Kirkwood-Dirac Quasiprobability

• KDQ is a quasiprobability framework representing quantum states through ordered products of noncommuting projectors, preserving normalization despite allowing negative or complex values.
• It generalizes traditional two-basis constructions to arbitrary projector families and POVMs, linking to weak values, resource theories, and work statistics in quantum thermodynamics.
• KDQ provides operational insights for state reconstruction, nonclassical diagnostics, and topology discrimination, making it a vital tool in understanding quantum measurement dynamics.

Kirkwood–Dirac quasiprobability (KDQ) is a quasiprobabilistic representation of quantum states and sequential measurement statistics built from ordered products of generally noncommuting operators. In its standard form, for two projective measurements with projectors $\Pi_i^a$ and $\Pi_j^b$, it is written as

$p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$

or, in the equivalent conjugate convention often used for pre-/post-selected measurements,

$Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$

KDQ preserves normalization and Born-rule marginals, yet can assume negative or complex values because operator ordering matters and the relevant observables need not commute. Recent work has broadened KDQ from a two-basis construction to ordered products of arbitrary projector families, connected it to weak values, resource theories, work statistics, topology diagnostics, and Gaussian continuous-variable processes, and shown that it can be informationally complete under suitable nonorthogonality conditions (Tan et al., 2024, Zhang et al., 25 Jan 2026, Hernández-Gómez et al., 2024).

1. Formal definitions and generalizations

The standard two-basis KDQ is defined for resolutions of the identity $\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$ by

$p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$

which sums to $1$ over $(i,j)$ but may be negative or complex because $\Pi_i^a$ and $\Pi_j^b$ need not commute. In the alternative convention used in measurement-theoretic analyses,

$\Pi_j^b$0

with marginals

$\Pi_j^b$1

If $\Pi_j^b$2 for all $\Pi_j^b$3, then KDQ is informationally complete and reconstructs the state through

$\Pi_j^b$4

These formulas place KDQ simultaneously in state representation and in sequential-measurement theory (Zhang et al., 25 Jan 2026).

A major generalization replaces the two-projector product by an ordered family of projector sets $\Pi_j^b$5, with

$\Pi_j^b$6

For binary projector sets $\Pi_j^b$7, this becomes

$\Pi_j^b$8

where $\Pi_j^b$9 and $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$0. This “KD-type” extension strictly enlarges the expressive power of standard KDQ and is central in resource-theoretic applications (Tan et al., 2024).

A different generalization replaces projective measurements by POVMs. For a POVM $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$1, the quantities

$p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$2

are treated as generalized KD quasiprobabilities. In the rank-one tight-frame case, with

$p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$3

they become

$p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$4

This POVM-based form links KDQ to channel unravelings and frame geometry rather than only to pairs of orthonormal bases (Rastegin, 2023).

2. Nonclassical values, incompatibility, and resource-theoretic meaning

KDQ retains correct marginals but relaxes positivity. Its nonclassicality appears as nonreal values or negative real parts. A recurring criterion is that nonclassical KD values arise from incompatibility among the state and the measured operators; if the relevant operators commute appropriately, KDQ reduces to an ordinary joint probability or to the two-point-measurement distribution in thermodynamic settings (Budiyono, 2024, Pratapsi et al., 2024).

In the generalized ordered-product setting, negativity acquires a universal resource-theoretic role. For any quantum resource theory whose free states form a closed convex set $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$5, there exists a real-valued KD-type quasiprobability $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$6 such that $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$7, with at least one negative entry for every $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$8, and no negative entries for $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$9. The construction starts from a Hermitian witness $Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$0 separating $Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$1 from $Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$2, embeds it into an extended Hilbert space as $Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$3, and factors $Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$4 as a product of projectors. Negativity is then a resource witness in quasiprobability form. The same framework yields KD-type quasiprobabilities whose total negativity is proportional to the Frobenius distance from $Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$5 to the nearest free state, and it shows that incompatible measurements are necessary for negativity. This directly motivates the claim that measurement incompatibility may underlie any quantum advantage associated with convex quantum resources (Tan et al., 2024).

