Anomalous Negative Weak Value

• Anomalous negative weak values are defined as weak measurement outcomes whose real parts lie outside the eigenvalue spectrum, defying classical bounds.
• They arise when nearly orthogonal pre- and post-selected states cause interference terms to dominate, amplifying small denominators in the von Neumann measurement framework.
• Experimental studies using photons and interferometers validate these effects, linking negative quasiprobabilities with contextuality and nonclassical causation.

An anomalous negative weak value is a weak value whose real part is negative even though the measured observable has a nonnegative spectrum, or more generally a spectrum whose convex hull excludes such negativity. In the standard Aharonov–Albert–Vaidman framework, the weak value of an observable $A$ between a pre-selected state $|\psi_i\rangle$ and a post-selected state $|\psi_f\rangle$ is $$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$; for a projector $\Pi$, anomalous negativity means $\mathrm{Re}(\Pi_w)<0$, outside the interval $[0,1]$. The subject now encompasses not only the conventional post-selected weak-measurement protocol, but also sequential weak measurements without post-selection, intensity-based extraction schemes without meter states, contextuality witnesses, and several competing interpretive accounts of why such anomalies occur (Abbott et al., 2018, Masiello et al., 23 Jun 2026).

1. Formal framework and basic definitions

Weak values arise in the von Neumann measurement model, where the system observable $A$ is coupled impulsively to a meter by

$H_{\text{int}}(t)=g(t)\,A\otimes p,$

with $p$ the pointer momentum. For a Gaussian pointer of width $|\psi_i\rangle$0, the weak regime yields the standard readout relations

$|\psi_i\rangle$1

and, in the normalization used in one standard treatment,

$|\psi_i\rangle$2

for the generalized weak value

$|\psi_i\rangle$3

The pure-state expression is recovered when $|\psi_i\rangle$4 and $|\psi_i\rangle$5 (Abbott et al., 2018).

Anomalousity is defined relative to the spectrum of $|\psi_i\rangle$6. For a Hermitian observable with eigenvalues in $|\psi_i\rangle$7, a weak value is anomalous when $|\psi_i\rangle$8; a nonzero imaginary part also lies outside the convex hull of the spectrum in the generalized sense used in operational contextuality results (Kunjwal et al., 2018). For projectors, whose eigenvalues are $|\psi_i\rangle$9 and $|\psi_f\rangle$0, a negative real weak value is the canonical anomalous negative case (Masiello et al., 23 Jun 2026).

A crucial elementary fact is that a single weak measurement without nontrivial post-selection cannot produce an anomaly. If one “post-selects” the identity, then

$|\psi_f\rangle$1

which is just the ordinary expectation value and remains within the spectral bounds of $|\psi_f\rangle$2 (Abbott et al., 2018). This is why the historical association of anomalous weak values with post-selection became so strong.

2. Mechanisms generating anomalous negativity

The standard route to anomalous weak values is a small denominator $|\psi_f\rangle$3, typically produced by nearly orthogonal pre- and post-selected states. In this regime, the coherent numerator $|\psi_f\rangle$4 is divided by a suppressed overlap, and $|\psi_f\rangle$5 can become arbitrarily large in magnitude, including large negative values for observables such as projectors or Pauli operators (Sokolovski, 2015, Dressel, 2014).

Several formulations sharpen this mechanism. One decomposition writes

$|\psi_f\rangle$6

so that

$|\psi_f\rangle$7

In this representation, the anomalous part is entirely the interference term between the post-selected state and a component orthogonal to the pre-selected state. This yields a necessary and sufficient condition for anomaly in that framework: the post-selected state must overlap the orthogonal component generated by $|\psi_f\rangle$8 (Pati et al., 2014).

A stronger structural statement is that anomalous weak values require coherence in the eigenbasis of $|\psi_f\rangle$9. If the pre- and post-selection states are both diagonal in that basis, the weak value becomes a convex combination of eigenvalues and cannot be anomalous. In the mixed-state form

$$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$0

the associated quasiprobability weights

$$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$1

must cease to be ordinary probabilities for anomaly to occur; this requires coherence (Wagner et al., 2023).

A complementary linear-algebraic viewpoint expresses any weak value as the expectation value of a suitable non-normal operator, such as

$$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$2

Because non-normal expectations are not constrained by the spectral bounds of the original Hermitian observable, this framework explains why weak values can be negative or otherwise anomalous. In that analysis, the departure from normality scales as

$$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$3

linking amplification directly to uncertainty and small overlap (Ferraz et al., 2023).

These mechanisms are closely tied to quasiprobability negativity. Negative anomalous weak values coincide with negative elements of Kirkwood–Dirac or related quasiprobability distributions in several treatments, and the same connection is made to pseudo-probabilities built from noncommuting projectors (Masiello et al., 23 Jun 2026, Asthana et al., 2020).

