Access Official Technical PDF

Share Summary:LinkedInEmail

The photoinduced inverse spin Hall effect offers a way to investigate electron spin transport in semiconductors using optical spin injection and electrical detection. In Pt/semiconductor junctions, circularly polarized light creates spin-polarized electrons in the semiconductor. Some of those electrons reach a thin platinum layer, where the inverse spin Hall effect (ISHE) converts spin-current information into a measurable voltage. How that voltage changes with photon energy can provide information about the semiconductor’s electron spin diffusion length.

A one-dimensional diffusion model applied to Pt/GaAs, Pt/Ge and Pt/Si junctions shows how optical absorption and spin transport jointly shape the measured signal. The approach is useful because the three materials have different spin-generation and transport mechanisms, yet the same basic model can describe their ISHE spectra.

From optical spin injection to an electrical signal

Circularly polarized light can generate an imbalance between electron spin populations through optical transition selection rules. In GaAs and Ge, transitions near the direct band gap can produce substantial spin polarization. In silicon, optical spin injection is possible at its indirect gap through phonon-assisted transitions, but the initial polarization is much smaller.

The photogenerated electrons move through the semiconductor toward the Pt interface. When a spin current enters platinum, its strong spin-orbit interaction enables conversion into a transverse charge response through the ISHE. Electrodes on the Pt layer measure the resulting voltage. In the described geometry, the detected signal depends on the spin current reaching the interface and on the orientation of the spin polarization relative to the measurement direction.

This makes the voltage more than an indicator of whether optical spin injection occurred. Its photon-energy dependence reflects the relationship between where electrons are generated, how polarized they are, and whether they retain their spin while traveling to the interface.

Why photon energy reveals spin diffusion

Light is absorbed as it travels into a semiconductor, so photon energy affects the depth distribution of photogenerated electrons. The absorption coefficient, α, describes how rapidly the light intensity falls with depth; its inverse, 1/α, is the absorption length. The electron spin diffusion length, Ls, describes the characteristic distance over which spin information survives during diffusive transport.

If electrons are generated much farther from the Pt interface than they can diffuse while retaining spin, relatively few contribute effectively to the spin current at the boundary. If the absorption length becomes comparable to or shorter than Ls, more of the generated spins can reach the interface before depolarizing. The balance changes with photon energy because both optical absorption and the initial electron spin polarization vary across the spectrum.

A compact diffusion model for Pt/semiconductor junctions

The model treats spin transport as one-dimensional, perpendicular to the semiconductor surface, and assumes steady-state diffusion. It relates the spin current at the interface to the photon flux, the initial spin polarization, the absorption coefficient and Ls. In simplified form, the energy-dependent part of the interface spin current is proportional to:

Photon flux × spin polarization × αLs / (1 + αLs)

This expression highlights two competing influences. The initial spin polarization determines how much spin is created, while the factor involving α and Ls captures the likelihood that spin-polarized electrons reach the interface. The measured voltage is proportional to the interfacial spin current in the experimental arrangement.

A real metal-semiconductor interface also includes a Schottky barrier. The model discussion allows for finite transfer and recombination at the surface; under the conditions considered, these effects change the predicted signal’s magnitude through a prefactor without changing its photon-energy dependence. Consequently, fitting normalized spectra can estimate Ls without treating the absolute voltage as a direct measure of spin-current magnitude.

Different spectral responses in GaAs, Ge and Si

GaAs and silicon

For GaAs and Si, the signal is relatively small near the relevant absorption edge and increases at higher photon energies. Near the edge, the absorption length is much greater than the fitted spin diffusion length. Many electrons are therefore generated too far from the Pt interface to arrive with their spin information intact. At higher photon energies, the absorption length approaches Ls, improving the contribution from electrons that reach the interface, even though the initial spin polarization may be lower.

Fitting the spectra gives Ls = 30 ± 5 nm for GaAs and Ls = 9 ± 2 µm for Si. These values are estimates within the assumptions of the diffusion model, rather than measurements independent of those assumptions.

Germanium

In Ge, the ISHE spectrum more closely follows the initial electron spin polarization. The absorption length is no greater than the spin diffusion length across the measured energy range, so the model is less sensitive to the precise value of Ls. The analysis therefore supports a lower-bound estimate of about 1 µm, rather than a precise value.

Spin generation and transport also differ among the materials. In GaAs, spin-polarized electrons relax and diffuse around the Γ valley. In Ge, optical spin orientation occurs near Γ before rapid scattering to the L valleys, where transport occurs. In Si, electrons are generated and diffuse around the Δ valleys. These distinctions matter when interpreting a fitted spectrum, particularly because the simplified model assumes that the initial spin polarization is retained after carrier relaxation.

Practical interpretation and limitations

For engineers and technical decision-makers assessing spin-based devices, the method illustrates how a spectroscopic measurement can probe a transport property that is otherwise difficult to infer from a single electrical reading. Comparing the photon-energy response with known optical spin polarization and absorption behavior helps distinguish a weak signal caused by low spin generation from one limited by spin loss during transport.

The model has important boundaries. It assumes diffusion rather than drift, reduces transport to one dimension, and focuses on spin-polarized electrons rather than holes. The analysis indicates that diffusion is the dominant contribution for the systems considered, while acknowledging that a Schottky-barrier field may influence transport. At photon energies well above a semiconductor’s gap, scattering of hot electrons may alter spin polarization; the simplified treatment does not include the relevant energy-dependent spin-relaxation processes.

Conclusion

Photon-energy-dependent photoinduced ISHE measurements connect optical absorption and spin polarization to the transport of electrons toward a Pt interface. A simple diffusion model reproduces the reported spectral behavior across Pt/GaAs, Pt/Ge and Pt/Si and provides useful spin diffusion length estimates where the spectra are sufficiently sensitive. The practical implication is that optical excitation can serve as a probe of semiconductor spin transport, provided that fitted values are interpreted in light of interface effects, material-specific band structure and the model’s assumptions.

Access Official Technical PDF

About The Author