---
title: Probabilistic Projective Measurement (PPM)
url: https://www.emergentmind.com/topics/probabilistic-projective-measurement-ppm
type: topic
---

# Probabilistic Projective Measurement (PPM)

to=arxiv_search.search  微信里的天天中彩票json
{"query":"all:probabilistic projective measurement OR ti:\"Optimal probabilistic measurement of phase\" OR ti:\"Simulating all quantum measurements using only projective measurements and postselection\" OR ti:\"Optimal quantum-programmable projective measurement with linear optics\"","max_results":10,"sort_by":"relevance"}
to=arxiv_search.search  彩神争霸邀请码json
{"query":"ti:\"Optimal probabilistic measurement of phase\" OR ti:\"Simulating positive-operator-valued measures with projective measurements\" OR ti:\"Pretty-good simulation of all quantum measurements by projective measurements\"","max_results":10,"sort_by":"relevance"}
to=arxiv_search.search  五分彩json
{"query":"id:1307.3070 OR id:1807.08449 OR id:1609.06139 OR id:1805.02546 OR id:2501.09339 OR id:2507.02201 OR id:1501.01339 OR id:1004.4313","max_results":20,"sort_by":"relevance"}
Probabilistic Projective Measurement (PPM) denotes a family of measurement constructions in which a projective, projective-simulable, or projective-target operation is realized only probabilistically. In the literature, the term is not used with a single universal definition. Depending on context, it may refer to a projective measurement with an inconclusive outcome, a postselected simulation of a POVM by randomized projective measurements, a heralding projector used for conditional state preparation, a programmable approximation to a target projector, or a classical randomization between projective measurement and identity in sequential protocols. A common structural feature is that the desired projective action is available only on a successful branch, while the complementary branch is discarded, flagged, or treated as a failure/inconclusive event [1307.3070], [1807.08449], [1805.02546], [2507.02201], [2510.02710].

## 1. Terminological scope and representative meanings

The literature suggests that “PPM” functions more as an umbrella label than as a single formalism. The main usages identified in recent work are summarized below [1307.3070], [1609.06139], [1805.02546], [2507.02201], [2510.02710].

| Setting | Operational meaning | Representative source |
|---|---|---|
| Phase estimation | Filter \(F\) plus canonical phase measurement, with inconclusive outcome | [1307.3070] |
| POVM simulation | Classical randomization, projective measurements, post-processing, often postselection | [1609.06139], [1807.08449], [2501.09339] |
| Programmable measurement | Quantum program state approximates a target projector | [1805.02546] |
| Heralded state engineering | Projector onto a detected outcome prepares a conditional state | [2507.02201] |
| Sequential correlation tests | Coin-toss mixture of projective measurement and identity | [2510.02710] |

This heterogeneity is operationally significant. In some settings the probabilistic element is intrinsic to the measurement itself, because an extra outcome \(\Pi_0\) collects failed trials. In others it is generated by postselection after a larger measurement protocol, or by heralding on a detector click, or by classical randomization over projective instruments. Accordingly, “projective” may denote an actual orthogonal projector, an approximated target projector, or membership in the convex hull of projective measurements.

## 2. Filtered and inconclusive-outcome realizations

A canonical optics-based usage appears in phase estimation. There, a probabilistic measurement is explicitly defined as a measurement that may return an inconclusive outcome with nonzero probability, but, upon success, implements a sharper-than-deterministic measurement [1307.3070]. For phase, the construction extends the canonical phase POVM by a success filter \(F\),
\[
\Pi^{(P)}_\phi = \frac{1}{2\pi}F|\phi\rangle\langle\phi|F^\dagger,\qquad
\Pi^{(P)}_0 = 1-\int \Pi^{(P)}_\phi\,d\phi,
\]
with \(F=\mathrm{diag}(f_0,f_1,\dots)\), \(0\le f_n\le 1\). In this formulation, PPM is literally “filter + canonical phase measurement.”

The target of the optimization is phase precision. The paper uses
\[
V=|\mu|^{-2}-1,\qquad \mu=\langle e^{i\theta}\rangle,
\]
and, for a finite-dimensional state \(|\psi\rangle=\sum_{n=0}^N c_n|n\rangle\),
\[
\mu=\left|\sum_{n=0}^{N-1}c_n c_{n+1}^*\right|.
\]
Larger \(|\mu|\) means smaller phase variance \(V\). The ideal single-shot phase measurement is described as projection onto phase states
\[
|\theta\rangle=\sum_{k=0}^{\infty} e^{ik\theta}|k\rangle,
\]
which are unphysical because they are not normalizable and not orthogonal. The probabilistic filter therefore functions as a physically admissible route toward the ideal canonical phase measurement, but only on a successful branch.

