---
title: Uhlmann Fidelity in Quantum Information
url: https://www.emergentmind.com/topics/uhlmann-fidelity
type: topic
---

# Uhlmann Fidelity in Quantum Information

Uhlmann fidelity quantifies the similarity between two quantum states, generalizing the transition probability of pure states to mixed-state density matrices. Defined as the trace norm of the geometric mean between square roots of two density operators, Uhlmann fidelity occupies a central role in quantum information theory, quantum statistical mechanics, and condensed matter physics, serving as a metric foundation, a probe for phase transitions, and a computationally tractable information distance with strong operational justifications.

## 1. Formal Definitions and Equivalent Characterizations

For two density matrices $\rho$ and $\sigma$ on a separable (possibly infinite-dimensional) Hilbert space, the Uhlmann fidelity is
\[
F(\rho, \sigma) = \left[\text{Tr} \sqrt{ \sqrt{\rho} \, \sigma \, \sqrt{\rho} } \right]^2.
\]
This is often equivalently written as
\[
F(\rho, \sigma) = \| \sqrt{\rho} \sqrt{\sigma} \|_1^2,
\]
where $\|\cdot\|_1$ denotes the trace norm. In the case of pure states, this reduces to the squared overlap, $F(|\psi\rangle\langle\psi|, |\varphi\rangle\langle\varphi|) = |\langle\psi|\varphi\rangle|^2$.

Uhlmann’s theorem provides a geometric interpretation: it asserts that for any purifications $|\psi_\rho\rangle, |\psi_\sigma\rangle$ of $\rho$ and $\sigma$ in some extended Hilbert space, the maximal overlap over all choices of purifications satisfies
\[
F(\rho, \sigma) = \max_U |\langle \psi_\rho | (I \otimes U) |\psi_\sigma\rangle|^2,
\]
where the maximization is over all unitaries $U$ acting on the purifying space. This result extends to infinite-dimensional systems, under trace-class and separability assumptions [1107.0354].

Mathematically, Uhlmann fidelity satisfies:
- Symmetry: $F(\rho,\sigma) = F(\sigma,\rho)$.
- $0 \leq F(\rho,\sigma) \leq 1$, with equality to 1 if and only if $\rho=\sigma$.
- Joint concavity in both arguments.
- Monotonicity under completely positive trace-preserving (CPTP) maps.
- Metric property: $\arccos F(\rho,\sigma)$ is a distance [1107.0354, 1408.3462, 1704.04033].

## 2. Variational, Optimization, and SDP Formulations

Several variational and convex-optimization characterizations exist:
- Trace-norm (primal): $F(\rho,\sigma) = \|\sqrt{\rho}\sqrt{\sigma}\|_1^2$ [1408.3462].
- Maximization over unitaries: $F(\rho,\sigma) = \max_U | \text{Tr}(U\sqrt{\rho}\sqrt{\sigma})|$ [1408.3462, 1704.04033].
- SDP primal: maximize $\text{Re}[\text{Tr}(X)]$ subject to $\begin{pmatrix} \rho & X \\ X^\dagger & \sigma \end{pmatrix} \geq 0$ [1704.04033].
- SDP dual: minimize $\tfrac12(\text{Tr}(A\rho) + \text{Tr}(B\sigma))$ subject to $\begin{pmatrix} A & I \\ I & B \end{pmatrix} \geq 0$.

In convex duality language, Uhlmann fidelity emerges as the maximal quantum extension of classical fidelity for measurement channels, and the polar (dual) inherits convexity and homogeneity [1408.3462].

Measured $f$-divergences unify these forms: for $f(t) = -\sqrt{t}$ (i.e., $\alpha=1/2$), the measured divergence reduces to the negative logarithm of Uhlmann fidelity, $D_{Meas,1/2}(\rho \|\sigma) = -2 \log F(\rho,\sigma)$, and the variational characterization becomes a maximization over positive operators $\gamma$ [2502.07745].

## 3. Computational Approaches and Practical Algorithms

Direct computation of $F(\rho, \sigma)$ is tractable for moderate dimensions:
- The canonical method requires two matrix square-roots and two matrix multiplications.
- An alternative, proven by Baldwin and Jones [2312.12438, 2211.02623], demonstrates the spectrum of $\sqrt{\rho}\sigma\sqrt{\rho}$ coincides with $\rho\sigma$, so it suffices to diagonalize the latter:
  \[
  F(\rho,\sigma) = \left(\sum_j \sqrt{\lambda_j}\right)^2,
  \]
  where $\lambda_j$ are eigenvalues of $\rho\sigma$.
  This reduces computational overhead by up to an order of magnitude for large matrices.
- For large many-body systems represented as matrix product density operators (MPDOs), scalable polynomial-time algorithms yield certified lower and upper bounds by variational optimization over sequential quantum circuits, accurately tracking fidelity scaling laws in critical regimes [2601.13333].
- In tensor network settings (MPS, TTN), subsystem fidelities can be efficiently computed for regions cutting $O(1)$ bonds, leveraging the Schmidt decompositions and transfer matrix constructions [1807.01640].

