When using Dirac notation, how do we distinguish the difference between the inner product and the action of a covector on a vector?

As I understand it, $\vert x \rangle$ is an element of say a Hilbert space $\mathcal{H}$ and when we write $\langle x \vert$ we denote the corresponding linear map in $\mathcal{H}^*$ which takes a vector in $\mathcal{H}$ to $\mathbb{C}$.

Here's where my confusion lies: $\langle x \vert y \rangle$ is supposed to denote the "inner product" defined by the metric/orthonormality relation $\langle e_i \vert e_j \rangle = \delta _{ij}$, where $e_i$ denotes the $i^{ith}$ basis vector. But how do we distinguish this from the action of a covector $\langle x \vert$ on a vector $\vert y \rangle$?

Moreover, it seems like this notation only works in the case where the basis of $\mathcal{H}^*$ is defined such that $e^i(e_j) = \delta^{i}_j$. If we had chosen an arbitrary metric $\langle e_i \vert e_j \rangle = g_{ij}$ then this notation wouldn't make sense anymore.

  • $\begingroup$ What do you mean by "the action of a covector on a vector" ? Usually, this action is exactly the inner product so there is no ambiguity ! $\endgroup$
    – Frotaur
    Commented Mar 27, 2018 at 22:50
  • 4
    $\begingroup$ Indeed, Dirac notation assumes by default that all bases are orthonormal and breaks down otherwise, since the factors of the metric are missing. I remember getting all kinds of contradictions before learning this when I started. $\endgroup$
    – knzhou
    Commented Mar 27, 2018 at 22:57
  • 1
    $\begingroup$ @knzhou That sounds like the basis of a good answer $\endgroup$
    – David Z
    Commented Mar 28, 2018 at 3:41

1 Answer 1


The point of Dirac notation is that there is no need to distinguish them. There is a canonical isomorphism between vectors and covectors given by the inner product, so that if we denote the inner product by $(\cdot,\cdot)$, given $|\phi\rangle$ and $|\psi\rangle$ we define $\langle \phi |$ by $\langle \phi | \psi \rangle = (|\phi\rangle, |\psi\rangle)$.

The notation is independent of the basis. What does depend on the basis is the relation between the components: the components of a covector are the complex conjugates of the components of the vector only if the basis is orthonormal; otherwise you have to lower/raise indices with the metric.


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.