betaFeedback

WorksPhysics and Astronomy

Raman Photogalvanic Effect: photocurrent at inelastic light scattering

Abstract

We show theoretically that electromagnetic waves propagating in the transparency region of a non-centrosymmetric medium can induce a dc electric current. The origin of the effect is the Raman scattering of light by free carriers in the system. Due to the photon scattering, electrons undergo real quantum transitions resulting in the formation of their anisotropic momentum distribution and in shifts of electronic wavepackets giving rise to a steady state photocurrent. We present microscopic theory of the Raman Photogalvanic effect (RPGE) focusing on two specific situations: (i) generic case of a bulk gyrotropic semiconductor and (ii) a quantum well structure where the light is scattered by intersubband excitations. We uncover the relation of the predicted RPGE and the traditional photogalvanic effect at the light absorption.

Introduction

Photogalvanic and photon drag effects (PGE and PDE) resulting in the dcdc electric current generation under steady-state illumination belong to a class of non-linear high-frequency transport phenomena and bridge optics and transport [1, 2, 3, 4, 5] [6, 7, 8]. The dcdc current magnitude and direction depend on the light intensity, propagation direction, and polarization. The processes of the photocurrent generation are highly sensitive to the symmetry of the system, fine structure of the electron energy spectrum, and microscopic processes of the optical transitions and scattering [4, 5]. Furthermore, the photocurrent generation can be related in some cases to the topological properties of the charge carriers Bloch functions [9–13]. It makes polarization-dependent photocurrents an important tool to study the delicate features of the electronic spectrum and kinetic properties of charge carriers in metals and semiconductors and opens up prospects to develop polarization detectors based on these effects [14, 15].

Usually, photogalvanic and photon drag currents are observed under the conditions of light absorption, see Figure 1(a). The PGEs are studied in conventional bulk semiconductors like Te and GaAs [4] [16], in low-dimensional structures such as quantum wells [17, 18], and in a wide range of emergent material systems including topological insulators [19–24], Weyl semimetals [25, 26], graphene based nanosystems [8, 27–29], and transition metal dichalcogenides [30–32]. Both inter- and intraband optical transitions can be involved in the dcdc current generation.

Schematic of traditional and Raman photogalvanic effects

Figure 1. (a) Scheme of the absorption process leading to circular PGE. (b) Illustration of the scattering process resulting in real electronic transition. (c,d) Basic diagrams describing the effects depicted in panels above (see SI for details). Green wavy lines denote electromagnetic field, jj denotes the current vertex with magenta filling describing the sum of the ladder diagrams accounting for the electron scattering by impurities and phonons. Magenta rectangle denotes the sum of the ladder diagrams accounting for the valence band hole scattering.

It is commonly assumed that, if the light propagates in the transparency region of the crystal, no dcdc current is formed [4] [33, 34]. It is indeed the case provided real electronic transitions and corresponding changes of the electromagnetic field are absent in the system. In such a situation, the irradiation results solely in renormalization of the energy dispersion. Thus, after a transient process the current vanishes [33]; otherwise, in violation of the energy conservation law, such current could generate the Joule heating in the external circuit.

Here we show that even in the absence of photon absorption, the dcdc electric current can be generated if the light is scattered by the free carriers in the medium. The

Raman scattering of light leads to the electronic transitions, Figure 1(b), resulting in the asymmetry of the electron distribution in the steady state and, eventually, in the dcdc current. A similar idea has been put forward in Ref. [35] without detailed analysis, here we present an explanation of the effect and transparent microscopic model. We develop the microscopic theory of the Raman photogalvanic effect (RPGE) in noncentrosymmetric semiconductors and semiconductor nanostructures. We mainly focus on the case of circular RPGE where the current reverses its sign under reversal of the radiation helicity. We address the situation where the photon energy is smaller than the fundamental energy gap. We take as examples (i) a non-resonant Raman scattering in bulk semiconductors and (ii) the intersubband resonant Raman scattering in quantum wells.

General description

We recall that the dc current linear in the radiation intensity II arising in the noncentrosymmetric media can be written in the most general form as [4, 5]

jα=γαβi[e×e∗]βI+χαβμ(eβeμ∗+eμeβ∗)I,(1)j_\alpha= \gamma_{\alpha\beta}i[e \times e^*]_\beta I + \chi_{\alpha\beta\mu}(e_\beta e_\mu^* + e_\mu e_\beta^*)I, \tag*{(1)}

where ee is the complex polarization vector of the incident electromagnetic field, i[e×e∗]=Pcircn^i[e \times e^*] = P_{\mathrm{circ}}\hat{n} with PcircP_{\mathrm{circ}} being the circular polarization degree and n^\hat{n} being the unit vector along the light propagation axis describes the light helicity. Tensors γαβ\gamma_{\alpha\beta} and χαβμ=χαμβ\chi_{\alpha\beta\mu} = \chi_{\alpha\mu\beta} describe circular and linear photocurrents, respectively; α,β,μ\alpha,\beta,\mu are the Cartesian components. Notably, the first and second terms in Eq. (1) have different properties under time reversal, t→−tt \to-t. Particularly, since both current and light helicity change their signs at the time reversal, tensor γαβ\gamma_{\alpha\beta} is even and the tensor χαβμ\chi_{\alpha\beta\mu} is odd at t→−tt \to-t.

We assume that the light propagates in the transparency region of the direct-gap semiconductor

ℏω<Eg,ωτp≫1,(2)\hbar\omega< E_g,\qquad\omega\tau_p \gg1, \tag*{(2)}

where ω\omega is the frequency of radiation, EgE_g is the band gap, and τp\tau_p is the conduction electron momentum scattering time. Here, for definiteness, we assume that the system is nn-doped. First condition in Eq. (2) ensures that the real interband transitions are forbidden, while the second one allows us to neglect the intraband Drude-like absorption.

Under condition Eq. (2) absorption of light is absent and the only possible real processes are the free-carrier light scattering as illustrated in Figure 1(a). The incident photon with the frequency ℏω\hbar\omega, polarization ee, and wavevector qq scatters and gives rise to a secondary photon with the frequency ℏω′\hbar\omega', polarization e′e' and wavevector q′q' while a resident electron undergoes a transition from the k′k' to kk state. Thus, the dc current density can be readily expressed as

j=eV0Tr⁡{v^ρ(2)}=jb+js,(3)j = \frac{e}{V_0}\operatorname{Tr}\{\hat{v}\rho^{(2)}\} = j_b + j_s, \tag*{(3)}

with the ballistic

jb=2eV0∑k,k′,q′[vkτp(Ec,k)−vk′τp(Ec,k′)]Wk,k′sc(e)f0(Ec,k′)[1−f0(Ec,k)],(4)j_b = \frac{2e}{V_0}\sum_{k,k',q'} [v_k\tau_p(E_{c,k}) - v_{k'}\tau_p(E_{c,k'})] W^{\mathrm{sc}}_{k,k'}(e)f_0(E_{c,k'})[1-f_0(E_{c,k})], \tag*{(4)}

and shift

js=2eV0∑k,k′,q′Rk,k′Wk,k′sc(e)f0(Ec,k′)[1−f0(Ec,k)],(5)j_s = \frac{2e}{V_0}\sum_{k,k',q'} R_{k,k'} W^{\mathrm{sc}}_{k,k'}(e)f_0(E_{c,k'})[1-f_0(E_{c,k})], \tag*{(5)}

contributions, respectively [4, 5] [36, 37]. Here ee is the electron charge, V0V_0 is the normalization volume, v^\hat{v} is the velocity operator and ρ(2)\rho^{(2)} is the electron density matrix calculated in the second-order in the incident electromagnetic field amplitude. In Eqs. (4), factors ‘2’ account for the spin degeneracy, Ec,kE_{c,k} is the electron dispersion, f0(E)f_0(E) is the equilibrium Fermi-Dirac distribution function, vk=ℏ−1∂Ec,k/∂kv_k = \hbar^{-1}\partial E_{c,k}/\partial k is the electron velocity, Wk,k′sc(e)W^{\mathrm{sc}}_{k,k'}(e) is the probability of electron scattering k′→kk' \to k at the incident light polarization ee averaged over the polarization and propagation direction of the final photon, and Rk,k′R_{k,k'} is the electron shift at the quantum transition k′→kk' \to k.

