Plasmon-to-exciton spin conversion in semiconductor-metal hybrid nanostructures

Optical spin control is the basis for ultrafast spintronics: circularly polarized light in combination with spin-orbit coupling enables spin manipulation of electronic states in condensed matter. However, the conventional approach is limited to longitudinal spin initialization along one particular axis that is dictated by the direction of light propagation. Here, plasmonics opens new possibilities, allowing one to tailor light polarization at the nanoscale. We demonstrate ultrafast optical excitation of electron spin on femtosecond timescales via plasmon-to-exciton spin conversion. By time resolving the THz spin dynamics in a hybrid (Cd,Mn)Te quantum-well structure covered with a metallic grating, we unambiguously determine the orientation of the photoexcited electron spins which is locked to the propagation direction of the optically excited surface plasmon polaritons. Using the spin of the incident photons as an additional degree of freedom, one can adjust not only the longitudinal, but also the transverse electron spin components normal to the light propagation at will.

Figure 1

Optical orientation of electron spin. Conventional optical orientation of electron spin $\mathit{S}$ along the (a) $z$ or (b) $y$ axes with circularly polarized light. (c) Optical orientation of electron spin along the $y$ axis via surface plasmon polaritons (SPPs) with transverse spin, which are launched by linearly $p$-polarized light. The direction of the propagating SPP is indicated with a yellow arrow, while red arrows pointing towards the cycloid indicate the rotation of the SPP electric-field vector ${\mathit{E}}_{\mathit{p}}$.

Figure 2

Selection rules for optical transitions. Schematic presentation of optical transitions and selection rules for excitation with circularly polarized light in a semiconductor quantum well. The valence-band states with angular momentum projection of $\pm 3/2$ (heavy holes) and $\pm 1/2$ (light holes) are split due to confinement along the $z$ direction. The conduction-band states have spin projections of $\pm 1/2$. Arrows show excitation with ${\sigma }^{+}$ circularly polarized light incident along the (a) $z$ axis and (b) $y$ axis. For excitation along the $y$ axis, the rotated basis of eigenstates $|\pm \widetilde{1/2}\rangle , |\pm \widetilde{3/2}\rangle$ with the corresponding spin projections along the $y$ axis is used (for details, see the main text). Numbers near arrows indicate the relative absorption probability, normalized in the same way for all transitions.

Figure 3

Experimental realization and spin transients. (a) Scheme of pump-probe experiment (O is objective; PBS is polarization beam splitter; PD is photodiode; and LI is lock-in amplifier). The pump and probe beams are incident in the horizontal and vertical planes, respectively. Pump-probe transients measured for (b) bare QW and (c) hybrid plasmonic structures using circularly polarized and $p$-polarized pump, respectively. Experimental data are shown with dots. Solid lines are fits with damped oscillations.

Figure 4

Optical orientation in hybrid structure for different polarization states of pump excitation. The polarization of the pump excitation pulse and resulting spin orientation ${\mathit{S}}_{\mathit{0}}$ are shown on (a) Poincaré and (b) Bloch spheres; the ${\xi }_i$ are the Stokes parameters. Colored lines and dots correspond to theory and experiment, respectively. (c), (d) Initial amplitude ${\theta }_{K0}$ and phase $\phi$ of Kerr rotation transients in a hybrid plasmonic structure plotted as a function of angle $\alpha$ defining the polarization state of the pump pulse as shown in (a). Data are presented for magnetic fields $B=\pm 620\mathrm{mT}$.

Figure 5

Magnetic field dependences of the (a) Larmor precession frequency ${\nu }_L$ and (b) amplitude ${\theta }_{K0}$ for the bare semiconductor QW (black circles) and hybrid structure using $p$ (blue squares) and ${\sigma }^{+}$ (red triangles) polarized excitation. $T=2\mathrm{K}$. All data are taken with a spectrally broad laser without amplitude shaping. For the bare QW structure, no bandpass filter has been used in detection.

Figure 6

Pump-probe transient signals for different directions of the polarization plane of the probe beam. V and H correspond to $p$- and $s$-polarized probe beams, respectively, while the D is the mixture of both. The pump beam is $p$ polarized. Temperature $T=2\mathrm{K}$.

Figure 7

Excitation with linearly polarized pump beam. The polarization plane of the pump is defined by the angle $\alpha$, which is taken with respect to the $x$ axis. (a), (b) The dependence of signal amplitude ${\theta }_{K0}\left(\alpha \right)$ and phase $\phi \left(\alpha \right)$, respectively. Dots show the experimental data and solid curves are the results of numerical modeling. $T=2$ K.

Figure 8

Temperature and power dependences of the spin precession frequency in the bare QW structure. The measurements are taken without using an interference bandpass filter in the detection (for details, see Appendix pp1). (a) Pump-probe transients for $B=620\mathrm{mT}$ at temperature $T=2$ (black squares) and 6 K (green triangles). Solid curves are fit with Eq. (D3). Dashed curves show the signal corresponding to the first two terms in Eq. (D3). (b) Dependence of the precession frequency in the bare QW structure on the excitation power of the pump beam. $T=6\mathrm{K}, B=640\mathrm{mT}$. The red dashed line corresponds to the expected dependence of ${\nu }_L$ for an additional linear increase of lattice temperature $T_c=T+\gamma P_{\mathrm{Pump}}$ with $\gamma =1.2$ K/mW.

Figure 9

The dependence of precession frequency as a function of pump polarization state. The polarization is controlled (a) by the $\lambda /4$ plate between the circularly ${\sigma }^{\pm }$- and linearly $p$-polarized pump and (b) by the $\lambda /2$ plate between the linearly $s$- and $p$-polarized pump. The pure polarization states are labeled in the top axis. $B=\pm 620\mathrm{mT}, T=2\mathrm{K}$.

Figure 10

Dependence of the average (a) transverse and (b) longitudinal polarizations of electromagnetic field $P_y$ and $P_z$ on the distance from the metallic grating. The structure is excited by a circularly polarized light. Curves have been calculated numerically for different angles of incidence $\mathrm{\Theta }$ at the same excitation energy, 1.7 eV. The thin vertical line corresponds to the experimental distance of $\approx 37\mathrm{nm}$.

Figure 11

Dependence of the average transverse polarizations of electromagnetic field $P_y$ on the thickness of the metallic grating. The structure is excited by a $p$-polarized light. Curves have been calculated numerically for different angles of incidence $\mathrm{\Theta }$ at the same excitation energy 1.7 eV. The thin vertical line corresponds to the thickness of the grating in the experimental sample equal to 45 nm.