Eigenpair definitionFind the special nonzero vectors
Av = λv, v ≠ 0
An eigenvector is a nonzero vector that is only scaled by A; λ is the eigenvalue. The zero vector is never an eigenvector. Any nonzero scalar multiple of an eigenvector has the same eigenvalue.
An eigenspace contains all eigenvectors for one eigenvalue together with zero. Its dimension is called the geometric multiplicity. The number of repetitions of a root of the characteristic polynomial is its algebraic multiplicity. Always 1 ≤ geometric multiplicity ≤ algebraic multiplicity.
Inner product & normal matricesUse conjugation for complex vectors
⟨u, v⟩ = u†v ‖v‖² = v†v
For a complex vector, use conjugate transpose in the inner product. Orthogonal vectors satisfy u†v = 0; a normalized vector has v†v = 1.
Spectral theorem: a complex square matrix is normal if and only if it can be diagonalized by a unitary matrix. Eigenvectors of a normal matrix belonging to distinct eigenvalues are orthogonal. Within a repeated eigenspace, choose an orthonormal basis.
Hermitian & anti-Hermitian spectraReality and imaginary eigenvalues
For H† = H, every eigenvalue is real. If Hv = λv, then v†Hv = λv†v is real, so λ is real. Eigenvectors with distinct eigenvalues are orthogonal.
For K† = −K, eigenvalues are purely imaginary or zero: λ̅ = −λ. One can write λ = iα with α real. If H is Hermitian, iH is anti-Hermitian and its eigenvalues are i times those of H, with the same eigenvectors.
Unitary & general spectraKnow what is and is not guaranteed
For U†U = I and Uv = λv, preservation of length gives ‖Uv‖² = |λ|²‖v‖² = ‖v‖². Therefore |λ| = 1. Unitary eigenvalues lie on the unit circle; they need not be real.
An arbitrary n × n complex matrix has n characteristic roots counted with algebraic multiplicity, but may have fewer than n independent eigenvectors. Distinct eigenvalues give independent eigenvectors. A real matrix can have complex eigenvalues, occurring in conjugate pairs.
The sum and product of all eigenvalues equal tr A and det A, respectively, with algebraic multiplicity counted.
Simultaneous eigenvectorsDiagonalize compatible matrices together
A common eigenvector v obeys Av = λv and Bv = μv. A common eigenbasis means n independent vectors that are eigenvectors of both matrices.
Two commuting Hermitian matrices have a common orthonormal eigenbasis. More generally, two complex normal matrices are simultaneously unitarily diagonalizable if and only if AB = BA. If A and B are diagonalizable over the same field, commuting is equivalent to having a common eigenbasis over that field.
Commuting arbitrary matrices need not have a complete common eigenbasis: a defective matrix and I already give a counterexample. Sharing just one eigenvector does not imply that the matrices commute.
DegeneracyResolve the freedom inside an eigenspace
If Av = λv and AB = BA, then A(Bv) = B(Av) = λBv. Thus B maps the λ-eigenspace of A into itself.
When that eigenspace is one-dimensional, Bv must be proportional to v. When it is degenerate, an arbitrary eigenvector of A need not be an eigenvector of B. Diagonalize B within that eigenspace to choose common eigenvectors. Worked example 18 shows the steps.