General description

The lack of the inversion center allows for odd in the electron wavevector terms in Wk,k′sc(e)W^{\mathrm{sc}}_{k,k'}(e) and even in the wavevector terms in the Rk,k′R_{k,k'}. It makes contributions (4) non-zero. The contribution (4) has a physical interpretation: In the course of quantum transitions electrons acquire an “average” velocity vˉ\bar{v} depending on the light polarization. The velocity generation rate is given by the rate of electron transitions N˙\dot{N} and the velocity relaxation rate is given by the momentum relaxation rate τp−1\tau_p^{-1}. As a result, the dc current according to this mechanism is formed during the ballistic propagation of electrons between the scattering events and given by the balance of the generation and relaxation processes [4]

j=evˉτpN˙.(6)j = e\bar{v}\tau_p\dot{N}. \tag*{(6)}

The shift photocurrent in Eq. (4b) can be estimated in the same manner with the replacement in Eq. (6) vˉτp\bar{v}\tau_p by the “average” shift of the wavepacket Rˉ\bar{R} in the course of scattering [4]. Equation (6) underlies that the current generation requires real electronic transitions and N˙\dot{N} can be expressed via the light intensity and the extinction coefficient K\mathcal{K} related to the scattering process N˙=KI/ℏω\dot{N} = \mathcal{K}I/\hbar\omega.

Microscopic model of the impurity or phonon-assisted RPGE

Now we turn to the microscopic description of the scattering processes. We focus on the photocurrent generation process in the bulk gyrotropic semiconductor under assumption that the incident photon energy is smaller but close to the fundamental bandgap Δ=Eg−ℏω≪ℏω\Delta= E_g - \hbar\omega\ll\hbar\omega. In this situation the main contribution to the free-carrier scattering of light is provided by virtual states in the valence band [5]. Accordingly, the scattering can be described as a three-stage process, where as shown in Figure 2(a) (i) the incident photon is absorbed (virtually) and creates an electron-hole pair by promoting the electron with the wavevector kk from the valence band to the conduction band; (ii) the hole in the valence band scatters (by phonon or impurity) in such a way that the state kk in the valence band becomes filled with electron and the state with the wavevector k′k' becomes unoccupied; and (iii) the hole recombines with the resident electron, so that finally the valence band remains unperturbed (all the states are filled) and in the conduction band the state with the wavevector k′k' is empty and the state with the wavevector kk is filled. The corresponding scattering rate is given by

Scheme of the light scattering process and spectral dependence of the photocurrent

Figure 2. (a) Scheme of the light scattering process with allowance for the momentum relaxation. (b) Spectral dependence of the photocurrent for a bulk semiconductor (solid line) and quantum wells (dotted line). Inset shows the relevant diagram, see SI.

Wk,k′sc(e′,e)=2πℏδ(ℏω−ℏω′−Ec,k+Ec,k′)I∣Mk′em(e′)Vk,k′Mkabs(e)(ℏω−Ec,k+Ev,k)(ℏω−Ec,k+Ev,k′)∣2,(7)W_{k,k'}^{\mathrm{sc}}(e',e)=\frac{2\pi}{\hbar}\delta(\hbar\omega-\hbar\omega'-E_{c,k}+E_{c,k'})I\left|\frac{M_{k'}^{\mathrm{em}}(e')V_{k,k'}M_k^{\mathrm{abs}}(e)}{(\hbar\omega-E_{c,k}+E_{v,k})(\hbar\omega-E_{c,k}+E_{v,k'})}\right|^2, \tag*{(7)}

where Mkabs(e)IM_k^{\mathrm{abs}}(e)\sqrt{I} and Mk′em(e′)M_{k'}^{\mathrm{em}}(e') are the interband transition matrix elements describing the absorption and emission of photons, respectively. It is convenient to present

∣Mkabs(e)∣2=∣M0∣2(1+Dαβkαi[e×e∗]β),(8)\left|M_k^{\mathrm{abs}}(e)\right|^2=|M_0|^2(1+D_{\alpha\beta}k_\alpha i[e\times e^*]_\beta), \tag*{(8)}

where the real second-rank tensor DαβD_{\alpha\beta} is responsible for the gyrotropy of the system [4] and M0M_0 is a constant. The presence of kk-linear terms in Eq. (8) makes an odd in the wavevector contribution to the electron transition rate Wk,k′(e)=∑q′,e′Wk,k′sc(e′,e)W_{k,k'}(e)=\sum_{q',e'}W_{k,k'}^{\mathrm{sc}}(e',e) and eventually results in the non-zero photocurrent, Eq. (4b). Equations (4) and (7) correspond to the diagram in Figure 2(b). Assuming that the carrier’s momentum relaxation is caused by short-range impurities, introducing ξ\xi as the ratio of the conduction and valence band elastic scattering matrix elements squared and γr\gamma_r as the radiative decay rate of the photoexcited electron-hole pair we obtain the following expression for the circular photocurrent (see SI for details of derivation)

jα=eneγrDαβn^βPcircI∣M0Δ∣2ξΦ(ν)3π.(9)j_\alpha=en_e\gamma_rD_{\alpha\beta}\hat{n}_\beta P_{\mathrm{circ}}I\left|\frac{M_0}{\Delta}\right|^2\frac{\xi\Phi(\nu)}{3\pi}. \tag*{(9)}

Here nen_e is the electron density in the conduction band, ν=1+me/mh\nu=1+m_e/m_h with mem_e and mhm_h being the electron and hole effective masses, and Φ(ν)=[(ν+1)ln⁡ν−2(ν−1)](ν−1)−3\Phi(\nu)=[(\nu+1)\ln\nu-2(\nu-1)](\nu-1)^{-3}. In derivation of Eq. (9) we assumed degenerate electrons with their Fermi energy EF≪ΔE_F\ll\Delta. The photocurrent in Eq. (9) increases with decreasing the detuning Δ\Delta because the smaller Δ\Delta is the more efficient is the light scattering, see Figure 2(b).

The ballistic photocurrent Eq. (9) changes its sign at reversal of the radiation helicity and Eq. (9) describes the circular RPGE. To obtain the ballistic linear RPGE one has to go beyond the three-stage process described above and take into account additional scattering processes to ensure the correct properties of the current under time reversal or evaluate the shift contribution, Eq. (4b), see Ref. [38] for evaluation of Rk,k′R_{k,k'} for multiquantum transitions. In any case, the linear RPGE will have an additional smallness ∼(Δτp/ℏ)−1\sim(\Delta\tau_p/\hbar)^{-1}, (ωτp)−1(\omega\tau_p)^{-1}, (EFτp/ℏ)−1(E_F\tau_p/\hbar)^{-1} depending on the particular mechanism of the effect.

