From Spin Connection in 5 dimensions I can define a massless fermion's covariant derivative on a curved manifold as $$ \nabla_\mu \psi = (\partial_\mu - {i \over 4} \omega_\mu^{ab} \sigma_{ab}) \psi \tag{1} $$ where $\sigma_{ab}$ are the dirac bilinears and $\omega_\mu^{ab}$ is the spin connection with three indices.

In 5 dimensions I have a $4\times 4$ spinor space, giving me three sets of irreducible matrices: $I$ as identity, $\gamma^a$ as monolinears, and $\sigma_{ab}=[\gamma_a,\gamma_b]$ as bilinears. This give me a total of $1+5+10=16$ matrices forming a complete set.

In 9 dimensions I can have $9=2(4)+1$, giving me a spinor space of $2^{(4)}\times 2^{(4)}=16\times 16$ creating additional irreducibles: $\sigma^{abc}=[\gamma^a,\gamma^b,\gamma^c]$ as trilinears and $\sigma^{abcd}=[\gamma^a,\gamma^b,\gamma^c,\gamma^d]$ as quadrilinears. This gives me a total of $1+9+36+84+126=256$. These numbers were calculated from the binomial coefficients ( binomial[d,k] ) for the total number of kth-linears in $d$ spatial dimensions.

Since there are additional irreducibles in $9$ dimensions, not found in 5 dimensions, does my covariant derivative in Eq. 1 have additional terms? For example $$ \nabla_\mu \psi = \left(\partial_\mu - {i \over 4} \omega_\mu^{ab} \sigma_{ab} - {i \over 48} \omega_\mu^{abcd} \sigma_{abcd} \right) \psi \tag{2} $$ where $\omega_\mu^{abcd}$ is a new spin connection of 5 indices or is Eq. 1 still valid in $9$ dimensions?

  • $\begingroup$ That wikipedia article does use 3+1 dimensional specific vocabulary where it does not need to. $\endgroup$
    – AHusain
    Dec 15, 2015 at 7:33
  • $\begingroup$ 1. In odd dimension $n$, the Clifford volume element acts as $\pm 1$, so the products of up to $\frac{n-1}2$ elements span $\mathrm{End}\Sigma$, if $\Sigma$ is the spinor module. 2. If your connection comes from a metric connection of the underlying space, you only need to correct by elements of length $2$ in the Clifford algebra. $\endgroup$ Dec 15, 2015 at 10:20
  • 1
    $\begingroup$ I would suggest the book by Lawson and Michelssohn. It is quite long though, so maybe you find a shorter one that does the job. $\endgroup$ Dec 15, 2015 at 13:51
  • 1
    $\begingroup$ @SebastianGoette I have a question. How does $\psi$ transform under coordinate transformation. Can we simplify its transformation using an infinitesimal transformation? $\endgroup$ Dec 15, 2015 at 17:58
  • 6
    $\begingroup$ Meta discussion: meta.mathoverflow.net/questions/2656/learn-from-my-mistake $\endgroup$
    – Todd Trimble
    Dec 18, 2015 at 20:09

1 Answer 1


To answer this question, we have to make a few rather natural assumptions, which I give in a slightly different notation. Let the spinor bundle $S$ be associated to a Euclidean vector bundle $E$. Assume that both bundles are locally trivialised over some open coordinate patch $U$ by orthonormal / unitary frames $e_1,\dots, e_n$, $\psi_1,\dots,\psi_N$, such that Clifford multiplication has constant coefficients ("Pauli matrices"/"Dirac matrices") with respect to these frames, so $e_i\cdot\psi_\alpha=\sum_\beta c_{i\alpha}^\beta\psi_\beta$.

Let $\nabla^E$ be a metric connection on $E$, so the metric satisfies a Leibniz rule. Then $\nabla^E$ differs from partial derivation with respect to the given trivialisation by a one-form with values in skew-symmetric matrices, so $\nabla^E_{\partial_i}e_j|_x=\sum_k\Gamma_{ij}^k(x)e_k$. The Christoffel symbols $\Gamma_{ij}^k\colon U\to\mathbb R$ are smooth functions with $\Gamma_{ij}^k=-\Gamma_{ik}^j$.

Assume that the spin connection $\nabla^S$ is compatible with $\nabla^E$ in the sense that Clifford multiplication satisfies a Leibniz rule. This determines $\nabla^S$ uniquely. Then $\nabla^S$ differs from partial derivation in the given trivialisation by a one-form with values in the degree-2-part of the clifford algebra, more precisely $$\nabla^S_{\partial_i}\psi_\alpha=\frac14\sum_{j,k}\Gamma_{ij}^k(x)e_j\cdot e_k\cdot\psi_\alpha\;.$$ No terms of higher degree in the Clifford algebra are needed.

We check the Leibniz rule\begin{align} \nabla^S_{\partial_i}(e_\ell\cdot \psi_\alpha)&=\frac14\sum_{j,k}\Gamma_{ij}^ke_j\cdot e_k\cdot e_\ell\cdot \psi_\alpha\\ &=-\frac12\sum_j\Gamma_{ij}^\ell e_j\cdot\psi_\alpha +\frac12\sum_k\Gamma_{i\ell}^ke_k\cdot\psi_\alpha +\frac14\sum_{j,k}\Gamma_{ij}^ke_\ell\cdot e_j\cdot e_k\cdot \psi_\alpha\\ &=(\nabla^E_{\partial_i}e_\ell)\cdot\psi_\alpha +e_\ell\cdot(\nabla^S_{\partial_i}\psi_\alpha)\;. \end{align}

You find the whole story in Lawson-Michelsohn, section II.4 or in Berline-Getzler-Vergne, Section 3.3.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.