3. Sequential weak measurements and anomalies without post-selection

A major revision of the conventional picture is that anomalous weak values do not, in general, require discarding data. For two sequential measurements of $$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$4 followed by $$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$5, the no-post-selection sequential weak value is

$$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$6

If the first measurement is weak, then the joint pointer correlations satisfy

$$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$7

while all runs are retained. The second measurement can be arbitrarily strong, and in this setting it acts as an effective post-selection factor inside the correlator rather than as an explicit data filter (Abbott et al., 2018).

This mechanism permits anomalous negative values for products of nonnegative observables. For two projectors $$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$8 and $$A_w=\langle\psi_f|A|\psi_i\rangle/\langle\psi_f|\psi_i\rangle$$9, one has the general lower bound

$\Pi$0

and that bound also holds for mixed states. Thus the most negative anomalous value achievable without post-selection for two $\Pi$1 measurements is $\Pi$2 (Abbott et al., 2018).

The standard qubit example prepares $\Pi$3 and performs sequential weak measurements of two projectors $\Pi$4 and $\Pi$5 built from

$\Pi$6

The exact joint pointer mean is

$\Pi$7

which depends only on the first measurement strength. In the weak limit $\Pi$8,

$\Pi$9

Since each projector is $\mathrm{Re}(\Pi_w)<0$0-valued, the classically allowed range of the product is $\mathrm{Re}(\Pi_w)<0$1, so the negative limit is anomalous. The individual pointer means remain non-anomalous; the anomaly appears only in the correlator (Abbott et al., 2018).

For more than two measurements, the relation between pointer products and sequential weak values becomes more intricate. For $\mathrm{Re}(\Pi_w)<0$2, the mean product of pointer positions is a mixture of $\mathrm{Re}(\Pi_w)<0$3 different permutations rather than a single ordered product. For three measurements without post-selection,

$\mathrm{Re}(\Pi_w)<0$4

This complicates interpretation and separates pointer anomalies from sequential weak values themselves (Abbott et al., 2018).

4. Explicit examples and experimental realizations

The first direct experimental observation of an anomalous weak value without post-selection used single-photon polarization as the qubit and two transverse spatial degrees of freedom as independent meters. The sequential projectors

$\mathrm{Re}(\Pi_w)<0$5

were implemented by polarization-dependent blazed gratings on a spatial light modulator. No polarization post-selection was performed; all photons were used. The joint mean $\mathrm{Re}(\Pi_w)<0$6 became negative for $\mathrm{Re}(\Pi_w)<0$7, with maximum negativity at $\mathrm{Re}(\Pi_w)<0$8, and returned positive in the strong-measurement regime, in agreement with the theory that the anomaly disappears when the first interaction becomes effectively projective (Yang et al., 2019).

A distinct 2026 realization extracted anomalous path weak values in a generalized Mach–Zehnder interferometer directly from output intensities, without meter states and without weak interactions. For a path qubit

$\mathrm{Re}(\Pi_w)<0$9

a controlled phase $[0,1]$0 in one arm yields output intensities $[0,1]$1 whose sinusoidal dependence determines the complex path weak values $[0,1]$2. In an unbalanced neutron interferometer with $[0,1]$3, the theory predicts

$[0,1]$4

and at $[0,1]$5,

$[0,1]$6

The measured data agreed with these predictions within error bars. In the balanced configuration, by contrast, $[0,1]$7 and no anomalous real parts occur (Masiello et al., 23 Jun 2026).

Another line of work emphasized that the physical effect of a negative projector weak value can mirror that of a positive unit weak value with opposite sign. For a path projector with weak value $[0,1]$8, the polarization of a photon probe is rotated by the same magnitude and opposite sign as for weak value $[0,1]$9. This symmetry was confirmed by two-photon Hong–Ou–Mandel interference through the compensating half-wave-plate angle needed to restore indistinguishability (Yokota et al., 2016).

A single-click experiment on a spatially resolving detector established that anomalous weak values need not be interpreted purely as many-shot statistical artifacts. In a seven-stage robust weak measurement of $A$0, the reported anomalous run had theoretical weak value $A$1 for an observable with spectrum $A$2, and a single-click readout $A$3. The paper’s central anomaly was positive, but the same formalism gives explicit negative anomalous constructions, for example

$A$4

which equals $A$5 when $A$6 (Rebufello et al., 2021).

5. Contextuality, quasiprobability, and competing interpretations

Negative anomalous weak values occupy an unusual position in quantum foundations because they are simultaneously operationally measurable and interpretively contested. One influential line of work proves that anomalous weak values witness contextuality under explicit operational assumptions. In a noise-robust formulation for projector weak values, a noncontextual model implies

$A$7

where $A$8 is the tail probability of negative pointer outcomes followed by successful post-selection, $A$9 is the post-selection probability, and $H_{\text{int}}(t)=g(t)\,A\otimes p,$0 quantifies disturbance. Violating this inequality certifies contextuality. The same framework extends to imaginary weak values and to qubit-pointer implementations (Kunjwal et al., 2018).

