For any idempotent linear map , each vector decomposes asIf , write ; then . Thus the kernel of a linear map and image of a linear map give a direct sum.
For the self-adjoint map in the question, and satisfy . Hence the direct sum is an orthogonal decomposition:
For a normal operator , the adjoint operator identity and givePositive definiteness implies . Also , since orthogonality to every means for every . In finite dimension this gives .
The idempotent linear map decomposes as in part (a), now orthogonally, and acts as zero on its kernel and identity on its image. Writing and in these orthogonal summands,Thus is self-adjoint, and together with idempotence this proves is an orthogonal projection.
The finite-dimensional inner product space has the orthogonal decomposition . Define for , . This is a linear map with image , satisfies , and is self-adjoint by taking inner products between decomposed vectors. Hence it is an orthogonal projection.
Any other orthogonal projection with image has kernel by part (a); it is identity on by idempotence and zero on its kernel. Its values are therefore forced to agree with , proving uniqueness.
Take any positive definite inner product on and define the inner product adapted to an idempotent linear mapIt is symmetric and bilinear. If its quadratic value at is zero, positive definiteness of the original inner product forces both and , hence . Thus it is positive definite.
Using and givesThe map is therefore self-adjoint for this new inner product and remains idempotent, so it is an orthogonal projection for the explicitly constructed inner product. Geometrically the construction makes its complementary kernel and image orthogonal.
Articles by others on the same topic
There are currently no matching articles.