Approaching the Zero‐Power Operating Limit in a Self‐Coordinated Organic Protonic Synapse

Abstract High‐performance artificial synapse with nonvolatile memory and low power consumption is a perfect candidate for brainoid intelligence. Unfortunately, due to the energy barrier paradox between ultra‐low power and nonvolatile modulation of device conductances, it is still a challenge at the moment to construct such ideal synapses. Herein, a proton‐reservoir type 4,4′,4″,4'''‐(Porphine‐5,10,15,20‐tetrayl) tetrakis (benzenesulfonic acid) (TPPS) molecule and fabricated organic protonic memristors with device width of 10 µm to 100 nm is synthesized. The occurrence of sequential proton migration and interfacial self‐coordinated doping will introduce new energy levels into the molecular bandgap, resulting in effective and nonvolatile modulation of device conductance over 64 continuous states with retention exceeding 30 min. The power consumptions of modulating and reading the device conductance approach the zero‐power operating limits, which range from 16.25 pW to 2.06 nW and 6.5 fW to 0.83 pW, respectively. Finally, a robust artificial synapse is successfully demonstrated, showing spiking‐rate‐dependent plasticity (SRDP) and spiking‐timing‐dependent plasticity (STDP) characteristics with ultra‐low power of 0.66 to 0.82 pW, as well as 100 long‐term depression (LTD)/potentiation (LTP) cycles with 0.14%/0.30% weight variations.


Disadvantages of charge transfer and conductive filament models
As mentioned in the main text, ultra-low power and nonvolatile conductance modulation is a natural paradox for organic neuromorphic devices working on either charge transfer (CT) or metallic conductive filament (CF) mechanisms.Although the small effective mass and large charge-to-mass ratio of electrons usually lead to strong electrostatic driving force, fast response speed and low-power modulation characteristics in CT based processes, they also encounter easy backward relaxation of spontaneous electron-hole recombination and therefore short life time problems (Figure S1a).On the other hand, the relatively large volume and mass of metal cations can guarantee long-term stability of thick conductive filament formed upon ion migration and redox reaction.The high electric field used to inject abundant metal ions from electrode into organic switching layer, as well as the excellent metallic conduction of the as-formed filament that carries high current during device operation, nevertheless, give rise to significant power consumption that is obviously undesired for large scale integration and applications (Figure S1b).erence electrode, and a Pt wire as the counter electrode, respectively.Tetrabutyl ammonium hexafluorophosphate was recrystallized using chloroform to remove the low-content impurity before using.The UV-visible absorption spectral measurement was conducted on PerkinElmer Lambda 750S.The fluorescence spectrum was measured on Renishaw inVia Qontor with fluorescence mode.Grazing incidence wide angle X-ray scattering (GIWAXS) experiments were performed using a Xenocs Xeuss 3.0 beamline at the Vacuum Interconnected Nanotech Workstation (NANO-X) of Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences.Xeuss 3.0 (GI-)SAXS/WAXS/USAXS beamline was equipped with dual X-ray sources, an Excillum high flux gallium metal jet source and a Xenocs GeniX 3D micro-focus copper source.An X-ray wavelength of λ = 1.3414Å and a beam size of (0.9×0.9) mm 2 were employed for the measurements.The scattered intensities were collected using a Dectris EI-GER2 R 1M detector at a sample to detector distance of 130 mm, with an instrument resolution of ~0.02 Å -1 .GIWAXS data were collected with an exposure time of 1200 s and X-ray incident angle of 0.2°.Atomic force microscopy (AFM), conductive atomic force microscopy (C-AFM) and piezo force microscopy (PFM) measurements were performed on an FastScan Bio (Bruker Co., American) microscope to monitor the surface morphology, local conduction and polarization characteristics of the TPPS film deposited on Au coated SiO2/Si substrate.
It should be noted that the chemical shift of δ=4.07 can be attributed to the cation of H3O + produced from the coordination between H2O molecule and proton (H2O + H + → H3O + ), which exchanges hydrogen element with deuterated DMSO and thus results in an obvious shift to the low field.MALDI TOF-MS (C44H30N4O12S4) m/z: calcd.934.989, found: 935.03 (Figure S3a).
According to the FT-IR spectrum of TPPS, the peak at ca 1228 cm -1 , 1192 cm -1 and 1122 cm -1 can be ascribed to the SO3 -vibrations, which indicates a successful synthesis with sulfonic acid group (Figure S3b). Figure S2. 1 H NMR spectrum of TPPS.

