I see definitions (in Wikipedia, for example) about inner product spaces over arbitrary fields. But I don't understand how positivity makes any sense for fields which are not ordered? Am I missing something?

Clarification: Let me elaborate, to make clear what I already know. An inner product $g$ is a $\mathbb F$-sesquilinear map from $V\times V\to \mathbb F$, that is conjugate symmetric, non-degenerate and positive. My qualms are regarding positivity: that $g(v,v)\geq0$ for all $v\in V$. That inequality means nothing if $\mathbb F$ has no order structure on it.

  1. A suitable solution is to actually restrict our definition to ordered fields. But, this clearly is a strong condition- even the complex numbers are not ordered.

  2. In the real case, the above definition corresponds to the familiar symmetric, non-degenerate, positive bilinear forms; as $\mathbb R$ has a trivial $*$-operation.

  3. Let us look at the complex case, where the first such non-trivial definition arises. By using conjugate symmetry, $g_{\mathbb C}(v,v)=g_{\mathbb C}(v,v)^*$, so $g_{\mathbb C}(v,v)\in \mathbb R$, and since $\mathbb R$ has a canonical order- one can talk about positivity. So, the definition seems to be consistent here. A similar argument seems to works for quaternions as well. This suggests that if your field is a normed division algebra over $\mathbb R$ (by Hurzewicz theorem, there are only four such examples), the above construction may be extendable.

  4. Does it work for $*$-fields, in general? While conjugate symmetry and sesquilinearity generalise well to $*$-fields, positivity remains an issue. For the above argument to work out in a general case, one would require the conjugate symmetric elements of a *-field to have an order structure. Any concrete counter-example to that here would be helpful.
  5. This is still a far away from general fields.
  • $\begingroup$ Cf. here and here $\endgroup$ – J. W. Tanner Feb 21 at 14:44
  • $\begingroup$ $<x,x>$ is real-valued. You may be thinking of dot product. In a real vector space, the dot product is an inner product. $\endgroup$ – saulspatz Feb 21 at 14:45
  • $\begingroup$ @J.W.Tanner Those answers are more along the lines of what I was thinking. So I refined the question a bit more to elucidate what I couldn't find in those answers. $\endgroup$ – Sandesh Jr Feb 22 at 12:07

Because an inner product induces a norm $\|x\|=\sqrt{\langle x|x\rangle},$ and a norm, by definition, is trying to measure the "size" of something, we require the inner product of a vector with itself to be non-negative. So it makes sense in the reflexive case. If you're computing $\langle x|y\rangle,$ then all bets are off. The result could be a complex number, but that's fine. Geometrically, $\langle x|y\rangle$ is related to measuring the "angle" between $x$ and $y$, though that may not be in the Euclidean sense.

  • 1
    $\begingroup$ I'm not particularly interested in the motivation; what bugs me is the inconsistency of the definition itself. $\endgroup$ – Sandesh Jr Feb 22 at 12:00
  • $\begingroup$ $\langle x|y\rangle=\frac{1}{4}\sum_{n=0}^3 i^{-n}\Vert x+i^n y\Vert^2$ (or in the real case, $\langle x|y\rangle=\frac{1}{4}(\Vert x+y\Vert^2-\Vert x-y\Vert^2)$). $\endgroup$ – J.G. Feb 22 at 12:04

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.