CC BY · PubMed Central
Figure 11 System model interface with e-skin on prosthetic hand (A) Flexible and transparent touch sensitive e-skin (B) Image of the prosthetic/robotic hand with e-skin (C) Snapshot of SimBrain model showing mechanoreceptors layer and associated sensory processing level 1 NN cells are connected to higher hierarchical level 2 NN block. Video of the demonstration in Supplementary Video 1.
Figure 11 System model interface with e-skin on prosthetic hand (A) Flexible and transparent touch sensitive e-skin (B) Image of the prosthetic/robotic hand with e-skin (C) Snapshot of SimBrain model showing mechanoreceptors layer and associated sensory processing level 1 NN cells are connected to higher hierarchical level 2 NN block. Video of the demonstration in Supplementary Video 1.CC BY · PubMed Central
Figure 12 (A) SEM image of NiCr nanoline in low magnification. This pattern is subsequently etched into trapezoidal NWs (Inset: High magnification). Three-dimensional AFM scan of SiNW taken (B) before atomic layer deposition (ALD) and (C) after ALD & thermal annealing processes; (D) Line profile corresponding to the AFM image (B) where a ~45° trapezoidal structure is observed. (E) Optical microscopy image of the fabricated υ-NWFET. (F) Symbolic representation and weights of the fabricated υ-NWFET.
Figure 12 (A) SEM image of NiCr nanoline in low magnification. This pattern is subsequently etched into trapezoidal NWs (Inset: High magnification). Three-dimensional AFM scan of SiNW taken (B) before atomic layer deposition (ALD) and (C) after ALD & thermal annealing processes; (D) Line profile corresponding to the AFM image (B) where a ~45° trapezoidal structure is observed. (E) Optical microscopy image of the fabricated υ-NWFET. (F) Symbolic representation and weights of the fabricated υ-NWFET.CC BY · PubMed Central
Figure 13 Electrical characteristics of Si υ-NWFET. (A) Ideal (blue) and experimental (red) C-Vg relationship. (B) VGS vs. IDS characteristics for G5/14 to G2/14 at VDS = 4 V (C) VDS vs. IDS characteristics while applying 6 V to each gate, one by one.
Figure 13 Electrical characteristics of Si υ-NWFET. (A) Ideal (blue) and experimental (red) C-Vg relationship. (B) VGS vs. IDS characteristics for G5/14 to G2/14 at VDS = 4 V (C) VDS vs. IDS characteristics while applying 6 V to each gate, one by one.CC BY · PubMed Central
Figure 14 (A) Schema to study impact on neural function of the NN due to the deviation in weights. (B) Effect of weight deviations on the NN's performance.
Figure 14 (A) Schema to study impact on neural function of the NN due to the deviation in weights. (B) Effect of weight deviations on the NN's performance.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Cited by

Sorted by citations

See all 127

Related

See all

Metadata

Authors
Taube Navaraj, W., García Núñez, C., Shakthivel, D., Vinciguerra, V., Labeau, F., Gregory, D. H., & Dahiya, R.
Affiliations
  • University of Glasgow (GB)
  • STMicroelectronics (Italy) (IT)
  • McGill University (CA)
Corresponding author
Dahiya, Ravinder
ORCID
Venue
Frontiers in Neuroscience
Publisher
Frontiers Media SA
Published
2017-09-20
Volume
11
Pages
501-501
ISSN
1662-453X (ISSN-L)
DOI
10.3389/fnins.2017.00501
PMC
PMC5611376
OpenAlex
W2755222108
Cited by
127
References
70
Citations by year
2026: 4 · 2025: 9 · 2024: 10 · 2023: 7 · 2022: 16 · 2021: 21 · 2020: 28 · 2019: 15 · 2018: 13 · 2017: 4
FWCI
6.82
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • Frontiers in Neuroscience (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Nottingham Trent University's Institutional Repository (Nottingham Trent Repository) (acceptedVersion)
  • ENLIGHTEN (Jurnal Bimbingan dan Konseling Islam) (acceptedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
  • eScholarship@McGill (McGill) (submittedVersion, cc-by)
Funders
Engineering and Physical Sciences Research Council (EP/M002527/1)
Topics
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Memory and Neural Computing
  • Physical Sciences › Materials Science › Electronic, Optical and Magnetic Materials › Nanowire Synthesis and Applications
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Advancements in Semiconductor Devices and Circuit Design
Keywords
nanowire field-effect transistors, silicon nanowires, electronic skin, tactile sensing, robotic hand, synaptic weights, floating gate, capacitive tactile sensor, neurorobotics, flexible electronics, high-k dielectrics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review