Spatial evolution of the proton-coupled Mott transition in correlated oxides for neuromorphic computing

Abstract

The proton-electron coupling effect induces rich spectrums of electronic states in correlated oxides, opening tempting opportunities for exploring novel devices with multifunctions. Here, via modest Pt-aided hydrogen spillover at room temperature, amounts of protons are introduced into SmNiO₃-based devices. In situ structural characterizations together with first-principles calculation reveal that the local Mott transition is reversibly driven by migration and redistribution of the predoped protons. The accompanying giant resistance change results in excellent memristive behaviors under ultralow electric fields. Hierarchical tree-like memory states, an instinct displayed in bio-synapses, are further realized in the devices by spatially varying the proton concentration with electric pulses, showing great promise in artificial neural networks for solving intricate problems. Our research demonstrates the direct and effective control of proton evolution using extremely low electric field, offering an alternative pathway for modifying the functionalities of correlated oxides and constructing low-power consumption intelligent devices and neural network circuits.

Fig. 1.. Fabrication of the asymmetric devices and characterization of the hydrogenated films.

(A) In situ variation of the device resistance in the hydrogenation process measured with a readout voltage of 50 mV. The inset is the schematics of the SNO-based devices. Through the hydrogen doping procedure on one side of the SNO film using Pt-assisted hydrogen spillover, the device with the structure of Au/SNO/Pt transforms to Au/SNO/H-SNO/Pt. (B) Depth profiles of hydrogen (H₁) concentration in the pristine (SNO) and hydrogenated (H-SNO) films measured with TOF-SIMS. (C) Raman spectrums of the SNO and H-SNO films. The red and the black dashed lines indicate the Raman peaks of SNO and LaAlO₃ (LAO) substrate, respectively. a. u. is the abbreviation for arbitrary unit. (D) Optical image of the device near the Pt electrode after hydrogenation (top) and the contour map of Raman line scanning results (bottom). Each Raman spectrum was normalized to the intensity of the LAO Raman mode at 487 cm⁻¹. (E) Variation of the Raman intensity obtained by integrating signals in the range of 400 to 475 cm⁻¹ and 520 to 720 cm⁻¹ of each spectrum in (D).

Fig. 2.. Atomic-scale modification of the SNO lattice caused by hydrogen-ion intercalation.

(A to C) HAADF-STEM and corresponding ABF-STEM images of the pristine SNO thin film along the [1-10] projection. (D to F) HAADF-STEM and corresponding ABF-STEM images of the hydrogenated H-SNO film. In HAADF images [(A), (B), (D), and (E)], the atom columns with the stronger intensity are Sm columns, while the Ni columns have a weaker intensity, and the O columns are barely visible. Orange lines in (B) and (E) mark the zigzag distortion of Sm atoms along the [001] direction. In addition to HAADF, ABF imaging is relatively sensitive to light elements, enabling the identification of oxygen atoms [(C) and (F)]. Insets in (C) and (F) show the calculated lattice structure with oxygen octahedral rotation and tilting, which display the very consistent results to the measured STEM. (G to I) Calculated structure of SNO with different doping contents of the hydrogen ions. The cyan, gray, magenta, and purple atoms represent the Sm, Ni, O, and H atoms, respectively. The angles produced by three neighboring Sm ions are labeled with black lines.

Fig. 3.. Electrical behavior of the asymmetric device.

(A) Schematic showing the electric field–controlled migration of protons in the device. The cyan, gray, magenta, and purple atoms represent the Sm, Ni, O, and H atoms, respectively. (B) Conductance variation of the Pt/H-SNO/SNO/Au device during the dc bias electric field sweeping process (0 → 0.2 → 0 → −0.25 → 0 mV/nm). (C) Nonvolatility test at LRS formed by a positive sweep and HRS formed by a negative sweep. (D) IV curves after application of electric pulses. (E) Endurance test under alternating positive and negative pulses. (F) Multiresistance states realized by the application of continuous pulses. The readout voltage in (C), (E), and (F) is 50 mV.

Fig. 4.. Theoretical calculation and mechanism discussion.

(A) First-principles calculation of density of states of SNO for different proton doping concentrations. The purple dashed circle indicates the intermediate band. The purple dashed arrow indicates the Mott transition with enlarged bandgap, accompanied by the disappearance of the intermediate band. (B) Energy landscape and atomic-scale pathway of proton migration in the SNO lattice. The potential energy is shown along the most preferred migration pathway, together with selected configurations along this pathway (labeled as I₁ to I₆). (C) Schematic of the resistance switching enabled by the hopping of protons in our device. The purple and orange octahedrons indicate that their Ni sites have two and one occupied e_g states, respectively. The right panel shows the schematics of Ni 3d orbitals with two and one occupied e_g states. U represents the on-site electron-electron correlation. Strong Coulomb repulsion in doubly occupied e_g orbitals above the t_2g orbitals causes the electrons to become localized.

Fig. 5.. Simulation of synaptic functions.

(A) Schematic illustration of a biological chemical synapse. The release of neurotransmitters driven by presynaptic action potential is very similar to electric field–controlled proton migration in the nickelate device, and the neurotransmitter-induced opening of ligand-gated channels is equivalent to the proton coupling effect. (B) Continuous LTP and LTD properties simulated in the nickelate device by continuous positive (0.125 mV/nm, 1 ms) and negative (−0.15 mV/nm, 1 ms) pulses with a pulse interval of 10 ms. A readout voltage of 500 mV is applied between each pulse, as shown in the inset. (C) Cycle tests of LTP and LTD properties in (B). (D) Conductance variation of potentiation process under consecutive pulses with different amplitudes. (E) Schematic of synaptic weight as a function of stimulus number showing a tree-like structure with the behavior of habituation and reactivation in biological synapses. (F) Variation of the device conductance to consecutive electrical pulses. The concrete application form of continuous pulses is shown in fig. S20. The different colors represent different pulse amplitudes. The reactivation behavior was observed under the stronger pulse stimulus after the same consecutive pulses. (G) Multiconductance states with a branch structure in the potentiation process. (H) Tree-like structure of the multiconductance states generated by applying consecutive positive or negative pulses, with characteristics similar to the synaptic behavior shown in (E). The same colored data points correspond to a fixed pulse stimulus. The pulse width and interval of each pulse in [(D), (F), (G), and (H)] are fixed at 1 and 10 ms, respectively.