betaFeedback

WorksPhysics and Astronomy

Theory of circular dichroism in angle- and spin-resolved photoemission from the surface state on Bi(111)

Abstract

Circular dichroism (CD) of photoemission from a surface state on Bi(111) is considered within an ab initio one-step theory of photoemission. The symmetry of the azimuthal distribution of CD for different spin directions is discussed in terms of the spin structure of the initial states. Both normal and off-normal light incidence are considered. Strong photon energy dependence of photoemission in the ultraviolet range is observed. The principal difference between the initial state spin polarization and the spin photocurrent is emphasized and problems with revealing the orbital angular moment with photoemission are pointed to.

Introduction

Photoemission has been long established as the most straightforward method to infer the electronic structure of a crystal [1, 2]. The energy and angular distribution of photoelectrons brings information about the surface-projected spectral function. A further degree of freedom is the spin of the photoelectron. For nonmagnetic solids a spin dependence of the photocurrent in vacuum characterizes relativistic effects [5, 6, 7, 8] [3, 4]. The theoretical background is given by the one-step theory of photoemission [9, 10, 11, 12, 13], which was applied to various aspects of spin-resolved photoemission [14, 15, 16, 17]; in particular to the symmetry analysis of the spin photocurrent [18, 19]. For a specific initial state, the photocurrent depends on the geometric parameters of the photoemission setup—light polarization and incidence geometry, as well as the photon energy ℏω\hbar\omega. By varying these parameters one can reveal certain aspects of the structure of the initial state wave function.

In particular, the idea was put forward that measurements with circularly polarized light may provide information about orbital angular moment (OAM) [20, 21, 22, 23, 24]. The photoemission CD was claimed to be related to OAM for the surface states in Bi2_2Te3_3 [20], Bi2_2Se3_3 [21, 25], Cu(111) and Au(111) [22]. This interpretation was questioned by Scholz et al. [26] based on the apparent photon energy dependence of CD—in particular, a reversal of the sign was observed in Bi2_2Te3_3. Indeed, ℏω\hbar\omega-dependence of CD has been widely observed experimentally for various materials: Bi2_2Te2_2Se [27], Bi2_2Se3_3 [28], BiTeX, X = I, Br, Cl [29]. Furthermore, a theoretical analysis of CD of the surface states on Au(111) with one-step photoemission theory [30] showed rapid changes of CD with ℏω\hbar\omega.

Regarding the spin photocurrent, some authors expected it to follow the spin polarization of the initial states [31, 32, 33]. At the same time, the spin photocurrent is known to depend on the light polarization and experimental geometry [34, 35, 36, 37, 38]. Ref. [39] admits the important role of final-state effects but concludes that CD is a suitable method to analyze spin texture. In a simple pseudospin model [40], CD can be related to the spin polarization of the initial states, and the azimuthal dependence of circular dichroism from the topological surface states (TSS) in Bi2_2Se3_3 was experimentally observed in refs. [21] and [40] to follow the spin texture of the TSS model. On the contrary, the theoretical analysis in Ref. [26] established a final-state origin of CD in Bi2_2Te3_3 and demonstrated that the spin photocurrent was not related to the sign of CD. This point of view was confirmed experimentally for Bi2_2Se3_3 in Ref. [28]. There, over a wide ℏω\hbar\omega interval the spin texture was found to be independent of light polarization, from which the authors concluded that the photoelectron spin polarization was “representative of the initial TSS”.

The present paper theoretically addresses the above issues for the surface state on the Bi(111) surface with the aim to reveal possible difficulties that may prevent CD from being interpreted solely in terms of properties of initial states. Here, the azimuthal distribution of spin-resolved CD and photoelectron spin polarization in the photon-energy interval ℏω=7−19\hbar\omega= 7 - 19 eV is studied using an ab initio one-step theory of photoemission. In this proof-of-principle calculation we are interested in the symmetry of the spin-dependent CD distribution and its ℏω\hbar\omega dependence. The Bi(111) surface is interesting because of the very strong spin-orbit coupling (SOC), which makes it of practical importance for spin-to-charge current conversion [41]. It is particularly instructive owing to its C3vC_{3v} symmetry, which is common to many topological insulators, transition-metal dichalcogenides, as well as elementary metals—popular objects of spectroscopic studies.

Theoretical background and computational methodology

Angle-resolved photocurrent is calculated based on the one-step theory of photoemission [9, 10, 11, 12, 13], in which the optical excitation of the electron, its scattering by the crystal-vacuum interface, and transport to the detector are subsumed into the time reversed LEED (low energy electron diffraction) state. The photocurrent in vacuum excited by light of frequency ω\omega from the initial state ∣ik∥⟩|i\mathbf{k}_{\parallel}\rangle reads

I(ω,k∥)∼Ef−Evac∣⟨fk∥∣o^∣ik∥⟩∣2.(1)I(\omega,\mathbf{k}_{\parallel}) \sim\sqrt{E_f-E_{\mathrm{vac}}}\left|\langle f\mathbf{k}_{\parallel}|\hat{o}|i\mathbf{k}_{\parallel}\rangle\right|^2 . \tag*{(1)}

Here ⟨r∣fk∥⟩≡Φ(r)\langle\mathbf{r}|f\mathbf{k}_{\parallel}\rangle\equiv\Phi(\mathbf{r}) is the time reversed LEED wave function at energy Ef=Ei+ℏωE_f=E_i+\hbar\omega and surface-parallel crystal momentum k∥\mathbf{k}_{\parallel}. EvacE_{\mathrm{vac}} is the vacuum level (kinetic energy zero). Equation (1) stems from the well-established time-dependent perturbation theory, so the accuracy of the photocurrent depends on the accuracy of the wave functions of the quasi-particle states ∣fk∥⟩|f\mathbf{k}_{\parallel}\rangle and ∣ik∥⟩|i\mathbf{k}_{\parallel}\rangle. Here, both the initial and the final states are eigenfunctions of the Kohn-Sham Hamiltonian obtained within the local density approximation of the density functional theory.

