Papers
Topics
Authors
Recent
Search
2000 character limit reached

Direct Weak Measurement in Quantum Mechanics

Updated 14 July 2026
  • Direct weak measurement is a quantum protocol that leverages weak interactions and a complementary strong measurement to directly access the wavefunction coefficients.
  • The technique has been expanded to include mixed states and density matrices through methods like the Dirac quasi-probability distribution.
  • Experimental implementations in photonic optics, neutron interferometry, circuit QED, and NMR demonstrate its broad applicability and potential for in situ quantum state analysis.

Direct weak measurement denotes a family of quantum measurement protocols in which the complex coefficients of a state representation are obtained from weakly coupled measurement records, rather than inferred only after a global tomographic reconstruction. In the original formulation due to Lundeen and collaborators, the wavefunction is operationally defined as the average result of a weak measurement of a variable followed by a strong measurement of the complementary variable; subsequent work generalized the method to mixed states, density matrices, Dirac quasi-probability distributions, measurements, and processes, and also clarified that the same direct-measurement target can be realized with couplings of arbitrary strength (Lundeen et al., 2011, Lundeen et al., 2011).

1. Origin of the concept

The modern notion of direct weak measurement emerged from the proposal that an unknown pure state can be read out coefficient by coefficient in a chosen basis. Let {∣a⟩}\{|a\rangle\} be an orthonormal basis for an observable AA, and let {∣b⟩}\{|b\rangle\} be a complementary, mutually unbiased basis for an observable BB, so that ⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N} up to a phase in an NN-dimensional Hilbert space. If one weakly measures the projector

πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|

and then strongly measures BB, post-selecting a fixed outcome ∣b0⟩|b_0\rangle, the weak value becomes

⟨πa⟩W=⟨b0∣πa∣Ψ⟩⟨b0∣Ψ⟩=⟨b0∣a⟩⟨a∣Ψ⟩⟨b0∣Ψ⟩.\langle \pi_a\rangle_W = \frac{\langle b_0|\pi_a|\Psi\rangle}{\langle b_0|\Psi\rangle} = \frac{\langle b_0|a\rangle\langle a|\Psi\rangle}{\langle b_0|\Psi\rangle}.

Because AA0 is known and independent of AA1 for a suitable choice of complementary bases, scanning AA2 directly yields the expansion coefficients AA3, and hence

AA4

for an AA5-independent constant AA6 (Lundeen et al., 2011).

This formulation was presented as a contrast with conventional quantum state tomography. In tomography, one measures many different observables or bases and then reconstructs the state by linear inversion, maximum likelihood, or related post-processing. In direct weak measurement, the raw signal at each apparatus setting is proportional to the real and imaginary parts of the target wavefunction component, up to normalization. The protocol therefore gives the wavefunction a specific operational definition rather than treating it only as an inferred object (Lundeen et al., 2011).

2. Weak values and the pure-state protocol

The underlying measurement model is the von Neumann interaction

AA7

with AA8 the system observable, AA9 a pointer momentum, and {∣b⟩}\{|b\rangle\}0 a small coupling. In the weak regime, each individual outcome carries little information and the disturbance to the system is correspondingly small, but averaging over many trials yields a stable signal. For a pre-selected pure state {∣b⟩}\{|b\rangle\}1 and post-selection onto {∣b⟩}\{|b\rangle\}2, the weak value is

{∣b⟩}\{|b\rangle\}3

It can be complex and can lie outside the spectrum of {∣b⟩}\{|b\rangle\}4; in the pointer model, the real and imaginary parts appear as shifts in conjugate pointer variables (Lundeen et al., 2011).

For the continuous-variable version of the original protocol, one weakly measures the position projector

{∣b⟩}\{|b\rangle\}5

and post-selects a momentum eigenstate {∣b⟩}\{|b\rangle\}6. The resulting weak value is

{∣b⟩}\{|b\rangle\}7

Choosing {∣b⟩}\{|b\rangle\}8 gives

{∣b⟩}\{|b\rangle\}9

so the wavefunction at each position is proportional to the measured weak value. In the optical realization, the system degree of freedom was the transverse spatial coordinate of a single photon, while polarization served as the pointer; the real and imaginary parts of BB0 were obtained from complementary polarization observables after post-selection on BB1 (Lundeen et al., 2011).

