Electric-field-induced submicrosecond resistive switching

Abstract

Electric-field-induced resistive switching in metal-oxide interfaces has attracted extensive recent interest. While many agree that lattice defects play a key role, details of the physical processes are far from clear. There is debate, for example, regarding whether the electromigration of pre-existing point defects or the field-created larger lattice defects dominates the switch. We investigate several Ag-Pr 0.7 Ca 0.3 MnO 3 samples exhibiting either submicrosecond fast switching or slow quasistatic dc switching. It is found that the carrier trapping potentials are very different for the pre-existing point defects associated with doping ͑and/or electromigration͒ and for the defects responsible for the submicrosecond fast switching. Creation/removal of the defects with more severe lattice distortions and spatial spreading ͑trapping potential Ն0.35 eV͒, therefore, should be the dominating mechanism during submicrosecond switching. On the other hand, the shallow defects ͑trapping potential Ӷ0.2 eV͒ associated with doping/annealing are most likely responsible for the resistance hysteresis ͑slow switch͒ during quasistatic voltage sweep.

Kinetics and relaxation of electroresistance in transition metal oxides: Model for resistive switching

“…It is interesting to note that switching can even be observed under quasistatic I-V sweeps for the samples never showing noticeable switching under 1 -2 s pulses up to 100 V. 7 Sample B, for example, has shown no noticeable switching with T w Յ 10 s and V 0 Յ 100 V, i.e., a horizontal ⌬R͑T W ͒ = 0 line in a ⌬R vs T W plot similar to Fig. 4.…”

“…Many probes have been used to explore the issue, but the picture is still unclear at this moment. Recent studies also have hinted 7 that there might be no single unified mechanism even for the very limited cases of the metal-PCMO interfaces alone.…”

“…8 It confirms that such sub-s switching has an extremely large ratio of R / S on the order of 10 15 ͑ S is fast switching time ϳ2 ϫ 10 −7 s͒, which is far larger than that allowed for migration/diffusion models under the NE constraint. 7 It is also interesting to note that such a slow rate of decay ͑long retention time͒ will imply an energy barrier on the order of kT ln͓v 0 . R ͔Ϸa few eV, a value more suitable for the process of bonds breaking, 14 where v 0 Ϸ 10 10 / s is the characteristic frequency of ion oscillations.…”

“…6 The situation may be further complicated by the fact that different samples ͑even with similar materials/configurations͒ show rather different switching/transport properties. 7 For example, it has been demonstrated that oxygen electromigration may be responsible for the switching of single-crystal SrTiO 3 under quasistatic I-V sweep. 3 Similar migration/ diffusion of oxygen ion/vacancy, 8,9 however, can hardly be accommodated with the nanosecond switching observed across metal-oxide interfaces.…”

“…3,5 As previously discussed, however, such models may have difficulty to ac- commodate the NE correlation in the case of sub-s switching. 7 This correlation directly relates the defect migration rate under field with its thermal diffusion rate based on the assumption that both can be simplified as the same random-walk process of the pre-existing defects across the lattice ͑as an array of barriers͒; a reasonable assumption if the energy gain during each walk step is low enough to avoid irreversible changes in the lattice. The two rates, therefore, differ merely on the driving forces: concentration gradient, kT⌬n / ⌬x in diffusion but an effective nZeE in electromigration, where n, ⌬n / ⌬x, e, E, and Z are the defect density, its gradient, the electronic charge, the local electric field, and an effective charge number, respectively.…”