Matrix product states (MPS) are a way of representing a (many-body) wavefunction. The method has been described, for example, in

The density-matrix renormalization group in the age of matrix product states. U. Schollwoeck. Ann. Phys 326, 96 (2011), arXiv:1008.3477.

However, would it be possible to see a concrete example for a small system? How will the MPS look like for the following two cases?

  1. A two spin system: $\frac{1}{\sqrt{2}}(| \uparrow \downarrow \rangle + |\downarrow \uparrow \rangle)$
  2. For a three-spin system: $\frac{1}{\sqrt{2}}(| \uparrow \uparrow \uparrow \rangle + |\downarrow \downarrow \downarrow \rangle)$

You can think of an MPS as being built up by objects with three indices. How to easily represent such an object? We can think of this as a matrix where each entry is a vector (in particular the vector will be a vector in the on-site Hilbert space, so e.g. for a spin 1/2 system it will be of the form $\alpha |\uparrow \rangle + \beta |\downarrow \rangle$).

So for a chain where the on-site Hilbert space is a spin 1/2, each site will have such an object associated to it, like this: $$ A = \left( \begin{array}{ccc} \alpha_{11} \;| \uparrow \rangle + \beta_{11} \; |\downarrow \rangle & \alpha_{12} \;| \uparrow \rangle + \beta_{12} \; |\downarrow \rangle & \cdots \\ \alpha_{21} \;| \uparrow \rangle + \beta_{21} \; |\downarrow \rangle & \alpha_{22} \;| \uparrow \rangle + \beta_{22} \; |\downarrow \rangle & \cdots \\ \vdots & \vdots & \ddots \end{array} \right)$$

If it is a translationally invariant system, we can have that every site has the same $A$ associated to it. The physical state of our system is then obtained by multiplying all these $A$ matrices (where we `multiply' the matrix elements via the tensor product structure), more precisely: $$|\psi \rangle = \textrm{Tr}(A^N)$$ where we have assumed periodic boundary conditions. Here $N$ is the number of sites. Note that this object is now purely a vector in the tensor product of the on-site Hilbert spaces. So for example in the simple case that $$ A = \left( \begin{array}{cc} |\uparrow \rangle & 0 \\ 0 & |\downarrow \rangle \end{array} \right)$$ then $$ A^N = \left( \begin{array}{cc} |\uparrow \rangle \otimes |\uparrow \rangle \otimes \cdots \otimes |\uparrow \rangle & 0 \\ 0 & |\downarrow \rangle \otimes |\downarrow \rangle \otimes \cdots \otimes |\downarrow \rangle\end{array} \right)$$ So with periodic boundary conditions we get the cat state $|\psi \rangle = |\uparrow \cdots \uparrow \rangle + |\downarrow \cdots \downarrow\rangle$.

That answers your (ii). I'll leave (i) as an exercise ;)

  • $\begingroup$ Here is a more systematic way of doing it, but more cumbersome. This is how a computer would do it: physics.stackexchange.com/questions/251522/… $\endgroup$ – Ruben Verresen Jul 6 '16 at 15:59
  • $\begingroup$ Many thanks for the answer :) Taking cue from your comment, the solution for (i), i guess, should be $2\times 2$ matrix with elements $(0, 1/\sqrt{2}, 1/\sqrt{2}, 0)$. $\endgroup$ – Adhip A. Jul 7 '16 at 7:31
  • $\begingroup$ Based on the link you shared, the solution seems to be the one i had mentioned in my comment above. While, when i think about the same solution in the way you mentioned in the detailed answer here, there should be a $A$ matrix corresponding to each site and therefore the MPS state should be a tensor product of two matrices. Should these be, \begin{equation} |\Psi\rangle = Tr(A_1 A_2) \end{equation} where $A_1 = \begin{bmatrix} |\uparrow \rangle & 0 \\ 0 & |\downarrow \rangle \end{bmatrix}$ and $A_2 = \begin{bmatrix} |\downarrow \rangle & 0 \\ 0 & |\uparrow \rangle \end{bmatrix}$ ? Thanks. $\endgroup$ – Adhip A. Jul 7 '16 at 7:54

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.