Self-assembling crystalline peptide microrod for neuromorphic function implementation

•Memristor exhibits repeatable switching behavior and flexible relaxation process•Ion transport confinement was observed via in situ electrostatic force microscopy•Memristor emulates common nociceptive behaviors neuropathic pain Resistive memory technologies have gained significant attention due to their potential for creating bio-inspired intelligent systems that memorize learn locally. However, a major bottleneck of memristors lies the reliability issue uncontrollable filament dynamic induced by inherent stochastic nature ion migration, hindering applications neuromorphic computing. Here we introduce convenient approach which, planar structured cell, solution-processable diphenylalanine microrod (good ionic permeability) is used as medium facilitate quasi-one-dimensional channel Ag diffusion. The volatile feature, highly behavior, process our peptide-based memristor render it promising candidate perform function implementation. For example, tuning external electrical stimuli, device can emulate damaged tissue. Memristive devices offer desirable voltage-regulated conductance promise address zettabyte storage challenges big-data era. Generally, most these reported use amorphous solids, where structural compositional inhomogeneity regarded origin variability. Self-assembling peptide crystals with solution-processed fabrication, controllable morphologies, stability are therefore issues. report diffusive possesses reliable characteristics based on crystallized material: (FF) (MR). This element offers preferential migration path, confining conductive filaments defined crystalline surface reducing programming stochasticity. Ion along FF MR microscopy technique. Additionally, high uniformity reproducible provides an ideal hardware platform nociceptor emulation identification four-bit decimal numbers. development artificial intelligence (AI) analytics driving revolution methods data processing storage. Confronting speed energy consumption issues, fields require new computing system concept parallel retrieve, process, store massive amounts data.1Sebastian A. Le Gallo M. Khaddam-Aljameh R. Eleftheriou E. Memory in-memory computing.Nat. Nanotechnol. 2020; 15: 529-544Crossref PubMed Scopus (301) Google Scholar, 2Ielmini D. Wong H.S.P. In-memory resistive devices.Nat. Electron. 2018; 1: 333-343Crossref (638) 3Fuller E.J. Keene S.T. Melianas Wang Z. Agarwal S. Li Y. Tuchman James C.D. Marinella M.J. Yang J.J. et al.Parallel floating-gate array scalable computing.Science. 2019; 364: 570Crossref (253) 4Xia Q. crossbar arrays brain-inspired Mater. 18: 309-323Crossref (497) 5Lv Chen J. Zhou Han S.-T. Semiconductor quantum dots memories systems.Chem. Rev. 120: 3941-4006Crossref (82) 6van de Burgt Malliaras G. Salleo Organic electronics 386-397Crossref (345) Scholar Beyond complementary metal oxide semiconductor (CMOS) technology, two-terminal stand out computing, offering attributes including subnanosecond times, picojoule multi-bit programmability, integratable structure.7Guo L. Liu Zhao Zhu Lin P. Jia C. Lee S.-J. al.Highly low-voltage memristive synapse enabled van der Waals integration.Matter. 2: 965-976Abstract Full Text PDF (24) 8Xu W. Min S.-Y. Hwang H. T.-W. core-sheath nanowire synapses femtojoule consumption.Sci. Adv. 2016; e1501326Crossref (290) 9Shi Liang X. Yuan B. V. Hui F. Yu Pop al.Electronic made layered two-dimensional materials.Nat. 458-465Crossref (293) 10Liu Jiang Y.-G. Zhang D.W. Two-dimensional materials next-generation technologies.Nat. 545-557Crossref (180) 11Wang Cai Pan Lian Zhuo Xu K. Cao T. al.Robust 130-136Crossref (324) 12Xue Gao Shang Yi R.-W. Recent advances memristors.Adv. 5: 1800854Crossref (27) 13Wu Lu Shi Wu A habituation sensory nervous 32: 2004398Crossref 14Lee H.-S. Sangwan V.K. Rojas W.A.G. Bergeron Jeong H.Y. Su Hersam M.C. Dual-gated MoS2 memtransistor array.Adv. Funct. 30: 2003683Crossref (33) 15Hwang J.-S. hybrid materials.Adv. 1800519Crossref (59) 16Goswami Rath S.P. Thompson Hedström Annamalai Pramanick Ilic B.R. Sarkar Hooda Nijhuis C.A. al.Charge disproportionate molecular redox discrete memcapacitive switching.Nat. 380-389Crossref (29) generally consist simple metal-insulator-metal configuration two or more switchable resistance states application “set” “reset” operation. typical relies formation rupture one relatively result localized defect concentration. Defects could stem from (for anion vacancies oxides random-access [RRAM]) active electrodes mass across insulating media bridge (CBRAM).17Nili Adam G.C. Hoskins Prezioso Kim Mahmoodi M.R. Bayat F.M. Kavehei O. Strukov D.B. Hardware-intrinsic security primitives analogue state nonlinear variations integrated memristors.Nat. 197-202Crossref (84) 18Jiang Yan Holcomb Xia provable key destruction scheme arrays.Nat. 548-554Crossref 19Herder Koushanfar Devadas Physical unclonable functions applications: tutorial.Proc. IEEE. 2014; 102: 1126-1141Crossref (647) Unfortunately, three-dimensional ion/atom there great cycle set reset voltages, high-resistance (HRS), low-resistance (LRS), etc., which lead uncertainty, nonuniformity, unreliability power value, its computing.20Alaghi Qian Hayes J.P. challenge computing.IEEE Trans. Comput. Aid. Des. Integr. Circ. Syst. 37: 1515-1531Crossref (103) 21Adam Khiat Prodromakis Challenges going mainstream.Nat. Commun. 