The uniqueness that defines a direct sum
A sum U + W is called a [[direct-sum|direct sum]], written U ⊕ W, when the decomposition is unique: every v in U + W can be written as u + w with u in U and w in W in exactly one way. The single algebraic condition that guarantees this is U ∩ W = {0}.
Why U ^ W = {0} forces uniqueness:
suppose v = u1 + w1 = u2 + w2
then u1 - u2 = w2 - w1
left side in U, right side in W, so the common value
lies in U ^ W = {0}.
=> u1 - u2 = 0 and w2 - w1 = 0 => u1=u2, w1=w2. Unique!Complements: every subspace has a partner
Given a subspace U of a finite-dimensional V, a [[complementary-subspace|complementary subspace]] W is any subspace with V = U ⊕ W. Complements always exist — extend a basis of U to a basis of V, and the new vectors span a W that works. Note: complements are not unique. In R^2, a line U has *many* complementary lines.
Worked split of M_2 and many summands
Split M_2 (2x2 matrices) into symmetric + skew-symmetric:
S = { A : A^T = A } dim 3 (basis [1,0;0,0],[0,0;0,1],[0,1;1,0])
K = { A : A^T = -A } dim 1 (basis [0,1;-1,0])
Every matrix splits uniquely:
A = (A + A^T)/2 + (A - A^T)/2
^symmetric ^skew
S ^ K = {0} and dim S + dim K = 3 + 1 = 4 = dim M_2
=> M_2 = S (+) K.Direct sums chain: V = U_1 ⊕ U_2 ⊕ ... ⊕ U_k means every vector splits uniquely across all k pieces. This is the structural engine behind diagonalization — a diagonalizable map breaks V into a direct sum of eigenspaces. Splitting a space is rarely the end goal; it is the *tool* for understanding a transformation one block at a time.