Optical spin pumping in silicon

Technical Summary: Optical Spin Pumping in Silicon

Problem Statement
The generation of out-of-equilibrium spin-polarized carrier populations is a fundamental requirement for quantum technologies and spintronics. While "optical spin orientation"—the generation of spin polarization via circularly polarized light—is an established technique for direct band-gap semiconductors (e.g., GaAs, GaSb), it has proven ineffective for indirect band-gap materials like silicon (Si). In Si, the weak oscillator strengths of phonon-mediated optical transitions result in two critical limitations: (i) a negligible spin-injection rate that washes out the average electron spin, and (ii) an electron lifetime in the indirect valleys that is significantly longer than the spin relaxation time. Consequently, spin relaxation occurs before radiative recombination, yielding unpolarized photoluminescence (PL). Previous attempts to measure spin properties in Si using all-optical methods have been limited by a fundamental detection floor of approximately $10^{-4}$ for the degree of polarization under continuous-wave excitation.

Methodology
To circumvent the intrinsic limitations of direct optical excitation in Si, the authors propose an all-optical analog of the "spin pumping" technique. Instead of directly exciting Si, they utilize a Ge-on-Si heterostructure where a Germanium (Ge) epilayer acts as a spin injector and Si acts as a spin sink.

1. System Design: A p-type Ge film (initially 1.3 µm thick) is epitaxially grown on an n-type Si substrate. The band alignment at the Ge/Si interface is Type-II, facilitating the transfer of electrons from Ge to Si while confining holes in the Ge layer.
2. Excitation: The sample is excited with circularly polarized light (1.165 eV) resonant with the direct band gap of Ge. This optically orients the spins of electrons excited from the heavy-hole (HH) and light-hole (LH) bands of Ge.
3. Spin Transfer: These spin-polarized, high-energy electrons diffuse across the heterojunction into the Si substrate. Due to the favorable band lineup, they are funneled into the $\Delta$-valley minimum of Si, effectively thermalizing there.
4. Progressive Etching: To isolate the mechanism, the Ge absorbing layer was progressively thinned via selective wet etching (down to 0 µm) while monitoring the PL polarization.
5. Characterization: The study employed low-temperature (4 K) PL spectroscopy, polarization analysis (using a retarder and polarizer), and magneto-optical Hanle effect measurements to quantify the degree of circular polarization and spin lifetime.

Key Contributions and Results

• Observation of Polarized Emission: The study reports the observation of circularly polarized luminescence from the Si substrate with a degree of polarization ($\rho$) reaching 9%. This value represents an improvement of nearly five orders of magnitude over the limits of conventional direct excitation in Si.
• Mechanism Validation via Etching:
  • In the pristine (thick Ge) sample, the Si emission showed negligible polarization.
  • As the Ge layer was thinned to approximately 0.55 µm, a sinusoidal modulation of the PL intensity appeared.
  • At a thickness of 0.05 µm, the polarization reached its maximum (9%). The sign of the polarization was opposite to that of the Ge direct-gap emission, consistent with the transfer of spin angular momentum from HH/LH-excited electrons in Ge to Si.
  • Upon complete removal of the Ge layer, the polarization dropped significantly but remained non-zero (approx. -2%), suggesting a residual contribution from other mechanisms.
• Role of Defects and Carrier Lifetime: Magneto-optical Hanle effect measurements revealed an extremely short carrier lifetime in the Si near the interface ($\tau \sim 200$ ps), several orders of magnitude shorter than in bulk Si. The authors attribute this to the presence of extended defects (dislocations) at the heteroepitaxial interface, which act as efficient non-radiative recombination centers.
  • Significance of Short Lifetime: The authors argue that this ultrafast recombination is crucial. It allows the spin-polarized electrons to recombine radiatively before they lose their spin orientation via relaxation, thereby preserving the high degree of polarization.
• Defect Identification: Low-energy PL spectra revealed a peak at 0.82 eV, identified as the D1 line associated with dislocations. This confirms that extended defects penetrate from the Ge layer into the Si substrate, providing the necessary non-radiative channels to shorten the carrier lifetime.
• Theoretical Modeling: Drift-diffusion simulations and tight-binding calculations supported the experimental findings. The models confirmed that spin-polarized electrons generated in Ge can diffuse into Si and that the Type-II band alignment facilitates this injection without external bias. The simulations estimated a maximum spin polarization transfer of $\approx 30\%$ from Ge to Si, which, when combined with the selection rules for phonon-assisted recombination in Si, aligns with the observed 9% polarization.

Significance and Claims
The paper claims to demonstrate a viable strategy for injecting and detecting spin-polarized carriers in silicon, a material whose optical spin exploitation has historically been hindered by its indirect band structure. By mimicking spin pumping optically, the authors show that:

1. Spin Injection is Feasible: Spin-polarized carriers can be generated in a direct-gap absorber (Ge) and effectively transferred to an indirect-gap semiconductor (Si).
2. Defects as Enablers: Contrary to the usual view of defects as detrimental, the extended defects at the Ge/Si interface play a constructive role by shortening the carrier lifetime, thereby preventing spin relaxation and enabling the observation of polarized emission.
3. Overcoming Fundamental Limits: The approach bypasses the fundamental detection limits of direct optical spin orientation in Si, achieving a polarization degree (9%) that is practically significant for spintronic applications.

The authors conclude that this method opens new research directions for harnessing spin-dependent phenomena in technologically relevant materials like silicon, potentially enabling the integration of spintronics and quantum technologies with standard electronic and photonic circuits. They suggest that further optimization, such as using low-dimensional Ge structures (e.g., quantum wells) to lift HH/LH degeneracy, could potentially increase the polarization degree further.