For an input \(|\psi\rangle=\sum_{n=0}^{\infty} c_n|n\rangle\) with \(c_n>0\), the success probability is
\[
P=\sum_{n=0}^{\infty} f_n^2 c_n^2.
\]
The optimization problem is: for fixed \(P\), choose \(f_n\) to maximize \(\mu\) of the filtered state. The central conclusion is a bounded success–precision tradeoff: lower success probability can improve phase precision, the improvement is bounded, and for each success probability \(P\) there is an optimal filter. For coherent states, the paper derives a semianalytic optimal solution and identifies a highly nonlinear filter as the optimal measurement device.

Operationally, this class of PPM is equivalent to state preprocessing by filtering followed by a standard phase measurement. The paper treats direct probabilistic measurement and probabilistic manipulation of the state first, including noiseless amplification, as essentially the same framework at the level of analysis.

## 3. Projective simulation of generalized measurements

A second major meaning of PPM is projective simulation of POVMs by classical randomization, post-processing, and sometimes postselection. In this line of work, a POVM \(M=(M_1,\ldots,M_n)\) is PM-simulable if it can be obtained from projective measurements using classical randomization over different projective measurements and classical post-processing. A basic structural theorem states
\[
S(d,n)=P(d,n)^{\mathrm{conv}},
\]
so the set of PM-simulable measurements is exactly the convex hull of projective measurements [1609.06139]. The same work proves that every measurement on a \(d\)-dimensional system can be realized by classical processing of projective measurements on the system plus an ancilla of the same dimension, and shows that simulability for qubits and qutrits can be decided by semidefinite programming. In the qubit case, a notable benchmark is the tetrahedral POVM, whose critical visibility is
\[
t_{\mathrm{tetra}}=\sqrt{\frac{2}{3}}\approx 0.8165.
\]

An ancilla-free probabilistic realization of arbitrary POVMs is given by a postselection scheme based solely on classical randomization, projective measurements on the system of interest, and postselection on a designated failure outcome [1807.08449]. For a POVM \(M=(M_1,\dots,M_n)\), the postselected simulation is encoded by
\[
M_q=(qM_1,\dots,qM_n,(1-q)\mathbb{1}),
\]
and the paper proves that every POVM on \(\mathbb{C}^d\) can be simulated this way with success probability
\[
q=\frac{1}{d},
\]
which is optimal in general. The construction reduces to rank-one effects \(M_i=\alpha_i|\psi_i\rangle\langle\psi_i|\), samples index \(i\) with probability \(p_i=\alpha_i/d\), performs the two-outcome projective measurement \(\{|\psi_i\rangle\langle\psi_i|,\mathbb{1}-|\psi_i\rangle\langle\psi_i|\}\), outputs \(i\) on the “\(+\)” branch, and assigns the “\(-\)” branch to failure. Conditioning on success reproduces the target POVM exactly.

A stronger asymptotic simulability statement appears in the depolarized setting. For every POVM \(M\) on \(\mathbb{C}^d\),
\[
[\Phi_c(M)]_i=c\,M_i+(1-c)\frac{\operatorname{tr}(M_i)}{d}I_d,
\]
with \(c=0.02\), belongs to the set of projectively simulable measurements \(S(\mathbb{C}^d)\) [2501.09339]. This means that a dimension-independent amount of depolarizing noise renders every POVM implementable as a randomized convex combination of projective measurements on the original system, with no ancilla at all. In this sense, one strand of PPM theory studies how far generalized measurement can be reduced to randomized projective measurement under either probabilistic overhead or controlled noise.

## 4. Programmable and approximate projective measurements

Another usage concerns projective measurements whose basis is specified by a quantum program state. The target is the two-outcome projective measurement
\[
\{\ket{\psi}\!\bra{\psi},\,I-\ket{\psi}\!\bra{\psi}\},
\]
where the device receives one copy of \(\ket{\phi}\) and \(M-1\) copies of the program state \(\ket{\psi}\) [1805.02546]. Exact implementation is impossible with a finite number of program copies, so the relevant object is an \(\epsilon\)-approximate projective measurement with one-sided error.