The interaction with the photon field is calculated in the non-relativistic dipole approximation, i.e., the spin-orbit related term [15] is neglected. The validity of this approximation is discussed in Ref. [7]. Thus, the perturbation reads o^=−i∇⋅ϵ\hat{o}=-i\nabla\cdot\boldsymbol{\epsilon}, where ϵ\boldsymbol{\epsilon} is a complex polarization vector. Circular dichroism is the difference between the intensities for circular right (CR) and circular left (CL) light polarizations, with ϵCR=ξ−iη\boldsymbol{\epsilon}_{\mathrm{CR}}=\boldsymbol{\xi}-i\boldsymbol{\eta} and ϵCL=ξ+iη\boldsymbol{\epsilon}_{\mathrm{CL}}=\boldsymbol{\xi}+i\boldsymbol{\eta}, where ξ\boldsymbol{\xi} and η\boldsymbol{\eta} are two orthogonal unit vectors in the plane perpendicular to the light propagation, see Figure 1(a) and Figure 1(b), and it arises from the interference of two transition amplitudes: ϵ=ξ\boldsymbol{\epsilon}=\boldsymbol{\xi} and ϵ=η\boldsymbol{\epsilon}=\boldsymbol{\eta}. Note that here the dichroism is defined as the difference D=ICR−ICLD=I_{\mathrm{CR}}-I_{\mathrm{CL}}, which is not normalized to the total intensity I=ICR+ICLI=I_{\mathrm{CR}}+I_{\mathrm{CL}}. This is convenient when one considers the k∥\mathbf{k}_{\parallel} distribution of the dichroism because it gives a better idea of what would be the result of the angular averaging.

Figure 1. Light incidence geometry and local coordinate frame. (a) Crystal surface projection of the light incidence plane at the azimuth ϕ\phi and ss-polarization vector ξ\boldsymbol{\xi}. (b) Light incidence angle θ\theta and pp-polarization vector η\boldsymbol{\eta}. (c) Local coordinate frame at the k∥\mathbf{k}_{\parallel} azimuth α\alpha.

The LEED wave function describes the scattering of the plane wave exp⁡(−ikr)\exp(-i\mathbf{k}\mathbf{r}) incident from vacuum on the surface, where k=k∥+n2mE/ℏ2−k∥2\mathbf{k}=\mathbf{k}_{\parallel}+\mathbf{n}\sqrt{2mE/\hbar^2-k_{\parallel}^2}. In the semi-infinite crystal Φ\Phi satisfies the Schrödinger equation H^Φ=EΦ\hat{H}\Phi=E\Phi, with the scalar-relativistic Hamiltonian H^\hat{H}, which contains the imaginary potential −iVi-iV_i to allow for the inelastic scattering of the outgoing electron. Here, a constant value Vi=1V_i=1 eV is assumed.

In the LEED calculation the electron is incident from the right, and in the crystal half-space z<0z<0 the wave function Φ(r)\Phi(\mathbf{r}) is represented by a set of the Bloch eigenstates, both propagating and evanescent. The set of the Bloch solutions for a given EfE_f and k∥\mathbf{k}_{\parallel} is obtained with the inverse k⋅p\mathbf{k}\cdot\mathbf{p} method in terms of augmented plane waves [42]. At the matching plane z=0z=0, see Figure 2, the sum of the bulk solutions is matched in value and slope to a sum of the incident and reflected plane waves in vacuum, as explained in Ref. [43]. Thus, the time-reversed LEED wave function fully takes into account the singularity of the crystal potential at the nuclei and its variation in the interstitial. Finally, the Laue representation of the final state is obtained by a straightforward expansion of the wave functions in terms of plane waves:

Figure 2. (a) G∥=0\mathbf{G}_{\parallel}=0 Fourier component of the time-reversed LEED function for k∥=0\mathbf{k}_{\parallel}=0 and Ef=10E_f=10 eV. (b) Probability density profile ρ(z)\rho(z): red line is total density, see Eq. (3), and dashed blue line is the G∥=0\mathbf{G}_{\parallel}=0 contribution ∣ϕG∥=0(z)∣2|\phi_{\mathbf{G}_{\parallel}=0}(z)|^2. Orange shaded area shows the contribution from the G∥≠0\mathbf{G}_{\parallel}\ne0 surface Fourier harmonics. Circles show the location of the Bi(111) atomic planes. The matching plane between the semi-infinite crystal and vacuum is at z=0z=0.

Φ(r∥,z)=∑G∥ϕG∥(z)exp⁡[i(k∥+G∥)r∥],(2)\Phi(\mathbf{r}_{\parallel},z)=\sum_{\mathbf{G}_{\parallel}}\phi_{\mathbf{G}_{\parallel}}(z)\exp[i(\mathbf{k}_{\parallel}+\mathbf{G}_{\parallel})\mathbf{r}_{\parallel}], \tag*{(2)}

where r∥\mathbf{r}_{\parallel} is the position vector parallel to the surface and G∥\mathbf{G}_{\parallel} are surface reciprocal lattice vectors. An example of the scattering solution for normal emission at E=10E=10 eV relative to the Fermi energy is presented in Figure 2 as the zeroth Fourier component ϕ0(z)\phi_0(z) and density profile

ρ(z)=∫∣Φ(r∥,z)∣2 dr∥.(3)\rho(z) = \int\left|\Phi(\mathbf{r}_{\parallel},z)\right|^2 \,d\mathbf{r}_{\parallel}. \tag*{(3)}

Note that in the interior of the crystal the contribution from the Fourier harmonics with reciprocal lattice vectors G∥≠0\mathbf{G}_{\parallel} \ne0 (orange) is comparable to the G∥=0\mathbf{G}_{\parallel}=0 contribution, which points to possible problems with applying a free-electron approximation for the final state [44].