9: 5267Crossref (39) 22Midya Asapu Joshi Song Upadhay N. Rao al.Artificial neural network (ANN) spiking (SNN) converters 1900060Crossref (42) In general, intrinsically variability.23Wang Burr G.W. C.S. K.L. information processing.Nat. 173-195Crossref (266) 24Wang Savel’ev S.E. Midya Hu Ge Strachan al.Memristors dynamics synaptic emulators 2017; 16: 101-108Crossref (1100) 25Lu Alvarez Kao C.-H. Bow I.W. An electronic silicon-based uniformity.Nat. 66-74Crossref (26) 26Zhao Ma Xiao Shao Niu al.Breaking current-retention dilemma cation-based utilizing graphene controlled defects.Adv. 1705193Crossref (156) Crystalline candidates issues practical computing.27Milano Luebben Dunin-Borkowski Boarino Pirri C.F. Waser Ricciardi Valov I. Self-limited single combining all-in-one functionalities.Nat. 5151Crossref (70) 28Jiang Belkin al.A novel true random number generator memristor.Nat. 8: 882Crossref (177) 29Lee Kwak Moon Choi Park Yoo Lim Sung Banerjee al.Various threshold integrate fire neuron applications.Adv. 1800866Crossref (46) 30Fu Ward J.E. Yin Walker D.J.F. al.Bioinspired bio-voltage 11: 1861Crossref (63) 31Zhu W.D. Memristor networks real-time activity analysis.Nat. 2439Crossref 32Yoon J.H. K.M. Ravichandran 417Crossref (171) 33Choi Tan S.H. P.-Y. Yeon SiGe epitaxial performance engineered dislocations.Nat. 17: 335-340Crossref (316) 34Yeon al.Alloying conducting channels 574-579Crossref Although widely hardware, conventional inorganic crystals/semiconductors display several disadvantages, such harsh conditions complex crystal growth utilization toxic elements. overcome some shortcomings, but poor oxidation and, cases, complicated synthetic processes, may hinder commercial translation. Peptide self-assemblies, typically aromatic short peptides, owing low-cost biocompatibility biodegradability, strong stability, well-crystallized structure. Besides, bio-machine interface. Diphenylalanine (FF), most-studied peptide, self-assembles into various architectures nanofiber nanosphere nanoscale microtube (MR) microscale, controlling solution conditions.35Fleming Ulijn R.V. Design nanostructures amphiphiles.Chem. Soc. 43: 8150-8177Crossref These supramolecular structures possess same crystallographic lattice structures, indicating facile low-temperature method develop abundant regulating dewetting solution. this regard, serve scale down size beyond limits top-down lithographic patterning, suitable regulate movement customized probe mechanism effect. addition, recent studies indicated self-assemblies (e.g., MR) ferroelectric photoluminescent properties, potentially granting unique in-sensor computing.36Gan Shen Light-induced ferroelectricity bioinspired self-assembled nanotubes/microtubes.Angew. Chem. Int. Ed. 2013; 52: 2055-2059Crossref (66) Hence, article, formation/rupture short-peptide self-assembling structure (FF MR), Ag/FF MR/Ag shows voltage temporal variation (σ/μ) lower than 0.09, uniform cell-to-cell (five cells constructed short-term features compared Ag/evaporated film/Ag Ag/SiO2/Ag. low activation thus promotes formation. effect directly microscopy, sensitive sufficiently resolution. memristor, nonvolatile be achieved simply adjusting electrode spacing, discovered basic (2-μm spacing) dynamics. Furthermore, nociceptor's numbers (0–15) been demonstrated virtue characteristic. bioresorbable ultrafast dissolution (within 100 s) upon relative humidity, implying suitability further degradable systems. MRs were realized bottom-up finely optimizing ratio concentration 1,1,1,3,3,3-hexafluoro-2-propanol (HFP)/deuteroxide (D2O) (Figure 1A, see experimental procedures), according previously developed methods.37Li Dai Controlled rod nanostructured assembly optical waveguide properties.ACS Nano. 2015; 2689-2695Crossref (158) obtained smooth diameters ∼200–1,000 nm S1). We selected diameter 500–600 mechanical next-stage fabrication. Mechanical damage events during transferring likely happen non-optimal <400 nm, Figure as-synthesized morphology studied scanning electron (SEM) atomic (AFM) depicted Figures 1B 1C. ordered nearly linear