How to create simple (tensor) product spaces?

Question

Is it possible to work with simple tensor product spaces, like multiplying product states from quantum mechanics?

I basically have a simple two dimensional vector space, whose elements are represented by two dimensional vectors: $(1,0)$ and $(0,1)$ are the basis vectors. And I have $2\times 2$ matrices acting on them.

What I want to do is to combine two or more of such vector space to a bigger one, so that I get states like this: $(1,0)\otimes (1,0)$, $(1,0)\otimes (0,1)$, $(0,1)\otimes (1,0)$ and $(0,1)\otimes (0,1)$ are the basis states of the combined vector space. Matrices should combine similarly like: $((1,2),(3,4))\otimes((1,2),(3,4))$

I tried different ways to achieve this, using the \[CircleTimes] or the \[TensorProduct] symbols to combine them, but in the end I do not get anything to work.

Answer 1

Here is the basic method, illustrated with the combination of two spin-1/2 particles. (Hopefully, the physics language is familiar or accessible; I don't really have an idea of where else this kind of construction is useful.)

sigmaZ = {{1, 0}, {0, -1}}
id = IdentityMatrix[2]
(* {{1, 0}, {0, -1}} *)
(* {{1, 0}, {0, 1}} *)

Let's suppose we want to compute the $z$ component of the total spin of two particles. The individual operators and the sum are

(sz[1] = KroneckerProduct[sigmaZ, id]) // MatrixForm
(sz[2] = KroneckerProduct[id, sigmaZ]) // MatrixForm
(sz[1, 2] = sz[1] + sz[2]) // MatrixForm

Now, we can make single-particle states (written in the eigenbasis of sigmaZ) using

state[1] = {s[1][1], s[1][-1]}
state[2] = {s[2][1], s[2][-1]}
(* {s[1][1], s[1][-1]} *)
(* {s[2][1], s[2][-1]} *)

(I am using this abstract indexing in order to keep track of the states and particles that each amplitude is associated with: s[2][-1], for instance, stands for the amplitude of particle 2 to be in the "down" state.)

Then, we can take the tensor product of these two states to form the two-particle state

state[1, 2] = KroneckerProduct[state[1], state[2]] // Flatten
(* {s[1][1] s[2][1], s[1][1] s[2][-1], s[1][-1] s[2][1], s[1][-1] s[2][-1]} *)

(Below, I will present an alternative to Flattening, but this is the way I prefer to do this). Note the ordering of the two-particle basis:

Particle 1, up; Particle 2 up

Particle 1, up; Particle 2 down

Particle 1, down; Particle 2 up

Particle 1, down; Particle 2 down

We can see by direct matrix multiplication that this is consistent with the ordering used in KroneckerProduct when constructing the two-particle operators:

sz[1].state[1, 2]
sz[2].state[1, 2]
(* {s[1][1] s[2][1], s[1][1] s[2][-1], -s[1][-1] s[2][1], -s[1][-1] s[2][-1]} *)
(* {s[1][1] s[2][1], -s[1][1] s[2][-1], s[1][-1] s[2][1], -s[1][-1] s[2][-1]} *)

We can see that when Particle 1 is in the down state, the amplitude gets multiplied by -1 when being acted on by sz[1], and so on. Note that we can also do

state[1, 2].sz[2]

with no problems.

Finally, we can construct our states in alternative ways. The ordering in Mathematica is consistent with many different ways of constructing the states. For instance,

Table[state[i, j], {i, 1, -1, -2}, {j, 1, -1, -2}] // Flatten
state @@@ Tuples[{1, -1}, 2]
(* {state[1, 1], state[1, -1], state[-1, 1], state[-1, -1]} *)
(* {state[1, 1], state[1, -1], state[-1, 1], state[-1, -1]} *)

Here, the first and second positions in the argument of state are, respectively, the spin of particle 1 and 2, and we can see that the ordering of the basis is respected.

---

Alternative to Flattening

If we feel like making column vectors and row vectors separately, then do the following:

columnState[1, 2] = KroneckerProduct[state[1], List /@ state[2]]
rowState[1, 2] = KroneckerProduct[List@state[1], state[2]]
columnState[1, 2] // MatrixForm
rowState[1, 2] // MatrixForm
(* {{s[1][1] s[2][1]}, {s[1][1] s[2][-1]}, {s[1][-1] s[2][1]}, {s[1][-1] s[2][-1]}} *)
(* {{s[1][1] s[2][1], s[1][1] s[2][-1], s[1][-1] s[2][1], s[1][-1] s[2][-1]}} *)

And of course,

rowState[1, 2] == Transpose@columnState[1, 2]
(* True *)

Now, Dotting the matrix with the vectors can only be done one way, and Mathematica will complain if we try to do things in the wrong order.

sz[1].columnState[1, 2] // MatrixForm
rowState[1, 2].sz[1] // MatrixForm