Resistive Random Access Memory (RRAM) has been widely studied as a promising next-generation memory with advantageous characteristics such as high-speed and low-power operation^[1]. In particular, the SiO₂/Cu conductive-bridge random access memory (CBRAM), where electric-field-driven formation/dissolution of Cu metal filaments provides resistance switching, has been drawing much attention because of its compatibility with the backend interconnect process. Recently, we have demonstrated negative photo conductivity (NPC) or light-induced resistance reset (LIR) in the SiO₂/Cu stack by disrupting the electrically-formed filamentary breakdown path^[2]. It is considered that the NPC is caused by recombination of light-excited oxygen ions and oxygen vacancies created in the breakdown path during electrical forming. Although the exact mechanism is still elusive, the finding suggests a potential to realize an electro-optical control of memory devices.

In our pursuit of the mechanism(s) behind the abovementioned NPC phenomenon, we reveal here an impact on NPC by Ar plasma surface treatment, which is reported to increase oxygen vacancies in the oxide electrolyte^[3]. The effect of light illumination on the electrical property of the Ar-plasma-treated Cu sample (SiO₂ 10nm/Cu) and control sample (SiO₂ 5nm/Si) is investigated via ultra-high vacuum conductive atomic force microscope or CAFM (Fig. 1). During forming, a negative bias voltage is supplied to the probe with a preset current compliance of 1.1 nA. The area of the probe-oxide contact is estimated to be ~26nm^2[4].

From the current-voltage (I-V) curves for the Cu or control sample depicted in Fig. 2, it is evident that the switching voltage from the high resistance state (HRS) to the low resistance state (LRS) in the Cu sample is shifted from -6V to -4V by 33% after the Ar plasma treatment, while the breakdown voltage in the control sample is almost unchanged. Therefore, the electrical formation of Cu filaments is considered to be accelerated by Ar plasma treatment.

Fig. 3 shows typical response to light in the post-forming Cu sample and percentage of NPC occurrence calculated from 30 tested locations in each of the Cu and control sample. The light intensity used was 1 mW/cm². Clearly, Ar plasma treatment suppresses the NPC in both samples and especially enlarges the proportion of no-NPC cases in the Cu sample.

The FT-IR spectra (Fig. 4) of the pre- and post-plasma-treated SiO₂ film show the intensities of HO-H stretching (3300cm^-1) and SiO-H stretching (3650cm^-1) become higher after the Ar plasma treatment. This rise implies the destruction of the Si-O network near the surface and the increase of various defects like Si dangling bonds and oxygen vacancies where water in the air may be adsorbed.

Based on the results, we propose a possible explanation for the mechanism of the lower switching voltage in the Cu sample and the decreased percentage of NPC occurrence with Ar plasma treatment in both samples. For the control sample, Ar plasma treatment may sputter oxygen ions and create more oxygen vacancies near the surface of SiO₂^[3], which helps forming a larger conducting filament composed of oxygen vacancies. This can be confirmed by a higher current in the reverse I-V curve after Ar plasma exposure (Fig. 2 (c)(d)). Since oxygen ions are depleted, the NPC response (due to recombination of light-excited oxygen ions and vacancies) is suppressed. In the case of the Cu sample, the oxygen vacancies created in the surface of SiO₂ by Ar sputtering may adsorb water molecules. The resultant hydroxyl groups (OH^-) reportedly promote Cu ion (Cu⁺) diffusion into SiO₂^[5] and reduce the switching voltage during forming (Fig. 5). When more Cu ions migrate into SiO₂, part of the conducting filament made up of oxygen vacancies there is replaced by a Cu filament (Fig. 5 (e)), which is not readily disrupted due to the stronger metal-metal bond (which requires higher energy for cleavage than that of the light used^[2]). Therefore, the conducting path composed of more Cu metal filaments remain after light illumination (Fig.5 (f)), leading to the suppression of the NPC response.

References: [1] Y. Li et al., Small, vol. 13, issue 35, 1604306, 2017; [2] T. Kawashima et al., Appl. Phys. Lett., vol. 111, 113505, 2017; [3] Y. Shuai et al., IEEE Elec. Dev. Lett., vol. 34, no. 1, pp. 54-56, 2013; [4] B. Cappella and G. Dietler, Surf. Sci. Rep., vol. 34, no. 1-3, pp. 1-3, 5-104, 1999; [5] T. Tsuruoka et al., Adv. Funct. Mater., vol. 22, no. 1, pp. 70-77, 2012;