Hilbert Spaces - Key Proof
We present a detailed proof of the Projection Theorem, one of the most fundamental results in Hilbert space theory, which establishes the existence and uniqueness of best approximations in closed subspaces.
Let be a Hilbert space and a closed subspace. Then for every , there exists a unique such that Moreover, is characterized by the condition .
Existence: Let . Choose a sequence in such that as .
We show is Cauchy using the parallelogram law:
Since is a subspace, , so
As , we have , therefore
So is Cauchy. Since is complete and is closed, . By continuity of the norm, .
Orthogonality Characterization: For any and , we have , so
This gives for all .
Choosing for small yields . Dividing by and letting gives for all .
Uniqueness: If both satisfy , then by the parallelogram law:
Therefore .
-
Finite-Dimensional Subspaces: If is orthonormal, then
-
Fourier Series: The projection of onto trigonometric polynomials of degree is the partial Fourier sum
The Projection Theorem requires the subspace to be closed. For non-closed subspaces, the infimum may not be attained. This is why completeness of Hilbert spaces is essential.
This theorem is the foundation for least squares approximation, Fourier analysis, and variational methods in PDEs.