It is worth to mention that the main contribution to the Raman scattering of light by free charge carriers in semiconductors does not require an additional transition of the hole in the valence band. The electron wavevector can change due to the variation of the light wavevector in the course of scattering

k−k′=q−q′,(10)k-k'=q-q', \tag*{(10)}

see SI for details. This process is described by the diagram analogous to that in Figure 2(b) but without the dashed vertical line. Calculation shows that the resulting photocurrent differs from that derived above in Eq. (9) by a factor of ℏq2τp/me≪1\hbar q^2\tau_p/m_e\ll1. The smallness of such contribution is related to the fact that, in the absence of additional scattering of the hole, the initial and final wavevectors of electrons are close to each other, see Eq. (10): The electron wavevector cannot change more than by a radiation wavevector. It results in a significant reduction of the effect.

RPGE at the intersubband resonance in quantum well structures. Additional specifics of Raman scattering of light appears where the change of the photon frequency corresponds to a frequency of a resonant excitation in the system. Such situation can naturally arise in quantum well structures as shown in Figure 3 under conditions of the intersubband scattering [5]. We consider for simplicity a symmetric structure with lowest occupied conduction subband (c1). As before the frequency of incident photon corresponds to the transparency region of the structure, Eq. (2) and EgE_g in this case corresponds to the gap between the topmost valence subband v1 and the bottom conduction subband c1. A situation of interest occurs in the vicinity of the intersubband resonance

Quantum well structure and scheme of the intersubband resonant light scattering process

Figure 3. (a) Quantum well structure. (b) Scheme of the intersubband resonant light scattering process.

ℏω−ℏω′=Ec3,k−Ec1,k,(11)\hbar\omega-\hbar\omega' = E_{c3,k}-E_{c1,k}, \tag*{(11)}

where the process depicted in Figure 3(b) becomes possible. In this case the virtual photon absorption via v1→c3v1 \to c3 electron transition is followed by the photon emission process resulting from the c1→v1c1 \to v1 transition. As a result, an electron is promoted from c1→c3c1 \to c3 subband (in asymmetric quantum wells similar transitions involving c2c2 subband are also allowed). Corresponding transition rate is readily evaluated [cf. Eq. (7)]

W3←1sc(e′,e,k)=2πℏ∣Mkem(e′,e)Mkabs(e)ℏω−Ec3,k+Ev1,k∣2×δ(ℏω−ℏω′−Ec3,k+Ec1,k)I.(12)\begin{aligned} W^{\mathrm{sc}}_{3\leftarrow1}(e',e,k)=\frac{2\pi}{\hbar}\left|\frac{M^{\mathrm{em}}_{k}(e',e)M^{\mathrm{abs}}_{k}(e)}{\hbar\omega-E_{c3,k}+E_{v1,k}}\right|^{2} \\ &\times\delta(\hbar\omega-\hbar\omega'-E_{c3,k}+E_{c1,k})I. \tag*{(12)} \end{aligned}

We neglect the photon momentum and any additional phonon or impurity scattering processes, hence the transitions take place at the same electron wavevector kk. As above we focus on the ballistic circular RPGE because it dominates the photocurrent and using the same Eq. (8) for ∣Mkabs(e)∣2|M^{\mathrm{abs}}_{k}(e)|^{2} as in the bulk case we arrive at the following expression for the current density (see SI for details):

jα=−eneγrQWDαβn^βPcircI∣M0∣2τtrν31ΔQWℏχ(ε).(13)j_{\alpha}=-en_{e}\gamma_{r}^{\mathrm{QW}}D_{\alpha\beta}\hat{n}_{\beta}P_{\mathrm{circ}}I\frac{|M_{0}|^{2}\tau_{\mathrm{tr}}}{\nu_{31}\Delta_{\mathrm{QW}}\hbar}\chi(\varepsilon). \tag*{(13)}

Here γrQW\gamma_{r}^{\mathrm{QW}} is the recombination rate of the electron in c1c1 subband with the v1v1 hole, we assumed the parabolic dispersion Ec1,k=ℏ2k2/2mc1E_{c1,k}=\hbar^{2}k^{2}/2m_{c1}, Ec3,k=Δ31+ℏ2k2/2mc3E_{c3,k}=\Delta_{31}+\hbar^{2}k^{2}/2m_{c3} with mcim_{ci} (i=1,3i=1,3) being the effective mass in the iith subband, and Δ31\Delta_{31} being the intersubband energy gap. In Eq. (13) we have introduced ν31=mc1/mc3+mc1/mh\nu_{31}=m_{c1}/m_{c3}+m_{c1}/m_{h}, τtr\tau_{\mathrm{tr}} is the momentum scattering time of c1c1 electron at a Fermi surface, the detuning for quantum wells ΔQW=Eg+Δ31−ℏω\Delta_{\mathrm{QW}}=E_{g}+\Delta_{31}-\hbar\omega, and

ε=ΔQWν31EF,χ(ε)=ε(ln⁡1+εε−11+ε).(14)\varepsilon=\frac{\Delta_{\mathrm{QW}}}{\nu_{31}E_{F}},\qquad\chi(\varepsilon)=\varepsilon\left(\ln\frac{1+\varepsilon}{\varepsilon}-\frac{1}{1+\varepsilon}\right). \tag*{(14)}

Equation (13) is valid for degenerate electrons and under assumption that the momentum relaxation in the c3c3 subband is much faster than in c1c1 subband: It is typically the case because of optical phonon emission processes causing electrons to relax to the bottom subbands; the general case is considered in the SI. It follows from Eq. (13) that RPGE current at intersubband scattering tends to a constant at the absorption edge being much weaker function of the detuning as compared to the analogous photocurrent in the bulk, cf. Eq. (9) and Figure 2.

Comparison with circular PGE in the absorption region. It is instructive to compare the results for the circular photocurrent obtained here for the transparency region with the well-known results for the circular PGE at the direct interband transitions. Considering the bulk semiconductor in the model described above [see diagram in Figure 1(c)] we obtain the following expression for the conduction electron photocurrent at ℏω>Eg\hbar\omega>E_g:

jαabs=eAIℏωDαβn^βPcirc2∣Δ∣τp3ν2ℏ.(15)j_{\alpha}^{\mathrm{abs}}=e\frac{AI}{\hbar\omega}D_{\alpha\beta}\hat{n}_{\beta}P_{\mathrm{circ}}\frac{2|\Delta|\tau_{p}}{3\nu^{2}\hbar}. \tag*{(15)}

Here AA is the absorption coefficient of the semiconductor, the electron momentum scattering time τp\tau_{p} is taken at the energy ∣Δ∣/ν|\Delta|/\nu, and Δ=Eg−ℏω<0\Delta=E_{g}-\hbar\omega<0 in the case of direct optical transitions. One can recast Eq. (9) in a similar form via the extinction coefficient κ\kappa

jαscatt=eκIℏωDαβn^βPcirc2∣Δ∣τp3ℏΦ(ν)Φ~(ν).(16)j_{\alpha}^{\mathrm{scatt}}=e\frac{\kappa I}{\hbar\omega}D_{\alpha\beta}\hat{n}_{\beta}P_{\mathrm{circ}}\frac{2|\Delta|\tau_{p}}{3\hbar}\frac{\Phi(\nu)}{\widetilde{\Phi}(\nu)}. \tag*{(16)}

with Φ~(ν)=πν/[2(1+ν)3]\widetilde{\Phi}(\nu)=\pi\sqrt{\nu}/[2(1+\sqrt{\nu})^{3}]. The factors ν2\nu^{2}, Φ/Φ~∼1\Phi/\widetilde{\Phi}\sim1, as a result for the same value of detuning the traditional and Raman photocurrents differ by the factor ∼A/κ\sim A/\kappa which provides an estimate of the ratio of the electronic transition rates at the light absorption and scattering, respectively. Naturally, both expressions can be brought to the form of general Eq. (6) with

νˉα∼ΔℏDαβn^βPcirc.(17)\bar{\nu}_{\alpha}\sim\frac{\Delta}{\hbar}D_{\alpha\beta}\hat{n}_{\beta}P_{\mathrm{circ}}. \tag*{(17)}

Equations (6), (16), and (17) demonstrate that even for transparent media, real electronic transitions should occur to enable the photocurrent [39].

