Suppose we have a theory is covariant under the Spin group Spin(2n-1; 1). We consider the real vector space $V = R^{2n-1,1}$, which naturally comes with a Lorentzian inner product. On this vector space we introduce an orthonormal basis $e_0; e_1; ... ; e_{2n-1}$, where $e _0$ denotes the time direction.

To construct the Dirac representation of Spin(2n-1; 1) we take the complexified space $T = \mathbb{C} <e_1; ... ; e_n>$. My question is why is it that in order to construct the Dirac representation we complexify the space?

NOTE: Theory is even dimensional and of Lorentzian signature.

  • $\begingroup$ What do you mean, "why"? The Dirac representation is defined this way (and at least one of the representing matrices will have complex entries, so you can't just restrict to a real subspace)! $\endgroup$
    – ACuriousMind
    Commented Aug 22, 2015 at 23:41
  • 3
    $\begingroup$ "why" is clearly a request for motivations. He wants to understand what is gained by using that representation. Why this was proposed in the first place. Not hard to understand! $\endgroup$ Commented Aug 22, 2015 at 23:42
  • $\begingroup$ I don't understand what you are looking for, then. It's an allowed representation of the symmetry group of the theory, and a quite low-dimensional and hence natural one to look at. Why would we need specific motivation to look at representations of the symmetry group (our fields need something to transform in, after all)? $\endgroup$
    – ACuriousMind
    Commented Aug 22, 2015 at 23:48
  • $\begingroup$ I think maybe you only need to when your spacetime has a dimension that is a multiple of 4 but that it works just as well when it is other dimensions. $\endgroup$
    – Timaeus
    Commented Aug 23, 2015 at 0:13
  • $\begingroup$ @PhilosophicalPhysics I don't know why you insist on a representation of any kind whatsoever. You had some transformations that were a symmetry in a sense (you seem to want to use things that are covariant rather than invariant which is like using coordinates instead of vectors from a vector space) and you want a representation as matrices but again I don't know why having matrices is better than having the actual objects? $\endgroup$
    – Timaeus
    Commented Aug 23, 2015 at 1:03

1 Answer 1


We assume that OP asks apart from the facts that:

  1. Dirac representations by definition are complex;

  2. It is much easier to work with an algebraically closed field;

  3. Any real representation can be extended to a (possibly reducible) complex representation, so one is not missing anything by going complex.

In other words, OP is interested in why certain real Lie group representations cannot exist. Since it is well-known that every Lie group representation induces a corresponding Lie algebra representation, it will be enough for our purpose to show that certain real Lie algebra representations cannot exist.

So we are interested in whether there exists an $2^{[\frac{n}{2}]}$-dimensional$^1$ real spinor representation of $so(p,q)$, where $n=p+q\geq 2$?

A low dimension where this fails is $(p,q)=(3,0)$, i.e. 3D rotations, where we leave it as an exercise for the reader to check that the 1-dimensional pseudoreal/quaternionic spinor representation of the Lie algebra $so(3)\cong su(2)\cong u(1,\mathbb{H})$ has no real 2-dimensional irreducible subrepresentation.

OP only asks about even dimension $n$ with Minkowski signature. One may similarly show that $(p,q)=(5,1)$ fails, i.e that the direct sum of the 2-dimensional left and the 2-dimensional right pseudoreal/quaternionic Weyl spinor representations of the Lie algebra $so(5,1)\cong sl(2,\mathbb{H})$ has no real 8-dimensional irreducible subrepresentation.

Incidentally, Witten recently discussed real, pseudoreal and complex representations of fermions in arXiv:1508.04715.


$^1$ To understand where the dimension $2^{[\frac{n}{2}]}$ comes from, see e.g. this Phys.SE post.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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