A central feature of this early literature is that “direct” referred to a local relation between the measured weak average and the state coefficient. The protocol still requires an ensemble of identically prepared systems, but it does not require a global inversion algorithm beyond final normalization (Lundeen et al., 2011).

3. Mixed states, Dirac distributions, and direct density matrices

The pure-state protocol does not by itself reconstruct a general mixed state. For a mixed state BB2, weakly measuring BB3 and post-selecting BB4 yields

BB5

and scanning BB6 gives only BB7 real parameters, rather than the BB8 needed for a general density matrix. Lundeen and Bamber therefore generalized direct weak measurement to mixed states through the Dirac quasi-probability distribution (Lundeen et al., 2011).

In a discrete BB9-dimensional Hilbert space with complementary bases ⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}0 and ⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}1, the Dirac distribution is

⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}2

Defining the non-Hermitian operator

⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}3

one has

⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}4

A joint weak measurement of complementary variables therefore directly measures the Dirac distribution, point by point. When ⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}5 is chosen as the discrete Fourier transform of ⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}6, the density matrix is recovered by a discrete Fourier transform: ⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}7 This establishes the Dirac distribution as a complete, directly measurable representation of the state (Lundeen et al., 2011).

A second line of work gave a direct measurement of the density matrix elements themselves. Choosing a basis ⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}8 and a complementary state ⟨a∣b⟩=1/N\langle a|b\rangle = 1/\sqrt{N}9 with NN0, one defines

NN1

Its weak average is

NN2

Accordingly, each density-matrix element is associated with a specific weak–weak–strong sequence of complementary projectors. In a later photonic experiment, the first two measurements were made weak and the last strong, and the density matrix was obtained directly from pointer correlations without a global reconstruction step (Lundeen et al., 2011, Thekkadath et al., 2016).

These mixed-state developments also underwrote the claim that direct weak measurement can be performed “in situ”: by taking the weak-coupling limit, the disturbance can be made arbitrarily small while still determining a complete state representation (Lundeen et al., 2011).

4. Experimental realizations

Direct weak measurement has been implemented across several physical platforms. In the original optical wavefunction experiment, a narrow half-wave-plate sliver performed a weak position measurement by rotating polarization only within a selected spatial slice, a Fourier-transform lens mapped position to momentum, and a slit selected NN3; the final polarization readout gave the real and imaginary parts of the weak value, and hence of NN4 (Lundeen et al., 2011). In the direct density-matrix experiment, two walk-off crystals induced polarization-dependent spatial shifts that weakly measured two complementary projectors, followed by a strong polarization projection (Thekkadath et al., 2016).

Matter-wave implementations used neutron interferometry. One experiment directly characterized the neutron path state using both weak and strong interactions, with the neutron spin serving as the pointer; it reported that strong interactions outperform weak ones (Denkmayr et al., 2016). A related neutron experiment used the imaginary part of a single weak value of a projector to directly test the commutation relation between two dichotomic observables, with the path degree of freedom as the system and the spin/energy degree of freedom as ancilla (Wagner et al., 2020).

Subsequent work transferred the direct weak measurement program to ensemble quantum-information platforms. In circuit QED, the qubit was dispersively coupled to a cavity and the homodyne current played the role of the pointer; the analysis was carried out for finite-strength weak measurement beyond the bad-cavity and weak-response limits (Qin et al., 2016). On an NMR quantum processor, direct quantum state tomography and direct quantum process tomography were implemented by weak measurements of Pauli operators using an extra metre qubit and, for processes, an ancilla qubit, without involving any projective measurements (Gaikwad et al., 2023).