KDQ also admits a boundary theory. Classical joint probabilities form a strict subset of KD quasiprobabilities, and KD quasiprobabilities form a strict subset of a broader “postquantum quasiprobability” class defined only by unit sum and per-entry modulus constraints. For an $Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$6-dimensional distribution, the paper establishes

$Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$7

together with the nontrivial KD bound

$Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$8

The same $Q_{ij}(\rho)=\mathrm{Tr}(\Pi_{a_i}\rho \Pi_{f_j}).$9 bound extends to KDQ generated by an arbitrary number of measurements. This places KDQ strictly between classical probability and a larger postquantum class while showing that KDQ remains strongly constrained despite admitting negative and complex values (Liu et al., 12 Apr 2025).

3. Weak values, measurement regimes, and informational completeness

KDQ is closely tied to weak values. For postselection on $\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$0, the weak value of $\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$1 can be written as

$\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$2

so weak values are conditional first moments of KDQ. In this decomposition, anomalous weak values arise from coherence-dependent correction terms, while the projective-measurement limit retains only the classical-like sequential contribution. This gives KDQ a direct operational role in pre-/post-selected measurement theory (Zhang et al., 25 Jan 2026).

A recent measurement-regime analysis proposes that KDQ is not confined to weak measurements. In a von Neumann pointer model, the reduced system state acquires a decoherence factor

$\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$3

and the time-dependent KDQ deforms as

$\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$4

Here $\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$5 gives the full complex KDQ and weak-value regime, while $\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$6 yields the real nonnegative “Wigner formula” $\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$7 appropriate to projective measurements. In this account, pointer-induced decoherence simultaneously controls coherence loss, measurement strength, attenuation of KD nonclassicality, and the disappearance of weak-value anomalies (Zhang et al., 25 Jan 2026).

The generalized resource-theoretic KD-type framework sharpens the weak-value connection further. Whenever the KD-type quasiprobability is negative, there exist sufficiently strong anomalous weak values in a conical decomposition built from projector weak values; conversely, this provides a route to experimental access. The same work also proves that one can build a single KD-type representation that is both informationally complete and negative if and only if the state is resourceful, by concatenating an informationally complete family of rank-1 projectors with witness-derived projectors on an extended space (Tan et al., 2024).

Direct reconstruction has also been demonstrated in a two-time thermodynamic setting. For work statistics under unitary dynamics, the characteristic function

$\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$8

is the Fourier transform of the work KDQ. An ancilla-assisted Ramsey interferometric protocol reconstructs $\sum_i \Pi_i^a=\sum_j \Pi_j^b=\openone$9, with

$p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$0

and thereby reconstructs the full complex KDQ, including its imaginary part, in an electron–nuclear spin system based on a nitrogen-vacancy center in diamond (Hernández-Gómez et al., 2024).

4. Structural characterizations and state-space geometry

Among Born-compatible quasiprobability representations on the joint spectrum of two nondegenerate observables $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$1, KD is singled out by conditional expectation. For a general Born-compatible frame representation $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$2, one can define a conditional expectation $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$3. The distinctive result is that only the left and right KD representations have the property that this induced conditional expectation coincides, for every observable $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$4, with the quantum analogue of the classical least-squares best predictor by a function of $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$5. Equivalently, KD is the unique Born-compatible quasiprobability representation whose conditioning obeys the natural pull-through law and realizes the corresponding optimal predictor (Spriet et al., 3 Nov 2025).

For finite-state systems, KDQ has an additional structural advantage: it is supported only on tuples of actual eigenvalues of the observables. In the framework of quasi-classicalization, the KD distribution of any combination of observables in an $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$6-state system takes non-zero values only at the possible values of those observables. For spin-$p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$7 and spin-1 examples, this property is contrasted with alternative quasi-probabilities that assign weight to impossible values. The same work proves that in two-state and three-state systems the KD distribution of only two spin directions can completely distinguish the state, and for qubits the imaginary part is essential for that distinguishability (Umekawa et al., 2023).

The geometry of KD-classical states has been analyzed in the special case where the transition matrix between the two reference bases is a discrete Fourier transform. A state is KD-classical when all entries $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$8 are nonnegative. In this setting, the KD-classical pure states are known explicitly, and in a $p(i,j)=\mathrm{Tr}(\Pi_i^a\Pi_j^b\rho),$9-dimensional Hilbert space the full KD-classical state set is the convex hull of KD-classical pure states. In arbitrary dimension, a directed-graph construction organizes the KD-classical pure-state families and yields pathwise convex decompositions: $1$0 This shows that KD-classicality can have a rigid arithmetic and graph-theoretic structure even when a global convex-hull theorem fails (Cai et al., 14 Mar 2026).

