betaFeedback

WorksPhysics and Astronomy

Qubit-Qutrit Quantum Tomography of hadronic ΛϕΛϕ and ΛK∗0ΛK^{\ast 0} systems

Abstract

Quantum-information observables have emerged in recent years as new tools in nuclear and particle physics, from entanglement in top-quark pairs to spin correlations in ΛΛˉΛ\barΛ production. Extending these studies to unequal-spin hadronic final states poses a fundamental challenge: the 6×66\times6 density matrix of a qubit-qutrit system contains 35 independent spin parameters, but the decays of ΛVΛV pairs, with V=ϕV=ϕ or K∗0K^{*0}, provide access to only 23 due to the hidden vector polarization from the strong decay. In this Letter, we formulate a qubit-qutrit quantum tomography (QQQT) technique for these spin-12⊗1\tfrac{1}{2}\otimes1 systems and establish exact criteria for entanglement certification from the incomplete density matrix. Compared with the ΛΛˉΛ\barΛ system, QQQT of ΛϕΛϕ and ΛK∗0ΛK^{*0} provides a new probe of nonperturbative QCD hadronization, enabling a direct comparison of the spin evolution of entangled quark pairs produced from the vacuum as they hadronize into a baryon or a vector meson.

Introduction. How quarks and gluons form hadrons remains one of the least understood aspects of Quantum Chromodynamics (QCD). Known as hadronization, it has been modeled phenomenologically, through string or cluster fragmentation tuned to data, rather than derived directly from the first principles of confining dynamics. An especially open question concerns the role of spin: whether, and how, spin correlations generated among partons at short distances survive, evolve, or decohere as the system enters the nonperturbative regime and forms color-singlet hadrons.

Quantum-information observables, including entanglement, Bell nonlocality, and quantum-state tomography, have recently emerged as quantitative probes of such spin correlations [1, 2, 3, 4, 5, 6]. The ATLAS Collaboration has observed spin entanglement in top-quark pairs at the LHC [7], while complementary studies have reconstructed the top-quark spin density matrix and explored quantum-state tomography [8, 9, 10, 11, 12, 13]. Because top quarks decay before hadronization, these correlations are established at the partonic level. By contrast, the STAR Collaboration has reported an analogous spin-correlation measurement for ΛΛˉ\Lambda\bar{\Lambda} hyperon pairs produced in proton-proton collisions [14], extending such quantum observables to hadronic final states [15, 16, 17, 18, 19, 20, 21]. In this case, the correlations must persist through the nonperturbative formation of two baryons. Taken together, these measurements probe spin correlations on opposite sides of hadronization: before confinement can act on the spin state and after hadron formation is complete.

The systems studied so far, however, involve particles of equal spin: both ttˉt\bar{t} and ΛΛˉ\Lambda\bar{\Lambda} are spin-12⊗12\frac{1}{2}\otimes\frac{1}{2} systems. A natural next step is therefore a mixed-spin system, consisting of a spin-12\frac{1}{2} hyperon produced together with a spin-1 vector meson (VV), such as Λϕ\Lambda\phi or ΛK∗0\Lambda K^{*0}. There is also a physics motivation - such final states provide a qualitatively new probe of spin transport through hadronization. A potentially correlated quark pair extracted from the QCD vacuum with maximal entanglement [14, 22, 23] (or any other scenarios, e.g., jet fragmentation) can evolve through two structurally distinct hadron-formation channels, producing a baryon and a vector meson [22, 24, 25, 26, 27, 28, 29, 30]. Comparing these two channels offers a new way to investigate how quark-spin correlations are propagated into hadronic degrees of freedom during hadronization.

Challenge. The central experimental challenge is incomplete tomography. In principle, the relevant tools from Quantum Information Science (QIS) can be applied directly: for a qubit–qutrit state in a 2×32 \times3 Hilbert space, the positive-partial-transpose (PPT) criterion is both necessary and sufficient for separability. Experimentally, however, not all components of the density matrix are accessible. The weak decay of the Λ\Lambda is self-analyzing, whereas the strong decay of the vector meson is insensitive to part of its polarization. Consequently, only 23 of the 35 independent spin parameters of the full qubit–qutrit density matrix can be measured, leaving 12 experimentally inaccessible.

