Concentrating Majorana fermions


Ákos Nagy

QM3 Quantum Matter meets Maths

Monday, July 5, 2021

this talk can be viewed at akosnagy.com/talks/Majorana-GM3/talk.html

Outline of the talk




  • I introduce a Dirac-type equation in 2D, the Jackiw–Rossi equation.


  • Main Result #1: I generalize the Jackiw–Rossi equation to:

    1. Higher dimensions $\sim$ representation theory of Clifford algebras.

    2. Multi-spinors/spinors with flavors $\sim$ spinors with "twists".

    3. (Pseudo)-Riemannian manifolds $\sim$ spin geometry.


  • Main Result #2:

    1. I describe the "low energy spectrum" in 2D.

    2. Construct solutions in higher dimensions.

    3. (If time permits, then) explain connections to vortices and braids.

The classical Jackiw–Rossi equation




  • Describes electronic surface excitations in a topological insulator in contact with an $s$-wave superconductor.

  • Let $\psi = \begin{pmatrix} \psi_\uparrow \\ \psi_\downarrow \end{pmatrix} : \mathbb{R}^2 \mapsto \mathbb{C}^2$ be a spinor field.

  • Let $\Phi : \mathbb{R}^2 \mapsto \mathbb{C}$ be the superconducting order parameter.

  • The (static, massless) Jackiw–Rossi equation is: \begin{equation} \left( D + \mathcal{R}_\Phi \right) \psi := \begin{pmatrix} \left( \partial_x + i \partial_y \right) \psi_\uparrow + \Phi \overline{\psi}_\uparrow \\ \left( - \partial_x + i \partial_y \right) \psi_\downarrow + \Phi \overline{\psi}_\downarrow \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \end{pmatrix}. \end{equation}
  • Simple example (1-vortex): $\Phi = z = x + i y$, $\psi_\downarrow = 0$, and $\psi_\uparrow = e^{- |z|^2/2}$.

  • The "large" 1-vortex limit: $\Phi = t z$, with $t \gg 1$, $\psi_\downarrow = 0$, and $\psi_\uparrow = \sqrt{t} \: e^{- t |z|^2/2}$.

  • Important note: If $\psi$ is a (nonzero) solution, then $i \psi$ is not. $\Rightarrow$ Solutions only form a real (as opposed to complex) vector space!

  • Majorana fermions!

The generalized Jackiw–Rossi equations in higher dimensions




  • Let $CL (s, t)$ be the Clifford algebra of $\mathbb{R}^{s, t}$, and $S$ be the (unique) spinor representation.

  • There is always a(n essentially) unique invariant bilinear pairing $B: S \otimes S \rightarrow \mathbb{C}$. Moreover, $(S \otimes S)^* = \Lambda$, where $\Lambda$ is (a subspace of) the vector space of exterior forms on $\mathbb{R}^{s, t}$ and the bilinear pairings corresponds to the 0-forms in $\Lambda$.

  • The metric dual of $B$ yields a map $\mathcal{R} : S \rightarrow \overline{S}$, which can also be viewed as complex conjugate linear map of $S$.

  • Let $\psi : \mathbb{R}^{s, t} \mapsto S$ be a spinor field and $\Phi : \mathbb{R}^{s, t} \mapsto \mathbb{C}$ be the superconducting order parameter.

  • The generalized Jackiw–Rossi equation with mass $m \in \mathbb{R}$ is: \begin{equation} \left( D + \mathcal{R}_\Phi \right) \psi := \left( D + \Phi \mathcal{R} \right) \psi = m \psi. \end{equation}
  • When $s = 2, t = 0$, and $m = 0$, then the above equation recovers the original Jackiw–Rossi equation.

  • When $s = 2, t = 1$, then the above equation recovers the time dependent Jackiw–Rossi equation with mass $m$!

The generalized Jackiw–Rossi equations with twists




  • Let $CL (s, t)$ be the Clifford algebra of $\mathbb{R}^{s, t}$, and $M$ be any Clifford module.

  • Then $M = S \otimes E$, where $E$ is a vanilla complex vector space, the flavor space.

  • Question: Are there invariant bilinear pairings on $M$?

  • Answer: Yes! Let $F := E \otimes E$. Given any $\Phi \in F^*$ and using $M \otimes M = (S \otimes S) \otimes F$, we can define $B_\Phi : B \otimes \Phi$.
    Every invariant bilinear pairing on $M$ has this form.

  • The metric dual of $B_\Phi$ yields a map $\mathcal{R}_\Phi : M \rightarrow \overline{M}$, which can also be viewed as complex conjugate linear map of $M$.

  • Let $\psi : \mathbb{R}^{s, t} \mapsto M$ be a spinor field with flavors in $E$ and $\Phi : \mathbb{R}^{s, t} \mapsto F^*$ be a superconducting order parameter.

  • Given a vector potential (an EM field) $A : \mathbb{R}^{s, t} \mapsto \mathbb{R}^{s, t}$, we can also twist the Dirac operator: $D_A := D + \sum\limits_{i = 1}^{s + t} A_i \ \mathrm{cl} \left( \mathrm{d} x^i \right)$.
  • The generalized Jackiw–Rossi equation with flavors in $E$ and with mass $m \in \mathbb{R}$ is: $\left( D_A + \mathcal{R}_\Phi \right) \psi = m \psi$.

  • When $F = \mathbb{C}$ and $A = 0$, then we get back the Jackiw–Rossi equation from the previous slide.

  • Using $(S \otimes S)^* = \Lambda$, we can consider even more general "form-order parameters" $\Phi \in \Lambda \otimes F^*$.

