In a magnetic tunnel junction, thin material layers and their interfaces can determine how electrons with different spin orientations contribute to current. Research on WS2 spin filtering in vertical spin valves demonstrates how strongly that behavior can depend on the thickness of an atomically thin semiconductor. In devices incorporating chemical vapor deposited (CVD) tungsten disulfide, a monolayer produced a positive spin polarization, while bilayer and trilayer devices showed a reversed signal.
The result is relevant to engineers and technical decision-makers evaluating two-dimensional materials for electronic devices: layer count is not merely a geometric parameter. In this system, it was associated with a change in the WS2 band structure and a different spin-selection mechanism at the cobalt interface.
How the WS2 magnetic tunnel junction was configured
A magnetic tunnel junction (MTJ) uses two ferromagnetic electrodes separated by a thin barrier. One electrode can act as a spin source, while the other serves as a spin analyzer. The measured resistance depends on whether the magnetic orientations of the electrodes are parallel or antiparallel. This difference, known as tunneling magnetoresistance (TMR), provides a way to investigate spin-dependent transport through the junction.
The devices in the study used a Co/Al2O3/WS2/Co stack. The aluminum oxide layer formed a tunnel barrier in the reference Co/Al2O3 analyzer, and a second cobalt electrode completed the vertical structure. CVD-grown WS2 flakes were transferred onto the analyzer and characterized using Raman spectroscopy, atomic force microscopy (AFM), and photoluminescence (PL). These measurements helped distinguish monolayer, bilayer, and trilayer flakes before their electrical behavior was assessed.
This device approach matters because measuring a spin signal requires both a suitable structure and a reference for interpreting it. With a known polarization for the analyzer, the researchers used the TMR measurements to estimate the spin polarization associated with the WS2/cobalt interface.
WS2 spin filtering changes with layer thickness
At 4 K, the monolayer WS2 junction showed a TMR signal of about 8%. Using the reference analyzer and a phenomenological Julliere model, the study estimated a Co/WS2 interfacial spin polarization of +12%. The reported value was higher than previous results cited by the authors for transition-metal dichalcogenide-based MTJs.
The sign changed when the WS2 had more than one layer. The estimated interfacial polarizations were −6% for bilayer WS2 and −0.5% for trilayer WS2. In other words, the measured spin signal reversed between monolayer and multilayer devices, while its magnitude also declined as thickness increased.
The measurements do not mean that adding a layer simply reduces an otherwise unchanged signal. The reversal points to a change in which spin-dependent electronic states contribute most strongly to transport across the interface.
Band structure offers an explanation
WS2 changes its electronic band structure as its thickness increases. In the study, the monolayer was associated with a direct bandgap, with the relevant band-edge transition at the K-points of the Brillouin zone. Bilayer and thicker WS2 instead have an indirect bandgap: the valence-band maximum is at the Γ-point, while the conduction-band minimum lies between Γ and K, near the Q-point.
That shift matters because electron transport through the interface depends in part on crystal momentum—the wave-like character associated with an electron’s motion through a crystal. The authors’ model compares the momentum-space regions available in cobalt with the conduction-band minima of WS2. For a monolayer, the favored transport regions align with cobalt states dominated by majority-spin carriers. For a bilayer, the shift toward Q selects regions where minority-spin carriers dominate. This provides a possible explanation for the observed change in polarization sign.
Calculations support this interpretation, but the model is not a complete description of every interface. The detailed momentum-space alignment can depend on interface structure, relative crystal orientation, and strain. The neighboring ferromagnet may also affect the WS2 electronic structure. The proposed mechanism should therefore be understood as an interpretation of the experimental trend, rather than a universal prediction for every WS2 junction.
What the results mean for device engineering
For developers of 2D spintronic devices, the findings highlight several practical considerations:
- Verify layer count: Optical appearance alone is not sufficient to establish thickness. The study combined Raman spectroscopy and AFM, with PL providing information about direct and indirect bandgap behavior.
- Control and protect interfaces: The spin signal depends on the electronic states at the ferromagnet interfaces. Oxidation or other changes to those interfaces may affect device behavior, making fabrication and transfer choices important.
- Measure signal direction as well as magnitude: A reversed TMR signal indicates a change in spin polarization, not simply a weaker version of the monolayer response.
- Account for transport pathways: The authors discuss in-plane current diffusion within TMDC layers as a potential contributor to spin depolarization in thicker flakes, alongside possible spin scattering.
These considerations are important when comparing devices or designing experiments: material thickness, interface preparation, and measurement configuration all influence how a spin-transport result should be interpreted.
Conclusion
CVD-grown WS2 in vertical magnetic tunnel junctions exhibited a clear thickness-dependent spin-filtering effect. The monolayer produced an estimated positive interfacial polarization of 12%, while bilayer and trilayer devices showed negative values. The proposed explanation links this reversal to the transition from a direct to an indirect bandgap and the associated change in momentum-selective transport.
For practical device development, the central lesson is that atomically precise thickness and interface control can shape both the magnitude and direction of spin polarization. WS2 therefore offers a way to investigate tunable spin transport, while also underscoring the need to characterize each layer stack and treat theoretical spin-selection models with appropriate attention to interface-specific effects.