# Metadata **Title:** Spin Polarization of Photoelectrons from Topological Insulators **Authors:** Park, Cheol-Hwan; Louie, Steven G. **Year:** 2012 **Tags:** #TI #ARPES [URL](http://dx.doi.org/10.1103/physrevlett.109.097601)|[Bookends](bookends://sonnysoftware.com/18744) --- # Maximum spin Polarization and Photocurrent The measured [[Physics/Spin|spin]]-polarization in spin-[[Angle-Resolved Photoemission Spectroscopy|ARPES]] experiments is defined by: $ P_{\text{max}} = \max_{\hat{t}} \frac{I_{\hat{t}} - I_{-\hat{t}}}{I_{\hat{t}} + I_{-\hat{t}}} $ Where $I_\hat{t}$ and $I_{-\hat{t}}$ are the intensities for the electron spin aligned and anti-aligned with $\hat{t}$ respectively. Here, the unit [[vector]] $\hat{t} = \hat{t}_{\max}$ that maximizes the equation above defines the direction of the electron spin-polarization. The photocurrent from a detector (i.e. an apparatus that selects a specific 3D wavevector and hence an energy) with the spin quantization axis aligned with $\hat{t}$ is: $ I_\hat{t} \propto \left| \sum_{\{f|E_f = E_i + h\nu\}} \langle \hat{t}, \pmb{R}_D | f \rangle \langle f | H^{\text{int}} | i\rangle \right|^2 $ Here $|i\rangle$ is the initial TSS state, $|f\rangle$ is the final photoexcited state, $E_i$ and $E_f$ are their energies. $h\nu$ is the photon energy, and $H^\text{int}$ is the light-matter interaction [[Hamiltonian]] connecting the two states. Finally, $|\hat{t}, \pmb{R}_D\rangle$ is the detected state. Its spin part is the [[eigenstate]] of $\pmb{s}.\hat{t}$ with [[eigenvalue]] +1 and its spatial part is localized at $\pmb{R}_D$, the location of the detector far away from the sample. Here we use $\pmb{s}$ to denote the [[vector]] of [[Pauli matrices]] for [[Spin-Half|spin-1/2]] systems. ## Deriving $I_\hat{t}$ Consider an electronic system described by a single particle [[Hamiltonian]] $H_0$ and a time dependent pertubation of the form $H_{\text{int}}(t) = H^{\text{int}}(\pmb{A}) e^{-2 \pi i \nu t}$, where $\pmb{A}$ is the [[Vector Potential]] and $h\nu$ is the [[photon]] [[energy]]. We are interested in the evolution of an electron which is at [[eigenstate|state]] $|i\rangle$ before turning on the time-dependent perturbation into a final state $|f\rangle$. Thus, the total [[Hamiltonian]] is: $ H(t) = H_0 + H^{\text{int}}(\pmb{A})e^{-2\pi i \nu t} $ Assuming the perturbation is turned on at $t = 0$, the normalized [[wavefunction]] of the electron at time $t$ can be written, to first order, as $ |\psi(t)\rangle = e^{-i E_i t/\hbar} |i\rangle + \sum_f c_f(t) e^{-i E_f t/\hbar}|f\rangle $ where $|f\rangle$ are the possible final states and $c_f(t)$ is the probability amplitude of finding the electron in the state $|f\rangle$ at time $t$. Clearly, $c_f(0) = 0 \, \forall \, f$. Solving the time-dependent [[Schrödinger equation]], we get: $ c_f(t) = -\frac{i}{\hbar}e^{it \frac{\Delta E_f}{2\hbar}} \frac{\sin \frac{\Delta E_f}{2\hbar}}{\frac{\Delta E_f}{2\hbar}} \langle f | H^\text{int} | i \rangle $ where $\Delta E_f \equiv (E_f - E_i) - h\nu$. The transition rate is given by [[Fermi's Golden Rule]] as follows: $ \Gamma_{f \leftarrow i} = \lim_{t\rightarrow \infty} P_f(t)/t = \frac{2 \pi}{\hbar} \delta(E_f - E_i - h\nu) |\langle f | H^{\text{int}}|i\rangle|^2 $ Now, in a spin-resolved [[photoemission]] experiment, the photocurrent is given by $ I_\hat{t} \propto \lim_{t \rightarrow \infty} | \langle \hat{t}, \pmb{R}_D | \psi(t) \rangle|^2/t $ Now, we can write the [[inner product]] $ \begin{align} \langle \hat{t}, \pmb{R}_D | \psi(t) \rangle &= \sum_E \sum_{\{f | E_f = E\}} -\frac{i}{\hbar}e^{it \frac{\Delta E_f}{2\hbar}} \frac{\sin \frac{\Delta E_f}{2\hbar}}{\frac{\Delta E_f}{2\hbar}} \langle f | H^\text{int} | i \rangle e^{-i E_f t/\hbar} \langle \hat{t}, \pmb{R}_D|f\rangle,\\ &= \sum_E -\frac{i}{\hbar}e^{it \frac{\Delta E - 2E}{2\hbar}} \frac{\sin \frac{\Delta E}{2\hbar}}{\frac{\Delta E}{2\hbar}} \sum_{\{f | E_f = E\}} \langle f | H^\text{int} | i \rangle\langle \hat{t}, \pmb{R}_D|f\rangle \end{align} $ where $\Delta E = E - E_i - h\nu$. Here we have used the fact that $\langle \hat{t}, \pmb{R}_D | i\rangle = 0$ since the initial state is localized at the crystal and far away from the detector. ==Obviously==, a product of the terms in the equation above corresponding to different $E$ values does not contribute to the photocurrent calculated from the equation for $I_\hat{t}$ Plugging back into $I_\hat{t}$, and using the equation: $ \lim_{a\rightarrow \infty} \frac{1}{\pi} \frac{\sin^2(ax)}{ax^2} = \delta(x) $ We end up with: $ \begin{align} I_\hat{t} &\propto \sum_E \delta(E - E_i - h\nu) \sum_{\{f | E_f = E\}} |\langle \hat{t}, \pmb{R}_D|f\rangle\langle f | H^\text{int} | i \rangle|^2 \\ &\propto \sum_{\{f | E_f = E_i + h\nu\}} |\langle \hat{t}, \pmb{R}_D|f\rangle\langle f | H^\text{int} | i \rangle|^2 \end{align} $ ## Interpretation of $I_\hat{t}$ The detector reads the spin character of the photoexcited state at $\pmb{R}_D$, far away from the surface sample, thus the near-surface part of the [[wavefunction]] only affects the measured spin indirectly through the [[Matrix Element]] $\langle f | H^{\text{int}}| i \rangle$. Moreover, a summation of the transition amplitude over [[degeneracy|degenerate]] photoexcited states is necessary because we cannot tell, even in principle, which state $|f\rangle$ is involved in detection. We can write $ |f'\rangle = \sum_{\{f|E_f = E_i + h\nu\}} |f\rangle\langle f|H^\text{int}|i\rangle $ Then $ I_\hat{t} = |\langle \hat{t}, \pmb{R}_D|f'\rangle|^2 $ Since at $\pmb{R}_D$, where the measurement is performed, the $|f\rangles are eigenstates of the free-electron [[Hamiltonian]], then we can regard the Bloch periodic part of the state $|f'\rangle$ as a position-independent, two-component [[Spinor]]. Then, we can always find a [[vector]] $\hat{t}'$ such that $|f'\rangle$ is an [[eigenstate]] of $\pmb{s}\cdot \hat{t'}$ with [[eigenvalue]] $+1$ (and thus the [[eigenstate]] of $\pmb{s}\cdot (-\hat{t'})$ with [[eigenvalue]] -1). Recalling that the spin part of $|\hat{t'}, \pmb{R}_D\rangle$ is the [[eigenstate]] of $\pmb{s}\cdot \hat{t'}$ with [[eigenvalue]] +1, we conclude that $I_\hat{t} \neq 0$ and $I_{-\hat{t}}= 0$ for this orientation, and thus, looking back at the definition of $P_\text{max}$, we get that $P_\max = 100\%$. Thus, the degree of spin-polarization of photoejected electrons is always 100% regardless of the initial electronic state. It is important to note that the non-[[degeneracy]] of $|i\rangle$ (the TSS of the TI) is critical for this result. If it is