Proposition: (The Gram-Schmidt process) Let be a real inner product space of dimension , and let be a linearly independent set in . Then there is a set of vectors such that
if .
the span of is the same as the span of for .
Proof: Let and Then the span of and is the same as that of and , and by construction. Now suppose we have constructed with the desired property. Set
Then and the span property is preserved.
Orthogonal complements
If is a subspace of , define .
Proposition: is a subspace of . Furthermore:
(so .)
.
if , then .
Proof: Suppose and . Use Gram-Schmidt to construct an orthogonal basis for whose first elements are an orthogonal basis for . A vector
is in if and only if for .
Proposition: Suppose that is a self adjoint operator and . Then .
Proof: Suppose and . Then
since .
Proof of the (real) spectral theorem
We have a self-adjoint map . Pick a basis for and use Gram-Schmidt to construct an orthonormal basis (an orthogonal basis where the elements all have norm ).
The -matrix for this basis is the identity, and so the inner product is just the dot product.
If is the matrix representation of in this basis, then the matrix representation of is the transpose of . So since is self-adjoint, is symmetric.
We know that a symmetric matrix has a real eigenvalue with eigenvector .
Let be the orthogonal complement to the one-dimensional space spanned by . Since is an eigenvector, . Therefore . Furthermore, if , then
so is self-adjoint as a linear map from to itself. Thus we can continue by induction to construct an orthogonal basis of eigenvectors for .
Orthogonal matrices
Let be a symmetric matrix. As such it is a self adjoint map from to itself with respect to the usual dot product. Therefore there is a basis of consisting of orthonormal eigenvectors for the dot product - eigenvalues .
Let be the matrix whose columns are the vectors written in the standard basis of . Since the are an orthonormal basis, the matrix satisfies .
At the same time,
where is the diagonal matrix with entries . Since the are linearly independent, the matrix is invertible and is diagonalizable:
The bilinear map defined by
is an inner product provided that with equality only one . If we write in terms of the orthogonal basis :
then we get
which will be positive provided that all .
Example
Let be the symmetric matrix
Its eigenvalues are and with eigenvectors
and
The norm of a vector in the inner product given by is
The level curves of this are ellipses, and the eigenvectors point in the directions of the major and minor axes of the ellipse.