In this Letter, we formulate a qubit–qutrit quantum tomography (QQQT) technique for ΛV\Lambda V systems, construct the most general spin density matrix, and determine precisely which of its components are experimentally accessible. We then establish exact conditions under which entanglement can be inferred from the incomplete density matrix and quantify the limitations imposed by the unmeasured sector. These results provide both the theoretical foundation and the experimental framework for entanglement measures in hadronic qubit–qutrit systems, which establish Λϕ\Lambda\phi and ΛK∗0\Lambda K^{*0} production as new probes of spin evolution in nonperturbative QCD.

Formalism. We define the spin quantization axes in the ΛV\Lambda V pair rest frame using the orthonormal triad (k^,r^,n^)(\hat{k},\hat{r},\hat{n}) shown in Figure 1, with k^\hat{k} along the Λ\Lambda momentum, n^\hat{n} normal to the production plane, and r^=n^×k^\hat{r}=\hat{n}\times\hat{k}. Throughout, we identify (x^,y^,z^)≡(r^,n^,k^)(\hat{x},\hat{y},\hat{z})\equiv(\hat{r},\hat{n},\hat{k}). This is the standard helicity basis used in ttˉt\bar{t} analyses [4, 8, 10]. Since the two hadrons are back-to-back in the pair rest frame, a single triad suffices to describe both spin states.

Coordinate system for the $\Lambda V$ pair rest frame

Figure 1. Coordinate system, defined in the ΛV\Lambda V pair rest frame, where the two hadrons are back to back. The orthonormal triad is k^=p^Λ\hat{k}=\hat{p}_{\Lambda} (the Λ\Lambda momentum direction), n^∝p^beam×k^\hat{n}\propto\hat{p}_{\mathrm{beam}}\times\hat{k} normal to the production plane, and r^=n^×k^\hat{r}=\hat{n}\times\hat{k}, with k^×r^=n^\hat{k}\times\hat{r}=\hat{n}. The daughter directions q^1\hat{q}_{1} (proton from Λ→pπ−\Lambda\to p\pi^{-}) and q^2\hat{q}_{2} (K+K^{+} from ϕ→K+K−\phi\to K^{+}K^{-} or K∗0→K+π−K^{*0}\to K^{+}\pi^{-}) are each evaluated in the rest frame of their own parent but referred to this single shared triad.

We first consider the two particles separately. The density matrix of the spin-12\frac{1}{2} Λ\Lambda is

ρA=12(12+aiσi),\rho_A=\frac{1}{2}(\mathbb{1}_2+a_i\sigma_i),

where ai=⟨σi⟩a_i=\langle\sigma_i\rangle are the three components of the Λ\Lambda polarization. A spin-1 vector meson requires, in addition to its vector polarization, a rank-2 tensor describing its spin alignment. Introducing the spin-1 generators SiS_i in the helicity basis ∣+1⟩,∣0⟩,∣−1⟩\lvert+1\rangle,\lvert0\rangle,\lvert-1\rangle and the traceless quadrupole tensor

Tjk≡32(SjSk+SkSj)−2δjk13,T_{jk}\equiv\frac{3}{2}(S_jS_k+S_kS_j)-2\delta_{jk}\mathbb{1}_3,

the vector-meson density matrix can be written as,

ρB=1313+12PiSi+[ρ00Tzz+R10Txz+R1mTxx−yy+I10Tyz+I1mTxy],(1)\rho_B=\frac{1}{3}\mathbb{1}_3+\frac{1}{2}P_iS_i+\left[\rho_{00}T_{zz}+R_{10}T_{xz}+R_{1m}T_{xx-yy}+I_{10}T_{yz}+I_{1m}T_{xy}\right], \tag*{(1)}

where Pi=⟨Si⟩P_i=\langle S_i\rangle denotes the vector (dipole) polarization. The five parameters in brackets describe the rank-2 quadrupole alignment familiar from spin-alignment measurements of ϕ\phi, K∗0K^{*0}, and J/ψJ/\psi [31, 32, 34, 35] [33]. Up to normalization conventions, they correspond to the helicity-density-matrix elements ρ00\rho_{00}, Re⁡(ρ1,0)\operatorname{Re}(\rho_{1,0}), Re⁡(ρ1,−1)\operatorname{Re}(\rho_{1,-1}), Im⁡(ρ1,0)\operatorname{Im}(\rho_{1,0}), and Im⁡(ρ1,−1)\operatorname{Im}(\rho_{1,-1}) of the Schilling–Wolf parametrization. Together, the three dipole and five quadrupole components provide the eight independent parameters of a spin-1 density matrix.