degenerate, a hole will be left in one of the degenerate states after photoexcitation. Thus, we can distinguish in principle between which initial TSS is involved in the measurement, and the detection probability amplitude corresponding to each $|i\rangle$ should be summed then squared, not the other way round. Thus, the photocurrent will be the sum of the contributions coming from all the degenerate initial TSS $|i\rangles, and when this happens, $P_\max \neq 100\%$. For example, in a spin non-degenerate material $I_\hat{t}$ and $I_{-\hat{t}}$ will always be the same regardless of the choice of $\hat{t}$, giving $P_{max} = 0$. ==^^ Dense parapgraph== # Averaged Spin Polarization of the TSS The averaged spin polarization of the TSS is: $ P_\text{ave}^\text{TSS} = |\langle\psi(n, \pmb{k})|\pmb{s}|\psi(n, \pmb{k})\rangle| $ where $|\psi(n, \pmb{k})\rangle$ is the two component [[Spinor]] [[wavefunction]] of the TSS. According to *ab initio* studies, this quantity is about 50%. Because $|\psi(n, \pmb{k})\rangle$ is not an [[eigenstate]] of $\hat{t}\cdot\pmb{s}$ for all $\hat{t}$ due to [[Spin-Orbit Coupling|SOC]], the degree of spin polarization in these calculations has to be defined as an average quantity. However, the degree of spin polarization of the photoelectrons $P_max$ is 100% because the detector probes the spin character of the photoexcited states precisely at $\pmb{R}_D$ instead of taking the average over all of space. Thus, it is not meaningful to compare $P_\max$ and $P_\text{ave}^\text{TSS}$, and this explains the discrepancy in the values between experiment and theory. # Spin Textures We employ the effective [[Hamiltonian]] for the Bloch periodic part $|\phi(n, \pmb{k})\rangle$ of the [[wavefunction]] of the TSS in a TI: $ \begin{align} H_{TI}^0 = \hbar v k (\sin \theta_{\pmb{k}} \sigma_x - \cos \theta_{\pmb{k}} \sigma_y) \end{align} $ Here $\pmb{k} = k_x \hat{x} + k_y \hat{y}$ is the in-plane Bloch wave-[[vector]] with $\hat{z}$ along the surface normal. $v$ is the band velocity and $\theta_{\pmb{k}}$ is the angle between $\pmb{k}$ and the $+k_x$ direction. $\sigma_x$ and $\sigma_y$ are the standard [[Pauli matrices]]. The [[basis]] states defining the $\begin{pmatrix} 1 \\ 0 \end{pmatrix}$ and $\begin{pmatrix} 0 \\ 1 \end{pmatrix}$ are constructed from the two degenerate states at $\pmb{k} = 0$, $|\phi_1\rangle$ and $|\phi_2\rangle$. They define [[pseudospin]] up and down. The symmetry properties at the surface of e.g. Bi$_2$Te$_3$ ([[Bi2Te3]]) can be used to fix these, for example: $ \begin{align} \Theta |\phi_1\rangle &= - | \phi_2 \rangle\\ \Theta |\phi_2\rangle &= | \phi_1 \rangle\\ M |\phi_1\rangle &= i| \phi_2 \rangle\\ M |\phi_2\rangle &= i| \phi_1 \rangle\\ C_3 |\phi_1\rangle &= e^{-i\pi/3}| \phi_1 \rangle\\ C_3 |\phi_2\rangle &= e^{+i\pi/3}| \phi_2\rangle\\ \end{align} $ Here, $\Theta$ is the time reveral operator, $M$ is the reflection operator $x \rightarrow -x$ where $\hat{x}$ is along the $\Gamma K$ direction, and $C_3$ is rotation by $2\pi/3$ around the z-axis. The [[eigenvalue]] and [[Bloch's Theorem|Bloch eigenstate]] are: $ E(n, \pmb{k}) = n \hbar v k $ $ |\phi(n, \pmb{k})\rangle = \frac{1}{\sqrt{2}} (|\phi_1\rangle - n i e^i\theta_k |\phi_2\rangle) = \frac{1}{\sqrt{2}}\begin{pmatrix} 1 \\ -n i e^{i \theta_k}\end{pmatrix}$ Here, $n = \pm 1$ is the band index. The [[pseudospin]] [[Expectation Value]] of the TSS is: $ \begin{align} \langle \pmb{\sigma} \rangle &= \langle \phi(n, \pmb{k}) | \pmb{\sigma} | \phi(n, \pmb{k})\rangle\\ &= \begin{pmatrix} 1 & n i e^{-i \theta_k}\end{pmatrix} (\sigma_x \hat{x} + \sigma_y \hat{y} + \sigma_z \hat{z}) \begin{pmatrix} 1 \\ -n i e^{i \theta_k}\end{pmatrix}\\ &= n (\sin(\theta_{\pmb{k}}) \hat{x} - \cos(\theta_{\pmb{k}}) \hat{y}) \end{align} $ It is well known that the actual spin expectation value is aligned with the pseudospin expectation value, i.e., $ \langle \pmb{s} \rangle \propto n (\sin(\theta_{\pmb{k}}) \hat{x} - \cos(\theta_{\pmb{k}}) \hat{y}) $ The orientation of the [[Spinor]] eigenstates in the n=+1 case (upper band) is shown below ![[Pasted image 20210116165958.png]] # The Photoemission Matrix Elements ## Assumptions and Theoretical Framework We use the theoretical framework of [Wang *et al.*](https://journals.aps.org/prl/pdf/10.1103/PhysRevLett.107.207602). We assume that the final photoemission states (i.e. the $|f\rangles) are spin-degenerate at small $k$ becuase they are well within the spin degenerate the bulk band continuum. This is because they are $h\nu$ in energy above the Dirac point. We will approximate the periodic part of the final state Bloch [[wavefunction]] $|\phi^f_\uparrow(\pmb{k}, k_\perp)\rangle$, $|\phi^f_\downarrow(\pmb{k}, k_\perp)\rangle$ by those with $\pmb{k = 0}$ and $k_\perp = \sqrt{\frac{2 m_e}{\hbar^2}(h\nu - E_D)}$. ==Not 100% clear but this clearly has something to do with a free electron ejected from very close to the Dirac point== Here, $\pmb{k}$ is the in-plane component of the wavevector, $k_\perp$ is the out of plane component of the the photoelectron wavevector and $m_e$ is the electron mass. $E_D$ is the energy of the [[Dirac Cone|Dirac point]], i.e. the energy of the energy of the two-fold degenerate [[Topological Surface State|TSSs]] at $\pmb{k} = 0$. It is important to note that $|\phi^f_\uparrow\rangle$ and $|\phi^f_\downarrow\rangle$ are the actual spin-up and spin-down states along $\hat{z}$ far away from the surface of the sample, in [[vacuum]], where the measurement is performed, because we have already imposed the symmetry constraints. The real spin character of these two [[basis]] states can in principle be very complicated at the surface, we can regard the chosen degenerate doublet as the actual spin-up and spin-down states, respectively, from a measurement point of view. We do *not* need to assume that the final states are [[free electron]]-like [[plane waves]] nor that they have zero [[Spin-Orbit Coupling|SOC]]. The only assumption we make is that they have negligible spin splitting at small $\pmb{k}$ so that they can be approximated by the two spin degenerate states at $\pmb{k} = 0$ These two $\pmb{k}$-independent states are denoted by $|\phi^f_\uparrow\rangle$, $|\phi^f_\downarrow\rangle$ and have the same symmetry relations of the surface states discussed before, i.e. $ \begin{align} \Theta |\phi^f_\uparrow\rangle &= - | \phi^f_\downarrow \rangle\\ \Theta |\phi^f_\downarrow\rangle &= | \phi^f_\uparrow \rangle\\ M |\phi^f_\uparrow\rangle &= i| \phi^f_\downarrow \rangle\\ M |\phi^f_\downarrow\rangle &= i| \phi^f_\uparrow \rangle\\ C_3 |\phi^f_\uparrow\rangle &= e^{-i\pi/3}| \phi^f_\uparrow \rangle\\ C_3 |\phi^f_\uparrow\rangle &= e^{+i\pi/3}| \phi^f_\uparrow\rangle\\ \end{align} $ Another assumption we make is that we can neglect the $k$ dependence of the velocity operator $\pmb{v}(\pmb{k}) = e^{-i \pmb{k} \cdot \pmb{r}} \pmb{v} e^{+i \pmb{k} \cdot \pmb{r}}$ when calculating the optical transition matrix element between periodic parts of the Bloch wave functions. This theoretical setup was previously used by Wang *et al.* to find the energy- and momentum-dependent spin polarization of the TSS in Bi$_2$Se_3$ ([[Bi2Se3]]) from time-of-flight [[Angle-Resolved Photoemission Spectroscopy|ARPES]] measurements with excellent agreement, so the results presented here are valid for low energy light sources. ==How low? What is the energy scale here?== ## Derivation The microscopic [[Hamiltonian]] of an electron with SOC is: $ H_D = \frac{\pmb{p}^2}{2m_e} + V + \frac{\hbar}{4m_e^2c^2}\left(\nabla V \times \pmb{p}\right) \cdot \pmb{s} $ Now we can use a [[Pierels substitution]] $\pmb{p} \rightarrow \pmb{p} - \frac{e}{c} \pmb{A}$, where $\pmb{A}$ is the [[Vector Potential]], to derive the [[electron-photon interaction]] [[Hamiltonian]]: $ \begin{align} H_D^\text{int}(\pmb{A}) &= H_D\left(\pmb{p} - \frac{e}{c}\pmb{A}\right) - H_D(\pmb{p}) \\ &= \frac{1}{2m_e} \frac{-e}{c}(\pmb{p}\cdot\pmb{A} + \pmb{A} \cdot \pmb{p} - \frac{e}{c} A^2) + \frac{\hbar}{4 m_e^2c^2} (\nabla{V} \cdot \frac{-e}{c} \pmb{A})\cdot\pmb{s}\\ &= \frac{-e}{c}\pmb{A}\cdot\left(\frac{\pmb{p}}{m_e} + \frac{\hbar}{4 m_e^2c^2} (\pmb{s} \times \nabla V)\right) + \mathcal{O}(A^2)\\ &= \frac{-e}{c}\pmb{A} \cdot \pmb{v} \end{align} $ ==Why did we assume $\pmb{p} \cdot \pmb{A} = \pmb{A} \cdot \pmb{p}$?== Where $\pmb{v}$ is the velocity operator: $ \pmb{v} =\frac{\pmb{p}}{m_e} + \frac{\hbar}{4 m_e^2c^2} (\pmb{s} \times \nabla V) $ Now we compute the velocity matrix elements: $ \pmb{v}_{s,i} = \langle{\phi_s^f}|\pmb{v}|\phi_i\rangle $ Here $s = \uparrow, \downarrow$ is the index of the photoexcited final state and $i = 1, 2$ is the index of the [[pseudospin]] [[basis]] of the [[Topological Surface State|TSSs]]. We define $v_{\pm} = v_x \pm i v_y$ and compute the 12 matrix elements: $ \begin{align} \langle \phi^f_\uparrow | v_+ | \phi_1 \rangle &= \langle \phi^f_\uparrow | v_+ C_3^3 | \phi_1 \rangle = \langle \phi^f_\uparrow | C_3^\dagger v_+ C_3 | \phi_1 \rangle = e^{2\pi i /3} \langle \phi^f_\uparrow | v_+ | \phi_1 \rangle = 0 \\ %%%% \langle \phi^f_\downarrow | v_+ | \phi_1 \rangle &= \langle \phi^f_\downarrow | v_+ \Theta^2 | \phi_1 \rangle = \langle \phi^f_\downarrow | \Theta ^\dagger v_- \Theta | \phi_1 \rangle = \langle \phi^f_\uparrow | v_- | \phi_2 \rangle^*\\ %%%%%%% \langle \phi^f_\uparrow | v_+ | \phi_2 \rangle &= 0 \quad \text{using} \, C_3\\ %%%%%%%% \langle \phi^f_\downarrow | v_+ | \phi_2 \rangle &= \langle \phi^f_\downarrow | v_+ C_3^3 | \phi_2 \rangle = \langle \phi^f_\downarrow | C_3^\dagger v_+ C_3 | \phi_2 \rangle = e^{-2\pi i /3} \langle \phi^f_\uparrow | v_+ | \phi_1 \rangle = 0 \\ %%%%%%% \langle \phi^f_\uparrow | v_- | \phi_1 \rangle &= \langle \phi^f_\uparrow | v_- C_3^3 | \phi_1 \rangle = \langle \phi^f_\uparrow | C_3^\dagger v_- C_3 | \phi_1 \rangle = e^{2\pi i /3} \langle \phi^f_\uparrow | v_- | \phi_1 \rangle = 0 \\ %%%%%%% \langle \phi^f_\downarrow | v_- | \phi_1 \rangle &= -\langle \phi^f_\uparrow | v_+ | \phi_2 \rangle \\ %%%%%%%%% \langle \phi^f_\uparrow | v_- | \phi_2 \rangle &= 0 \quad \text{using} \, C_3 \\ %%%%%%%%%%% \langle \phi^f_\downarrow | v_- | \phi_2 \rangle &= \langle \phi^f_\downarrow | v_- C_3^3 | \phi_2 \rangle = \langle \phi^f_\downarrow | C_3^\dagger v_- C_3 | \phi_2 \rangle = e^{-2\pi i /3} \langle \phi^f_\uparrow | v_- | \phi_1 \rangle = 0 \\ %%%%%%%%%%% \langle \phi^f_\uparrow | v_z | \phi_1 \rangle &= \langle \phi^f_\uparrow | v_z \Theta^2 | \phi_1 \rangle = \langle \phi^f_\uparrow | \Theta ^\dagger v_z \Theta | \phi_1 \rangle = -\langle \phi^f_\downarrow | v_z | \phi_2 \rangle^*\\ %%%%%%%%%% \langle \phi^f_\downarrow | v_z | \phi_1 \rangle &= 0 \quad \text{using} \, C_3 \\ %%%%%%%% \langle \phi^f_\uparrow | v_z | \phi_2 \rangle &= 0 \quad \text{using} \, C_3 \\ %%%%%%%%%%%%% \langle \phi^f_\downarrow | v_z | \phi_2 \rangle &= \langle \phi^f_\downarrow | v_z M^2 | \phi_2 \rangle = \langle \phi^f_\downarrow |M^\dagger v_z M | \phi_2 \rangle = -i^2 \langle \phi^f_\uparrow |v_z \phi_1 \rangle = \langle \phi^f_\uparrow | v_z | \phi_1 \rangle\\ \end{align} $ Combining the non-zero equations, we conclude: $ \begin{align} \langle \phi^f_\uparrow | v_+ | \phi_2 \rangle &= - \langle \phi^f_\downarrow | v_- | \phi_1 \rangle = i \alpha \\ \langle \phi^f_\uparrow | v_z | \phi_1 \rangle &= \langle \phi^f_\downarrow | v_z | \phi_2 \rangle = i \beta\\ \end{align} $ Here $\alpha$ and $\beta$ are real constants that can be calculated using e.g. DFT. Now we can calculate the interaction [[Hamiltonian]] connecting the two [[basis]] functions of the TSS to the spin-up and spin-down photoexcited states. $ \begin{align} H_\text{TI}^{int}(\pmb{A}) &= -\frac{e}{c} \begin{pmatrix} i\beta A_z & i\alpha A_x - i*i\alpha A_y \\ - i\alpha A_x - i*i\alpha A_y & i \beta A_z \\ \end{pmatrix} \\ &= \frac{-e}{c}\left(i \beta A_z \sigma_0 - \alpha A_x \sigma_y + \alpha A_y \sigma_x \right) \end{align} $ ==I am missing a factor of -1 here and another factor of 1/2 in the $\alpha$ terms. I will fudge it in what follows.== $ \begin{align} H_\text{TI}^{int}(\pmb{A}) &= \frac{e}{2c} \begin{pmatrix} 2i\beta A_z & i\alpha A_x - i*i\alpha A_y \\ - i\alpha A_x - i*i\alpha A_y & 2i \beta A_z \\ \end{pmatrix} \\ &= \frac{\alpha e}{2 c}\left(2i \beta/\alpha A_z \sigma_0 - A_x \sigma_y + A_y \sigma_x \right) \end{align} $ Here $\sigma_0$ is the 2x2 identity matrix. We will neglect the $A_z$ term to see the new physics clearly, noting that it does not contribute to the alteration of the spin direction since it is proportional to the identity matrix. # Different Light Polarization ## Linear Polarization Consider the case of linear polarization: $ \begin{align} H_\text{TI}^\text{int}(\pmb{A}) &= \frac{\alpha e}{2 c}\left(A_y \sigma_x - A_x \sigma_y\right)\\ &= \frac{\alpha e}{2 c} A \left(\sin\theta_{\pmb{A}} \sigma_x - \cos\theta_{\pmb{A}} \sigma_y\right) \end{align} $ Where $\theta_{\pmb{A}}$ is the angle between the [[Vector Potential]] and the $+\hat{x}$ direction. The photocurrent is: $ I_\hat{t} = |\langle \hat{t}, \pmb{R}_D|\phi_f'\rangle|^2 $ where: $ \begin{align} |\phi_f'\rangle &= \sum_{\{\phi^f_s|E_f = E_i + h\nu\}} |\phi_s^f\rangle\langle \phi_s^f|H^\text{int}_\text{TI}|\phi_i\rangle \\ &= \frac{1}{\sqrt{2}} \begin{pmatrix} 1\\ -n i e^{i(2\theta_{\pmb{A}} - \theta_{\pmb{k}})} \end{pmatrix} \end{align} $ ==Not sure how to derive this last equation, sorta silly.