The generalized Jackiw–Rossi equations on manifolds




  • Let $X$ be a (pseudo)-Riemannian manifold and $M \rightarrow X$ be a bundle of Clifford modules.

  • Locally, $M = S \otimes E$, where $E$ is a complex vector bundle. This picture need not hold globally! In particular, $X$ need not be spin, or even spin$^c$!

  • Nonetheless, $(M \otimes M)^*$ is still (canonically) isomorphic to $\Lambda \otimes F^*$, where $\Lambda$ is (a subbundle of) the bundle of differential forms on $X$.

  • Given $\nabla$, a compatible connection on $M$, and $\Phi$, a section of $F^*$ (or, more generally $\Lambda \otimes F^*$), we can define the Jackiw–Rossi operator on the sections of $M$ as \begin{equation} D_\nabla + \mathcal{R}_\Phi. \end{equation}

Side note: Bogoliubov--de Gennes equations on manifolds




  • Let $X$ be a (pseudo)-Riemannian manifold and $M \rightarrow X$ be a bundle of Clifford modules, as before.

  • Given $\nabla$, a compatible connection on $M$, and $\Phi$, a section of $F^*$ (or, more generally $\Lambda \otimes F^*$), we get the operators $D_\nabla$ and $\mathcal{R}_\Phi$.

  • While $\mathcal{R}_\Phi$ is not complex linear, by doubling the spinors, we can still associate a complex linear operator to this datum, the (generalized) Bogoliubov--de Gennes operator: \begin{equation} \begin{pmatrix} D_\nabla & \mathcal{R}_\Phi \\ \mathcal{R}_\Phi^* & \pm D_\nabla \end{pmatrix} \end{equation} considered as an operator on $S \oplus \overline{S}$.

  • This is a story for another time though...

Back to 2D: Solutions



  • First, let us return to the original Jackiw–Rossi equation (no gauging, flavors, or manifolds).

  • Theorem (Rauch '04):
    If $\Phi = t z^p \overline{z}^q$, then the space of all $L^2$ solutions is $|p - q|$ dimensional and purely (anti-)chiral if $p > q$ ($p < q$).

  • Theorem (N. '21):
    If $\Phi = t z^p \overline{z}^q + \mbox{"remainder terms"}$ and $A$ is smooth and bounded, then the space of all $L^2$ solutions to \begin{equation} \left( D_A + \mathcal{R}_\Phi \right) \psi = 0. \end{equation} is $|p - q|$ dimensional and purely (anti-)chiral if $p > q$ ($p < q$), all of which concentrate around the origin.

  • Furthermore, I studied the spectral properties of the operator $D_A + \mathcal{R}_\Phi$. In particular, I proved a gap result.

Main Theorem



  • Theorem (N. '21):
    Let $\Sigma$ be a closed surface, $S \rightarrow \Sigma$ be a spin$^c$ bundle, and $\left( \nabla, \Phi \right)$ as before.
    Assume that the zero locus of $\Phi$ is discrete and at each zero $z \in \Sigma$ it has the form $\Phi = z^{p_z} \overline{z}^{q_z} + \mbox{"remainder terms"}$.
    Then for $t \gg 1$ there are exactly $\max (\{ p_z - q_z, 0 \})$ linearly independent, chiral eigenspinors concentrating at $z$ with mass $o (1)$. All other eigenvalues are $O (t^{1/(m + 1)})$.
    (Similar statement holds for anti-chiral solutions.)

  • Proof: Glue-in $\oplus$ gap result.

  • Furthermore, if for all zeros, $z$, we have $q_z = 0$, then the above "low energy" solutions are in fact massless.
    This gives an independent proof of the Riemann–Roch Theorem!

  • One can pull back solutions to $X = \Sigma \times Y$, where $Y$ is any (pseudo)-Riemannian manifold, using harmonic spinors on $Y$.

  • Conjecture (vaguely): Solution to $D_\nabla + t \mathcal{R}_\Phi$ in higher dimensions are classified by harmonic spinors on $Y := \Phi^{- 1} (0)$, which is generically a codimension 2 subset.

Vortices



  • The input for the (generalized) Jackiw–Rossi theories is a pair $\left( \nabla, \Phi \right)$.

  • Canonical families of such pairs come from (for example) the Ginzburg–Landau theory.

  • The "nicest" Ginzburg–Landau fields are called $\tau$-vortices in 2D. (See Taubes, Bradlow, ...)

  • Theorem (Bradlow '90): For $\tau$ large enough, the moduli space of $\tau$-vortices of degree $d$ is $\mathrm{Sym}^d (\Sigma)$.

  • It is easy to see that each degree $2k$ vortex field, $(\nabla, \Phi)$, yields a Jackiw–Rossi operator $D_\nabla + \mathcal{R}_\Phi$ whose kernel has real dimension $2k$.

  • Problem: Due to the breaking of the gauge symmetry, these kernels only descend to the vortex moduli as a bundle of real projective spaces.

  • It is unclear whether this bundle comes from a vector bundle, because
    1. I don't really understand real Atiyah–Segal class (that would classify bundles of real projective spaces).
    2. No one else seem to know or care about them either.

  • In any case, away from the "diagonal" in $\mathrm{Sym}^d (\Sigma)$, the bundle "flattens" as $\tau \rightarrow \infty$, thus we get a (nontrivial) projective representation of surface braid group of $\Sigma$ with $2k$ braids.

Few future directions



  • Understanding solutions in higher dimensions.

  • New proofs of Atiyah–Singer type theorems.

  • Nonlinear generalizations: $\Phi = (\psi \psi)_0^\flat$.

  • Understanding the braid representation and the connections to vortices.

  • Understanding the generalized Bogoliubov--de Gennes equation.

Thank you for your attention!