We now combine the two spin states. In the operator basis 12,σi⊗13,Si,Tjk\mathbb{1}_2,\sigma_i\otimes\mathbb{1}_3,S_i,T_{jk}, the joint 2×3=62\times3=6 dimensional Hilbert space carries a Hermitian, unit-trace density matrix with 62−1=356^2-1=35 independent real parameters,

ρ=16[12⊗13+ai(σi⊗13)+12⊗(ρ00Tzz+R10Txz+R1mTxx−yy+I10Tyz+I1mTxy)+12⊗PiSi+Cij(σi⊗Sj)+Ki,α(σi⊗Tαeven)+Li,α(σi⊗Tαodd)],(2)\begin{aligned} \rho={}&\frac{1}{6}\big[\mathbb{1}_2\otimes\mathbb{1}_3+a_i(\sigma_i\otimes\mathbb{1}_3)+\mathbb{1}_2\otimes(\rho_{00}T_{zz}\\ &+R_{10}T_{xz}+R_{1m}T_{xx-yy}+I_{10}T_{yz}+I_{1m}T_{xy})\\ &+\mathbb{1}_2\otimes P_iS_i+C_{ij}(\sigma_i\otimes S_j)\\ &+K_{i,\alpha}(\sigma_i\otimes T_{\alpha}^{\mathrm{even}})+L_{i,\alpha}(\sigma_i\otimes T_{\alpha}^{\mathrm{odd}})\big], \tag*{(2)} \end{aligned}

with αeven∈{zz,xz,xx−yy}\alpha^{\mathrm{even}}\in\{zz,xz,xx-yy\} and αodd∈{yz,xy}\alpha^{\mathrm{odd}}\in\{yz,xy\}. The 35 parameters have a natural physical organization: three describe the Λ\Lambda polarization, eight describe the vector-meson polarization and alignment, and the remaining 24 encode correlations between the two spins. The latter consist of nine dipole–dipole coefficients CijC_{ij}, nine dipole–quadrupole coefficients Ki,αK_{i,\alpha}, and six dipole–quadrupole coefficients Li,αL_{i,\alpha}. The total count is therefore 3+3+3+2+9+6=353+3+3+2+9+6=35. Tracing over either subsystem recovers the corresponding single-particle density matrix above. See Appendix for details.

Two features of the 2×32\times3 Hilbert space are particularly important for entanglement. First, since min⁡(dA,dB)=2\min(d_A,d_B)=2, every pure state has Schmidt rank at most two, so its entanglement is bounded by one ebit. More importantly, the Horodecki theorem makes the positive-partial-transpose (PPT) criterion exact for 2×32\times3 systems: positivity of the partial transpose is not only necessary but also sufficient for separability [36, 37]. If the full density matrix were known, entanglement testing would therefore reduce to a direct eigenvalue test of ρBTB\rho_B^{T_B}.

Under partial transposition of the vector-meson subsystem (the operation denotes as TBT_B as a subscript of the density matrix), the operator basis separates naturally into TBT_B-even components Sx,Sz,Tzz,Txz,Txx−yyS_x,S_z,T_{zz},T_{xz},T_{xx-yy} and TBT_B-odd components Sy,Tyz,TxyS_y,T_{yz},T_{xy}. Because the Λ\Lambda-side operators are unaffected, TBT_B changes the sign of exactly 12 of the 35 parameters while leaving the remaining 23 unchanged. The question is therefore not simply whether the full ΛV\Lambda V state is entangled, but whether entanglement can still be established from the experimentally accessible sector of an intrinsically incomplete tomography. Note that mixed-dimension final states have been examined in Higgs decays [38, 39, 40], but there the state is pure and tomographically complete, so the incompleteness studied here does not arise.

Specifically, the parity-violating decay Λ→pπ−\Lambda\to p\pi^{-} is self-analyzing and provides access to all three components of the Λ\Lambda polarization, whereas the parity-conserving decay V→PPV\to PP is insensitive to the vector-meson dipole polarization PiP_i [41]. Consequently, the three PiP_i and nine dipole–dipole correlations CijC_{ij} are intrinsically inaccessible, leaving 23 measurable parameters (See Table I). Of these, I10I_{10}, I1mI_{1m}, and the six Li,αL_{i,\alpha} require the full two-dimensional (θ,ϕ)(\theta,\phi) decay distribution and are lost in the conventional one-dimensional Schilling–Wolf projection [31]. The justification usually offered is parity conservation in production, but that argument covers only four of the eight: I10I_{10}, I1mI_{1m}, Ln,yzL_{n,yz} and Ln,xyL_{n,xy} are parity-odd and must vanish, whereas Lr,αL_{r,\alpha} and Lk,αL_{k,\alpha} are parity-even and need not (Sec. E). Thus, complete reconstruction of the ΛV\Lambda V density matrix is impossible from these decays alone, motivating the incomplete-tomography problem addressed below.