The paper realizes this using the swap test of order \(M\). Its output probabilities are
\[
\Pr(0)=\frac{1}{M}+\frac{M-1}{M}|\langle \phi|\psi\rangle|^2,\qquad
\Pr(1)=\frac{M-1}{M}\bigl(1-|\langle \phi|\psi\rangle|^2\bigr).
\]
Interpreting outcome \(0\) as projection onto \(\ket{\psi}\), the deviation from the ideal Born rule is
\[
\epsilon=\frac{1}{M}.
\]
The scheme has one-sided error because \(\Pr(0)=1\) whenever \(\ket{\phi}=\ket{\psi}\).

The work proves two optimality statements. First, the order-\(M\) swap test is optimal among all one-sided-error identity tests for the asymmetric input structure \(\ket{\phi}\ket{\psi}^{\otimes(M-1)}\). Second, any device that uses \(N\) copies of the program state and achieves error \(\epsilon\) under one-sided error must satisfy
\[
N\ge \frac{1}{\epsilon}-1.
\]
Thus the construction is optimal in program-copy complexity. The same measurement statistics can be implemented with passive linear optics, and for the Hadamard interferometer the required postprocessing runs in time \(O(M\log M)\). In this branch of the literature, PPM denotes not postselection on failure, but programmable approximation of a target projector with controlled probabilistic error.

## 5. Heralding, conditional projection, and forced measurement

In nonlinear optics, PPM appears as a heralding operation that extracts a non-Gaussian conditional state from an entangled multimode resource. In the depleting-pump SPDC proposal for bright Schrödinger-cat generation, the signal mode \(\hat a\) and pump mode \(\hat b\) interact through
\[
\hat H_{\rm dim}=\hbar g\left(\hat a^2\hat b^\dagger+\hat a^{\dagger 2}\hat b\right),
\]
after which one performs a photon-number projective measurement on the pump output [2507.02201]. If the detector reports \(m\) photons, the pump is projected by
\[
\hat\Pi_m=|m\rangle_{\rm p}\langle m|,
\]
and the conditional signal state is obtained from \({}_{\rm p}\langle m|\psi_{2\text{mode}}\rangle\). The protocol is probabilistic because the outcome \(m\) is not predetermined, and only selected outcomes are retained. The main working point is \(m=0\), though any even number of pump photons can also herald cat-like states. For \(\beta=8\), the reported total probability of measuring an even number of pump photons is about \(87\%\). The retained signal state is approximately a squeezed even cat state with numerically found squeezing \(r=-\ln(\sqrt{2})\approx -0.35\).

A topological analogue appears in non-abelian anyon protocols. There, local projective measurement \(\operatorname{Proj}_a\) measures the topological charge of a single isolated quasiparticle, while interferometric measurement \(\operatorname{Int}_a\) measures the total charge enclosed by a loop [1501.01339]. The probabilistic element enters through “forced measurement”: after interferometric measurement introduces an \(\omega_0\)-type decoherence loop, one performs additional interferometric measurements until the vacuum channel \(0\) is obtained. The relevant \(F\)-symbol satisfies
\[
F^{\bar a,a}_{\bar a,a;0,0}=\frac{1}{d_a},
\]
so the vacuum recovery probability per trial is
\[
p=\left(\frac{1}{d_a}\right)^2.
\]
The failure probability after repeated trials decays exponentially with rate
\[
2\log_e\left(1-\frac{1}{d_a^2}\right).
\]
Here PPM denotes a repeated, postselected projection protocol rather than a single-shot measurement rule.

## 6. Sequential, reversible, and foundational variants

In ensemble NMR, a reversible version of projective measurement was demonstrated by combining reversible unitary dynamics, weak measurement of an intermediate state, and reverse unitary evolution [1004.4313]. The aim is to extract projective-measurement outcomes and probabilities non-destructively, with minimal net effect on the ensemble state. The sequence is schematically
\[
\rho_0 \xrightarrow{U} \rho_1 \xrightarrow{\text{weak measurement}} \rho_1+\delta\rho \xrightarrow{U^\dagger} \rho_0+\text{small error}.
\]
The paper explicitly connects this to a PPM picture in which only a small fraction of a large ensemble undergoes actual collapse, while the ensemble average is minimally disturbed.

