Geometric Quantum Discord (GQD)
- Geometric Quantum Discord (GQD) is a metric-based measure that quantifies non-classical correlations in bipartite and multipartite quantum states by minimizing the distance to zero-discord states.
- GQD employs diverse metrics—including Hilbert–Schmidt, trace, and Bures distances—to capture distinct operational, contractivity, and computational features, enhancing analytical tractability.
- Its applications span quantum state discrimination, metrology, and multipartite systems, offering insights into quantum correlations and dynamics under various noise channels.
Geometric quantum discord (GQD) is a distance-based quantifier of non-classical correlations in a bipartite or multipartite quantum state, defined by measuring how far the state lies from an appropriate set of zero-discord, or classical, states. In its original finite-dimensional form, GQD replaces the optimization over local measurements appearing in entropic quantum discord by an optimization over distances in state space, typically to the set of classical-quantum states (Hassan et al., 2010). Subsequent work showed that this geometric idea is not tied to a unique metric: Hilbert–Schmidt, trace, and Bures constructions lead to distinct discord functionals with different monotonicity, operational, and computational properties (Paula et al., 2013, Spehner et al., 2013).
1. Classical-quantum geometry and the basic definition
For a bipartite state , the zero-discord states with respect to measurements on subsystem are classical-quantum states of the form
or, equivalently,
with orthogonal projectors on the measured subsystem and arbitrary states of the unmeasured subsystem (Paula et al., 2013, Hassan et al., 2010). In the Bures formulation these are denoted -classical states and written as
emphasizing the asymmetry inherited from the choice of measured subsystem (Spehner et al., 2013).
The Hilbert–Schmidt version introduced in the early literature is
so that GQD is the minimum squared Hilbert–Schmidt distance to the zero-discord set (Paula et al., 2013). Equivalent measurement-based forms were also derived: which makes explicit that GQD can be understood as the minimal disturbance induced by an optimal local projective measurement (Doustimotlagh, 2014).
This geometric reformulation was motivated by the computational difficulty of entropic discord. In the formulation 0, the classical correlation term requires an optimization over local measurements, and exact formulas are available only for restricted state families. GQD replaces that information-theoretic optimization by a distance minimization to the nearest zero-discord state, which is often more tractable analytically and numerically (Doustimotlagh, 2014).
2. Hilbert–Schmidt GQD and analytically solvable families
For two-qubit states, the Hilbert–Schmidt GQD admits the well-known closed form
1
where 2, 3 with 4, and 5 is the largest eigenvalue of 6 (Qiang et al., 2014, Hassan et al., 2012). In Bloch-operator language for a general 7 bipartite system, the problem can be written as an optimization over the coefficient matrix 8,
9
which underlies several general bounds and exact results (Hassan et al., 2010).
For Bell-diagonal two-qubit states,
0
the Hilbert–Schmidt discord simplifies to
1
and the physically allowed region is the tetrahedron in 2-space, with the zero-discord Bell-diagonal states lying on the coordinate axes (Yao et al., 2013). This yields a geometric picture in which level sets of constant discord are three intersecting cylinders truncated by the tetrahedron (Yao et al., 2013).
Several later works pushed exact Hilbert–Schmidt formulas beyond the original two-qubit setting. A rigorous lower bound stronger than the earlier Luo–Fu bound was derived for arbitrary finite-dimensional bipartite states in terms of the coherence vector 3, correlation matrix 4, and the eigenvalues of
5
with the bound becoming exact for 6 systems measured on the qubit side (Hassan et al., 2010). More recently, an explicit exact analytical value for the GQD of an arbitrary two-qudit state was obtained via the Bloch vector of the measured subsystem, the correlation matrix, and the spectrum of
7
with the important conclusion that a previously known lower bound is in fact the exact value for every two-qudit state (Loubenets et al., 2024).
Further exact or closed formulas were found for structured state families. For two-qubit 8 states, both geometric discord and geometric global quantum discord were derived analytically, with the latter satisfying 9 for all 0 states (Qiang et al., 2014). For the multiqubit family
1
the multipartite geometric discord takes the exact form
2
showing an explicit 3 scaling for fixed correlation coefficients (Zhu et al., 2021).
