We can think of the group SU(2) as the group of unit quaternions: namely, 𝑞 with |𝑞| = 1.

We can think of the space of spinors, ℂ², as the space of quaternions, ℍ.

Then acting on a spinor by an element of SU(2) becomes multiplying a quaternion on the left by a unit quaternion!

But what does it mean to multiply a spinor by 𝑖 in this story? It's multiplying a quaternion on the *right* by the quaternion 𝑖. Note: this commutes with left multiplications by all unit quaternions.

(2/n)

But there are some subtleties here. For example: multiplying a quaternion on the right by 𝑗 *also* commutes with left multiplication by unit quaternions. But 𝑗 anticommutes with 𝑖:

𝑖𝑗 = −𝑗𝑖

So there must be an 'antilinear' operator on spinors which commutes with the action of SU(2): that is, an operator that anticommutes with multiplication by 𝑖. Moreover this operator squares to -1.

In physics this operator is usually called 'time reversal'. It reverses angular momentum.

(3/n)

You should have noticed something else, too. Our choice of right multiplication by 𝑖 to make the quaternions into a complex vector space was arbitrary: any unit imaginary quaternion would do! There was also arbitrariness in our choice of 𝑗 to be the time reversal operator.

So there's a whole 2-sphere of different complex structures on the space of spinors, all preserved by the action of SU(2). And after we pick one, there's a circle of different possible time reversal operators!

(4/n)

The facts in my last paragraph don't rely on describing spinors and SU(2) using quaternions. We could have noticed them in the usual approach using only complex numbers. It just wouldn't be so blindingly obvious.

So, please don't think I'm some nut arguing that spin-1/2 particles are really quaternions. Indeed it seem we need a specific complex structure on the quaternions to recover the usual quantum description of a spin-1/2 particle!

(5/n)