Neuromorphic computing applications have peaked research interest in the last decades, for energy-efficient information processing to carry out highly sophisticated tasks utilizing artificial intelligence. Recently emerged lithionic memristor (LM) devices were evaluated to mimic the biological synapses in the brain by regulating the information flow in tuning the electronic conductance states of thin film Li-based oxides with respect to their Li⁺ insertion/extraction [1]. Such Li-ion synaptic electrochemical transistor has been demonstrated using thin film LiCoO₂ (LCO) as a channel (cathode) [2] combined with amorphous silicon (Si) as a gate (anode) and LIPON as gate oxide (solid electrolyte). However, the demonstrated electrochemical transistor still suffers from slow kinetics due the crystal structure quality of the thin Li-based oxides integrated on the Si substrate, alongside the ethical, sustainability and abundancy issues regarding the Co [3]. In this work, we evaluate Li_xMn₂O₄ (LMO) and Li_4+3xTi₅O₁₂ (LTO) (0≤x≤1) as alternative LM channels where x is defined as state-of-charge (SOC). Low toxicity, high natural abundance, moderate practical capacities (120-160 mAh.g^-1) and fast Li-ion diffusion have already made LMO and LTO attractive materials for electrodes in Li-ion batteries [3]. As part of the evaluation, it is necessary to first quantify the electronic conductivity (σ_e) variation of LMO and LTO relative to the SOC by comparing both dense pellets and thin films (≤100 nm), which is a crucial step for an efficient resistive switching mechanism.

For this purpose, LTO and LMO pellets were synthesized via isostatic pressing at 350 MPa followed with sintering step to attain sufficiently high densities (≈70%). The SOC was varied by electrochemical (de-)lithiation at 0.008-0.024 mA.cm^-2 in LP30 carbonate-based electrolyte. XRD studies were conducted to determine the structural change of Li_xMn₂O₄ before and after cycling to various cut-off potentials. A Li-ion blocking cell using Ag paste metal contact was employed to perform chronoamperometry (CA) at [OCP, 0%, 25%, 50%, 75% and 100%] SOC. Our results demonstrate σ_e increase from 3.10^-6 S cm^-1 (x=1) to 2.10^-4 S cm^-1 (x=0) for LMO pellets. Meanwhile, LTO pellets demonstrate an increase in σ_e from 2.10^-6 S cm^-1 (x=0) to 1.10^-3 S cm^-1 (x=1). As a result of this σ_e increase and considering a 1 μS conductance [2] as switching steps, LMO and LTO offer the possibility to fulfill one of the key requirements [100-1000 conductance states] for neuromorphic computing applications. Based on those encouraging results, a similar systematic and comparative study is currently in progress for thin film crystalline LMO and LTO integrated on Si substrates. The higher density and crystallinity, together with a controlled interface are expected to improve the Li-ion kinetics thus providing a conclusive assessment of LMO and LTO as suitable channels for lithionic application.