We also note that in gyrotropic semiconductors and nanostructures kk-linear terms are present in the effective Hamiltonian of the charge carriers due to the spin-orbit coupling [5] [7, 40, 41]:

HSO=ℏβαμkασμ.(18)\mathcal{H}_{\mathrm{SO}}=\hbar\beta_{\alpha\mu}k_{\alpha}\sigma_{\mu}. \tag*{(18)}

where σ/2\sigma/2 is the electron spin operator. These terms provide an additive mechanism for the RPGE current generation. It can be also described by the general Eq. (6) with νˉα∼βαμn^μPcirc\bar{\nu}_{\alpha}\sim\beta_{\alpha\mu}\hat{n}_{\mu}P_{\mathrm{circ}}.

Conclusion. We have shown that the light scattering results in the steady state current in noncentrosymmetric media. The RPGE current is generated even if the light is propagating in the transparency spectral region of the crystal. While absorption is absent in this case, the current results from real electronic transitions owing to the Raman scattering of photons by free charge carriers. These transitions cause asymmetric distribution of electrons and also quantum shifts. We have identified key mechanisms of the Raman scattering induced circular photocurrent for the photon energies slightly below the band gap of a semiconductor and studied the photocurrent generation under intersubband scattering in quantum well structures.

Acknowledgements. We are grateful to Y. Onishi for valuable discussions. The authors acknowledge support from RSF: Project 22-12-00211 (general theory of RPGE, M.M.G.) and 22-12-00125 (calculations of RPGE currents, L.E.G.). L.E.G. also thanks the Foundation for the Advancement of Theoretical Physics and Mathematics “BASIS”.

Further details and comparison with recent Refs. [35, 42] are presented in SI.

Ballistic circular photocurrent in bulk gyrotropic semiconductor

Substituting the scattering rate (6) [of the main text] with the squared matrix element (7) into Eq. (4a) we obtain the ballistic photocurrent in the form

jα=eν0∣M0∣2∣MΣem∣2Dαβi[e×e∗]β8π3ℏ2Ini∣V∣2ν0∑k,k′,q′τp(Ec,k)Ec,kf0(Ec,k′)[1−f0(Ec,k)](ℏω−Ec,k+Ev,k)2(ℏω−Ec,k+Ev,k′)2δ(ℏω−ℏω′−Ec,k+Ec,k′)(19)j_{\alpha} = \frac{e}{\nu_{0}} \lvert M_{0}\rvert^{2} \lvert M^{\mathrm{em}}_{\Sigma}\rvert^{2}D_{\alpha\beta i}[e \times e^{*}]_{\beta}\frac{8\pi}{3\hbar^{2}}I\frac{n_{i}\lvert V\rvert^{2}}{\nu_{0}} \sum_{k,k',q'} \frac{\tau_{p}(E_{c,k})E_{c,k}f_{0}(E_{c,k'})[1-f_{0}(E_{c,k})]}{(\hbar\omega-E_{c,k}+E_{v,k})^{2}(\hbar\omega-E_{c,k}+E_{v,k'})^{2}}\delta(\hbar\omega-\hbar\omega'-E_{c,k}+E_{c,k'}) \tag*{(19)}

Here we assumed the parabolic dispersion Ec,k=ℏ2k2/2meE_{c,k}=\hbar^{2}k^{2}/2m_{e} with mem_{e} being the effective mass. Here ∣MΣem∣2\lvert M^{\mathrm{em}}_{\Sigma}\rvert^{2} is the emission matrix element summed over the secondary photon polarization. In calculation of ∣MΣem∣2\lvert M^{\mathrm{em}}_{\Sigma}\rvert^{2} we can disregard k′k'-linear terms because they are sensitive to e′e', we also disregard k′2k'^{2} and higher-order contributions to the matrix elements. Note that Mabs(e)M^{\mathrm{abs}}(e) describes the stimulated process (we consider classical electromagnetic wave incident on the sample), while emission process is spontaneous. That is why we use different normalization of MemM^{\mathrm{em}} and MabsM^{\mathrm{abs}}, see the main text below Eq. (7) for details.

For the following calculations we perform the summation over q′q' by means of the energy conservation δ\delta-function:

jα=eν0Dαβi[e×e∗]β43ℏIni∣M0V∣2ν0γr∑k,k′τp(Ec,k)Ec,kf0(Ec,k′)[1−f0(Ec,k)](ℏω−Ec,k+Ev,k)2(ℏω−Ec,k+Ev,k′)2(20)j_{\alpha} = \frac{e}{\nu_{0}}D_{\alpha\beta i}[e \times e^{*}]_{\beta}\frac{4}{3\hbar}I\frac{n_{i}\lvert M_{0}V\rvert^{2}}{\nu_{0}}\gamma_{r}\sum_{k,k'}\frac{\tau_{p}(E_{c,k})E_{c,k}f_{0}(E_{c,k'})[1-f_{0}(E_{c,k})]}{(\hbar\omega-E_{c,k}+E_{v,k})^{2}(\hbar\omega-E_{c,k}+E_{v,k'})^{2}} \tag*{(20)}

where we introduced the rate of emission of the secondary photon (i.e., the recombination rate of the electron-hole pair in the vicinity of the fundamental band gap)

γr=2πℏ∣MΣem∣2∑q′δ(ℏω−ℏω′−Ec,k+Ec,k′)\gamma_{r}=\frac{2\pi}{\hbar}\lvert M^{\mathrm{em}}_{\Sigma}\rvert^{2}\sum_{q'}\delta(\hbar\omega-\hbar\omega'-E_{c,k}+E_{c,k'})

Further calculations can be simplified as follows: we sum over k′k' (initial electron wavevector) under the assumption that EF,T≪Eg−ℏωE_{F},T \ll E_{g}-\hbar\omega (TT is the temperature expressed in the energy units). Hence, the k′k'-dependence of the denominators can be neglected in this case. This summation yields neV0/2n_e\mathcal{V}_0/2 with ne=(2/V0)∑kf0(Ec,k)n_e=(2/\mathcal{V}_0)\sum_{\boldsymbol{k}} f_0(E_{c,\boldsymbol{k}}) being the electron density. Under the same assumption one can omit 1−f0(Ec,k)1-f_0(E_{c,\boldsymbol{k}}) (but keep the k\boldsymbol{k}-dependence in the denominators otherwise the integral diverges at large k\boldsymbol{k}) and we finally have

jα=eV0Dαβi[e×e∗]β23ℏIneni∣M0V∣2γr∑kτp(Ec,k)Ec,k(ℏω−Ec,k+Ev,k)(ℏω−Ec,k+Ev,k′)2.(21)j_\alpha=\frac{e}{\mathcal{V}_0}D_{\alpha\beta i}[e\times e^*]_\beta\frac{2}{3\hbar}In_en_i|M_0V|^2\gamma_r\sum_{\boldsymbol{k}}\frac{\tau_p(E_{c,\boldsymbol{k}})E_{c,\boldsymbol{k}}}{(\hbar\omega-E_{c,\boldsymbol{k}}+E_{v,\boldsymbol{k}})(\hbar\omega-E_{c,\boldsymbol{k}}+E_{v,\boldsymbol{k}'})^2}. \tag*{(21)}

