Estimating Spin Hall Angle in Heavy Metal/Ferromagnet Heterostructures: Methods and Pitfalls
The spin Hall angle (SHA) in a heavy metal/ferromagnet (HM/FM) bilayer is inferred from measurements such as a damping-like torque, a switching current, or a spin-pumping voltage. These measurements do not all yield the same quantity. A bulk or intrinsic spin Hall angle describes spin-current generation in the HM relative to charge current in that material. A measured torque efficiency also depends on how much spin current reaches and is absorbed by the FM. The extraction model, material parameters, and interface treatment can therefore affect reported values.
When estimating the spin Hall angle in HM/FM heterostructures, first identify the measured quantity and the current used to normalize it. Then state how the measured torque is related to spin-current generation and how interface effects are treated.
The core conversion: from observable to angle
A torque measurement can be expressed as a damping-like torque efficiency, often denoted ξ_DL. For a common convention,
[ \xi_{\mathrm{DL}} = \frac{2e}{\hbar}\frac{\mu_0 M_s t_F H_{\mathrm{DL}}}{j_{\mathrm{HM}}} ]
where (e) is the elementary charge, (\hbar) is the reduced Planck constant, (M_s) is the FM saturation magnetization, (t_F) is the FM thickness, (H_{\mathrm{DL}}) is the damping-like effective field, and (j_{\mathrm{HM}}) is the charge-current density in the HM. This expression uses (H_{\mathrm{DL}}) in A/m, with (\mu_0 H_{\mathrm{DL}}) in tesla. Definitions and signs can vary by convention, so a report should state the units and normalization used.
The measured (\xi_{\mathrm{DL}}) is a torque efficiency. It is not automatically the bulk SHA: spin diffusion, interface transmission, spin-memory loss, and other interface effects can influence the torque delivered to the FM. Converting torque efficiency to a bulk or intrinsic SHA requires a model for those effects.
For a harmonic-Hall or other torque measurement, the conversion requires:
- Determining (H_{\mathrm{DL}}) from the measured signal using the method’s model and corrections.
- Measuring or specifying (M_s) and (t_F), then calculating (\xi_{\mathrm{DL}}) with the stated convention.
- Determining (j_{\mathrm{HM}}). If the experiment provides total device current, current partition among the layers must be estimated from their thicknesses and resistivities. If (j_{\mathrm{HM}}) is already known, thickness and resistivity are not additional conversion factors.
- Applying a stated spin-transport and interface model if the goal is to infer a bulk or intrinsic SHA from the measured torque efficiency.
Current shunting matters when the measured current is the total current through the device. Its effect on the extracted value depends on how the current is normalized and on the model used to convert torque efficiency to SHA. Report the current convention and the shunting model instead of assuming a universal direction of bias.
Methods compared
| Method | What you measure | Primary extraction risk |
|---|---|---|
| Harmonic Hall / spin-torque effective field | Harmonic Hall voltage used to estimate damping-like and field-like effective fields | Thermal voltages and mixing of field-like and damping-like contributions |
| Spin-torque switching current | Current associated with magnetization switching | Dependence on the switching model and damping value; thermal activation can affect the threshold |
| Spin pumping + inverse spin Hall voltage | Generated DC voltage at resonance | Spin backflow and uncertainty in interface-related parameters |
| Second-harmonic Hall | Second-harmonic Hall voltage | Hall contributions and thermal voltages can mix with torque-related signals |
| Magnetoresistance-based torque measurement | Current-dependent magnetoresistance signal | Separation of torque-related changes from other magnetoresistive and thermal contributions |
Each method requires material parameters and a model. Parameters such as damping, spin-mixing conductance, and saturation magnetization should be measured on the sample when possible; otherwise, report which values were assumed.
Where the estimates actually go wrong
Thickness and resistivity enter through the model
HM thickness affects both current distribution and spin transport. In a spin-diffusion model, spin-current generation and transport depend on HM thickness, spin diffusion length, interface boundary conditions, and any thickness dependence of resistivity. The resulting torque efficiency need not increase linearly with HM thickness. A fit across a thickness series can test a specified model, but the fitted SHA depends on that model and its assumptions.
Thin-film resistivity can differ from a bulk value and can vary with deposition conditions. If total device current is used, resistivity affects estimates of how current is divided between the HM and FM. State the measured resistivity and explain how it enters the current normalization.
Interface transparency requires a model
An HM/FM interface can affect how much spin current is transmitted to and absorbed by the FM. Spin pumping can help estimate an effective spin-mixing conductance, but separating backflow, spin-memory loss, and other interface contributions requires a model and assumptions. A value obtained from spin pumping does not automatically equal the transparency relevant to a torque measurement. If an interface parameter is used to convert torque efficiency to SHA, describe how it was obtained and how it enters the model.
Thermal and magnetoresistive artifacts
Harmonic Hall measurements can contain thermal voltages, including contributions associated with temperature gradients in the bilayer. These signals can overlap with torque-related components. Second-harmonic Hall measurements require analysis of the relevant Hall and thermal contributions. Magnetoresistance-based approaches use different measured signals and require their own separation of torque-related and non-torque-related effects.
Reversing current or magnetization can provide useful symmetry checks, but the expected signs depend on the geometry, axis definitions, field direction, and signal channel. Thermal contributions can also share symmetries with torque-related signals. Use a measurement model suited to the experiment, describe the symmetry assumptions, and include control measurements. Current and magnetization reversal alone may not uniquely separate all contributions.
Practical checklist before you quote a number
- Measure resistivity and magnetization where possible. Report the values and the current-normalization convention.
- State the FM and HM thicknesses. Explain how current partition and thickness enter the analysis.
- Name the extracted quantity. Distinguish a torque efficiency from a bulk or intrinsic SHA.
- Report the extraction model. Include its spin-diffusion and interface assumptions.
- Describe interface parameters. State whether they were measured or assumed and how they were used.
- Explain artifact controls. Report the measurements and assumptions used to assess thermal and magnetoresistive contributions.
- State the measurement temperature. Temperature can affect material parameters and measurement conditions, so comparisons across temperatures should account for those changes.
A reported SHA depends on what was measured and how the result was converted. Two studies can report different values because they use different observables, current normalizations, or spin-transport models. Check those choices before comparing the numbers.
Key Takeaways
- A torque measurement yields a torque efficiency under a stated convention. Inferring a bulk or intrinsic SHA requires a model for spin transport and interface effects.
- Current shunting matters when the measured current is shared among device layers. Its effect depends on the current normalization and extraction model.
- Harmonic Hall, switching, spin-pumping, and magnetoresistance-based methods measure different signals and have distinct sources of uncertainty.
- A reproducible report states the current convention, measured parameters, thicknesses, extraction model, interface assumptions, temperature, and artifact controls.
