0
$\begingroup$

One may define vectors as tuples satisfying the vector transformation law, or as elements of a vector space. On the latter, the superposition of two quantum states is another state, and the product of two states is non-state scalar. The vacuum state $|\psi\rangle=0$ exists and, for a scalar $a$ and a state $|\psi\rangle$, the product $a|\psi\rangle$ is still a state vector. Briefly, the axioms of a vector space are satisfied. My question regards the idea that state vectors transform like vectors.

Since states may not be position states, or states with any sort of directionality, I am not sure how to apply the concept of coordiante transformation in an abstract vector space. For instance, if I say,

"Let $\hat R(\hat n,\phi)$ be a rotation operator and $|\psi\rangle=|a\rangle+|b\rangle$ so that $\hat R|\psi\rangle=\hat R|a\rangle+\hat R|b\rangle$ obviously preserves the `angle' between $a$ and $b$, meaning that $\hat R|a\rangle$ and $\hat R|b\rangle$ are still orthogonal,"

then I am not satisfied that I have shown the proper vector transformation. Since the states themselves are the spanning basis of the state space, I am not quite getting the gist of what a coordinate transformation would be such that I could make a demonstration like the one below copied from Wolfram. What is the general proof that quantum states transform like vectors even when they are not states directly related to $\mathbb{R}^3$ coordinates? Following the form below, if I start with

$$|\psi\rangle=\sum_k a_k|\psi_k\rangle\quad\implies\quad\psi_i=a_{ik}\psi_k$$

then how should I proceed? Namely, what are my $x_i$ and $x_j$ in the vector space of quantum states?

enter image description here

$\endgroup$

2 Answers 2

3
$\begingroup$

I wonder if you are confusing several things here. Vectors are defined by the axioms of the vector space.

When you talk about vectors transforming in a certain way, I think you are talking about the vectors in the tangent space of some differentiable manifold. This an additional concept. Not all vectors must have a tangent space in which they 'live'. For example, differential one-forms (e.g. $dx,\,dy,\,dz$) are vectors, dual to vectors you are more used to, and they certainly do not transform in the way you describe. By definition, one-forms will do the opposite. A more life-like example is magnetic field. It is an axial vector, and as a result it transforms under reflection in a rather curious way.

As far as state-vectors transforming in a certain way. I feel like the natural way to go here is Wigner's theorem

$\endgroup$
10
  • $\begingroup$ When I talk about vectors transforming in certain way, I mean that an N-dimensional state space is spanned by $\hat e_1,\hat e_2,...\hat e_N$ in the same way that $\mathbb{R}^3$ is spanned by $\hat x,\hat y, \hat z$. You can do coordinate transformations on the linearly independent spanning basis of an N-dimensional state space exactly as if it were $\mathbb{R}^n$. I ended up figuring out my question. I just need to write $x_i$ and $x_j$ in terms of the $\hat e_n$ instead of the $\mathbb{R}^3$ basis. $\endgroup$ Nov 15, 2021 at 5:17
  • $\begingroup$ Usually people don't think of state spaces as geometric spaces, but they are that. It's just weird to think of spin-up, for example, serving the same purpose in state space as $\hat x$ does in $\mathbb{R}^2$. $\endgroup$ Nov 15, 2021 at 5:17
  • 1
    $\begingroup$ @hodopsmith state spaces can be infinite dimensional, in which case, I think your approach to spanning them may need to be more subtle. In reply to "usually people don't think of state spaces as geometric spaces"... perhaps it is a blank in my education but I don't know this term. I get the meaning, I think, but I don't think I know the definition. Topological spaces, metric spaces, manifolds, differentiable manifolds, vector spaces, all make sense. What is a geometric space? $\endgroup$
    – Cryo
    Nov 15, 2021 at 8:53
  • $\begingroup$ @hodopsmith, I am pretty sure that to talk about rotations in the way you talk about them you need some sort of a manifold, i.e. you need a point -set with a defined topology and then a differentiable map that maps this topological space into itself (which then can be rotation under some restrictions). Now each point on that manifold could have a tangent space, and could therefore be the state space you describe, but each tangent space will be completely different, unless you have some connection between them. What I am getting at is that having $N$ basis vectors does not mean you are in R^N $\endgroup$
    – Cryo
    Nov 15, 2021 at 8:58
  • $\begingroup$ You need a lot more structure which seems unnecessary for state vectors in a Hilbert space. $\endgroup$
    – Cryo
    Nov 15, 2021 at 8:59
1
$\begingroup$

States live in a Hilbert space which is typically infinite-dimensional where standard coordinate transformations don't apply. However, unitary representations of the rotational group on elements of the Hilbert space can be applied. These representations follow the following rules ($R$ -- $R_1$ and $R_2$ too -- is a rotation from the rotation group SO(3)):

$$U(R) U(R)^\dagger =1 \quad\text{and}\quad U(R_1)U(R_2) = U(R_1 R_2)$$

Trivially one has also (no rotation) : $$U(1) = 1$$

However, it depends strongly on the state what effect they might have. If for instance it acts on a momentum state $|\mathbf{p} \rangle$ one gets:

$$U(R)|\mathbf{p} \rangle = |\mathbf{ R p} \rangle$$

This means that if

$$\hat{p}_z |\mathbf{p} \rangle = p_z |\mathbf{p} \rangle$$

then the rotated state will an eigenvector to a different eigenvalue:

$$\hat{p}_z |R\mathbf{p} \rangle = Rp_z |R\mathbf{p} \rangle$$

Of course the action of the rotation $R$ on $p_z$ is one we are used to in classical physics.

A Hilbert space is a vector space which moreover has a scalar product and is complete. So states indeed live in a vector space, but not the one we used to from classical physics. In order to explain what the vector space means I give a simple example:

One can decompose a member $f$ -- we assume for this very simple example $f$ as periodic with a period of $2\pi$ -- of the Hilbert space (typically a function $\in\, L_2$, but other mathematical spaces are possible (with differently defined scalar product)) in a Fourier series:

$$f(\phi) = \sum_{n=0,\ldots,\infty} a_n E_n \equiv \sum_{n=0,\ldots,\infty} a_n e^{in \phi}$$

Then the $e^{in \phi}$ are the base vectors $E_n$ and the $(a_n)$ are the components of the Hilbert space vector $f$. (I denoted the base vectors with a capital $E$ in order to avoid confusion with the $e$ from the exponential function). The components can, as it is well known, computed by projecting the vector $f$ on the different base vectors by means of the scalar product:

$$a_n = <f, E_n> = < f, e^{in\phi}>$$

which in a Hilbert space is an integral:

$$ <f,g> = \int f(\phi) \overline{g(\phi)} d\phi$$

Therefore:

$$ a_n = \int_0^{2\pi} f(\phi) e^{-in\phi} d\phi$$

$\endgroup$

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.