Here all energies are counted from the bottom of the conduction band. Using

ni∣V∣2τp(Ec)=ξℏ2πgc(Ec),∑k⋯=V0∫0∞dEcgc(Ec)⋯ ,Ec,k−Ev,k=νEc,k+Eg,ν=1+memh,(22)n_i|V|^2\tau_p(E_c)=\xi\frac{\hbar}{2\pi g_c(E_c)},\qquad\sum_{\boldsymbol{k}}\cdots=\mathcal{V}_0\int_0^\infty dE_cg_c(E_c)\cdots,\qquad E_{c,\boldsymbol{k}}-E_{v,\boldsymbol{k}}=\nu E_{c,\boldsymbol{k}}+E_g,\qquad\nu=1+\frac{m_e}{m_h}, \tag*{(22)}

where ξ\xi is the ratio of the conduction and valence band elastic scattering matrix elements squared: ξ=∣Vv∣2/∣Vc∣2\xi=|V_v|^2/|V_c|^2, gc(Ec)g_c(E_c) is the density of states in the conduction band, we get

jα=eDαβi[e×e∗]βIne∣M0∣2γrξ3π∫0∞dEcEc(ℏω−Eg−νEc)2(ℏω−Eg−Ec)2.(23)j_\alpha=eD_{\alpha\beta i}[e\times e^*]_\beta In_e|M_0|^2\gamma_r\frac{\xi}{3\pi}\int_0^\infty dE_c\frac{E_c}{(\hbar\omega-E_g-\nu E_c)^2(\hbar\omega-E_g-E_c)^2}. \tag*{(23)}

Calculating the integral

∫0∞dEcEc(ℏω−Eg−νEc)2(ℏω−Eg−Ec)2=Φ(ν)Δ2,Δ=Eg−ℏω,\int_0^\infty dE_c\frac{E_c}{(\hbar\omega-E_g-\nu E_c)^2(\hbar\omega-E_g-E_c)^2}=\frac{\Phi(\nu)}{\Delta^2},\qquad\Delta=E_g-\hbar\omega,
Φ(ν)=∫0∞dxx(1+νx)2(1+x)2=(ν+1)ln⁡ν−2(ν−1)(ν−1)3,Φ(1)=16,(24)\Phi(\nu)=\int_0^\infty dx\frac{x}{(1+\nu x)^2(1+x)^2}=\frac{(\nu+1)\ln\nu-2(\nu-1)}{(\nu-1)^3},\qquad\Phi(1)=\frac{1}{6}, \tag*{(24)}

we finally get Eq. (9) of the main text:

jα=eneγrDαβi[e×e∗]βI∣M0∣2Δ2ξΦ(ν)3π.(25)j_\alpha=en_e\gamma_rD_{\alpha\beta i}[e\times e^*]_\beta I\frac{|M_0|^2}{\Delta^2}\frac{\xi\Phi(\nu)}{3\pi}. \tag*{(25)}

Light extinction and RPGE

Let us introduce the extinction coefficient K\mathcal{K} [cm−1\mathrm{cm}^{-1}] related to the light scattering

K=2∑k,k′,q′Wk,k′sc(e)f0(Ec,k′)[1−f0(Ec,k)]N(c/n).(26)\mathcal{K}=\frac{2\sum_{\boldsymbol{k},\boldsymbol{k}',\boldsymbol{q}'}W^{\mathrm{sc}}_{\boldsymbol{k},\boldsymbol{k}'}(e)f_0(E_{c,\boldsymbol{k}'})[1-f_0(E_{c,\boldsymbol{k}})]}{\mathcal{N}(c/n)}. \tag*{(26)}

It describes light attenuation in the system due to the scattering. Here nn is the refractive index of the crystal and N\mathcal{N} is the number of photons in the electromagnetic wave. Taking into account that I=Nℏωc/nI=\mathcal{N}\hbar\omega c/n and performing summation over q′\boldsymbol{q}' to obtain γr\gamma_r, over k′\boldsymbol{k}' to obtain nen_e, and over k\boldsymbol{k} by virtue of

∫0∞dEcgc(Ec)(ℏω−Eg−νEc)2(ℏω−Eg−Ec)2=Φ~(ν)gc(Eω)Eω3ν3,Eω=Eg−ℏων,Φ~(ν)=πν2(1+ν)3,\int_0^\infty dE_c\frac{g_c(E_c)}{(\hbar\omega-E_g-\nu E_c)^2(\hbar\omega-E_g-E_c)^2}=\widetilde{\Phi}(\nu)\frac{g_c(E_\omega)}{E_\omega^3\nu^3},\qquad E_\omega=\frac{E_g-\hbar\omega}{\nu},\qquad\widetilde{\Phi}(\nu)=\frac{\pi\sqrt{\nu}}{2(1+\sqrt{\nu})^3},

and using gc(Eω)ni∣V∣2=ξℏ/[2πτp(Eω)]g_c(E_\omega)n_i|V|^2=\xi\hbar/[2\pi\tau_p(E_\omega)], we have

K=neγrξℏ2πτpΦ~(ν)ℏω∣M0∣2Δ3.(27)\mathcal{K}=n_e\gamma_r\frac{\xi\hbar}{2\pi\tau_p}\widetilde{\Phi}(\nu)\frac{\hbar\omega|M_0|^2}{\Delta^3}. \tag*{(27)}

As a result for the Raman scattering-induced photocurrent we have Eq. (16) of the main text:

jαscatt=eKIℏωDαβi[e×e∗]β2Δτp(Eω)3ℏΦ(ν)Φ~(ν).(28)j_\alpha^{\mathrm{scatt}}=e\frac{\mathcal{K}I}{\hbar\omega}D_{\alpha\beta i}[e\times e^*]_\beta\frac{2\Delta\tau_p(E_\omega)}{3\hbar}\frac{\Phi(\nu)}{\widetilde{\Phi}(\nu)}. \tag*{(28)}

Relation to the photogalvanic effect at light absorption

Let us calculate the (ballistic) CPGE current in the same system generated at ℏω>Eg\hbar\omega> E_g. We calculate the electron contribution only (this is the net electric current if the relaxation time in the valence band is very short). The CPGE current density at light absorption is given by

jαabs=2e∑kWkabs(e)τp(Ec,k)vα(k)δ(Ec,k−Ev,k−ℏω),(29)j_{\alpha}^{\mathrm{abs}}=2e\sum_{k}W_{k}^{\mathrm{abs}}(e)\tau_{p}(E_{c,k})v_{\alpha}(k)\delta(E_{c,k}-E_{v,k}-\hbar\omega), \tag*{(29)}

where the asymmetric contribution to the light absorption probability at direct optical transition is

Wkabs(e)=2πℏI∣M0∣2Dαβi[e×e∗]βkα.(30)W_{k}^{\mathrm{abs}}(e)=\frac{2\pi}{\hbar}I|M_{0}|^{2}D_{\alpha\beta i}[e\times e^{*}]_{\beta}k_{\alpha}. \tag*{(30)}

