The Adjoint of an Operator
~5 min read · Tier toggle changes depth (T1 plain · T2 standard · T3 rigour)
You should be comfortable with the definition of an inner product space, the linearity conventions (linear in the first slot, conjugate-linear in the second), and the Riesz representation theorem, which guarantees that every linear functional can be represented as an inner product with a unique vector.
Why this matters
The adjoint operator is the mathematical embodiment of "reversing the flow" in a space with geometry.
In physics, observables must be self-adjoint to ensure real eigenvalues, and time evolution is governed by unitary operators, which are defined precisely by their relationship to their adjoints. Without the adjoint, we cannot distinguish between a general linear map and a symmetry of the space. It is the bridge between the algebraic structure of linear maps and the geometric structure of the inner product.
The Intuition: Reversing the Arrow
Imagine a linear operator \(A\) as a machine that takes a vector \(u\) and outputs \(Au\). If you want to know how much this output overlaps with another vector \(v\), you compute the inner product \(\langle Au, v \rangle\). The adjoint \(A^{\dagger}\) is the unique machine that allows you to move the "action" from the first slot to the second slot without changing the value of the overlap. It is the operator that satisfies \(\langle Au, v \rangle = \langle u, A^{\dagger}v \rangle\). In plain language, \(A^{\dagger}\) is the operator that "undoes" the effect of \(A\) on the inner product, provided we also conjugate the result if we are in complex space. It is not the inverse; it is the geometric dual.
Definition and Existence
Let \(V\) be a finite-dimensional inner-product space over \(\mathbb{F}=\mathbb{R}\) or \(\mathbb{C}\). The adjoint of a linear operator \(A:V\to V\) is defined as the unique linear operator \(A^{\dagger}:V\to V\) such that for all \(u,v\in V\):
The existence and uniqueness of this operator is guaranteed by the Riesz representation theorem. For a fixed \(v\), the map \(u \mapsto \langle Au, v \rangle\) is a linear functional in \(u\). By Riesz, there exists a unique vector, which we call \(A^{\dagger}v\), such that \(\langle u, A^{\dagger}v \rangle = \langle Au, v \rangle\). The linearity of \(A^{\dagger}\) follows from the linearity of the inner product and \(A\).
Algebraic Properties
The adjoint operation is an anti-homomorphism of the algebra of linear operators. This means it reverses the order of multiplication and conjugates scalars. For any operators \(A, B\) and scalar \(\lambda \in \mathbb{F}\):
The reversal of order in \((AB)^{\dagger}\) is a common source of error. It arises because \(\langle ABu, v \rangle = \langle Bu, A^{\dagger}v \rangle = \langle u, B^{\dagger}A^{\dagger}v \rangle\). Note that if \(A\) is self-adjoint, \(A^{\dagger}=A\), and the order reversal becomes invisible, which is why self-adjoint operators are so central in quantum mechanics.
Matrix Representation
In a finite-dimensional space with an orthonormal basis, the adjoint has a simple matrix representation. If \(A\) is represented by the matrix \(M\), then \(A^{\dagger}\) is represented by the conjugate transpose \(M^{\dagger} = \overline{M}^T\). This is often called the Hermitian conjugate. For a real vector space, the conjugate is trivial, so \(A^{\dagger}\) is simply the transpose \(M^T\). The condition for an operator to be self-adjoint is that its matrix is Hermitian (complex) or symmetric (real): \(M = M^{\dagger}\).
The derivation
Existence and Uniqueness of the Adjoint
Let \(V\) be a finite-dimensional inner-product space over \(\mathbb{F}=\mathbb{R}\) or \(\mathbb{C}\), with inner product \(\langle\,\cdot\,,\,\cdot\,\rangle\) linear in its first slot and conjugate-linear in its second. For every linear operator \(A:V\to V\) there exists one and only one linear operator \(A^{\dagger}:V\to V\), the adjoint, satisfying \(\langle Au,v\rangle=\langle u,A^{\dagger}v\rangle\) for all \(u,v\in V\).
A common misconception is that the adjoint is the inverse. The inverse \(A^{-1}\) satisfies \(A A^{-1} = I\), whereas the adjoint satisfies \(\langle Au, v \rangle = \langle u, A^{\dagger}v \rangle\). An operator can be invertible but not self-adjoint, and it can be self-adjoint but not invertible (e.g., the zero operator). Only when \(A\) is unitary do we have \(A^{-1} = A^{\dagger}\). C
Worked example
Problem. Let \(V = \mathbb{C}^2\) with the standard inner product. Let \(A\) be the linear operator defined by the matrix \(M = \begin{pmatrix} 1 & i \\ 0 & 2 \end{pmatrix}\). Find the matrix representation of \(A^{\dagger}\) and determine if \(A\) is self-adjoint.
- In the standard orthonormal basis the matrix of \(A\) is \(M = \begin{pmatrix} 1 & i \\ 0 & 2 \end{pmatrix}\), and the adjoint is represented by the conjugate transpose.
- Conjugate: \(\overline{M} = \begin{pmatrix} 1 & -i \\ 0 & 2 \end{pmatrix}\); transpose: \(M^{\dagger} = \begin{pmatrix} 1 & 0 \\ -i & 2 \end{pmatrix}\).
- Compare: the off-diagonal entries differ (\(M_{12} = i\) but \((M^{\dagger})_{12} = 0\)), so \(M^{\dagger} \neq M\) and \(A\) is not self-adjoint.
Units check. The entries are dimensionless scalars in \(\mathbb{C}\). The operation involves complex conjugation and transposition, which are algebraic operations on the coefficients. The result is a valid matrix in \(\mathbb{C}^{2\times 2}\).
Three quick tests
- What is the adjoint of the identity operator \(I\)?
Answer
\(I^{\dagger} = I\), since \(\langle Iu, v \rangle = \langle u, v \rangle = \langle u, Iv \rangle\). - If \(A\) is a projection operator (i.e., \(A^2 = A\)), is \(A\) necessarily self-adjoint?
Answer
No. A projection is self-adjoint if and only if it is an orthogonal projection. Oblique projections are not self-adjoint. - Compute \((A^{\dagger})^{\dagger}\).
Answer
\(A\). The adjoint operation is an involution.
Problem set for PU-104 →
Each derivation in this unit ends with five graded problems with full solutions.
Lecture notes drafted with a local model and checked line by line by the editor; every number in the worked example was recomputed.