Spin structure of the Bi(111) surface state

Spin-orbit effects at surfaces are often discussed in terms of the spin direction of the electronic states. In the spirit of the Rashba model, much attention is paid to the details of the spin-momentum locking and its relation to spin-polarized photocurrent [45]. However, in real crystals the wave functions are generally not spin eigenfunctions, i.e., the spin direction varies from one real space point to another. A direct consequence of this is the dependence of the spin photocurrent on the geometrical setup of the photoemission measurement and on the photon energy [26, 30, 38], as we will see in the next section. Both the spin polarization of the photocurrent and dichroism depend on the spinor structure of the initial state owing to the different dependence on ϵ\epsilon of the matrix elements ⟨fk∥∣o^∣ik∥↑⟩\langle f_{\mathbf{k}_{\parallel}}|\hat{o}|i^{\uparrow}_{\mathbf{k}_{\parallel}}\rangle and ⟨fk∥∣o^∣ik∥↓⟩\langle f_{\mathbf{k}_{\parallel}}|\hat{o}|i^{\downarrow}_{\mathbf{k}_{\parallel}}\rangle. It is most convenient to consider the properties of the spinor components in the local coordinate frame, see Figure 1(c): for a given k∥\mathbf{k}_{\parallel} azimuth we introduce unit vectors κ\boldsymbol{\kappa} along k∥\mathbf{k}_{\parallel} and τ=n×κ\boldsymbol{\tau}=\mathbf{n}\times\boldsymbol{\kappa}, where n\mathbf{n} is the surface normal. Then, for the spin quantization direction s∥τ\mathbf{s}\parallel\boldsymbol{\tau}, as zeroth approximation, one may start with the solution for a semi-infinite jellium (Rashba model) and consider the spin-↑\uparrow component to be an even function under the transformation τ→−τ\boldsymbol{\tau}\to-\boldsymbol{\tau} and the spin-↓\downarrow one to be negligible. Figure 3 demonstrates that the spin structure of the Bi(111) surface state substantially deviates from this simple picture.

Figure 3. Density profiles and k∥\mathbf{k}_{\parallel} dispersion of the surface states on Bi(111). (a) Spin-density profiles of the surface state I in the ΓˉMˉ\bar{\Gamma}\bar{M} direction at k∥=0.1k_{\parallel}=0.1 and 0.30.3 Å−1^{-1} [circles in graph (d)]. Spin axis s\mathbf{s} is along yy, see Figure 1(c). (b) Total density distribution for this state for k∥k_{\parallel} along ΓˉMˉ\bar{\Gamma}\bar{M} and ΓˉKˉ\bar{\Gamma}\bar{K}. (c) Net-spin density distribution for s⊥k∥\mathbf{s}\perp\mathbf{k}_{\parallel}. Blue color means that the minority spin density is larger than majority at this zz. The ticks in the zz axis of graphs (b) and (c) denote the atomic planes. (d) Energy dispersion E(k∥)E(\mathbf{k}_{\parallel}) of the two surface states.

The surface states are calculated as eigenfunctions of a repeated-slab crystal, with a 70 Å wide supercell that comprises a 32-layer Bi(111) slab and a 9 Å vacuum clearance. The crystal potential is artificially perturbed at one of the surfaces in order to separate the surface states at the opposite surfaces and create physically relevant non-degenerate states localized at the other surface, see Figure 3(a). The calculations are performed with the self-consistent full-potential linear augmented plane waves method [46]. Spin-orbit coupling is included within the second-variational two-component Koelling-Harmon approximation [47, 48], with the potential gradient being taken into account only in the muffin-tin spheres.

For a given k∥\mathbf{k}_{\parallel}, the slab eigenstates are a set of discrete levels, which are related to bulk or to surface states of the true semi-infinite crystal depending on how strongly they are localized. Figure 3(d) shows the dispersion E(k∥)E(\mathbf{k}_{\parallel}) of

the two levels that represent the two spin-orbit-split surface states. The E(k∥)E(\mathbf{k}_{\parallel}) curves along ΓˉMˉ\bar{\Gamma}\bar{M} and ΓˉKˉ\bar{\Gamma}\bar{K} agree well with the seminal calculation of Ref. [49]. In the vicinity of the Γˉ\bar{\Gamma} point the surface states merge with the bulk continuum, which in the slab calculation is reflected in the extended character of the density profile ρ(z)\rho(z) in Figure 3(b) at small k∥k_{\parallel}. In the ΓˉKˉ\bar{\Gamma}\bar{K} direction the upper state disperses steeply to merge with bulk continuum at around k∥=0.2k_{\parallel}=0.2 Å−1^{-1}, while along ΓˉMˉ\bar{\Gamma}\bar{M} it remains localized over much wider k∥k_{\parallel} range.

Figure 3(a) shows the spin density profiles for two states in the ΓˉMˉ\bar{\Gamma}\bar{M} line. It lies in the mirror plane xzxz of the (111) surface, see Figure 1(c), and for the spin axis along yy the two spinor components are eigenfunctions of different parity of the reflection operator y→−yy\to-y [37].

For the higher energy state in Figure 3(d) the majority spin component ψ↑\psi^{\uparrow} is even and the minority one ψ↓\psi^{\downarrow} is odd; the spin density profiles are shown in Figure 3(a). Minority spin comprises some 10 to 17% of the total charge, and the spin polarization

S=∫∣ψ↑(r)∣2−∣ψ↓(r)∣2 dr(4)S = \int\left\lvert\psi^{\uparrow}(\mathbf{r}) \right\rvert^{2} - \left\lvert\psi^{\downarrow}(\mathbf{r}) \right\rvert^{2}\,d\mathbf{r} \tag*{(4)}

does not exceed 80%. Similar values occur for the Γ‾K‾\overline{\Gamma}\overline{K} direction, i.e., for k∥\mathbf{k}_{\parallel} along yy and spin axis along −x-x. At the same time, locally the spin minority density may exceed the majority one, as illustrated in the color map Figure 3(c), e.g., around k∥=0.3\mathbf{k}_{\parallel} = 0.3 Å−1^{-1} along Γ‾M‾\overline{\Gamma}\overline{M}.