shape height 542 408 respectively. X-ray diffraction (XRD) measurement reveals present much better characteristic evaporated thin film 1D).37Li provide good localization event, expected reduce variability corresponding performance.27Milano Then, MR-based bridging (i.e., 5-nm Cr adhesion followed 60-nm electrode; 1, 2, 5 μm) isolated MR, resting SiO2 substrate, schematic diagram circuit (Ag/FF MR/Ag) displayed 1E (see procedures, S2 S3 fabrication details). Cu usually electrochemical accompanied transport.38Waser Dittmann Staikov Szot Redox-based – nanoionic mechanisms, prospects, challenges.Adv. 2009; 21: 2632-2663Crossref (3989) fundamental factor affecting performance. Thus, first evaluated barrier materials. 1F depicts Arrhenius plot conductivity (thermal evaporation)/Ag partially electroformed subthreshold pulse (amplitude, V; width, 8 S4). both increases when temperature 298 K 373 From temperature-dependent plot, Ag+ 0.23 eV 0.30 achieved. material's structure, ion-jumping distance, radius, charge ions. Difference between distinct materials’ distances. value significantly 0.77–1.24 eV),39Pi Ghadiri-Sadrabadi Bardin J.C. Nanoscale radiofrequency switches.Nat. 6: 7519Crossref (79) Scholar,40McBrayer J.D. Swanson R.M. Sigmon T.W. Diffusion metals silicon dioxide.J. Electrochem. 1986; 133: 1242-1246Crossref (488) path. Therefore, built determine if would 1G). To initial current-voltage (I-V) Ag/SiO2/Ag Au/FF MR/Au device, presented 2A–2C (electrode 2 μm). metallization low-conductive functioned dielectric S5). original needs prior cyclical events. forming shown 2A sufficient applied ∼7 V) abruptly switched high-conductance (compliance current, μA), spontaneously relaxed back low-conductance at close-to-zero voltage, suggesting transient behavior. After electroforming, symmetric hysteresis loops detected positive negative sweeps, bipolar dramatically different cell 2B), especially step. Due property 0.043 S/m, S5), large on/off ∼105 steep switch-on slope ∼40 mV/dec. independent compliance settings, remained ∼1.6 V currents, field-driven S6). contrast remain configurations even under high-voltage sweep 20 V), 2C, originated synergistic effort electrodes. It worth noting closely related spacing Control prepared compare capabilities. type varying ranging 1 μm (Figures 2D–2F). second Ag/thermal fixed 2G). greatly depends spacing. Similar also shorter after electroforming 2D. 5-fold device. change operation explained because shortening enhance electric field intensity (electric E = V/day), ions through dielectrics, accelerate “dead” stuck LRS just three cycles permanent pathway reduction (below means higher integration density applications, 2-μm applications. contrast, longer (5 μm), maintained HRS 0 V, difficulty brings separation distance 2F). struggled cycle-to-cycle particularly precarious 3V 6V, randomly distributed unsmooth I-V 2G. 2E). Compared film/Ag, improved mainly originates one-dimensional (1D) MR. evaluate repeatability analyzed (cell 3, 4 5; (100 cycles) five 2H). As 2I 2J, as-fabricated all start HRS, show comparable narrow distributions LRS/HRS well voltages sweeps (forming 6.5–8.0 1.4–1.8 consistency devices. Repeatable 30 cells, each contains S7). consecutive positive/negative sweeping over observed, 2K. 2L (bottom panel) histogram 1. exhibited distribution ∼1.64 (−1.65 quite degree According Gaussian fits histograms, variabilities (defined σ/μ, σ being standard deviation μ mean distribution) 0.095 0.060 (top panel), respectively, other material-based devices.27Milano Scholar,41Fang Huang Improvement HfOx-based RRAM inserting ALD TiN buffer layer.IEEE Electron Device Lett. 39: 819-822Crossref (32) 42Parreira Paterson McVitie MacLaren D.A. Stability, bistability instability ZrO2 devices.J. Phys. D Appl. 49: 095111Crossref 43Fantini Goux Degraeve Wouters D.J. Raghavan Kar Belmonte Govoreanu Jurczak Intrinsic HfO2 RRAM.in: 5th IEEE International Workshop. IEEE, 2013: 30-33Google Temporal term elucidated tests cycles). limited delay time LRS, requires return HRS. pulses yield 2M), reproducibility using larger amplitude (8 ms), turn-on reduced 400 μs S8). similar feature range humidity (RH) (15%–65%), S9). nanostructure, stable weeks RH ∼80% S10), fully degrade fast manner humidifying treatment S9 S11 Video https://www.cell.com/cms/asset/c66993f8-f34f-45c5-8db4-91abf03c5cfe/mmc2.mp4Loading ... Download .mp4 (12.31 MB) Help files S1. Degradation Dynamic properties fading (an state) then investigated applying stimulation monitoring response currents. Under amplitudes (2.5V, 3.5 4.5 duration, 25 showed (with current 10−4 A). input ended, finite time, 3A). (or decay time) increased ms increasing amplitude. uniformity, intervals 200 (larger successively, on-state cur