Then we obtain

jαabs=2e2πℏI∣M0∣2Dαβi[e×e∗]β∑k23ℏEc,kτp(Ec,k)δ(νEc,k+Eg−ℏω)=eIDαβi[e×e∗]β8π∣M0∣23ℏ2gc(∣Eω∣)∣Eω∣τp(∣Eω∣)ν2Θ(ℏω−Eg).(31)\begin{aligned} j_{\alpha}^{\mathrm{abs}}=2e\frac{2\pi}{\hbar}I|M_{0}|^{2}D_{\alpha\beta i}[e\times e^{*}]_{\beta}\sum_{k}\frac{2}{3\hbar}E_{c,k}\tau_{p}(E_{c,k})\delta(\nu E_{c,k}+E_{g}-\hbar\omega) \\ &=eI D_{\alpha\beta i}[e\times e^{*}]_{\beta}\frac{8\pi|M_{0}|^{2}}{3\hbar^{2}}g_{c}(|E_{\omega}|)\frac{|E_{\omega}|\tau_{p}(|E_{\omega}|)}{\nu^{2}}\Theta(\hbar\omega-E_{g}). \tag*{(31)} \end{aligned}

It is convenient to introduce the light absorption coefficient A(ω)\mathcal{A}(\omega) with the dimension of cm−1\mathrm{cm}^{-1} related to the direct optical transitions as

A=ℏωIWcv,(32)\mathcal{A}=\frac{\hbar\omega}{I}W_{cv}, \tag*{(32)}

where the direct interband transition rate

Wcv=22πℏ∑kI∣M0∣2δ(Ec,k−Ev,k−ℏω)=4πℏI∣M0∣2gc(∣Eω∣)νΘ(ℏω−Eg),(33)W_{cv}=2\frac{2\pi}{\hbar}\sum_{k}I|M_{0}|^{2}\delta(E_{c,k}-E_{v,k}-\hbar\omega)=\frac{4\pi}{\hbar}I|M_{0}|^{2}\frac{g_{c}(|E_{\omega}|)}{\nu}\Theta(\hbar\omega-E_{g}), \tag*{(33)}

and

A=4πω∣M0∣2gc(∣Eω∣)νΘ(ℏω−Eg).(34)\mathcal{A}=4\pi\omega|M_{0}|^{2}\frac{g_{c}(|E_{\omega}|)}{\nu}\Theta(\hbar\omega-E_{g}). \tag*{(34)}

Finally, we obtain from Eq. (31) the circular photocurrent caused by the CPGE in the form of (15) of the main text:

jαabs=eAIℏωDαβi[e×e∗]β2∣Eω∣τp(∣Eω∣)3νℏ.(35)j_{\alpha}^{\mathrm{abs}}=e\frac{\mathcal{A}I}{\hbar\omega}D_{\alpha\beta i}[e\times e^{*}]_{\beta}\frac{2|E_{\omega}|\tau_{p}(|E_{\omega}|)}{3\nu\hbar}. \tag*{(35)}

Allowance for the radiation wavevector in RPGE

In previous section we studied the RPGE photocurrent that is generated in a course of a three-step process of virtual photon absorption, valence band hole scattering, and virtual photon emission. In this way the electron wavevectors in the initial, k′k', and final, kk, states are decoupled. Let us now discuss the contribution to RPGE where the valence band hole scattering is absent. Such process takes place with allowance for the photon wavevector, see Figure S1, this process is the main process of light scattering by free carriers in semiconductors for ℏω≲Eg\hbar\omega\lesssim E_g [3]. Corresponding transition probability is given by

Scheme of the light scattering process with allowance for the radiation wavevector

Figure S1. Scheme of the light scattering process with allowance for the radiation wavevector.

Wk,k′q(e′,e)=2πℏδ(ℏω−ℏω′−Ec,k+Ec,k′)δk+q′,k′+q∣Mk′em(e′)Mkabs(e)ℏω−Ec,k+Ev,k∣2.(36)W_{k,k'}^{q}(e',e)=\frac{2\pi}{\hbar}\delta(\hbar\omega-\hbar\omega'-E_{c,k}+E_{c,k'})\delta_{k+q',k'+q}\left|\frac{M_{k'}^{\mathrm{em}}(e')M_{k}^{\mathrm{abs}}(e)}{\hbar\omega-E_{c,k}+E_{v,k}}\right|^{2}. \tag*{(36)}

Here Kronecker-δ\delta accounts for the momentum conservation in the course of light scattering [Eq. (10) of the main text]

k′+q=k+q′.k'+q=k+q'.

The photocurrent is given by [cf. Eq. (4a) of the main text]

jb=2eν0∑k,k′,q′[vkτp(Ec,k)−vk′τp(Ec,k′)]Wk,k′q(e)f0(Ec,k′)[1−f0(Ec,k)],(37)j_b = \frac{2e}{\nu_0}\sum_{k,k',q'} \left[v_k\tau_p(E_{c,k})-v_{k'}\tau_p(E_{c,k'})\right]W^q_{k,k'}(e)f_0(E_{c,k'})\left[1-f_0(E_{c,k})\right], \tag*{(37)}

Let us assume for simplicity that the energy dependence of the electron momentum scattering time can be disregarded. In this case the velocity-dependent term in square brackets can be recast as

[vkτp(Ec,k)−vk′τp(Ec,k′)]=τpℏme(q′−q).\left[v_k\tau_p(E_{c,k})-v_{k'}\tau_p(E_{c,k'})\right]=\tau_p\frac{\hbar}{m_e}\left(q'-q\right).

Note that the difference of velocities in the initial and final states is now small and related to the photon wavevector. It results in suppression of the photocurrent. Disregarding in Eq. (37) the q,q′q,q'-dependence of all remaining terms we obtain the polarization independent photocurrent related to the photon drag effect at the Raman scattering:

jb=2eν0ℏme(−q)∑k,q′Wk,kq(e)f0(Ec,k)[1−f0(Ec,k)].(38)j_b=\frac{2e}{\nu_0}\frac{\hbar}{m_e}(-q)\sum_{k,q'}W^q_{k,k}(e)f_0(E_{c,k})\left[1-f_0(E_{c,k})\right]. \tag*{(38)}

According to the time-reversal symmetry this effect is independent of the circular polarization of light. Furthermore, to obtain the polarization-dependent current we need to extract kk- and qq-linear contributions from the occupancy factors, otherwise the sum over kk in Eq. (37) vanishes. The resulting current contains the factor ∼ℏ2q2/me\sim\hbar^2q^2/m_e instead of ∼ξℏ/τp\sim\xi\hbar/\tau_p that appears in Eq. (21) due to the scattering in the valence band. As a result, the RPGE without scattering in the valence band is smaller than the current in Eq. (28) by a factor of

∼ℏq2τpme.\sim\frac{\hbar q^2\tau_p}{m_e}.

For typical conditions this factor is small.