Results and discussion

Let us focus on the light-polarization dependence of the emission intensity from the upper surface state. In spite of the considerable deviation from the spin structure in the Rashba model, for the spin axis perpendicular to k∥\mathbf{k}_{\parallel} the spin-up component is significantly larger than spin-down. In the next sections we will consider the circular dichroism for the three directions of the spin axis: τ\boldsymbol{\tau}, κ\boldsymbol{\kappa}, and n\mathbf{n}.

Normal light incidence

The k∥\mathbf{k}_{\parallel} dependence of the dichroism of photoemission at normal incidence for three spin directions is shown in Figure 4 and Figure 5 for ℏω=10.5\hbar\omega= 10.5 and 18 eV, respectively. The color maps are seen to have different symmetry depending on the spin quantization axis s\mathbf{s}: for s∥τ\mathbf{s} \parallel\boldsymbol{\tau} the maps are antisymmetric on reflection in the Γ‾M‾\overline{\Gamma}\overline{M} plane: Dτ↑↓(α)=−Dτ↑↓(−α)D_{\tau}^{\uparrow\downarrow}(\alpha) = -D_{\tau}^{\uparrow\downarrow}(-\alpha) (here subscript τ\tau indicates the choice of the spin quantization axis), and the maps for opposite spins are independent. For s∥κ\mathbf{s} \parallel\boldsymbol{\kappa} and s∥n\mathbf{s} \parallel\mathbf{n} the maps do not possess the mirror symmetry, but the dichroic signals for the opposite spin directions are related: Ds↑(α)=−Ds↓(−α)D_{s}^{\uparrow}(\alpha) = -D_{s}^{\downarrow}(-\alpha).

Figure 4. Normal incidence CD for ℏω=10.5\hbar\omega= 10.5 eV for three spin directions in the local k∥\mathbf{k}_{\parallel}-dependent frame, see Figure 1(c): s∥τ\mathbf{s} \parallel\boldsymbol{\tau} (left column), s∥n\mathbf{s} \parallel\mathbf{n} (central column), and s∥κ\mathbf{s} \parallel\boldsymbol{\kappa} (right column). Upper row shows spin-↑\uparrow and lower row spin-↓\downarrow CD. Maps are normalized to the maximum magnitude.

Figure 5. The same as in Figure 4 for ℏω=18\hbar\omega= 18 eV.

The total photocurrent CD, i.e., the sum of the maps D=Ds↑+Ds↓D = D_{s}^{\uparrow} + D_{s}^{\downarrow} is independent of the s\mathbf{s} direction, and it has the same symmetry as the Dτ↑↓D_{\tau}^{\uparrow\downarrow} maps. For k∥\mathbf{k}_{\parallel} sufficiently close to Γ‾\overline{\Gamma} these maps are approximately antisymmetric across the Γ‾\overline{\Gamma} point (resembling the symmetry of the ground-state spin texture), but at larger k∥\mathbf{k}_{\parallel} the deviation from the relation D(α+π)≈−D(α)D(\alpha+ \pi) \approx-D(\alpha) becomes ever stronger.

For all the choices of the spin axis CD is seen to strongly depend on the magnitude of the k∥\mathbf{k}_{\parallel} vector: it may change sign, and this happens at different k∥\mathbf{k}_{\parallel} for different α\alpha. Generally, the structure of all the maps is rather complicated, which reflects the complicated spatial and spin structure of the surface state. Furthermore, the maps strongly depend on the photon energy ℏω\hbar\omega. Nevertheless, gross symmetry features of the maps can be understood from a simplistic model that neglects the zz-coordinate of the wave functions. Then the general expression for the initial state spinor along the spin-axis τ\tau of the local coordinate frame (κ,τ)(\kappa,\tau) reads

ψα(κ,τ)=[(gα++uα+)exp⁡(+i2α)(gα−+uα−)exp⁡(−i2α)]τ(5)\psi_{\alpha}(\kappa,\tau)=\begin{bmatrix}(g_{\alpha}^{+}+u_{\alpha}^{+})\exp\left(+\frac{i}{2}\alpha\right)\\(g_{\alpha}^{-}+u_{\alpha}^{-})\exp\left(-\frac{i}{2}\alpha\right)\end{bmatrix}_{\tau} \tag*{(5)}

where the functions gα±(κ,τ)g_{\alpha}^{\pm}(\kappa,\tau) and uα±(κ,τ)u_{\alpha}^{\pm}(\kappa,\tau) are even and odd, respectively, upon the reflection τ→−τ\tau\to-\tau. Here the superscripts ±\pm are used for the spinor components along τ\tau to distinguish from the general notation ↑↓\uparrow\downarrow. On a surface of C3vC_{3v} symmetry, gα±g_{\alpha}^{\pm} and uα±u_{\alpha}^{\pm} are 2π/32\pi/3-periodic even functions of the azimuthal angle α\alpha, and gα±g_{\alpha}^{\pm} and uα±u_{\alpha}^{\pm} are odd functions of α\alpha. In order to construct a minimal model to approximately parametrize the variety of the data for the different spin directions let us restrict ourselves to the first non-vanishing terms in the Fourier expansion of the azimuthal dependence of the functions g±g^{\pm} and u±u^{\pm}, i.e., a constant for the even and sin⁡3α\sin3\alpha for the odd functions of α\alpha:

ψα(κ,τ)=[(g++u+sin⁡3α)exp⁡(+i2α)(g−sin⁡3α+u−)exp⁡(−i2α)]τ(6)\psi_{\alpha}(\kappa,\tau)=\begin{bmatrix}(g^{+}+u^{+}\sin3\alpha)\exp\left(+\frac{i}{2}\alpha\right)\\(g^{-}\sin3\alpha+u^{-})\exp\left(-\frac{i}{2}\alpha\right)\end{bmatrix}_{\tau} \tag*{(6)}

For example, in a circularly symmetric Rashba system the spin-orbit-split constant energy contours form two concentric circles, one of which is described by Eq. (6) with g+g^{+} being the only non-zero term, and the other one has the same structure of the spinor for the opposite spin direction. On a C3vC_{3v} surface, for k∥k_{\parallel} along the mirror plane ΓˉMˉ\bar{\Gamma}\bar{M}, the spinor components are of pure parity, i.e., in the general equation (5) we have u0+=0u_{0}^{+}=0 and g0−=0g_{0}^{-}=0. This means that for the light incident in the xzxz plane a spin flip occurs on the change from pp to ss linear polarization [37].

Assuming the final state Φα(κ,τ)\Phi_{\alpha}(\kappa,\tau) to be an α\alpha-independent scalar function even on reflection τ→−τ\tau\to-\tau we arrive at the angular dependence of the dichroism of the form Dτ(α)∼sin⁡3αD_{\tau}(\alpha)\sim\sin3\alpha. It should be however noted that the function Φ\Phi cannot be approximated by a single plane wave exp⁡(−ik∥r∥)\exp(-i\mathbf{k}_{\parallel}\mathbf{r}_{\parallel}) because in this case the circular dichroism at normal incidence vanishes. For the spin directions nn and κ\kappa this model yields the dichroism maps of a pure C3C_{3} symmetry, in agreement with the ab initio calculations in Figure 4 and Figure 5.

Note that the simplistic model also has the property Dτ(α)=−Dτ(α+π)D_{\tau}(\alpha)=-D_{\tau}(\alpha+\pi), which is missing in the real surface. The violation of this relation is caused by the fact that the final states Φ\Phi at k∥\mathbf{k}_{\parallel} and −k∥-\mathbf{k}_{\parallel} are not related by symmetry, nor are they related by time reversal because Φ\Phi and Φ∗\Phi^{*} propagate in opposite directions perpendicular to the surface. In other words, the additional symmetry is brought about by ignoring the zz-component of the wave functions. In this regard it should be noted that the suggestion in Ref. [34] to use for the final states the same time-reversal relation as for the initial states is incompatible with the one-step-theory understanding of the final state ∣fk∥⟩\lvert f_{\mathbf{k}_{\parallel}}\rangle as time-reversed LEED state [12]. (We note also that a nearly-free-electron approximation for the final state, which assumes Φ\Phi to be a single plane wave, does not change anything in this respect.) The “apparent time-reversal symmetry breaking” was thoroughly experimentally studied in Ref. [36] for Bi2_2Se3_3 and analyzed allowing for the three-dimensional nature of the surface states, and it was concluded to be caused by a “layer-dependent spin-orbital entanglement”. In fact, this effect is quite general and occurs also without SOC.

This aspect has important implications also on the angular distribution of the spin photocurrent. For example, in the measurement of Bi thin film on Si(111) [31] the in-plane spin-polarization of the photocurrent was found to be “remarkably suppressed on a half of six elongated hole pockets” in apparent contradiction to the behavior of the polarization of the initial states, see Eq. (4), whose spinor wave functions for k∥\mathbf{k}_{\parallel} and −k∥-\mathbf{k}_{\parallel} are related by time-reversal symmetry, so Sτ(−k∥)=Sτ(k∥)S_{\tau}(-\mathbf{k}_{\parallel}) = S_{\tau}(\mathbf{k}_{\parallel}). As is clear from the above discussion, the result of Ref. [31] is in full accord with the one-step theory, which we illustrate in the spin-polarization maps in Figure 6. In order to retain the crystal symmetry in the numerical experiment, let us calculate the sum of the net-spin photocurrents for CR and CL polarizations for a normal light incidence. The s∥τ\mathbf{s} \parallel\boldsymbol{\tau} map is clearly seen to have the C3vC_{3v} symmetry, and the other two maps are odd on reflection in the mirror planes. Interestingly, at ℏω=18\hbar\omega= 18 eV for s∥τ\mathbf{s} \parallel\boldsymbol{\tau} the spin-minority photocurrent dominates just at the mirror planes, but the picture rapidly changes with photon energy: photocurrent from the same state may be spin-up or spin-down depending on ℏω\hbar\omega. Another observation in Ref. [31] was that the out-of-plane net spin photocurrent was comparable to the in-plane one. The present calculations are compatible with this result, but the actual intensity ratio depends on k∥\mathbf{k}_{\parallel} and on ℏω\hbar\omega: For example, for k∥=0.14k_{\parallel} = 0.14 Å−1^{-1} the net spin amplitudes along τ\boldsymbol{\tau} and n\mathbf{n} are approximately equal at ℏω=8\hbar\omega= 8–10 eV, and around 20 eV the out-of-plane photocurrent becomes 10 times weaker, while the in-plane spin amplitude remains of the same order of magnitude. Thus, our analysis fully supports the interpretation in the spin asymmetry Ref. [50] as due to final-states effects.

Figure 6. Spin polarization of the photocurrent at normal light incidence with a “symmetrized” polarization: S=ICR↑+ICL↑−ICR↓−ICL↓S = I_{\mathrm{CR}}^{\uparrow} + I_{\mathrm{CL}}^{\uparrow} - I_{\mathrm{CR}}^{\downarrow} - I_{\mathrm{CL}}^{\downarrow}. Left to right: s∥τ\mathbf{s} \parallel\boldsymbol{\tau}, s∥n\mathbf{s} \parallel\mathbf{n}, and s∥κ\mathbf{s} \parallel\boldsymbol{\kappa}. Maps are normalized to the maximum magnitude.