Closely related approaches identify anomalous weak values with negative quasiprobabilities. In the generalized Mach–Zehnder analysis, the Kirkwood–Dirac element

$H_{\text{int}}(t)=g(t)\,A\otimes p,$1

is directly reconstructed from intensity normalization and weak values; $H_{\text{int}}(t)=g(t)\,A\otimes p,$2 for a projector implies negative $H_{\text{int}}(t)=g(t)\,A\otimes p,$3 (Masiello et al., 23 Jun 2026). In the pseudo-probability framework, negative entries of pseudo-probability schemes are precisely measured as negative weak values of suitable projectors, linking anomalous negativity to tests of non-Boolean logic, nonlocality, entanglement, and discord (Asthana et al., 2020).

At the same time, there is no consensus on whether anomaly itself is uniquely quantum. One operational account generalizes the phenomenon to anomalous post-selected shifts and argues that such anomalies require correlations between the intermediate outcome and the post-selection, whether the underlying model is quantum or classical. The paper gives a simple classical model with a maximally anomalous negative post-selected average $H_{\text{int}}(t)=g(t)\,A\otimes p,$4 and no disturbance, concluding that correlation alone can supply the anomaly in that operational sense (Ferrie et al., 2014). A different critique argues that the disturbance caused by the weak interaction, though small, is sufficient to bias the post-selected ensemble and thereby generate anomalous weak values; in that view, post-selection-conditioned disturbance is the most reasonable explanation of the anomaly (Ipsen, 2021).

Opposing accounts emphasize instead that weak values are fundamentally interference phenomena. On that view, the anomaly reflects coherent amplitude ratios and negative quasiprobabilities rather than classical disturbance, and classical disturbance models cannot reproduce the full functional dependence of weak values without effectively simulating interference (Dressel, 2014). Another paper states the point more strongly, arguing that the phenomenon has no nontrivial classical analogue because the anomalous average is built from negative weights derived from probability amplitudes rather than ordinary probabilities (Sokolovski, 2015). The coexistence of these positions has made anomalous negative weak values a focal case in debates over contextuality, disturbance, and the operational meaning of post-selection.

6. Broader developments, applications, and open problems

Anomalous negative weak values now appear in several adjacent contexts. In causal-structure analysis, an anomalous negative mean product of two projector readouts without post-selection certifies a direct-cause scenario, because a common-cause model yields an ordinary expectation value $H_{\text{int}}(t)=g(t)\,A\otimes p,$5 for projectors and therefore cannot be anomalous. This suggests using anomalous negativity as a witness of direct causation and motivates open questions about witnessing indefinite causal order through enhanced anomalies (Abbott et al., 2018).

In optics, weak values of local momentum have been interpreted as the canonical Poynting vector normalized by energy density. In that setting, negative local momentum values correspond to optical backflow, while superluminal local momenta arise near vortices and in evanescent waves. The formal weak value

$H_{\text{int}}(t)=g(t)\,A\otimes p,$6

connects negative anomalies to local energy-flow reversal in structured classical fields (Bliokh et al., 2013).

Several thought experiments push anomalous negativity into more dramatic territory. In one pre- and post-selected Mach–Zehnder setup, the path weak values are $H_{\text{int}}(t)=g(t)\,A\otimes p,$7 and $H_{\text{int}}(t)=g(t)\,A\otimes p,$8, so the left arm behaves as if it contains “$H_{\text{int}}(t)=g(t)\,A\otimes p,$9” atoms while the right behaves as if it contains “4,” with the sum still equal to $p$0. The resulting stimulated-emission and spontaneous-emission predictions include a “phantom arm” effect and motivate the notion of “counter-particles,” namely pre- and post-selected entities acting as if they had negative physical variables such as mass or energy (Aharonov et al., 2017).

The generalized Quantum Cheshire Cat and its dual extend the same logic to path–polarization separation and “phantom-arm” interference. In one formulation, the photon is weakly present in arm $p$1 but the polarization weak value is entirely assigned to arm $p$2; in the dual setup, a horizontally polarized single-arm interferometer exhibits vertically polarized interference arising from the interplay of pre- and post-selection. Negative and purely imaginary weak values appear by phase tuning, with the sign of the weak value controlling the sign of the pointer shift (Quach, 2017).

Open problems remain explicit in the literature. Sequential weak values beyond two measurements mix permutations in experimentally accessible pointer products; the physical interpretation of those mixed correlators is not settled. Numerical evidence cited in the no-post-selection theory suggests that pointer anomalies may not surpass $p$3 even when sequential weak values themselves approach $p$4 in longer projector sequences. Other unresolved directions include contextuality proofs without conventional post-selection, witnessing indefinite causal order, and sharpening the distinction between pointer anomalies and the underlying sequential weak values (Abbott et al., 2018).

In contemporary usage, the term “anomalous negative weak value” therefore denotes more than a curious sign reversal in a weak pointer. It names a family of quantum-conditioned quantities that can violate spectral intuition, coincide with negative quasiprobabilities, certify nonclassical causal or contextual structure under specific assumptions, and, in some experimentally accessible settings, arise even when no data are discarded.