== We find that the net effect of photoexcitation on the detected electron spin polarization direction defined by the first equation is a rotation in direction through a change: $ \theta_{\pmb{k}} \rightarrow 2\theta_{\pmb{A}} - \theta_{\pmb{k}} $ This is obtained by comparing $|\phi_f'\rangle$ to: $ |\phi\rangle = \frac{1}{\sqrt{2}}\begin{pmatrix} 1 \\ -n i e^{i \theta_k}\end{pmatrix} $ Which we derived before. Finally, using the equation for the spin polarization and writing $\Delta\theta_{\pmb{A, k}} \equiv \theta_{\pmb{A}} - \theta_{\pmb{K}}$ $ \langle \pmb{s} \rangle \propto n (\sin(\theta_{\pmb{k}}) \hat{x} - \cos(\theta_{\pmb{k}}) \hat{y}) \rightarrow n (\sin(\theta_{\pmb{k}} + 2 \Delta\theta_{\pmb{A, k}}) \hat{x} - \cos(\theta_{\pmb{k}}+ 2 \Delta\theta_{\pmb{A, k}}) \hat{y}) $ ![[Pasted image 20210117182148.png]] If $\theta_{\pmb{k}} = \theta_{\pmb{A}}$ then the direction of the photoemitted electron polarization is the same as that of the TSS. If $|\Delta \theta_{\pmb{A}, \pmb{k}}| = \pi/2$ then the initial and final spins are anti-parallel. ## Circularly Polarized Light Now, $\pmb{A} = A(\hat{x} \pm i \hat{y})/\sqrt{2}$ for RHCP and LCHP respectively. We end up with: $ \begin{align} H_\text{TI}^\text{int}(\pmb{A}) &= \frac{\alpha e}{2 c}\left(A_y \sigma_x - A_x \sigma_y\right)\\ &= \frac{\alpha e}{2\sqrt{2}c} A \left(\pm i\sigma_x - \sigma_y\right) \end{align} $ Proceeding to compute $|\phi_f '\rangle$ as before, we get: $ \begin{align} |\phi_f'\rangle_{\text{LHCP}} = \begin{pmatrix} 1 \\ 0 \end{pmatrix} = |\phi_\uparrow^f\rangle \\ |\phi_f'\rangle_{\text{RHCP}} = \begin{pmatrix} 0 \\ 1 \end{pmatrix} = |\phi_\downarrow^f\rangle \end{align} $ And thus the spin polarization direction of the photon is in the $\pm \hat{z}$ direction for LHCP and RHCP respectively. ![[Pasted image 20210117185142.png]] # Conclusions * The degree of spin polarization of [[photoemission|photoejected electrons]] as defined in typical spin-[[Angle-Resolved Photoemission Spectroscopy|ARPES]] measurements with fully polarized light is much higher (up to 100%) than that of the [[Topological Surface State|TSSs]] themselves (~50% according to first-principles calculations.) * The degree of spin polarization of [[photoemission|photoelectrons]] from the TSSs of a [[topological insulator|TI]] is 100% in principle, as long as the TSS are non-[[degeneracy|degenerate]] and fully polarized light is used. * It is not meaningful to compare $P_\max$ and $P_\text{ave}^\text{TSS}$, and this explains the discrepancy in the values between experiment and theory. * The degree of spin polarization measured experimentally in Bi$_2$Te$_3$ ([[Bi2Te3]]) and Bi$_2$Se$_3$ ([[Bi2Se3]]) ranges from 20% for the TSS itself, and from 60%-85% for the photoejected electrons. This is surprising because this is much higher than the result obtained from theory (~50%) because of e.g. spin independent background signals that reduce the measured value compared to the calculation.