Abstract
Conventional giant magnetoresistance (GMR) in spin valves is current-independent, so the resistance of a device depends only on the relative orientation of the magnetic layers. In dual spin valves consisting of three ferromagnetic (FM) layers separated by nonmagnetic (NM) spacers (i.e., a FM1/NM/FM2/NM/FM1), GMR can be current-dependent if spin can accumulate in FM2 when outer FM1 layers are aligned antiparallel. Currently the underlying physics is poorly understood, although spin accumulation in FM2 is likely to depend on the gradient in the density of states at the Fermi energy of the ferromagnet. To investigate this hypothesis, we have measured a series of dual spin valves with Ni$_{1-x}$Fe$_x$ as FM2 layers of varying composition. We show that both the magnitude and sign of the nonlinear GMR depend strongly on the Fe content and thus on the band structure of the ferromagnet FM2.
Introduction
In a standard spin valve, electrical resistance changes based on whether two magnetic layers are aligned parallel or antiparallel. This effect, giant magnetoresistance (GMR), is the cornerstone of modern hard drive technology. Crucially, in these conventional devices, the resistance is independent of the bias current.
However, in "dual spin valves" (DSVs)—which incorporate a third magnetic layer—the resistance becomes nonlinear, meaning it changes as the current increases. While this effect was previously observed, the underlying physical mechanism remained elusive. Researchers hypothesized that the effect stems from "spin accumulation" in the central layer, specifically linked to the gradient of the Density of States (DOS) at the Fermi energy. This study provides definitive evidence for this link by systematically varying the nickel-iron ($\text{Ni}_{1-x}\text{Fe}_x$) composition of the central layer.
Experimental Methods
The researchers fabricated nanopillar dual spin valves with a specific stack sequence: $\text{Cu} / \text{Co}_{90}\text{Fe}_{10} / \text{Cu} / \text{Ni}_{1-x}\text{Fe}_x / \text{Cu} / \text{Co}_{90}\text{Fe}_{10} / \text{IrMn} / \text{Cu}$. The $\text{Co}_{90}\text{Fe}_{10}$ outer layers act as fixed magnets, while the central $\text{Ni}_{1-x}\text{Fe}_x$ layer (FM2) is the active region where composition $x$ was varied.
The devices were fabricated using sputter deposition followed by optical lithography and Focused Ion Beam (FIB) milling to create nanopillars approximately $250 \times 250 \text{ nm}^2$ in size. Electrical transport was characterized by measuring differential resistance as a function of DC bias current ($I_{dc}$) using a lock-in amplifier. To ensure the observed effects were purely electronic and not due to current-induced Oersted fields, finite-element simulations were performed.
Results and Discussion
The study focused on four primary alloy compositions for the middle layer: $\text{Ni}_{70}\text{Fe}_{30}$, $\text{Ni}_{80}\text{Fe}_{20}$ (Permalloy), $\text{Ni}_{85}\text{Fe}_{15}$, and $\text{Ni}_{90}\text{Fe}_{10}$.
The most significant finding is the direct dependence of nonlinear GMR on the DOS gradient ($dg(E)/dE$). When current flows, it shifts the chemical potential of the spins. If the DOS has a steep slope at the Fermi level, this shift causes a significant change in scattering, manifesting as a change in resistance.
As shown in Figure 5, the researchers observed a "sign reversal" in the nonlinear GMR. While most $\text{NiFe}$ alloys exhibited a decrease in resistance with increasing current, the $\text{Ni}_{90}\text{Fe}_{10}$ alloy showed an increase. This reversal corresponds to the Fermi level crossing a sharp peak in the minority spin subband of the $\text{NiFe}$ band structure. By defining a parameter $\alpha$ to represent the magnitude of the nonlinear effect, they found that $\text{Ni}_{80}\text{Fe}_{20}$ showed the highest magnitude, while $\text{Ni}_{90}\text{Fe}_{10}$ was the only composition to yield a positive value.
Conclusion
This work successfully links the phenomenological parameters of the Valet-Fert theory to the physical band structure of ferromagnetic alloys. While the effect is currently limited by the spin resistivity mismatch between the $\text{NiFe}$ layers and the Copper spacers, it demonstrates a powerful method for probing the electronic properties of materials within functional devices. Future enhancements, such as using insulating barriers to create Magnetic Tunnel Junctions (MTJs), could significantly amplify spin accumulation, paving the way for purely electronic control of magnetic states.
Summary updated