Intersubband scattering-assisted photocurrent

In noncentrosymmetric quantum well systems, the intersubband scattering probability of light W3↔1sc(e,k)W^{\mathrm{sc}}_{3\leftrightarrow1}(e,k) contains, in general, an asymmetric part. As a result, there is a ballistic contribution to the photocurrent:

j=2eS0∑k,k′,q′[vc3,kτp(Ec3,k)−vc1,kτp(Ec1,k)]W3↔1sc(e,k)f0(Ec1,k)[1−f0(Ec3,k)],(39)j=\frac{2e}{S_0}\sum_{k,k',q'}\left[v_{c3,k}\tau_p(E_{c3,k})-v_{c1,k}\tau_p(E_{c1,k})\right]W^{\mathrm{sc}}_{3\leftrightarrow1}(e,k)f_0(E_{c1,k})\left[1-f_0(E_{c3,k})\right], \tag*{(39)}

where S0S_0 is the normalization area. Substitution of the scattering rate given by Eq. (12) of the main text yields

jα=eS0Dαβi[e×e∗]β4πℏ2∣M0∣4∑k,q′ℏ2k22τp(Ec3,k)/mc3−τp(Ec1,k)/mc1(ℏω−Ec3,k+Ev,k)2f0(Ec1,k)[1−f0(Ec3,k)]δ(ℏω−ℏω′−Ec3,k+Ec1,k).(40)j_\alpha=\frac{e}{S_0}D_{\alpha\beta i}[e\times e^*]_\beta\frac{4\pi}{\hbar^2}|M_0|^4 \sum_{k,q'}\frac{\hbar^2 k^2}{2}\frac{\tau_p(E_{c3,k})/m_{c3}-\tau_p(E_{c1,k})/m_{c1}}{(\hbar\omega-E_{c3,k}+E_{v,k})^2}f_0(E_{c1,k})[1-f_0(E_{c3,k})]\delta(\hbar\omega-\hbar\omega'-E_{c3,k}+E_{c1,k}). \tag*{(40)}

For the following calculations we perform the summation over q′q' by means of the energy conservation δ\delta-function:

jα=eS0Dαβi[e×e∗]β2ℏ∣M0∣2γrQW∑kℏ2k22τp(Ec3,k)/mc3−τp(Ec1,k)/mc1(ℏω−Ec3,k+Ev,k)2f0(Ec1,k)[1−f0(Ec3,k)],(41)j_\alpha=\frac{e}{S_0}D_{\alpha\beta i}[e\times e^*]_\beta\frac{2}{\hbar}|M_0|^2\gamma_r^{\mathrm{QW}}\sum_k\frac{\hbar^2 k^2}{2}\frac{\tau_p(E_{c3,k})/m_{c3}-\tau_p(E_{c1,k})/m_{c1}}{(\hbar\omega-E_{c3,k}+E_{v,k})^2}f_0(E_{c1,k})[1-f_0(E_{c3,k})], \tag*{(41)}

where we introduced the rate of emission of the secondary photon (i.e., the recombination rate of the electron-hole pair in the vicinity of the fundamental band gap)

γrQW=2πℏ∣M0∣2∑q′δ(ℏω−ℏω′−Ec3,k+Ec1,k).\gamma_r^{\mathrm{QW}}=\frac{2\pi}{\hbar}|M_0|^2\sum_{q'}\delta(\hbar\omega-\hbar\omega'-E_{c3,k}+E_{c1,k}).

Further calculations can be simplified as follows: we assume τp(Ec3,k)≪τp(Ec1,k)\tau_p(E_{c3,k})\ll\tau_p(E_{c1,k}), f0(Ec3,k)≪1f_0(E_{c3,k})\ll1. Then we have

jα=−eS0Dαβi[e×e∗]β2ℏ∣M0∣2γrQW∑kEc1,kτp(Ec1,k)(ℏω−Ec3,k+Ev,k)2f0(Ec1,k).(42)j_\alpha=-\frac{e}{S_0}D_{\alpha\beta i}[e\times e^*]_\beta\frac{2}{\hbar}|M_0|^2\gamma_r^{\mathrm{QW}}\sum_k\frac{E_{c1,k}\tau_p(E_{c1,k})}{(\hbar\omega-E_{c3,k}+E_{v,k})^2}f_0(E_{c1,k}). \tag*{(42)}

For crude estimation we can assume EF,T≪Eg3−ℏωE_F,T\ll E_{g3}-\hbar\omega, where Eg3=Eg+Δ31E_{g3}=E_g+\Delta_{31}, and obtain

jα≈−eneγrQWDαβi[e×e∗]β2⟨E⟩τtrℏ∣M0∣2(Eg3−ℏω)2.(43)j_\alpha\approx-en_e\gamma_r^{\mathrm{QW}}D_{\alpha\beta i}[e\times e^*]_\beta\frac{2\langle E\rangle\tau_{\mathrm{tr}}}{\hbar}\frac{|M_0|^2}{(E_{g3}-\hbar\omega)^2}. \tag*{(43)}

More precisely, using

∑k⋯=S0gc1∫0∞dEc1⋯ ,Ec3,k−Ev,k=ν31Ec1,k+Eg3,ν31=mc1mc3(1mc3+1mh),(44)\sum_k\cdots=S_0g_{c1}\int_0^\infty dE_{c1}\cdots,\qquad E_{c3,k}-E_{v,k}=\nu_{31}E_{c1,k}+E_{g3},\qquad\nu_{31}=\frac{m_{c1}}{m_{c3}}\left(\frac{1}{m_{c3}}+\frac{1}{m_h}\right), \tag*{(44)}

where gc1g_{c1} is the density of states in the 1st conduction subband, we get

jα=−eDαβi[e×e∗]β2ℏ∣M0∣2γrQWℏ2πni∣V∣2∫0∞dEE(ℏω−Eg3−ν31E)2f0(E).(45)j_\alpha=-eD_{\alpha\beta i}[e\times e^*]_\beta\frac{2}{\hbar}|M_0|^2\gamma_r^{\mathrm{QW}}\frac{\hbar}{2\pi n_i|V|^2}\int_0^\infty dE\frac{E}{(\hbar\omega-E_{g3}-\nu_{31}E)^2}f_0(E). \tag*{(45)}

At low temperatures we have:

∫0EFdEE(ℏω−Eg3−ν31E)2=1ν312(ln⁡1+εε−11+ε),ε=Eg3−ℏων31EF.(46)\int_0^{E_F}dE\frac{E}{(\hbar\omega-E_{g3}-\nu_{31}E)^2}=\frac{1}{\nu_{31}^2}\left(\ln\frac{1+\varepsilon}{\varepsilon}-\frac{1}{1+\varepsilon}\right),\qquad\varepsilon=\frac{E_{g3}-\hbar\omega}{\nu_{31}E_F}. \tag*{(46)}

while the electron concentration is given by ne=2gc1EFn_e=2g_{c1}E_F. Therefore we get for Fermi statistics (when τtr=τp(EF)\tau_{\mathrm{tr}}=\tau_p(E_F)) Eqs. (13), (14) of the main text:

jα=−eneγrQWDαβi[e×e∗]β∣M0∣2τtrν312ℏEF(ln⁡1+εε−11+ε).(47)j_\alpha=-en_e\gamma_r^{\mathrm{QW}}D_{\alpha\beta i}[e\times e^*]_\beta\frac{|M_0|^2\tau_{\mathrm{tr}}}{\nu_{31}^2\hbar E_F}\left(\ln\frac{1+\varepsilon}{\varepsilon}-\frac{1}{1+\varepsilon}\right). \tag*{(47)}

At Boltzmann statistics we have:

∫0∞dEEf0(E)(ℏω−Eg3−ν31E)2=exp⁡(μ/T)ν312∫0∞dxxexp⁡(−x)(x+b)2=exp⁡(μ/T)ν312[ebEi⁡(−b)(1+b)−1],(48)\int_0^\infty dE\frac{Ef_0(E)}{(\hbar\omega-E_{g3}-\nu_{31}E)^2}=\frac{\exp(\mu/T)}{\nu_{31}^2}\int_0^\infty dx\frac{x\exp(-x)}{(x+b)^2}=\frac{\exp(\mu/T)}{\nu_{31}^2}\left[e^b\operatorname{Ei}(-b)(1+b)-1\right], \tag*{(48)}

where Ei⁡(x)=∫−x∞t−1exp⁡(−t) dt\operatorname{Ei}(x)=\int_{-x}^{\infty}t^{-1}\exp(-t)\,dt is the exponential integral,

b=Eg3−ℏων31T,ne=2gc1Texp⁡(μ/T).(49)b=\frac{E_{g3}-\hbar\omega}{\nu_{31}T},\qquad n_e=2g_{c1}T\exp(\mu/T). \tag*{(49)}