3. Metric choices, consistency, and the criticism of the Hilbert–Schmidt geometry
A central development in the subject was the realization that the Hilbert–Schmidt version is not a good measure of quantum correlations, because it may increase or change under local reversible operations on the unmeasured subsystem (Paula et al., 2013). For the local ancilla map
4
one has
5
so simply attaching or removing a mixed local ancilla on the unmeasured side changes the discord value by the ancilla purity factor (Paula et al., 2013). This pathology was traced to the noncontractivity of the Hilbert–Schmidt norm under trace-preserving completely positive maps (Paula et al., 2013).
This motivated alternative geometric discords based on better-behaved distances. In the Schatten-6 family,
7
the multiplicativity
8
implies
9
Because 0, the trace norm is the only Schatten norm for which the value is invariant under this class of local reversible operations, making the 1-norm geometric discord the only consistent Schatten-2 version in that framework (Paula et al., 2013). The trace-distance discord is therefore defined as
3
and satisfies the contractivity property
4
for any trace-preserving local operation 5 on the unmeasured subsystem (Paula et al., 2013).
For Bell-diagonal states the trace-distance discord has the simple formula
6
the intermediate value among the absolute correlation coefficients, and coincides on that family with the negativity of quantumness (Paula et al., 2013). A hierarchy established in this setting is
7
where 8 is entropic discord and 9 is entanglement negativity (Paula et al., 2013).
The Bures-distance construction pursues the same consistency goal from a different direction. The Bures distance
0
is contractive under completely positive trace-preserving maps, Riemannian, and linked to the quantum Fisher information (Spehner et al., 2013). The corresponding Bures geometric discord is
1
and vanishes on exactly the same 2-classical states as standard discord (Spehner et al., 2013). This suggests that “geometric quantum discord” is not a single quantity but a family of metric-dependent correlation measures sharing the same zero set while differing in contractivity, ordering, and operational meaning.
4. Pure states, higher dimensions, multipartite systems, and Gaussian extensions
For pure states, several GQD variants collapse to especially transparent forms. In the Bures case, if
3
is in Schmidt form and 4 is the largest Schmidt coefficient, then
5
so the Bures-GQD coincides with the geometric measure of entanglement for pure states (Spehner et al., 2013). In the Hilbert–Schmidt framework used in a recent holographic analysis, the pure-state formula becomes
6
with 7 the Schmidt coefficients and 8 the second Rényi entropy (Banerjee et al., 2023).
Multipartite generalizations proceed in more than one way. One line defines discord with respect to a specific subsystem in an 9-partite system and yields a generic tensor formula
0
where 1 is the correlation tensor in local operator bases (Hassan et al., 2012). Another line defines total quantum correlations by sequential optimal local measurements on all parties, leading to
2
with explicit computable formulas for 3-qubit states (Hassan et al., 2012). A different multipartite extension, geometric global quantum discord, reduces for two-qubit 4 states to a closed formula and coincides with total quantum correlations in the sense of Hassan and Joag (Qiang et al., 2014).
Continuous-variable Gaussian systems require further restriction because Gaussian local measurements preserve Gaussianity. The Gaussian geometric discord is defined as
5
with the optimization restricted to Gaussian POVMs on one subsystem (Adesso et al., 2011). For two-mode squeezed thermal states, characterized by 6 in standard-form covariance matrices, the optimal Gaussian POVM is a noisy heterodyne measurement and the discord becomes
7
(Adesso et al., 2011). A crucial result in this setting is that, without an energy bound, there is no universal positive lower bound on Gaussian geometric discord at fixed entropic discord; one can have arbitrarily small geometric discord and arbitrarily large entropic discord (Adesso et al., 2011). This sharpens the distinction between geometric and entropic quantifiers in infinite-dimensional systems.
5. Operational interpretations and applications
The Bures formulation has a particularly direct operational meaning. For mixed states, the maximal fidelity 8 to the set of 9-classical states is equal to the optimal success probability of an ambiguous quantum state discrimination task associated with an ensemble 0 extracted from 1 (Spehner et al., 2013). In this picture, the closer 2 is to the classical set, the easier it is to discriminate the associated conditional states. The closest zero-discord states are obtained from the corresponding optimal measurements, and for qubit-measured systems the fidelity can be computed from the eigenvalues of a 3 Hermitian matrix 4 (Spehner et al., 2013).