Platform System and pointer Directly accessed quantity
Photonic spatial optics Transverse spatial mode and polarization Wavefunction or density matrix
Neutron interferometer Path qubit and spin/energy or spin pointer Path-state amplitudes or commutator expectation value
Circuit QED Qubit and homodyne current of a cavity output Qubit state via a post-selected weak value of NN5
NMR quantum processor System qubits and a metre qubit via scalar NN6-coupling Density-matrix or process-matrix elements

5. Generalizations beyond the original weak-coupling picture

A major conceptual revision came from the observation that direct state determination does not require infinitesimal coupling. Vallone and Dequal showed theoretically that direct measurement of the wavefunction can be obtained by using measurements of arbitrary strength, and in particular that strong measurements outperform weak measurements in both precision and accuracy for arbitrary quantum states in most cases; in this view, direct weak tomography is the weak-limit case of a more general direct strong tomography (Vallone et al., 2015). An experimental two-dimensional photonic realization later reported direct strong measurements of Gaussian and Laguerre–Gaussian (NN7) spatial wavefunctions with NN8-square values of NN9 and πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|0, respectively (Zhang et al., 2018).

Other generalizations extended the direct-measurement logic to new objects. A protocol for direct characterization of quantum measurements treated the POVM element itself as the object to be reconstructed, effectively taking the backward direction of the weak-measurement formalism; it worked rigorously with arbitrary measurement strength and used the completeness condition of the measurement operators to improve precision (Xu et al., 2020). Direct tomography of quantum states and processes on NMR used weak measurements of Pauli operators to directly access selected density-matrix and process-matrix elements (Gaikwad et al., 2023).

The original wavefunction protocol also did not apply straightforwardly to nonlocal entangled states, because direct weak measurement of nonlocal observables would require a nonlocal interaction Hamiltonian. A later photonic method replaced weak values of nonlocal observables by modular values obtained from local interactions, and used them to directly measure the nonlocal wavefunction of a bipartite system in a hyper-entangled state (Pan et al., 2019).

These developments suggest that the stable core of the subject is not simply “weakness” as a small-coupling approximation, but the existence of a direct analytic relation between a measured conditional average and the target state quantity.

6. Limitations, terminology, and conceptual status

Direct weak measurement retains several practical limitations. Weak measurements have inherently low signal-to-noise and require large ensembles; joint weak measurements of multiple observables have exponentially degraded signal-to-noise, with πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|1, and post-selection can strongly limit efficiency (Lundeen et al., 2011). In direct wavefunction measurement, the success probability of post-selection is a major bottleneck, and one proposed modification inserted a liquid crystal plate after the pointer coupling to increase post-selection efficiency; for the parameter set πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|2, πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|3, πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|4, πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|5, and πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|6, the analysis gave πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|7 and a magnification πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|8 (Wen et al., 2021).

The term itself has also broadened. In one optical interferometry paper, a protocol was described as a direct weak-measurement scheme because a small polarization phase πa=∣a⟩⟨a∣\pi_a = |a\rangle\langle a|9 was re-encoded into a larger effective phase and then measured directly as a phase, even though the authors explicitly stated that there does not exist a generally defined weak value in their scheme (Li et al., 2018). A later unified framework defined direct weak measurement as the case where the system evolves under a Hamiltonian BB0 with no explicit probe, so that the transition probability directly yields only BB1 under unitary dynamics; in that framework, direct measurement appears as a special case of indirect measurement (Ueda et al., 28 Sep 2025).

A further conceptual issue is the status of non-Hermitian observables. The Dirac-distribution and direct-density-matrix protocols rely on weak averages of non-Hermitian products of projectors. The operational answer given in this literature is that the experimental record consists of real pointer readings, while the complex weak average is a convenient grouping of two real measured averages; the complex number therefore has a direct experimental meaning even when the corresponding operator is not observable in the projective-measurement sense (Lundeen et al., 2011).

Taken together, these strands establish direct weak measurement as a methodological class rather than a single protocol. Its original form is the weak measurement of a projector followed by a strong complementary measurement, yielding wavefunction amplitudes directly. Its mature form includes direct access to Dirac distributions, density matrices, retrodicted measurement operators, process matrices, and certain nonlocal amplitudes, together with the recognition that directness can survive beyond the strict weak-coupling limit.

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

No one has generated a whiteboard explanation for this topic yet.

Follow Topic

Get notified by email when new papers are published related to Direct Weak Measurement.