Publication
MMM 2024
Talk

Spin transmission through NiO in Pt/NiO/NM/Py heterostructure

Abstract

Efficient operation of spin orbit torque magnetic random-access memories (SOT-MRAM) requires separation of the spin- and charge-current pathways [1] [2]. NiO is an electrically insulating, but spin conducting material. By interfacing it with a heavy metal spin current generator (Pt in this case), one can confine the charge current flow to Pt, while transmitting spin current through NiO into the adjacent layer [3] [4] [5] [6] [7] [8] [9]. We use d.c. bias spin torque ferromagnetic resonance (ST-FMR) on 600 nm × 1800 nm nano-bridge patterned from Pt/NiO/NM/Py (Py=permalloy) thin film stack to evaluate the spin current received by the Py layer. Here the non-magnetic metal (NM)=Cu,Ag, … is a layer for spin conduction but disrupts direct exchange coupling between the ferromagnetic detector (Py) and the antiferromagnetic NiO. The NiO thickness is varied while all other layers are kept at constant thickness. The samples considered are Ta(5) |Pt(50) |NiO (x=0,12.5,15,20,25) |NM(20) |NiFe(50) |MgO(15) |TaN(20) and, Ta(5) |Jn(20)|NiFe(50) |MgO(15) |TaN(20) where the numbers indicate thickness in angstroms. These samples are named S1 to S5 and S0 respectively. We observe robust spin transmission (>40%) even at 25 Å NiO thickness. The spin-transmissivity’s NiO thickness dependence is nearly exponential for samples with NM. Shown in Fig.1(a-b) are its SOT channel current dependence-based spin-transmissivity vs NiO thickness. For samples with direct NiO/Py interface however, a faster NiO-thickness dependence is seen. An insertion of a well grown NM between NiO and Py increases the decay length compared to Pt/NiO/Py (Figure 1(b)). These experiments provide an existence proof of a materials stack that allows charge- and spin-current separation without direct exchange coupling with the ferromagnet.