JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

The Bracket and the Payoff: SU(2), Spin, and Representations

The commutator measures how symmetries fail to commute and turns the tangent space into a Lie algebra. It explains the SU(2)-to-SO(3) double cover, quaternion rotations, and why representations rule physics.

The commutator that survives differentiation

The Lie bracket of two matrices is the commutator [A, B] = AB - BA. It is exactly the part of the group's structure that plain vector addition loses. The Baker-Campbell-Hausdorff formula makes this precise: exp(A)exp(B) = exp(A + B + (1/2)[A,B] + ...). When A and B commute the bracket is zero and exponentials add cleanly; otherwise [A,B] is the leading correction.

Crucially, if A and B are in a Lie algebra then [A, B] is too — the algebra is closed under the bracket. So a Lie algebra is a vector space with this extra bracket product, and it encodes the curved, non-commutative group structure entirely from flat data. That is what makes the linearization lossless.

# so(3) generators (rotations about x, y, z):
Lx = [0,0,0; 0,0,-1; 0,1,0]
Ly = [0,0,1; 0,0,0; -1,0,0]
Lz = [0,-1,0; 1,0,0; 0,0,0]

# brackets reproduce the axis cycle x -> y -> z -> x:
[Lx, Ly] = Lx*Ly - Ly*Lx = Lz
[Ly, Lz] = Lx
[Lz, Lx] = Ly

# these structure constants ARE the geometry of rotation.
The brackets of so(3) generators are the cross-product / rotation structure.

SU(2), quaternions, and the double cover

Here is the jewel. The algebra su(2) of the special unitary group SU(2) has the *same* bracket relations as so(3) — they are the same Lie algebra. Yet the groups differ: there is a smooth 2-to-1 map SU(2) -> SO(3), so SU(2) double-covers the rotation group. Both g and -g in SU(2) send to the same rotation, which is why a spin-1/2 particle must turn 720 degrees, not 360, to return to itself.

Concretely, unit quaternions ARE SU(2). A unit quaternion q rotates a vector by the sandwich v |-> q v q^-1, composes rotations by simple multiplication, and never suffers gimbal lock. This is why game engines and spacecraft store orientation as quaternions — the abstract double cover is the most practical rotation format known.

Representations and invariant measure

A representation realizes an abstract group as actual matrices acting on a vector space — letting the same symmetry act on scalars, vectors, tensors, or quantum states at once. The adjoint representation is the group acting on its own algebra by g . X = g X g^-1; differentiated, it is the bracket [A, X]. So the bracket is the group acting on itself, infinitesimally.

Linear algebra reborn as continuous symmetry

Look back at the arc. The invertible matrices of Volume I became groups; length-preservers became O(n), SO(n), and U(n); the exponential bridged each group to its flat Lie algebra; the bracket captured what plain linearity missed. The same SU(2) that classifies electron spin also formats every 3D rotation in your phone.

That is the payoff of the field. Linear algebra is not just solving systems and diagonalizing matrices — it is the language of continuous symmetry. The Standard Model of particle physics is built on U(1) x SU(2) x SU(3); general relativity on the Lorentz group; robotics and graphics on SE(3). Master matrix groups and Lie theory and you hold the grammar that unifies geometry, physics, and computation.