Since the discovery of memristive devices, the conduction mechanisms are highly discussed and are of high technological interest, e.g. for neuromorphic computing. Here, TiO_2-x based lateral memristive metal-insulator-metal (MIM) devices with inert Pt-electrode and active Ag-electrode were investigated by in situ electrical DC measurements in the scanning electron microscope (SEM). Afterwards the structures were also investigated by electrical impedance spectroscopy (EIS).
The lateral structures were produced by multiple exposure UV-lithography on a p-type (100) 4’’ Si-Wafer with a dry thermal SiO₂ as a diffusion and electrical barrier. TiO_2-x was sputtered on the whole wafer by a reactive pulsed DC magnetron process. The electrode materials were deposited by e-beam evaporation and structured with a lift-off process.
After the lift-off process some Ag particles remain in the gap between the electrodes. However, the particles seem to be a useful artefact helping to build percolation paths between the electrodes and to induce memristive effects. During the DC-measurements some particles are partly deforming out of the electrode into the gap and form a steadily changing electrical path to the Pt-electrode. This happens only locally in a small region of the gap. Additionally, the particles move and deform in direction of the electric field during the measurement. For investigation of the dynamic behaviour, 10 V DC SET-pulses with a duration of 1 s were applied and the current responses were measured at a small reading bias of 0.2 V for 15 s. This revealed a stepwise resetting to high resistance state.
In the impedance spectra, a “low frequency hook“ can be an indication for ion migration which should happen in the memristive layer with local material changes. Therefore, EIS was used here to investigate the frequency dependence of the device’s resistance and capacitance and eventually the occurrence of the “low frequency hook“.

Acknowledgments
This work was supported by the Carl Zeiss Foundation within project “MemWerk“. This work was also supported by BMBF project ForLab NSME (Grant. No. 16ES0939). Support by the Center of Micro- and Nanotechnologies (ZMN) (DFG RIsources reference: RI_00009), a DFG-funded core facility (Grant No. MU 3171/2-1 + 6-1, SCHA 632/19-1 + 27-1, HO 2284/4-1 + 12-1) of the TU Ilmenau, is gratefully acknowledged.