Results. Let ρmeas\rho_{\mathrm{meas}} denote the density matrix obtained by setting the 23 measured parameters to their observed values and the inaccessible 12 to zero, which is the most complete reconstruction available from the decay angular distributions alone. The key observation is that this operation has a simple symmetry interpretation. Zeroing a parameter can be viewed as an averaging operation: if a linear map GG reverses its sign while leaving all other parameters unchanged, then 12[ρ+G(ρ)]\frac{1}{2}[\rho+G(\rho)] removes that component. Since the missing 12 parameters constitute exactly the meson dipole sector, the required transformation must reverse every SiS_i while preserving every TjkT_{jk}. This is precisely the action of time reversal on the meson spin.

SectorAccessible?
aia_i (Λ\Lambda polarization, 3)yes, standard
ρ00\rho_{00}, R10R_{10}, R1mR_{1m} (VV alignment, 3)yes, standard
Ki,αK_{i,\alpha} (dipole⊗V\otimes V-quad., 9)yes, joint fit
I10I_{10}, I1mI_{1m}, Li,αL_{i,\alpha} (VV align. + corr., 8)yes, 2D fit only
PiP_i, CijC_{ij} (VV dipole ±\pm corr., 12)no

Table 1. Accessibility of the 35 joint parameters. Only the bottom row, 12 of 35, is beyond reach on kinematic grounds.

The usual time-reversal operator Θ=UK\Theta=UK, where KK is the complex conjugation operator, is antiunitary and therefore cannot itself be applied to only one factor of a bipartite density matrix. For a Hermitian operator OO, however, its action can be written as ΘOΘ−1=UOTU†\Theta O\Theta^{-1}=UO^{T}U^{\dagger}: a transposition followed by a unitary rotation. Defining UB(ρ)=(12⊗U)ρ(12⊗U)†U_B(\rho)=(\mathbb{1}_2\otimes U)\rho(\mathbb{1}_2\otimes U)^{\dagger}, with U=−e−iπSyU=-e^{-i\pi S_y} corresponding to a π\pi rotation of the meson spin about y^\hat{y}, averaging over Z2=1,UB∘TB\mathbb{Z}_2=1,U_B\circ T_B gives

ρmeas=12[ρtrue+UB(ρtrueTB)].(3)\rho_{\mathrm{meas}}=\frac{1}{2}\left[\rho_{\mathrm{true}}+U_B\left(\rho_{\mathrm{true}}^{T_B}\right)\right]. \tag*{(3)}

the state averaged with its own partial time reversal. The composite transformation UB∘TBU_B\circ T_B flips exactly PiP_i and CijC_{ij} while leaving all 23 accessible parameters invariant. Like the partial transpose operation, this transformation is positive but not completely positive and hence does not represent a physical quantum channel—a property that will be essential for its sensitivity to entanglement.

Lemma 1 (PPT reduces to positivity). For the reconstruction built from any input state, ρmeasTB=UB(ρmeas)\rho_{\mathrm{meas}}^{T_B}=U_B(\rho_{\mathrm{meas}}) exactly. Since UBU_B is a unitary conjugation, ρmeasTB\rho_{\mathrm{meas}}^{T_B} and ρmeas\rho_{\mathrm{meas}} share a spectrum, so the PPT test on ρmeas\rho_{\mathrm{meas}} reduces to asking whether ρmeas⪰0\rho_{\mathrm{meas}}\succeq0: a single diagonalization, with no partial transpose needed in practice.

The lemma follows directly from the transformation properties of the surviving operators: each is either invariant under both TBT_B and UBU_B, or changes sign under both. Thus the PPT test, which would ordinarily require constructing and diagonalizing the partial transpose, reduces here to the positivity of ρmeas\rho_{\mathrm{meas}} itself. We now ask the two questions relevant for an incomplete reconstruction: can the missing information generate a false entanglement signal, and conversely, when is the available information sufficient to establish entanglement?

