Fine structure

Fine structure splits atomic levels through relativistic corrections and interactions involving spin, such as spin-orbit coupling. Different total angular momenta can then have distinct energies. Even a small ground-term splitting can matter for density-sensitive metastable atomic levels, because weak radiative decay between the levels competes with collisional mixing.

Radiation from a transition between nearby atomic levels split by fine-structure interactions. Such lines often occur in the infrared and can be excited in gas cooler than that needed for optical transitions. Their critical densities and the ion's abundance determine which environments they diagnose.

For a point, fixed proton, the sum of the relativistic kinetic-energy correction, spin-orbit interaction including the Thomas precession factor, and Darwin term has the displayed leading correction. It depends on n and total angular momentum j, but not separately on l. Thus $2S_{1/2}$ and $2P_{1/2}$ remain degenerate while $2P_{3/2}$ lies above them. Radiative corrections, recoil, nuclear size and hyperfine structure require additional terms. The fine-structure constant is ${\alpha }_{\mathrm{fs}}=e^2/(4\pi {ϵ}_0\hslash c)$.

The spin-one-half low-energy expansion of the Dirac equation includes this local correction in a static potential energy V. For $V=-Z{e}^2/(4\pi {ϵ}_{0}r)$, ${\nabla }^{2}V=Z{e}^2{\delta }^{(3)}(\mathbf{r})/{ϵ}_0$, so the correction is positive and probes the wavefunction at the origin. In point-Coulomb hydrogen atom states it contributes only to S waves at leading nonrelativistic order. It is part of fine structure, distinct from the radiative Lamb shift. ^[ref]

Expanding the relativistic kinetic energy $\sqrt{m^2{c}^4+c^2{p}^2}-m{c}^2$ for $|p|\ll mc$ gives $p^2/(2m)-p^4/(8m^3{c}^2)+\cdots$. The first correction is negative and contributes to atomic fine structure. Its perturbative use requires small momentum compared with mc; the truncated negative quartic term is not a globally stable exact Hamiltonian.

The spin-orbit interaction couples orbital angular momentum to spin. In chemical atoms it helps determine fine structure and the coupling scale that separates weak-field Zeeman splitting from the Paschen-Back effect.

The Rashba effect, named for Emmanuel Rashba, is a momentum-dependent spin-orbit interaction associated with broken spatial inversion symmetry. In an isotropic two-dimensional model, choose $\lambda >0$ and write the Hamiltonian $H={\hslash }^2{k}^{2}I/(2m)+\lambda (\hat{\mathbf{z}}\times \mathbf{k})\cdot \mathbf{\sigma }$, where $I$ is the identity matrix and $\mathbf{\sigma }$ is the vector of Pauli matrices. The eigenvalues of the spin term give $E_{\pm }(k)={\hslash }^2{k}^2/(2m)\pm \lambda k$. Define $k_R=m\lambda /{\hslash }^2$ and $E_R={\hslash }^2{k}_R^2/(2m)$. The lower branch is $E_{-}(k)={\hslash }^2(k-k_R{)}^2/(2m)-E_R$ and has a ring of minima at $k=k_R$. A negative $\lambda$ reverses the spin directions; ordering the two energies still uses $|\lambda |$. The alternative parameter $q_R=2k_R$ must not be mistaken for the minimum radius.

For $H_R=\lambda (k_x{\sigma }_y-k_y{\sigma }_x)$ with $\lambda >0$, the lower- and upper-band spinors of unit norm can be chosen asTheir spin expectation values are $⟨\mathbf{S}{⟩}_{\pm }=\pm (\hslash /2)(-k_y,k_x,0)/k$. They are tangential to circles of constant momentum magnitude, opposite between branches, and reversed at opposite momenta as required by time-reversal symmetry. When $E_F>0$, the outer contour belongs to the lower branch and the inner one to the upper branch. For $E_F<0$ the two Fermi contours both belong to the lower branch and have the same tangential spin orientation at a given angle in polar coordinates. At $k=0$ the Hamiltonian has quantum degeneracy in spin, and no unique spin texture direction is selected.

For the Rashba effect with $k_R>0$, put $D=\sqrt{k_R^2+2m{E}_F/{\hslash }^2}$. If $-E_R<E_F<0$, both Fermi contours belong to the lower branch and bound the singly occupied annulus $k_R-D<k<k_R+D$; the central disk and exterior are empty. At $E_F=0$, the lower occupied disk has outer radius $2k_R$, while the origin is a degenerate point at the Fermi energy. For $E_F>0$, the lower Fermi contour has radius $k_{-}=D+k_R$ and the upper Fermi contour radius $k_{+}=D-k_R$. The inner disk has two occupied spin-split eigenstates at each momentum, the intervening annulus has one, and the exterior is empty. Consequently $k_{-}-k_{+}=2k_R$, whereas below the crossing the two lower-branch radii add to $2k_R$. At zero temperature, changing occupation values only on these Fermi contours affects a null set and leaves the integrated carrier density unchanged.

Spin-momentum locking means that a state's spin direction is constrained by its momentum and its energy branch. In the isotropic Weyl Hamiltonian with $v>0$, the upper and lower branches have mean spin parallel and antiparallel to the momentum. Their textures in an equatorial momentum plane are radial outwards and inwards. Other Hamiltonians can lock spin transverse to momentum, so radial locking is an example rather than the general definition. In an effective model one must check whether its Pauli matrices represent physical spin or pseudospin before interpreting the texture as a measured spin.

For $H=c(p_x{\sigma }_x+p_y{\sigma }_y)$ with $c>0$ and physical spin matrices, $⟨\mathbf{S}{⟩}_s=s(\hslash /2)\hat{\mathbf{p}}$. The positive-energy branch is parallel and the negative-energy branch antiparallel to momentum. Both are collinear; no preferred spin direction is defined at zero momentum.

For a central-potential Hamiltonian operator with the spin-orbit interaction $H_{\mathrm{so}}=[r^{-1}{V}^{\prime }(r)/(2m^2{c}^2)]\mathbf{L}\cdot \mathbf{S}$, use $\mathbf{J}=\mathbf{L}+\mathbf{S}$ andA coupled angular momentum eigenstate makes this angular factor a number. The radial operator commutes with all angular factors, giving the energy shift in the title. It is independent of the magnetic projection $m_j$, so a fixed-$j$ multiplet retains its $2j+1$ degeneracy.

For spin one-half, the angular bracket is $l$ when $j=l+1/2$ and $-(l+1)$ when $j=l-1/2$. Only the former exists at $l=0$, where it is zero. For the Coulomb potential energy, $r^{-1}{V}^{\prime }(r)=e^2/(4\pi {ϵ}_0{r}^3)$; the radial normalization convention must be respected when computing its expectation value.

For an isotropic three-dimensional quantum harmonic oscillator, $V(r)=m{\omega }^2{r}^2/2$ makes $r^{-1}{V}^{\prime }(r)=m{\omega }^2$. The spin-orbit energy shift in a central potential is therefore independent of the radial excitation. For spin one-half, the $l=0,j=1/2$ shift is zero, while the two $l=1$ shifts are $-{\hslash }^2{\omega }^2/(2m{c}^2)$ for $j=1/2$ and $+{\hslash }^2{\omega }^2/(4m{c}^2)$ for $j=3/2$.

These shifts are quadratic in trap frequency. Their degeneracy-weighted sum in the $l=1$ manifold is zero, since the multiplicities are two and four. They are perturbative corrections rather than full oscillator energies; a nonrelativistic treatment requires them to be small compared with the orbital spacings.