Section 4. Optical properties and energy band gap of TPPS film
Benefiting from these characterizations aforementioned to ensure a success synthesis of TPPS molecule, we subsequently evaluate its optical and structural properties in the condensed state.
A thin film of TPPS was formed on quartz substrate by spin-coating.Before spin-coating, the quartz substrate was pre-treated with oxygen plasma for 10 min to make the surface hydrophilic.
As shown in Figure S4, the TPPS film shows intrinsic absorption peaks at the wavelengths of 366, 428 and 490 nm, while the absorption peaks at 633 and 702 nm are ascribed to the π-π stacking of the planar molecules in the TPPS film. 1 The latter absorptions at the longer wavelengths of 633 and 702 nm indicate that TPPS film may have the ordered microstructure in solid state.When the TPPS film is excited by green light with the wavelength of 532 nm, a wide emission peak centered at 717 nm can be observed.The difference in the peak positions of fluorescence emission (717 nm) and maximum visible absorption (702 nm) gives rise to a small Stokes shift of 15 nm, again suggesting good crystallinity of the TPPS film that suppresses the vibrational relaxation of the activated-state molecules.With this plot, the energy band gap of TPPS can be calculated as the intersection of the first absorption peak's tangent with the X-axis, where α is the linear absorption coefficient of TPPS film, h=6.63×10 -34 J•s is the Planck's constant, ν is the photon frequency and B is a material related constant.α equals 2.303×A/d, where A and d are the absorption intensity and thickness of the TPPS film, respectively.n, with the value of 2 or 1/2, is related with electron hopping behavior and used to calculate the direct band gap or indirect band gap, respectively.Since most organic semiconductor materials have indirect band gaps, 2 herein we employ n=1/2 to derive the above function and plot in Figure S5a.Consequently, the energy band gap of TPPS film is read as the intersection of first absorption peak's tangent with the X-axis, which is 1.64 eV.

Section 5. Structural properties of TPPS film
Grazing incident wide-angle X-ray scattering (GIWAXS) measurement was also conducted to assess the crystallinity of TPPS films.As shown in Figure S6, the GIWAXS profile of TPPS film shows a scattering ring at the scattering vector of 1.48 Å, which can be ascribed to the inplane isotropic alignment of the grains' (001) lattice plane.The minor anisotropic signal observed at the scattering vector of 0.42 Å, on the other hand, is in good agreement with the ordered π-π stacking of the planar TPPS molecules in the out-of-direction.

Section 6. Morphological and local conduction properties of TPPS film
As depicted in Figure S7a, the atomic force microscope (AFM) profile of the TPPS thin film further proves featureless alignment of crystalline grains with a root-mean-square roughness of 1.632 nm.Such uniform distribution of the organic crystals can provide a three-dimensional continuous network for ion migration.Conductive atomic force microscope (C-AFM) scan over the area of 1 μm×1 μm shows a uniform tiny leakage current of 0.09 pA (read at 0.01 V), suggesting that the initial TPPS film is in the high resistance state (Figure S7b).When -1.5 V voltage stimuli were applied point-by-point onto the TPPS thin film through C-AFM cantilever, densely distributed local conductive regions with maximum current exceeding 25 pA can be observed (Figure S7c).It indicates that the TPPS film has been switched to a high conductance state.Zooming-in observation from 1 μm×1 μm to 0.5 μm×0.5 μm, 0.1 μm×0.1 μm and 50 nm×50 nm scanning areas reveals that the inter-spacing between the adjacent conductive regions can be as small as 10 nm, therefore manifesting the downscaling potential of TPPS-based OPM devices into nanometer range.

Section 7. Compatibility of TPPS with device fabrication process
To validate the compatibility of TPPS molecules with the device fabrication process, we firstly conducted solubility tests on the developing, fixing, and stripping solvents of methylisobutylketone (MIBK) and isopropanol (IPA) mixture in a 1:3 volume ratio, isopropanol (IPA), and acetone, respectively.20 mg powders of TPPS were added into each of the 3 mL above solvents.
After resting for 3 days, direct observation was conducted to check whether the TPPS powders were dissolved by these solvents or not.As shown in Figure S8, the upper liquid remained colorless while the solids were all precipitated in the bottom of the glass bottle.Upon discarding the liquids and drying the solids in vacuum oven at 60 °C overnight, it was found that the weight of each TPPS sample was in close proximity to 20 mg.These results suggest that TPPS does not dissolve in the solvents used in the e-beam lithography process.It is in good agreement with the "like-dissolves-like rule" in fundamental chemistry, wherein TPPS is highly polar with tethered strong electron withdrawing sulfonic acid groups, while the above solvents carry either low or moderate polarities.In order to confirm the validity of the material designing principle in this work, we fabricated 20 TPPS devices with the dimension of 10 μm and tested their electrical characteristics.
As plotted in Figure S11, all the devices show almost identical hysteresis in the current-voltage characteristics when sweeping between 0 V and -0.1 V.The device-to-device (D2D) uniformity of the conductances of these organic synapses is calculated as 93.49%, according to the following equation where δ and μ are the standard deviation and average values of the 20 devices' conductances read at -0.01 V, respectively.It suggests that the TPPS devices show promising uniformity and reliability.As shown in Figure S13a, the device currents read at -0.01 V increase from 0.36 nA to 1.0 nA accordingly.The power consumption of reading the device during this negatively biased modulation period varies from 3.64 pW to 10.06 pW, respectively (Figure S13b).The built-in field gradually varies from 0.44 V to 1.22 V (Figure S13c).A reserved sweeping from 0 V to 0.8 V consequently recovers the device conductance to the initial state as a reset operation (Figure S13d).When successive positively biased dual sweepings with the stopping voltages increasing from 0.1 V to 1.0 V with a ramping step of 0.01 V are performed, the device currents read at 0.01 V decreases from 0.26 nA to 22.51 pA, respectively (Figure S14a).The corresponding reading power consumption varies between 2.61 pW and 0.23 pW, while the built-in potential established across the TPPS film are between 0 V and -0.5 V (Figure S14b, c).The conductance evolution during the positive sweeping with the stopping voltages increasing from 0.1 V to 1.0 V is shown in Figure S15.Likely, a reverse sweeping from 0 to -0.2 V can erase the device returning to the initial state (Figure S14d).TPPS based OPM devices with small electrode linewidth of 5 μm, 1μm and 100 nm are also fabricated to assess low-power operation capabilities (Figure 2a).As shown in Figure S16a and 16b, both the 5 μm and 1 μm OPM devices exhibit continuous current modulation during negatively-biased voltage sweepings with the stopping voltages increasing from -0.1 V to -0.5 V with a ramping step of -0.01 V.When read at -0.5 V, the maximum currents of the 5 μm and 1 μm devices are 628.05pA and 42.62 pA, respectively.When the electrode linewidth further shrinks to 100 nm, the device currents drop dramatically to the detecting limit of the semiconductor parameter analyzer, which is ~ 648.0 fA as read at -0.5 V (Figure S16c).Coincidently, the current maxima read at -0.5 V for the 10 μm, 5 μm, 1μm and 100 nm devices show a good linear dependence on the device area (Figure S16d), suggesting that the conductance tuning in TPPS devices occurs as an interfacial phenomenon.

Section 1 . 2 . 4 .
Disadvantages of charge transfer and conductive filament models Section Reagents and characterization instruments for TPPS Section 3.Composition characterization of TPPS Section Optical properties and energy band gap of TPPS film Section 5. Structural properties of TPPS film Section 6. Morphological and local conduction properties of TPPS film Section 7. Compatibility of TPPS with device fabrication process Section 8. Electrical performance of Au/TPPS/Au OPM devices Section 9. Correlation between proton migration, self-coordination and conductance modulation of Au/TPPS/Au OPM devices Section 10.Experimental setup for in-situ electrochemical fluorescence measurements Section 1.

Figure S1 .
Figure S1.Energy band diagrams and working mechanisms of organic neuromorphic devices based on (a) charge transfer and (b) metal conductive filament models.

Section 2 .
Reagents and characterization instruments for TPPS 5,10,15,20-tetraphenylporphyrin (TPP) was purchased from Bide Pharmatech Ltd. and used without further purification.Concentrated sulphuric acid (H2SO4) was purchased from Sinopharm Group Co. Ltd. and used without further purification.The 1 H nuclear magnetic resonance (NMR) spectra were recorded on a Bruker AVANCE III HD 500 spectrometer at 500 MHz in deuterium substituted water with tetramethylsilane (TMS) as reference for the chemical shifts.The matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI TOF-MS) was measured on Bruker autoflex speed TOF/TOF.The Fourier transformed infrared (FT-IR) spectrometry was conducted on Thermo Fisher Nicolet 6700.The cyclic voltammetry (CV) measurements were performed on a model CHI 650D electrochemical workstation with tetrabutyl ammonium hexafluorophosphate (0.10 M) in acetonitrile as the supporting electrolyte, a platinum disk as the working electrode, an Ag/AgCl electrode as the ref-

Figure
Figure S5a converts the UV-visible absorption spectrum of the TPPS film into a Tauc plot, Figure S4.UV-visible absorbance and fluorescence spectra of TPPS thin films.

Figure S5 .
Figure S5.(a) Tauc plot of the TPPS film's UV-visible absorbance spectrum.(b) UPS measurement results of the (top) secondary electron cutoff and (bottom) ionization edge.

Figure S6 .
Figure S6.(a) 2D GIWAXS pattern and (b) out-of-plane direction X-ray scattering intensity profile of the TPPS thin film.

Figure S8 .
Figure S8.Images of MIBK/IPA, IPA and acetone solvents (left sample in each image) and mixture of TPPS powders with these solvents (right sample in each image), respectively.The images were taken 3 days after adding the TPPS powers into the solvents.

Figure S9 .
Figure S9.(a) UV-visible absorption, (b) fluorescence and (c) FTIR spectra of the TPPS thin films before (black curves) and after (red curves) being subjected to the device fabrication processes, respectively.

Figure
Figure S10 visualizes the cross-sectional view of the Au/TPPS/Au OPM device.It is clear that

Figure
Figure S12 replots the conductance-voltage characteristics of the TPPS device during

Figure S12 .
Figure S12.Conductance-voltage characteristics of TPPS device during sweeping between -0.1 V and -1.17 V, with the data retrieved from Fig. 2c.

Figure S13 .
Figure S13.(a) Current-voltage characteristics and (b) reading power consumptions of the Au/TPPS/Au device during the negatively biased voltage sweepings with the stopping voltages varying between -1.0 V and -5.0 V with a ramping step of -0.1 V. (c) Evolution of the built-in potential established across the TPPS film during the negatively biased voltage sweepings shown in (a).(d) Current-voltage characteristics of the device during the positively biased voltage sweeping between 0 V and 0.8 V.

Figure S14 .
Figure S14.(a) Current-voltage characteristics and (b) reading power consumptions of the device during the positively biased voltage sweepings with the stopping voltages varying between 0.1 V and 1.0 V with a ramping step of 0.01 V. (c) Evolution of the built-in potential established across the TPPS film during the positively biased voltage sweepings shown in (a).(d) Currentvoltage characteristics of the device during the negatively biased voltage sweeping between 0 V and -0.2 V.

Figure S15 .
Figure S15.Conductance-voltage characteristics of TPPS device during sweeping between 0.1 V and 1.0 V, with the data retrieved from Figure S14a.

Figure S16 .
Figure S16.Current-voltage characteristics of the Au/TPPS/Au devices with metal electrode linewidths of (a) 5 μm, (b) 1μm and (c) 100 nm, respectively.All these measurements are performed by negatively biased voltage sweepings with the stopping voltages increasing from -0.1 V to -0.5 V with a ramping step of -0.01 V. (d) Linear dependence of the device current maxima on the metal electrode linewidths.

Section 9 .
Figure S17.Current-voltage characteristics of the Au/TPPS/Au devices during negatively biased voltage sweepings between 0 V and -5.0 V. Regions showing slow and fast current-increasng behaviors, as well as negative differential resistance (NDR) effect, can be observed.

Figure S18 .
Figure S18.Evolution of the molecular geometry and dihedral angles between the sulfonic groups and planar porphyrin framework, along with the excitation and migration of protons from the TPPS molecule.

Figure S19 .
Figure S19.(a) Lateral structured Au/TPPS/Au device and experimental setup for in-situ electrochemical fluorescence measurements.(b) Distributions of molecular orbitals of TPPS in its neutral ground state and excited state.As shown, electron transfer from the neutral excited state TPPS to neutral ground state TPPS will emit strong fluorescence.(c) Distribution of molecular orbital of TPPS in its negatively charged excited state.The shift of electron cloud from the central porphyrin skeleton to the peripheral sulfonic moiety upon proton migration and intramolecular charge transfer will attenuate the intensity of TPPS fluorescence.