Ultra‐Low Power Multilevel Switching with Enhanced Uniformity in Forming Free TiO_2−x‐Based RRAM with Embedded Pt Nanocrystals

Abstract

Accurate control over the various resistance states is highly desired in order to attain reliable multilevel memory performance. However, due to the inherent random nature of oxygen vacancy creation, serious variability issues may arise. In this work, we demonstrate promising multilevel capability with ultra‐low power consumption in the range of nW, using sub‐200 nA operating current in a 45 nm TiO2−x‐based resistive random access memory (RRAM) with embedded small (≈5 nm in diameter) Pt nanocrystals (NCs). As the resistance values present a strong dependence on both the oxygen vacancy density and the diameter of the conducting filament (CF), hence the degree of the variability (or uniformity) can be improved by controlling one of these two parameters. It is shown here that the presence of NCs allows for the generation of multiple and dense CFs due to the concentrated electric field around the NCs which, in turn, enhance the consecutive cycling (temporal) uniformity, even at very low operating power conditions.

Resistive Random Access Memory (RRAM): an Overview of Materials, Switching Mechanism, Performance, Multilevel Cell (mlc) Storage, Modeling, and Applications

“…The resistance switching phenomenon has been observed in a variety of oxides, but binary metal oxides have been extensively studied as a preferred switching material for future non-volatile memory applications primarily due to their compatibility with the CMOS BEOL processing. Various metal-oxide-based materials exhibiting the non-volatile resistance switching such as hafnium oxide (HfO x ) [18][19][20][21][22][23], titanium oxide (TiO x ) [24][25][26][27][28][29][30][31], tantalum oxide (TaO x ) [32][33][34][35][36], nickel oxide (NiO) [37][38][39][40], zinc oxide (ZnO) [41][42][43][44][45][46], zinc titanate (Zn 2 TiO 4 ) [47], manganese oxide (MnO x ) [48,49], magnesium oxide (MgO) [50], aluminum oxide (AlO x ) [51][52][53], and zirconium dioxide (ZrO 2 ) [54][55][56][57][58] have drawn the most attention and have been studied extensively during the past several years. These metal oxides are deposited usually by pulse laser deposition (PLD), atomic layer deposition (ALD), and reactive sput- tering.…”

“…The maximum temperature reached into CF at x = 0, the middle of the filament is given in Eq. (27) and the equivalent electrical conductivity in the work area (σ eq ) is given in Eq. (28).…”

Tolerance against conducting filament formation in nanosheet-derived titania thin films

“…In addition, these films show poor switching as shown in figure 3(c), where the incomplete switching of the initial plot may be attributed to the incomplete movement of oxygen vacancies due to the short duration of the switching voltage application. On the contrary, Au-contacted titania thin films formed by other methods have been reported to show filamentary resistive switching with high switching ratios [29][30][31][32][33][34]. A plausible explanation of this difference is a lower expected density of oxygen vacancies in nanosheet-derived films.…”

Impact of Line Edge Roughness on ReRAM Uniformity and Scaling

“…Up to now, it has been mainly related to the stochastic nature of conducting a filament creation process. Remedies that have been proposed consider (a) engineering of electrode/oxide interface [12], (b) inserting seeds (nanoparticles) in the oxide bulk [13], and (c) reduction of the size of cell area [14]. The latter is related to the scaling trends of ReRAM devices and is a great advantage of this technology since scaling down the dimensions of a ReRAM cell goes with the increase in performance uniformity due to the mitigation of stochastic filament formation.…”