A distinct recent usage treats PPM as a randomized projective-measurement strategy for sequential entanglement sharing. In that setting, each intermediate observer flips a classical coin: with probability \(G=\alpha\) a projective measurement is applied, and with probability \(1-G\) the system is left untouched [2510.02710]. The paper emphasizes that this is physically different from weak measurement, where the unsharpness is intrinsic and the disturbance factor is \(F=\sqrt{1-\eta^2}\); for PPM, the disturbance is \(F=1-G\). Within the studied two-qubit sequential scenario, weak measurement is more favorable than PPM for exhibiting entanglement sharing, and the Pearson correlation criterion is reported as the most robust across the unilateral and bilateral settings.

At a more foundational level, repeated-measurement statistics have been proposed as a way to test whether instantaneous projective collapse is physically real. A many-body protocol based on repeated measurements after a quench in the transverse-field Ising model compares standard projective-collapse predictions with a continuous, collapse-free alternative [2506.20618]. The two theories give qualitatively different repeated-measurement statistics: projective collapse predicts strong confirmation of earlier outcomes for sufficiently frequent measurements, whereas the continuous model yields much faster decay of repeated-confirmation probabilities. The proposal targets analog quantum simulators such as Rydberg atom arrays and ultracold gases in optical lattices.

Generalized probabilistic theories provide a broader sequential framework. There, the coherent Lüders rule defines the least disturbing update for a projective-like effect \(f\) through a positive \(f\)-compatible map \(\phi\) satisfying
\[
\phi(g)=g\qquad \text{for all } g\in V^+ \text{ with } g\le f,
\]
and in quantum mechanics this reduces uniquely to Lüders updating \(F^\sharp(X)=FXF\) for a projection \(F\) [1402.3583]. This does not define PPM as an acronym, but it supplies an abstract state-update principle for sequential projective measurement beyond Hilbert-space quantum theory.

## 7. Broader theoretical surroundings

Several adjacent results clarify the projective component that PPM protocols inherit. For the nonselective projective measurement map
\[
\Pi(\rho)=\sum_k P_k\rho P_k,
\]
projective measurement cannot decrease the quantum Tsallis entropy:
\[
S_q(\rho')\ge S_q(\rho),
\qquad
\rho'=\sum_k P_k\rho P_k,
\]
with equality if and only if \(\rho=\rho'\) [0904.3794]. The same reasoning extends to quantum unified \((r,s)\)-entropy. This places the projective branch of many PPM constructions inside an entropy-nondecreasing, decohering channel when outcomes are ignored.

In probe-based measurement statistics, sequential projective measurements can diagnose algebraic nonclassicality. For a probe \(P\) coupled by pure dephasing to a system \(S\), the paper on Kolmogorov consistency shows that breaking consistency of sequential probe-measurement probabilities witnesses noncommutativity of the accessible algebra of \(S\) [2111.14694]. This is not a named PPM formalism, but it shows how probabilistic projective-measurement records can act as nonclassicality witnesses.

In conformal field theory, projective measurement in a conformal basis has been compared with the entanglement-of-purification minimization problem. For suitable limits, the difference between the holographic EoP and the entanglement entropy after projective measurement is
\[
E_{AB}-S_A=\frac{c}{3}\log 2
\]
up to boundary terms [1901.00330]. The measured state is therefore not exactly the optimal purification, but may approximately realize it in the specific CFT regimes studied.

A black-hole analogue appears in monitored quantum circuits and Hayden–Preskill recovery. There, local projective measurements are represented by projectors \(\Pi(m)\) associated with a random measurement record \(m\), and the post-measurement state is
\[
|\Psi(m)\rangle=\frac{1}{\sqrt{\mathrm{Prob}(m)}}\,\Pi(m)\,|\mathrm{in}\rangle
\]
with
\[
\mathrm{Prob}(m)=\langle \mathrm{in}|\Pi(m)^\dagger\Pi(m)|\mathrm{in}\rangle.
\]
The resulting correspondence identifies entanglement verification in monitored circuits with black-hole information recovery in the presence of projective measurements [2203.04968]. This broader literature suggests that PPM-related ideas are increasingly tied not only to measurement implementation, but also to postselection, recovery, and entanglement structure in many-body and gravitational settings.

Across these formulations, PPM does not denote a single axiomatized object. What persists is a restricted-access projective action: a sharper measurement, a target projector, or a heralded collapse is available only probabilistically, and the scientific content lies in quantifying the resulting tradeoff between success rate, disturbance, simulability, or task performance.

Source: https://www.emergentmind.com/topics/probabilistic-projective-measurement-ppm