Off-normal light incidence

In a real experimental setup, the light cannot be directed perpendicular to the surface, so the photoemission intensity depends on the location of the light incidence plane, angle ϕ\phi in Figure 1(a), and on the angle of incidence θ\theta, Figure 1(b). For θ≠0\theta\ne0, vector η\boldsymbol{\eta} acquires a surface-perpendicular component, and the symmetry of the angular distribution of the photocurrent becomes lower than in the θ=0\theta= 0 case. Figure 7 shows the spin-unresolved CD for ℏω=15\hbar\omega= 15 eV for the light incidence along ΓˉMˉ\bar{\Gamma}\bar{M} and ΓˉKˉ\bar{\Gamma}\bar{K}. In the former case, the experimental setup retains the mirror symmetry across the xzxz plane: D(α)=−D(−α)D(\alpha) = -D(-\alpha). Note, however, that with increasing θ\theta the angular distribution of CD gradually becomes positive everywhere in the upper half-plane and negative in the lower one, i.e., D(α)D(\alpha) approaches the symmetry of sin⁡α\sin\alpha.

Figure 7. Off-normal incidence total-current CD for ℏω=15\hbar\omega= 15 eV for two locations of the light incidence plane: xzxz plane (upper row) and yzyz plane (lower row). Four light-incidence angles are shown, left to right: θ=5∘\theta= 5^\circ, 15∘15^\circ, 45∘45^\circ, and 90∘90^\circ. Maps are normalized to the maximum magnitude.

For the light incidence along ΓˉKˉ\bar{\Gamma}\bar{K} the CD maps possess no symmetry, but again, with increasing θ\theta its shape becomes more and more close to the symmetry of sin⁡(α−π/2)\sin(\alpha- \pi/2), i.e., positive almost everywhere in the left half-plane and negative in the right one. Note that for an isotropic surface it holds D(α)∼sin⁡(α+ϕ)D(\alpha) \sim\sin(\alpha+\phi), so the deviation from this pattern characterizes the anisotropy of the electronic states involved. For example, the anisotropy of the surface states on Cu(111) and Au(111) is much lower, and D(α)D(\alpha) is rather close to the sin⁡α\sin\alpha shape [22]. At the same time, tangible anisotropy is observed in the D(α)D(\alpha) of the Fermi contour of Bi2_2Se3_3 [21]. Figure 7 shows that for both ϕ\phi angles, for θ=π/4\theta= \pi/4 the D(α)D(\alpha) maps are almost the same as for the grazing incidence, θ=π/2\theta= \pi/2. This means that the n-component of the momentum matrix element is much larger than the in-plane components, and that the θ\theta-dependence of the angular distribution of the azimuthal distribution of the photocurrent may give information about their relative magnitude.

Regarding the spin dependence of CD at θ≠0\theta\ne0, it follows from the spin structure analysis in section III that for the tangential spin s∥τ\mathbf{s} \parallel\boldsymbol{\tau}, the spin-majority CD+\mathrm{CD}^{+} follows rather closely the sin⁡(α+φ)\sin(\alpha+\varphi) pattern because the majority spinor component is dominated by the even function gα+g_{\alpha}^{+}, see Eq. (5), whereas the spin-minority CD−\mathrm{CD}^{-} has a more complicated angular dependence.

Conclusion

To summarize, the presented calculations demonstrate that circular dichroism in photoemission cannot be considered an intrinsic characteristic of the initial state even if the experimental setup does not break the symmetry of the surface. A conspicuous manifestation of the problem is the lack of a time-reversal relation between photoemission from k∥\mathbf{k}_{\parallel} and −k∥-\mathbf{k}_{\parallel} in contrast to its existence for the initial states. The present results indicate that the asymmetry between k∥\mathbf{k}_{\parallel} and −k∥-\mathbf{k}_{\parallel} observed in many experiments is not necessarily an artifact of the experimental setup but is in full accord with the rigorous and well-grounded ab initio one-step theory. The one-step formalism [9, 10, 11, 12, 13] makes quite transparent the origin of the asymmetry: the presence of the time-reversed LEED state in the matrix element, for which the relation Φ(k∥)=Φ∗(−k∥)\Phi(\mathbf{k}_{\parallel}) = \Phi^{*}(-\mathbf{k}_{\parallel}) does not hold.

Similarly, the strong photon energy dependence of the CD maps for all the spin directions considered and its symmetry being lower than the symmetry of the initial state spin texture points to the absence of a direct connection between CD and the spin polarization (which may appear as an artifact of a simplified theoretical model [40]). This agrees with the conclusion of Ref. [26] regarding a minor role of spin in CD.

Acknowledgements

This work was supported by the Spanish Ministry of Science and Innovation (Projects No. PID2022-139230NB-I00 and No. PID2022-138750NB-C22).