Experimental approaches for low-dimensional mixed quantum states have been devised using optical interference and measurement of first- and second-order overlaps, producing tight sub- and superfidelity bounds without recourse to full quantum state tomography [1308.5815].

Quantum circuits for Uhlmann transformations, employing block-encodings and quantum singular value transformations (QSVT), enable square-root fidelity estimation and have been shown to yield exponential quantum speedups over state tomography in certain access models [2509.03619].

## 4. The Uhlmann Connection, Bures Metric, and Geometric Structure

Uhlmann fidelity realizes the infinitesimal form of the Bures metric:
\[
ds^2_{\rm Bures} = 2\left[1 - F(\rho(t), \rho(t+dt))\right] \approx g_{Bures,ij} dx^i dx^j,
\]
where $g_{Bures,ij}$ is the Bures metric tensor [1702.07289]. The associated parallelism condition for amplitudes $w$ satisfying $\rho = ww^\dagger$ leads to the definition of the Uhlmann connection, a Hermitian matrix-valued one-form $A$ on state space, encoding holonomy and curvature that signal quantum phase transitions.

The polar decomposition $\sqrt{\rho_2}\sqrt{\rho_1} = |\sqrt{\rho_2}\sqrt{\rho_1}|U$ identifies the Uhlmann factor $U$, capturing changes of eigenbasis and reflecting the holonomy induced by adiabatic passage in parameter space [1803.05021, 1702.07289, 1609.00688].

## 5. Uhlmann Fidelity in Quantum Phase Transitions and Order

Uhlmann fidelity serves as a highly sensitive probe for quantum and topological phase transitions. Sharp drops in $F(\rho(\lambda),\rho(\lambda+\delta\lambda))$ under parameter variation $\lambda$ signal abrupt spectral or eigenbasis changes characteristic of critical points [1702.07289, 1803.05021, 1609.00688, 2105.05055]. This is robust in free-fermion models, topological insulators, and BCS superconductors.

Fidelity susceptibility $\chi_F$ (the second derivative of fidelity under parameter change) diverges at critical points, and its scaling exponents (e.g., $b \approx 2$ in 2D fermion models) can be extracted analytically and numerically [1803.05021, 2105.05055]. At finite temperatures, phase transition signatures in $F$ and the associated Uhlmann connection become analytic, consistent with the absence of finite-temperature topological transitions unless the Hamiltonian is explicitly temperature-dependent (as in BCS superconductivity) [1702.07289, 1609.00688].

Applications in many-body quantum information include using subsystem fidelities to map spatially local similarities and dynamics, extracting correlation lengths, tracking convergence in numerical simulations, and diagnosing codeword distinguishability in quantum error-correcting codes [2601.13333, 1807.01640].

## 6. Extensions, Bounds, and Special Cases

- In the classical limit where $\rho$ and $\sigma$ commute, Uhlmann fidelity reduces to the Bhattacharyya coefficient, $F_C(p,q) = \sum_i \sqrt{p_i q_i}$ [1408.3462].
- Sub- and superfidelity bounds provide tight, directly measurable constraints:
  \[
  E(\rho, \sigma) \leq F(\rho, \sigma) \leq G(\rho, \sigma)
  \]
  with explicit formulas in terms of overlaps and purities [1308.5815].
- For continuous-variable (Gaussian) states, Uhlmann fidelity depends explicitly on the covariance matrices and displacement vectors, with closed-form expressions for multi-mode, single-mode, and two-mode cases involving symplectic invariants [1111.7067].
- In infinite dimensions, all structural properties survive, but the relation to classical fidelity via measurement is replaced by an infimum, not a minimum, due to the lack of discrete POVM attainability in some cases [1107.0354].

## 7. Operational and Information-Theoretic Significance

Uhlmann fidelity admits several operational meanings:
- It quantifies the maximal transition probability between (possibly mixed) quantum states, relating to optimal quantum state conversion and discrimination [1704.04033].
- In hypothesis testing, it characterizes asymptotic error exponents and strong-converse rates, with the measured Rényi divergence at $\alpha=1/2$ reducing to $-2\log F(\rho,\sigma)$ [2502.07745].
- It governs the maximal overlap attainable by any pair of purifications, underpinning protocols in information transmission (entanglement transmission, state merging) and quantum algorithmic tasks such as Petz recovery [2509.03619].

Uhlmann fidelity provides a unique, robust, and computationally practical metric for mixed-state comparison, subsuming quantum generalizations of classical overlap, and equipping the geometric and operational framework that underpins much of modern quantum information theory.

Source: https://www.emergentmind.com/topics/uhlmann-fidelity