This machinery was applied to the DQC1 output state. If the eigenvalues of the unitary 5 are 6, then
7
and the geometric discord is largest when the eigenvalues of 8 are uniformly distributed on the unit circle modulo symmetry with respect to the origin (Spehner et al., 2013).
Another operational branch uses the modified 9-based geometric discord,
0
which behaves better under local operations on the unmeasured subsystem and, for all 1 systems, is equivalent up to a constant factor to local quantum uncertainty (Cordero et al., 2021). In a lossy N00N-state phase-estimation problem, this quantity satisfies
2
hence
3
so the discord exactly measures the fraction of ideal quantum Fisher information surviving the loss channel (Cordero et al., 2021).
Further applications are dynamical and thermodynamic. In two-spin-4 dimers in multiple quantum NMR, Hilbert–Schmidt GQD, entropic discord, and measurement-induced non-locality can all be evaluated analytically on an 5-state family, and they show very similar dependence on inverse temperature and evolution time (Doustimotlagh, 2014). In holographic settings, the pure-state Hilbert–Schmidt GQD of the thermofield double state becomes
6
so nonzero GQD signals non-factorization of the thermal partition function and is tied in that work to the presence of the Einstein–Rosen bridge (Banerjee et al., 2023). A more recent application to private-state-based quantum key distribution derives a lower bound on the distillable secret key rate in terms of the GQD of 7 and 8, showing that secret key can still be generated even when the relevant shield states are separable or PPT entangled (Jain et al., 5 Sep 2025).
6. Dynamics, ordering relations, and unresolved issues
Because GQD is metric-dependent, its dynamical behavior is likewise geometry-dependent. In structured bosonic reservoirs, both trace-distance and Bures-distance discords can be preserved well, and may even be improved and generated by a noisy common reservoir (Hu et al., 2014). For Bell-diagonal states under independent phase-flip channels, Hilbert–Schmidt GQD evolves as
9
with 0, 1, and 2, producing a nonanalytic sudden-change point when the maximizing coefficient switches branches (Yao et al., 2013). In that same Bell-diagonal setting, if one requires the Hilbert–Schmidt GQD to remain constant for a finite period under the phase-flip channel, the initial state must be separable (Yao et al., 2013).
Multipartite open-system studies reach a different conclusion about robustness. For three-qubit and six-qubit Werner–GHZ states under amplitude damping, phase damping, depolarizing, and flipping channels, multipartite geometric discord is more fragile than global entropic discord, and depolarizing noise is the most destructive channel among those considered (Ramzan, 2012). In another multiqubit family, local phase noise acting on a single qubit produces sudden changes of multipartite geometric discord at
3
when 4, reflecting a switch in the dominant coefficient in the optimization formula (Zhu et al., 2021).
Ordering relations between geometric and entropic measures are only partly stable. For Bell-diagonal states one has the hierarchy 5, but monotonic ordering between 6, 7, and entropic discord is preserved only on highly symmetric families such as 8-symmetric states with 9; it breaks down on less symmetric 00-symmetric families with 01 (Paula et al., 2013). In Gaussian systems, fixed entropic discord does not control Hilbert–Schmidt geometric discord without an energy constraint (Adesso et al., 2011). This suggests that discord orderings are geometry-sensitive rather than universal.
Two persistent misconceptions are therefore corrected by the literature. First, GQD is not a single canonical quantity; Hilbert–Schmidt, trace, Bures, and 02-based versions define distinct measures with distinct mathematical properties (Paula et al., 2013, Spehner et al., 2013, Cordero et al., 2021). Second, computational simplicity does not by itself justify a metric choice. The Hilbert–Schmidt geometry delivers many closed formulas and useful visualizations, but its noncontractivity under local operations on the unmeasured subsystem remains a decisive limitation for interpreting it as a bona fide measure of quantum correlations (Paula et al., 2013). The subsequent development of trace-distance and Bures-distance discords, together with exact formulas for arbitrary two-qudit states and operational links to discrimination, metrology, and non-factorization, has turned GQD from a convenient surrogate for entropic discord into a technically diverse framework for quantifying nonclassical correlations across finite-dimensional, Gaussian, and multipartite quantum systems (Loubenets et al., 2024, Spehner et al., 2013, Banerjee et al., 2023).