References

  1. [5]U. Heinzmann and J. H. Dil, Spin–orbit-induced photoelectron spin polarization in angle-resolved photoemission from both atomic and condensed matter targets, J. Phys.: Condensed Matter 24, 173001 (2012).DOI
  2. [6]T. Okuda and A. Kimura, Spin- and angle-resolved photoemission of strongly spin–orbit coupled systems, Journal of the Physical Society of Japan 82, 021002 (2013).DOI
  3. [7]E. E. Krasovskii, Spin-orbit coupling at surfaces and 2d materials, Journal of Physics: Condensed Matter 27, 493001 (2015).DOI
  4. [8]T. Okuda, Recent trends in spin-resolved photoelectron spectroscopy, Journal of Physics: Condensed Matter 29, 483001 (2017).DOI
  5. [9]I. Adawi, Theory of the surface photoelectric effect for one and two photons, Phys. Rev. 134, A788 (1964).DOI
  6. [10]G. D. Mahan, Theory of photoemission in simple metals, Phys. Rev. B 2, 4334 (1970).
  7. [11]C. Caroli, D. Lederer-Rozenblatt, B. Roulet, and D. Saint-James, Inelastic effects in photoemission: Microscopic formulation and qualitative discussion, Phys. Rev. B 8, 4552 (1973).DOI
  8. [12]P. J. Feibelman and D. E. Eastman, Photoemission spectroscopy – correspondence between quantum theory and experimental phenomenology, Phys. Rev. B 10, 4932 (1974).DOI
  9. [13]J. Pendry, Theory of photoemission, Surface Science 57, 679 (1976).
  10. [14]G. Borstel, Theoretical aspects of photoemission, Applied Physics A 38, 193 (1985).
  11. [15]R. Feder, Polarized Electrons in Surface Physics (Advanced Series in Surface Science) (Singapore: World Scientific, 1985).
  12. [16]J. Braun, The theory of angle-resolved ultraviolet photoemission and its applications to ordered materials, Reports on Progress in Physics 59, 1267 (1996).DOI
  13. [17]J. Braun, J. Minár, and H. Ebert, Correlation, temperature and disorder: Recent developments in the one-step description of angle-resolved photoemission, Physics Reports 740, 1 (2018).DOI
  14. [18]E. Tamura, W. Piepke, and R. Feder, New spin-polarization effect in photoemission from nonmagnetic surfaces, Phys. Rev. Lett. 59, 934 (1987).
  15. [19]J. Henk and B. Johansson, Magnetic dichroism in off-normal valence-band photoemission, Journal of Electron Spectroscopy and Related Phenomena 94, 259 (1998).DOI
  16. [20]W. Jung, Y. Kim, B. Kim, Y. Koh, C. Kim, M. Matsunami, S.-i. Kimura, M. Arita, K. Shimada, J. H. Han, J. Kim, B. Cho, and C. Kim, Warping effects in the band and angular-momentum structures of the topological insulator Bi₂Te₃, Phys. Rev. B 84, 245435 (2011).DOI
  17. [21]S. R. Park, J. Han, C. Kim, Y. Y. Koh, C. Kim, H. Lee, H. J. Choi, J. H. Han, K. D. Lee, N. J. Hur, M. Arita, K. Shimada, H. Namatame, and M. Taniguchi, Chiral orbital-angular momentum in the surface states of Bi₂Se₃, Phys. Rev. Lett. 108, 046805 (2012).
  18. [22]B. Kim, C. H. Kim, P. Kim, W. Jung, Y. Kim, Y. Koh, M. Arita, K. Shimada, H. Namatame, M. Taniguchi, J. Yu, and C. Kim, Spin and orbital angular momentum structure of Cu(111) and Au(111) surface states, Phys. Rev. B 85, 195402 (2012).DOI
  19. [23]S. Cho, J.-H. Park, J. Hong, J. Jung, B. S. Kim, G. Han, W. Kyung, Y. Kim, S.-K. Mo, J. D. Denlinger, J. H. Shim, J. H. Han, C. Kim, and S. R. Park, Experimental observation of hidden berry curvature in inversion-symmetric bulk 2h–wse₂, Phys. Rev. Lett. 121, 186401 (2018).DOI
  20. [24]M. Schüler, U. D. Giovannini, H. Hübener, A. Rubio, M. A. Sentef, and P. Werner, Local berry curvature signatures in dichroic angle-resolved photoelectron spectroscopy from two-dimensional materials, Science Advances 6, eaay2730 (2020), https://www.science.org/doi/10.1126/sciadv.aay2730.DOI
  21. [25]M. Bahramy, P. King, A. de la Torre, J. Chang, M. Shi, L. Patthey, G. Balakrishnan, P. Hofmann, R. Arita, N. Nagaosa, and F. Baumberger, Emergent quantum confinement at topological insulator surfaces, Nat. Commun. 3, 1159 (2012).
  22. [26]M. R. Scholz, J. Sánchez-Barriga, J. Braun, D. Marchenko, A. Varykhalov, M. Lindroos, Y. J. Wang, H. Lin, A. Bansil, J. Minár, H. Ebert, A. Volykhov, L. V. Yashina, and O. Rader, Reversal of the circular dichroism in angle-resolved photoemission from Bi₂Te₃, Phys. Rev. Lett. 110, 216801 (2013).
  23. [27]M. Neupane, S. Basak, N. Alidoust, S.-Y. Xu, C. Liu, I. Belopolski, G. Bian, J. Xiong, H. Ji, S. Jia, S.-K. Mo, M. Bissen, M. Severson, H. Lin, N. P. Ong, T. Durakiewicz, R. J. Cava, A. Bansil, and M. Z. Hasan, Oscillatory surface dichroism of the insulating topological insulator Bi₂Te₂Se, Phys. Rev. B 88, 165129 (2013).
  24. [28]J. Sánchez-Barriga, A. Varykhalov, J. Braun, S.-Y. Xu, N. Alidoust, G. Kornilov, J. Minár, K. Hummer, G. Springholz, G. Bauer, R. Schumann, L. V. Yashina, H. Ebert, M. Z. Hasan, and O. Rader, Photoemission of Bi₂Se₃ with circularly polarized light: Probe of spin polarization or means for spin manipulation?, Phys. Rev. X 4, 011046 (2014).
