Hermitian Operator

In quantum mechanics, every quantity you can physically measure, including energy, position, momentum, and spin, is represented by a Hermitian operator. The Hermitian property guarantees that measurement outcomes are always real numbers, which is a non-negotiable requirement for any physical observable. In quantum computing, Hermitian operators appear as Hamiltonians in variational algorithms, as the mathematical objects behind every measurement, and as the Pauli matrices that underpin nearly everything in the field.

The details

A matrix $H$ is Hermitian (or self-adjoint) if it equals its own conjugate transpose:

$H = H^\dagger$

where $H^\dagger$ is obtained by transposing $H$ and taking the complex conjugate of every entry. For a real symmetric matrix, this condition reduces to $H = H^T$.

Key properties:

1. Real eigenvalues. If $H|\lambda\rangle = \lambda|\lambda\rangle$, then $\lambda \in \mathbb{R}$. This is why measurement outcomes are real numbers. The proof is direct: $\lambda = \langle\lambda|H|\lambda\rangle = \langle\lambda|H|\lambda\rangle^* = \lambda^*$, so $\lambda$ must be real.

2. Orthonormal eigenbasis. The eigenstates of a Hermitian operator form a complete orthonormal basis for the Hilbert space. This means any quantum state can be decomposed as a superposition of eigenstates, which is exactly what happens during measurement.

3. Spectral decomposition. Every Hermitian operator can be written as $H = \sum_k \lambda_k |k\rangle\langle k|$, where $\lambda_k$ are eigenvalues and $|k\rangle$ are the corresponding eigenstates. This decomposition connects the abstract operator to its physical interpretation: measuring $H$ yields outcome $\lambda_k$ with probability $|\langle k|\psi\rangle|^2$.

Hermitian vs. unitary. Quantum gates are unitary ($U^\dagger U = I$); observables are Hermitian ($H = H^\dagger$). These are distinct properties. A matrix can be both Hermitian and unitary only if all its eigenvalues are $\pm 1$, which means $H^2 = I$. The Pauli matrices satisfy this: each is both Hermitian and unitary. The Hadamard gate is also both Hermitian and unitary. But most quantum gates (like the T gate, with eigenvalues $1$ and $e^{i\pi/4}$) are unitary without being Hermitian, and most Hamiltonians are Hermitian without being unitary.

The connection between the two is the matrix exponential: $U = e^{-iHt}$ converts a Hermitian operator $H$ into a unitary operator $U$, which is exactly how time evolution works in quantum mechanics. This relationship is the foundation of Hamiltonian simulation algorithms.

Role in VQE and quantum chemistry. The Variational Quantum Eigensolver estimates the ground state energy of a molecular Hamiltonian $H$. The Hamiltonian is Hermitian, and its smallest eigenvalue is the ground state energy. In practice, $H$ is decomposed as a weighted sum of Pauli strings:

$H = \sum_j c_j P_j$

where each $P_j$ is a tensor product of Pauli matrices (themselves Hermitian) and each $c_j$ is a real coefficient. The expectation value $\langle\psi|H|\psi\rangle$ is computed by measuring each Pauli string separately on the quantum computer and combining results classically.

Measurement formalism. When you measure a qubit “in the Z basis,” you are measuring the Hermitian operator $Z$. The eigenstates are $|0\rangle$ (eigenvalue $+1$) and $|1\rangle$ (eigenvalue $-1$), and the state collapses to whichever eigenstate the measurement outcome selects. Measuring in the X basis means measuring the operator $X$, whose eigenstates are $|+\rangle$ and $|-\rangle$. More generally, any measurement in quantum computing corresponds to projecting onto the eigenstates of some Hermitian operator.

Why it matters for learners

The Hermitian property is the bridge between quantum mathematics and physical reality. Every time you write down a Hamiltonian, choose a measurement basis, or decompose an operator into Pauli terms, you are working with Hermitian operators. Understanding the spectral theorem (real eigenvalues, orthonormal eigenstates) gives you the conceptual foundation for measurement, expectation values, and variational algorithms.

Common misconceptions

Misconception 1: All quantum gates are Hermitian. Most are not. The T gate, S gate, and general rotation gates are unitary but not Hermitian. Only gates that are their own inverse ($U^2 = I$) are both.

Misconception 2: Hermitian and symmetric mean the same thing. A symmetric matrix satisfies $A = A^T$. A Hermitian matrix satisfies $A = A^\dagger = \overline{A}^T$. For real matrices these are identical, but quantum mechanics uses complex matrices where the distinction matters. The Pauli-Y matrix, for instance, is Hermitian but not symmetric.

Misconception 3: The eigenvalues of a Hermitian operator must be integers. Eigenvalues must be real, but they can be any real number. The Hamiltonian of a hydrogen atom has eigenvalues $E_n = -13.6/n^2$ eV, which are neither integers nor rational numbers in general units.

See also

• Pauli Gates
• Unitary Operator
• Expectation Value
• Hamiltonian Simulation
• Variational Quantum Eigensolver