632 words
3 minutes
The Time-Reversal Operator for Spin-1/2

Of the discrete symmetries in quantum mechanics, time reversal is the one that behaves differently. Parity is a unitary operator. Time reversal is antiunitary, i.e., it contains a complex conjugation, and for spinors this leads to a rather odd result: applying it twice to a fermion returns minus the state you started with. In this post we write down the explicit 2×22\times2 form of the operator for spin-12\frac{1}{2}, and then see where that minus sign comes from. The same sign is behind Kramers degeneracy, which I come back to at the end.


The time-reversal operator#

For a spin-12\frac{1}{2} particle the time-reversal operator is

Θ=Ke−iπSy/ℏ=Ke−iπ2σy\Theta = K e^{-i\pi S_y/\hbar} = K e^{-i\frac{\pi}{2}\sigma_y}

where KK is complex conjugation and σy\sigma_y is the usual Pauli matrix. The factor e−iπSy/ℏe^{-i\pi S_y/\hbar} is a rotation of the spin by π\pi about the yy-axis. The reason a rotation shows up at all is that reversing time should flip the spin. Spin is an angular momentum, and orbital angular momentum L⃗=r⃗×p⃗\vec{L} = \vec{r}\times\vec{p} changes sign under t→−tt \to -t because p⃗\vec{p} does, so we expect the same of S⃗\vec S.


Expanding the exponential#

To get the matrix explicitly, we expand the exponential as a power series,

e−iπ2σy=∑j=0∞1j!(−iπ2σy)je^{-i\frac{\pi}{2}\sigma_y} = \sum_{j=0}^{\infty} \frac{1}{j!}\left(-\frac{i\pi}{2}\sigma_y\right)^j

The series is easy to sum since σy2=1\sigma_y^2 = \mathbf{1}, so even powers of σy\sigma_y are the identity and odd powers are σy\sigma_y itself,

σy2n=1,σy2n+1=σy\sigma_y^{2n} = \mathbf{1}, \qquad \sigma_y^{2n+1} = \sigma_y

Collecting the even and odd terms separately, the two brackets are the Taylor series of cosine and sine evaluated at π/2\pi/2,

e−iπ2σy=1(1−(π/2)22!+(π/2)44!−⋯ )⏟cos⁡(π/2) = 0−iσy(π2−(π/2)33!+(π/2)55!−⋯ )⏟sin⁡(π/2) = 1e^{-i\frac{\pi}{2}\sigma_y} = \mathbf{1}\underbrace{\left(1 - \frac{(\pi/2)^2}{2!} + \frac{(\pi/2)^4}{4!} - \cdots\right)}_{\cos(\pi/2)\,=\,0} - i\sigma_y\underbrace{\left(\frac{\pi}{2} - \frac{(\pi/2)^3}{3!} + \frac{(\pi/2)^5}{5!} - \cdots\right)}_{\sin(\pi/2)\,=\,1}

The cosine vanishes and the sine is one, which leaves

e−iπ2σy=−iσy=−i(0−ii0)=(0−110)e^{-i\frac{\pi}{2}\sigma_y} = -i\sigma_y = -i\begin{pmatrix}0 & -i \\ i & 0\end{pmatrix} = \begin{pmatrix}0 & -1 \\ 1 & 0\end{pmatrix}

and therefore

Θ=K(0−110)\boxed{\Theta = K\begin{pmatrix}0 & -1 \\ 1 & 0\end{pmatrix}}

Action on a general spinor#

Let us see what this does to a general state ∣ψ⟩=(αβ)|\psi\rangle = \begin{pmatrix}\alpha \\ \beta\end{pmatrix}. The matrix acts first and the conjugation acts last,

Θ(αβ)=K(0−110)(αβ)=K(−βα)=(−β∗α∗)\Theta\begin{pmatrix}\alpha \\ \beta\end{pmatrix} = K\begin{pmatrix}0 & -1 \\ 1 & 0\end{pmatrix}\begin{pmatrix}\alpha \\ \beta\end{pmatrix} = K\begin{pmatrix}-\beta \\ \alpha\end{pmatrix} = \begin{pmatrix}-\beta^* \\ \alpha^*\end{pmatrix}

So time reversal swaps the two components, conjugates them and puts a minus sign on one of them. The spin-up amplitude becomes the (conjugated) spin-down amplitude and vice versa, which is what we wanted, since angular momentum is odd under t→−tt \to -t and ∣↑⟩|\uparrow\rangle and ∣↓⟩|\downarrow\rangle should trade places.


Θ2=−1\Theta^2 = -1 for fermions#

Now apply Θ\Theta once more to the result. One has to be a bit careful with the order here, because KK conjugates the components that are already conjugated, which undoes the first conjugation,

Θ2(αβ)=Θ(−β∗α∗)=K(0−110)(−β∗α∗)=K(−α∗−β∗)=(−α−β)\Theta^2\begin{pmatrix}\alpha \\ \beta\end{pmatrix} = \Theta\begin{pmatrix}-\beta^* \\ \alpha^*\end{pmatrix} = K\begin{pmatrix}0 & -1 \\ 1 & 0\end{pmatrix}\begin{pmatrix}-\beta^* \\ \alpha^*\end{pmatrix} = K\begin{pmatrix}-\alpha^* \\ -\beta^*\end{pmatrix} = \begin{pmatrix}-\alpha \\ -\beta\end{pmatrix}

so that

Θ2∣ψ⟩=−∣ψ⟩\boxed{\Theta^2|\psi\rangle = -|\psi\rangle}

Two time reversals do not bring a spin-12\frac{1}{2} state back to itself, only back to minus itself. For bosons (integer spin) one finds Θ2=+1\Theta^2 = +1 instead, and for fermions (half-integer spin) Θ2=−1\Theta^2 = -1. This sign is not a curiosity, it has measurable consequences.

  • Kramers theorem. Take a system with half-integer total spin in a time-reversal invariant potential (so no magnetic field). Then every energy eigenstate is at least doubly degenerate. The states ∣ψ⟩|\psi\rangle and Θ∣ψ⟩\Theta|\psi\rangle have the same energy and are orthogonal (the orthogonality follows from Θ2=−1\Theta^2 = -1), so they really are two different states, and you cannot split them without breaking time-reversal symmetry.

  • Topological insulators. The Z2\mathbb{Z}_2 classification of time-reversal invariant topological insulators rests directly on the distinction Θ2=±1\Theta^2 = \pm 1.

As a last remark, this is the same mathematics as the 4π4\pi periodicity of spinors (a spin-12\frac{1}{2} particle has to be rotated by 4π4\pi, not 2π2\pi, to come back to itself). The statement Θ2=−1\Theta^2 = -1 says that two time reversals amount to a 2π2\pi rotation, and for fermions a 2π2\pi rotation gives −1-1.

The Time-Reversal Operator for Spin-1/2
https://rohankulkarni.me/posts/notes/time-reversal-spin-half/
Author
Rohan Kulkarni
Published at
2022-11-09
License
CC BY-NC-SA 4.0
Share Post
Comments