{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# On Atiyah's solution to the Berry-Robbins problem" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Previously, we attempted to gain insight into the connection between spin and statistics. On the one hand, we considered the behavior of spins under rotations, where half integer spins pick up a negative sign after a single rotation, but not for interger spins. On the other hand, we had the techniques of bosonic and fermionic second quantization, where we constructed multiparticle states of \"indistinguishable\" particles in either permutation symmetric or antisymmetric states, states where if you permute (in the tensor product) two subsystems, they either come back to themselves, or else to $-1$ times themselves. \n", "\n", "On the third hand, we have the idea of literally swapping the locations of the particles in space. \n", "\n", "Using ideas from relativistic quantum field theory, one can show a connection between spin and statistics: that in 3+1 dimensions, half integer spins must be quantized as antisymmetric fermions and integer spins must be quantized as symmetric bosons. Here instead we take inspiration from a paper of Berry and Robbins which attempts to show something like the connection between spin and statistics on more elementary, even geometrical, grounds. \n", "\n", "First, the Berry-Robbins oscillator construction unifies the idea of a rotation and a permutation. The idea is: we take two spins and split them into two $\\uparrow$ and $\\downarrow$ oscillators each, and then join the $\\uparrow$ oscillators and the $\\downarrow$ oscillators, across the original spins, into two new spins. If we look at what happens to the original spins under a simultaneous $Y$ rotation of these two new \"exchange\" spins by $\\pi$, the original spins permute their places in the tensor product. And moreover, they pick up a $-1$ if they're half integer, and $1$ if they're integer spin.\n", "\n", "So that's two out of the three ideas: rotations and permutations. But what about the idea of actually exchanging the positions of the particles? We want somehow the act of dragging two particles around each other, switching positions, to correspond to: a permutation of the tensor product of the spins which is also an exchange rotation, picking up the proper sign. It would be interesting if there were somehow a smooth map from the positions of the particles into the space of exchange rotations. Atiyah's claim is that there is.\n", "\n", "