As a result, we obtain

jα=−eneγrQWDαβi[e×e∗]β∣M0∣2τp(T)ν312ℏT[ebEi⁡(−b)(1+b)−1].(50)j_\alpha=-en_e\gamma_r^{\mathrm{QW}}D_{\alpha\beta i}[e\times e^*]_\beta\frac{|M_0|^2\tau_p(T)}{\nu_{31}^2\hbar T}\left[e^b\operatorname{Ei}(-b)(1+b)-1\right]. \tag*{(50)}

Diagrammatic approach to RPGE

It is instructive to consider the circular photocurrent generation in the diagrammatic approach.

Figure S2(a) shows the diagram relevant for the CPGE at the light absorption. Taking into account that in the absorption region of the spectrum ℏω≥Eg\hbar\omega\ge E_g the energy conservation law can be fulfilled, Figure S2 can be immediately calculated and represents the conduction band contribution to the photocurrent, Eq. (29). In the absence of absorption this diagram vanishes.

Diagrams relevant for the photocurrent generation (in-scattering terms)

Figure S2. Diagrams relevant for the photocurrent generation (in-scattering terms).

Diagrams S2(b-d) describe the RPGE: These diagrams take into account the photon scattering, i.e., the emission of the secondary photon, wavy line marked as ℏω′\hbar\omega'. The diagram Figure S2 and its counterpart, Figure S3, describe the RPGE with allowance for the photon momentum. Note that kk, k′′k'' and k′k' are related by the momentum conservation law. Figure S2(c) (and the diagram, analogous to that in Figure S3) describe the photocurrent with allowance for the scattering in the valence band. The sum of all relevant diagrams with (0,1,2,…)(0, 1, 2, \ldots) scattering is depicted in Figure S2.

Second diagram describing the RPGE with allowance for the photon wavevector (out-scattering term)

Figure S3. Second diagram describing the RPGE with allowance for the photon wavevector (out-scattering term).

Figure S4 shows the diagrams relevant for the RPGE at the intersubband resonant scattering in quantum well structures. Here extra scattering in the valence band is not required, see Eq. (12) of the main text.

Diagrams relevant for the RPGE at the intersubband scattering

Figure S4. Diagrams relevant for the RPGE at the intersubband scattering: (a) in-scattering and (b) out-scattering contributions.

Photocurrents in the transparency region: General remarks

It is instructive to make several comments about photocurrents for ℏω\hbar\omega in the transparency region of the crystal regarding recent preprints [1, 2]. We reiterate that in the absence of any real electronic transitions dcdc current is forbidden. It is obvious from general reasons: If a dcdc current is generated then this current results in a Joule heat in the sample or in the external circuit connected to the sample. It is forbidden by the energy conservation law in the absence of real transitions. On the microscopic level, the absence of the photocurrent follows from the arguments presented in Ref. [4]: In the absence of real transitions the only effect of the field is the renormalization of single particle dispersion Ec,k→E~c,kE_{c,\mathbf{k}} \to\widetilde{E}_{c,\mathbf{k}}. During the transient processes of energy and momentum relaxation the distribution function relaxes to an equilibrium function f0(E~c,k)f_0(\widetilde{E}_{c,\mathbf{k}}) of the renormalized dispersion. The dc current would be

jdc=2∑kv~kf0(E~c,k)=0,(51)j_{\mathrm{dc}} = 2 \sum_{\mathbf{k}} \widetilde{v}_{\mathbf{k}} f_0(\widetilde{E}_{c,\mathbf{k}}) = 0, \tag*{(51)}

where v~k=ℏ−1∂E~c,k/∂k\widetilde{v}_{\mathbf{k}} = \hbar^{-1}\partial\widetilde{E}_{c,\mathbf{k}}/\partial\mathbf{k}. Thus, real electronic transitions are crucial for the photocurrent generation.

In the case considered in our work the real transitions are induced by the light scattering (this point was mentioned in Ref. [1]). In recent preprint [2] the authors discussed the photocurrent in the optical gap of a metal related to the “Berry curvature dipole” and “jerk” effects. In this regard, two comments are due:

i) At linear polarization no current can be generated in the absence of absorption or scattering processes. It follows from the general relation (1) of the main text:

jα=γαβi[e×e∗]βI+χαβμ(eβeμ∗+eμeβ∗)I,(52)j_{\alpha} = \gamma_{\alpha\beta i}[e \times e^{*}]_{\beta}I + \chi_{\alpha\beta\mu}(e_{\beta}e^{*}_{\mu} + e_{\mu}e^{*}_{\beta})I, \tag*{(52)}

which explicitly shows that the tensor χ\chi responsible for the linear PGE is odd at the time reversal. Thus, it contains odd numbers of dissipative constants.

ii) The “Berry curvature dipole” contribution, at first glance, seems dissipationless: It appears at the circular polarization and does not vanish in the clean limit (Γ→0\Gamma\to0 in the terminology of the authors of Ref. [2]). However, as demonstrated in Ref. [5], this contribution is related to the interference of real electronic transition processes with the intermediate states in the same (conduction) band (Drude-like absorption) and with the intermediate states in the remote (valence) band. We also note that there is a side-jump contribution to the PGE which differs from the “Berry curvature dipole” contribution by a numerical factor only. Obviously, real transitions require dissipation, and as the authors of Ref. [2] explicitly check, these processes do not violate basic thermodynamic principles. Note that the absence of scattering rates in the expressions for the “Berry curvature dipole” and side-jump contributions is due to the cancellation of the rates in product of the Drude transition probability and in the momentum scattering time entering the general expression for the photocurrent [cf. Eq. (5)] of the main text:

jdc=evˉτpN˙,j_{\mathrm{dc}} = e\bar{v}\tau_p\dot{N},

where for Drude-like transitions N˙∝1/τp\dot{N} \propto1/\tau_p at ωτp≫1\omega\tau_p \gg1.

References

  1. [1]Y. Onishi, H. Watanabe, T. Morimoto, and N. Nagaosa, Photovoltaic effect in noncentrosymmetric material without optical absorption, arXiv: 2204.12727 (2022).
  2. [2]L. Shi, O. Matsyshyn, J. C. W. Song, and I. S. Villadiego, The Berry dipole photovoltaic demo and the thermodynamics of photo-current generation within the optical gap of metals, arXiv: 2207.03496 (2022).DOI
  3. [3]E. L. Ivchenko, Optical spectroscopy of semiconductor nanostructures (Alpha Science, Harrow UK, 2005).
  4. [4]V. I. Belinicher, E. L. Ivchenko, and G. E. Pikus, Transient photocurrent in gyrotropic crystals, Sov. Phys. - Semicond. 20, 558 (1986).
  5. [5]L. E. Golub, E. L. Ivchenko, and B. Spivak, Semiclassical theory of the circular photogalvanic effect in gyrotropic systems, Phys. Rev. B 102, 085202 (2020).DOI

Paper details

vlnc:W_myvfg9v8
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents