Unlocking Neuromorphic Vision: Advancements in IGZO-Based Optoelectronic Memristors with Visible Range Sensitivity

Optoelectronic memristors based on amorphous oxide semiconductors (AOSs) are promising devices for the development of spiking neural network (SNN) hardware in neuromorphic vision sensors. In such devices, the conductance state can be controlled by both optical and electrical stimuli, while the typical persistent photoconductivity (PPC) of AOS materials can be used to emulate synaptic functions. However, due to the large band gap of these materials, sensitivity to visible light (red/green/blue) is difficult to accomplish, which hinders applications requiring color discrimination. In this work, we report a 4 μm2 hydrogen-doped (H-doped) indium–gallium-zinc oxide (IGZO) optoelectronic memristor that emulates all of the important rules of SNNs such as short- to long-term memory transition (STM–LTM), paired-pulse facilitation (PPF), spike-time-dependent plasticity (STDP), and learning and forgetting capabilities. By the incorporation of hydrogen gas in the sputtering deposition of IGZO, visible sensitivity was achieved for green and blue wavelengths. Additionally, extremely high light/dark ratios of 179, 93, and 12 are demonstrated for wavelengths of 365, 405, and 505 nm, respectively, due to hydrogen-induced subgap states and device miniaturization. Therefore, the proposed device shows remarkable potential for integration with the pixel circuits of IGZO-based displays with extreme resolution for a true intelligent self-processing display.

In Figure S1, the atomic force microscope (AFM) image of the Ti/Au film with 7 nm is presented showing low root mean square (RMS) roughness of 402.2 pm.
In Figure S3, the transient response for 10 s of illumination, followed by 30 s in the dark, of the Mo/IGZO/ITO device with a VRead of 0.1 V is presented.There is no response to red light.For the other wavelengths, Ilight/Idark ratios of 2.8, 2.3 and 1.2 are achieved for 365, 405 and 505 nm irradiation, respectively.Higher Ilight/Idark ratios are accomplished for a VRead of -0.5 V, as discussed in the main article.However, the higher voltage applied forces electrons into VOs decreasing the photocurrent during illumination and fastening the PPC decay.For a VRead of 0.1 V this is solved, and the photocurrent continues to slightly increase during illumination.The reason why the photocurrent does not increase during irradiation as much as in the Ti/Au devices lies in the transmittance of the top electrodes.Since ITO is more transparent than the Ti/Au, the light intensity reaching the IGZO layer is higher and therefore the photoresponse is faster.The PPC decay is still fast, but this could be due to the very low Ilight/Idark ratios.
Moreover, the Idark is more than 1 order of magnitude higher for the ITO device than for the Ti/Au device for the same VRead.In Figure S4, the O 1s spectra is shown for the bulk hydrogen-doped (H-doped) IGZO film and the H-doped IGZO/Ti/Au interface.The peaks were fitted with a Gaussian-Lorentzian (G-L) function and a Shirley background subtraction.Similarly to the undoped IGZO results presented in Figure 1 of the main paper, the VOs percentage increases drastically from 34.6% to 67.7% at the interface with Ti/Au.This result confirms the Ti oxygen getter effect.In Figure S6, the H-doped IGZO device with ITO as top contact is analysed.Figure S5(a) shows the IV characteristic of the IV characteristic in the dark, between -0.5 and 0.5 V, of the undoped device and the doped device.The conductance state is substantially increased, especially for negative polarities which means the Idark is increased with doping.This can be related to H diffusion to the ITO which would increase the conductance of the top contact.In Figure S5(b), the transient response of the optoelectronic memristor can be found for 365, 405 and 505 nm wavelength irradiation.Ilight/Idark ratios of 11.6, 6.5 and 2.2 can be distinguished, respectively, which is not an improvement in relation to the undoped-device.However, the Ilight/Idark ratio is greatly reduced.Green sensitivity is also gradually lost as the memristor size increases.In fact, the 10000 µm 2 device shows no green detection and reduced ratios of 5 and 3 for UV and blue illumination, respectively.In Figure S8, the transient response to optical and electrical inputs on the doped memristor with ITO as top contact can be evaluated.Increasing current states are achieved with more energetic wavelengths.The Reset electrical pulses are also discriminated.The higher the current state, the higher the voltage or the longer the pulse required to perform a full Reset.doping and Ti/Au as top contact.The results can be found in Figure S9, where the transient response to 10 cycles of UV, blue and green illumination, followed by few seconds in the dark and a Reset pulse, is presented with no considerable change in device performance.

Figure S 1 .
Figure S 1. AFM image of Ti/Au surface.

Figure S 3 .
Figure S 3. Transient response of the Mo/IGZO/ITO device for 10 s of illumination followed by 30 s in the dark with VRead of 0.1 V for different wavelengths.

Figure S 4 .
Figure S 4. (a) and (b) Fitting of O 1s XPS spectra displaying the oxygen vacancies (VOs) percentage compared to the metal-oxygen (M-O) bonds percentage for the doped-IGZO bulk film and the doped-IGZO/Ti/Au interface, respectively.

Figure S 5 .
Figure S 5. Valence band spectra of the bulk IGZO:H film, used to calculate the valence band maxima (VBM).

Figure S 6 .
Figure S 6.(a) Comparison of the IV characteristic in the dark of the IGZO and the H-doped IGZO devices, with ITO as top contact.(b) Transient response with VRead of -0.5 V and 10 s of irradiation followed by 30 s in the dark of the H-doped device with ITO as top contact for 660, 505, 405 and 365 nm illumination.

Figure S 7 .
Figure S 7. (a) IV characteristic in the dark from -0.5 to 0.5 V for Mo/IGZO:H/Ti/Au devices with different sizes.(b) Idark and photocurrents after 10 s of illumination with 365, 405 and 505 nm wavelength for devices with different sizes and (c) Transient response of devices with different sizes for 10 s of illumination followed by 30 s in the dark with VRead of 0.1 V for different wavelengths.

Figure S 8 .
Figure S 8. (a) Increased conductance states achieved by decreased wavelengths illumination and respective electrical Reset pulse for the Mo/IGZO:H/ITO device.

Figure S 9 .
Figure S 9. 10 cycles of (a) UV, (b) blue and (c) green illumination, followed by a few seconds in the dark and a Reset pulse for the Mo/IGZO:H/Ti/Au device.

Figure S 11 .
Figure S 11.Different conductance states reached by (a) increased power, (b) increased time and (c) increased frequency of illumination with a 505 nm wavelength LED.Different conductance states reached by (d) increased power, (e) increased time and (f) increased frequency of illumination with a 365 nm wavelength LED.

Figure S 12 .
Figure S 12. Kohlrausch stretched exponential PPC fitting on different irradiation times for (a) blue, (b) green and (c) UV light irradiation.Characteristic relaxation time and stretch index for different irradiation times for (d) green light irradiation and (e) UV light irradiation.

Figure S 13 .
Figure S 13.Learning and Forgetting demonstration by pulsing.Learning is performed by 30 optical pulses of 150 ms, 405 nm wavelength, 0.1 mW/cm 2 of power and 1 Hz of frequency.Forgetting is performed in the dark by applying the VRead of 0.1 V for 20 s.

Table S 1
Comparison on energy consumption of the most energy efficient optoelectronic synaptic devices reported in the recent literature for visible light detection.