5. Applications in thermodynamics, dynamics, and topology

In quantum thermodynamics, KDQ replaces the two-point-measurement distribution when initial coherence in the energy basis must be preserved. For a process generated by $1$1, the work KDQ is

$1$2

and the induced work quasidistribution is

$1$3

Its first moment equals the unperturbed average energy change,

$1$4

whereas TPM generally differs because the first projective measurement dephases the initial state. In an NV-center experiment, the reconstructed work KDQ had a positive real part but a clearly nonzero imaginary part, showing that measuring only the Margenau–Hill part can miss nonclassicality tied to noncommutativity (Hernández-Gómez et al., 2024).

In quadratic fermionic models, KDQ work statistics quantify the correction induced by $1$5. The characteristic function is the two-time correlator

$1$6

and the difference between KDQ and TPM mean work isolates the off-diagonal contribution of $1$7 in the initial energy basis. In the transverse-field Ising model, nonclassical KDQ behavior—negative or complex values—tracks critical regions and accompanies enhanced work extraction relative to the dephased benchmark (Santini et al., 2023).

KDQ has also acquired a dynamical role through quantum speed limits. For

$1$8

Mandelstam–Tamm-type lower bounds are derived for the earliest time at which $1$9 can reach zero and potentially become negative, or $(i,j)$0 can exceed a chosen threshold. This turns KDQ nonclassicality into a time-constrained resource and links the onset of negativity to finite-time thermodynamic performance and power estimates (Pratapsi et al., 2024).

A distinct application concerns topology. A strange correlator between a trivial reference state $(i,j)$1 and a target state $(i,j)$2 can be rewritten in terms of KDQs

$(i,j)$3

with $(i,j)$4 and $(i,j)$5. In this form, the strange correlator is a weak value of an observable $(i,j)$6 that converts the initial state into an excited state relevant to topology discrimination. The small-$(i,j)$7 scaling of these KDQs reproduces the topology-sensitive distinction between algebraic and nonsingular behavior, and the paper proposes an interferometric KDQ-reconstruction protocol for topology discrimination after a sudden quench (Gherardini et al., 9 Jun 2026).

6. Continuous-variable, frame-theoretic, and Gaussian-process developments

In finite-dimensional frame theory, generalized KD quasiprobabilities arise naturally from tight-frame POVMs and quantum-channel unravelings. For a rank-one tight frame $(i,j)$8,

$(i,j)$9

and the corresponding channel-Kraus Gram matrix satisfies

$\Pi_i^a$0

For equiangular tight frames, one obtains exact Hilbert–Schmidt norm identities, spectral-norm bounds, and Rényi and Tsallis entropic uncertainty relations for arbitrary unravelings. In this approach, generalized KD matrices are Hermitian positive semidefinite at the channel level even though their individual entries may be complex (Rastegin, 2023).

For Gaussian continuous-variable processes, KDQ has been analyzed for ordered sequences of Gaussian measurements, optionally separated by Gaussian unitaries. The central quantity is the KD negativity

$\Pi_i^a$1

and for $\Pi_i^a$2 modes and $\Pi_i^a$3 Gaussian measurements with covariances $\Pi_i^a$4 the paper derives the universal bound

$\Pi_i^a$5

For Gaussian input states, $\Pi_i^a$6 is expressible through a complex covariance matrix $\Pi_i^a$7. In the single-mode, two-measurement case, the upper bound is saturated by the quadrature-eigenstate limit of infinitely squeezed pure Gaussian states, while a nontrivial minimum along the pure Gaussian manifold is also obtained. This indicates that, in the Gaussian-process setting considered, Gaussian states already suffice to realize extreme KD nonclassicality (Bianchi et al., 13 Jul 2026).

These developments also delimit scope. Universal resource-theoretic negativity results assume closed convex free-state sets and use witness factorization on an extended space; topology constructions depend on the strange-correlator framework; KD-classical-state geometry based on directed graphs is derived for discrete-Fourier-transform-related bases; and Gaussian extremality results are established for Gaussian measurements and Gaussian unitary processes. Taken together, however, they show that KDQ has become a unifying language for ordered quantum correlators, state representation, operational nonclassicality, and application-specific witnesses across finite-dimensional, many-body, and continuous-variable settings (Tan et al., 2024).