  25. [29]A. Crepaldi, F. Cilento, M. Zacchigna, M. Zonno, J. C. Johannsen, C. Tournier-Colletta, L. Moreschini, I. Vobornik, F. Bondino, E. Magnano, H. Berger, A. Magrez, P. Bugnon, G. Autès, O. V. Yazyev, M. Grioni, and F. Parmigiani, Momentum and photon energy dependence of the circular dichroic photoemission in the bulk Rashba semiconductors BiTeX (X = I, Br, Cl), Phys. Rev. B 89, 125408 (2014).DOI
  26. [30]M. Arrälä, J. Nieminen, J. Braun, H. Ebert, and M. Lindroos, Photon energy dependence of circular dichroism of the Au(111) surface state, Phys. Rev. B 88, 195413 (2013).DOI
  27. [31]A. Takayama, T. Sato, S. Souma, and T. Takahashi, Giant out-of-plane spin component and the asymmetry of spin polarization in surface Rashba states of bismuth thin film, Phys. Rev. Lett. 106, 166401 (2011).
  28. [32]Z.-H. Pan, E. Vescovo, A. V. Fedorov, D. Gardner, Y. S. Lee, S. Chu, G. D. Gu, and T. Valla, Electronic structure of the topological insulator Bi₂Se₃ using angle-resolved photoemission spectroscopy: Evidence for a nearly full surface spin polarization, Phys. Rev. Lett. 106, 257004 (2011).
  29. [33]P. Höpfner, J. Schäfer, A. Fleszar, J. H. Dil, B. Slomski, F. Meier, C. Loho, C. Blumenstein, L. Patthey, W. Hanke, and R. Claessen, Three-dimensional spin rotations at the Fermi surface of a strongly spin-orbit coupled surface system, Phys. Rev. Lett. 108, 186801 (2012).
  30. [34]C.-H. Park and S. G. Louie, Spin polarization of photoelectrons from topological insulators, Phys. Rev. Lett. 109, 097601 (2012).
  31. [35]C. Jozwiak, C.-H. Park, K. Gotlieb, C. Hwang, D.-H. Lee, S. G. Louie, J. D. Denlinger, C. R. Rotundu, R. J. Birgeneau, Z. Hussain, and A. Lanzara, Photoelectron spin-flipping and texture manipulation in a topological insulator, Nat. Phys. 9, 293 (2013).DOI
  32. [36]Z.-H. Zhu, C. N. Veenstra, G. Levy, A. Ubaldiini, P. Syers, N. P. Butch, J. Paglione, M. W. Haverkort, I. S. Elfimov, and A. Damascelli, Layer-by-layer entangled spin-orbital texture of the topological surface state in Bi₂Se₃, Phys. Rev. Lett. 110, 216401 (2013).
  33. [37]J. Henk, A. Ernst, and P. Bruno, Spin polarization of the L-gap surface states on Au(111), Phys. Rev. B 68, 165416 (2003).
  34. [38]H. Bentmann, H. Maaß, E. E. Krasovskii, T. R. F. Peixoto, C. Seibel, M. Leandersson, T. Balasubramanian, and F. Reinert, Strong linear dichroism in spin-polarized photoemission from spin-orbit-coupled surface states, Phys. Rev. Lett. 119, 106401 (2017).
  35. [39]H. Mirhosseini and J. Henk, Spin texture and circular dichroism in photoelectron spectroscopy from the topological insulator bi₂te₃: First-principles photoemission calculations, Phys. Rev. Lett. 109, 036803 (2012).
  36. [40]Y. H. Wang, D. Hsieh, D. Pilon, L. Fu, D. R. Gardner, Y. S. Lee, and N. Gedik, Observation of a warped helical spin texture in Bi₂Se₃ from circular dichroism angle-resolved photoemission spectroscopy, Phys. Rev. Lett. 107, 207602 (2011).
  37. [41]M. Isasa, M. C. Martínez-Velarte, E. Villamor, C. Magén, L. Morellón, J. M. De Teresa, M. R. Ibarra, G. Vignale, E. V. Chulkov, E. E. Krasovskii, L. E. Hueso, and F. Casanova, Origin of inverse rashba-edelstein effect detected at the cu/bi interface using lateral spin valves, Phys. Rev. B 93, 014420 (2016).
  38. [42]E. E. Krasovskii and W. Schattke, Surface electronic structure with the linear methods of band theory, Phys. Rev. B 56, 12874 (1997).
  39. [43]E. E. Krasovskii and W. Schattke, Calculation of the wave functions for semi-infinite crystals with linear methods of band theory, Phys. Rev. B 59, R15609 (1999).
  40. [44]E. E. Krasovskii, Character of the outgoing wave in soft x-ray photoemission, Phys. Rev. B 102, 245139 (2020).
  41. [45]D. Y. Usachov, I. A. Nechaev, G. Poelchen, M. Güttler, E. E. Krasovskii, S. Schulz, A. Generalov, K. Kliemt, A. Kraiker, C. Krellner, K. Kummer, S. Danzenbächer, C. Laubschat, A. P. Weber, J. Sánchez-Barriga, E. V. Chulkov, A. F. Santander-Syro, T. Imai, K. Miyamoto, T. Okuda, and D. V. Vyalikh, Cubic rashba effect in the surface spin structure of rare-earth ternary materials, Phys. Rev. Lett. 124, 237202 (2020).
  42. [46]E. E. Krasovskii, F. Starrost, and W. Schattke, Augmented Fourier components method for constructing the crystal potential in self-consistent band-structure calculations, Phys. Rev. B 59, 10504 (1999).
  43. [47]D. D. Koelling and B. N. Harmon, A technique for relativistic spin-polarised calculations, J. Phys. C 10, 3107 (1977).
  44. [48]A. H. MacDonald, W. E. Picket, and D. D. Koelling, A linearised relativistic augmented-plane-wave method utilising approximate pure spin basis functions, J. Phys. C 13, 2675 (1980).
  45. [49]Y. M. Koroteev, G. Bihlmayer, J. E. Gayone, E. V. Chulkov, S. Blügel, P. M. Echenique, and P. Hofmann, Strong spin-orbit splitting on Bi surfaces, Phys. Rev. Lett. 93, 046403 (2004).
  46. [50]K. Miyamoto, H. Miyahara, K. Kuroda, T. Maegawa, A. Kimura, and T. Okuda, Peculiar rashba spin texture induced by C₃ᵥ symmetry on the bi(111) surface revisited, Phys. Rev. B 97, 085433 (2018).

Paper details

vlnc:W_wtmhe3jh
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