Theorem 1 (No false positives). If ρtrue\rho_{\mathrm{true}} is separable, then ρmeas\rho_{\mathrm{meas}} is separable, for any separable ρtrue\rho_{\mathrm{true}}, pure or

Figure 2. Numerical verification of Theorem 1. For each trial a random input state is generated, the reconstruction ρmeas\rho_{\mathrm{meas}} is built from its 23 accessible parameters, and the partial transpose is diagonalized. Left: λmin⁡(ρmeasTB)\lambda_{\min}(\rho_{\mathrm{meas}}^{T_B}) for 100 separable states — 65 of full rank 6 (filled circles) and 35 of reduced rank (open circles) — together with 100 entangled states (triangles). This numerical analysis is generated by Claude Opus 5 model.

mixed. A negative eigenvalue of ρmeas\rho_{\mathrm{meas}} can therefore never be produced by a genuinely separable state.

Proof. We first note that if MM is a legitimate single-qutrit density matrix, so is UMTU†UM^{T}U^{\dagger}: MTM^{T} is Hermitian, unit-trace, and shares the eigenvalues of MM (identical characteristic polynomial), hence is positive semidefinite, and conjugation by the unitary UU preserves Hermiticity, trace and spectrum. Now write a separable state as ρtrue=∑kpkρA(k)⊗ρB(k)\rho_{\mathrm{true}}=\sum_k p_k\rho_A^{(k)}\otimes\rho_B^{(k)} with pk≥0p_k\geq0, ∑kpk=1\sum_k p_k=1. Both TBT_B and UBU_B are linear and act only on the BB factor, so UB(ρtrueTB)=∑kpkρA(k)⊗U(ρB(k))TU†U_B(\rho_{\mathrm{true}}^{T_B})=\sum_k p_k\rho_A^{(k)}\otimes U(\rho_B^{(k)})^{T}U^{\dagger}, in which each meson factor is a legitimate qutrit state by the preceding remark. Substituting into Eq. (3),

ρmeas=12∑kpkρA(k)⊗[ρB(k)+U(ρB(k))TU†].(4)\rho_{\mathrm{meas}}=\frac{1}{2}\sum_k p_k\rho_A^{(k)}\otimes\left[\rho_B^{(k)}+U(\rho_B^{(k)})^{T}U^{\dagger}\right]. \tag*{(4)}

an explicit convex combination of legitimate product states with weights pk/2≥0p_k/2\geq0 summing to one. By definition this is separable, hence positive semidefinite. ▪

Theorem 1 establishes that the reconstruction cannot manufacture an entanglement signal from a separable state. The stronger question is whether this test is also optimal given the information that is experimentally available.

Theorem 2 (Exact certifiability). Let C(D)={ρ physical:ρ reproduces the measured data D}\mathcal{C}(D)=\{\rho\ \text{physical}:\rho\ \text{reproduces the measured data }D\}. Then ∃\exists separable ρ∈C(D)\rho\in\mathcal{C}(D) if and only if ρmeas\rho_{\mathrm{meas}} is separable, if and only if ρmeas⪰0\rho_{\mathrm{meas}}\succeq0. Entanglement is therefore certifiable from DD if and only if ρmeas\rho_{\mathrm{meas}} has a negative eigenvalue, and no analysis of the same data can do better.

Proof. By construction ρmeas\rho_{\mathrm{meas}} is an affine function of the measured data DD alone: it fixes every accessible paparameter at its observed value and sets the remaining twelve to zero, and therefore does not depend on the unmeasured parameters of whichever physical state produced DD. Every ρ∈C(D)\rho\in C(D) consequently yields the same measured reconstruction, and whenever ρmeas\rho_{\mathrm{meas}} is a valid state it belongs to C(D)C(D).

For the first equivalence, if ρmeas\rho_{\mathrm{meas}} is separable then it is in particular a legitimate state, and hence a separable member of C(D)C(D). Conversely, if some separable ρs∈C(D)\rho_s \in C(D) exists, then averaging ρs\rho_s as in Eq. (3) returns ρmeas\rho_{\mathrm{meas}}, which Theorem 1 guarantees to be separable. Existence of a separable state compatible with DD is thus equivalent to separability of ρmeas\rho_{\mathrm{meas}} itself.

For the second equivalence the relevant criterion is the Peres–Horodecki criterion, which in 2×32 \times3 is both necessary and sufficient: ρmeas\rho_{\mathrm{meas}} is separable if and only if ρmeasTB⪰0\rho_{\mathrm{meas}}^{T_B} \succeq0. By Lemma 1, ρmeasTB=UB(ρmeas)\rho_{\mathrm{meas}}^{T_B} = U_B(\rho_{\mathrm{meas}}) is a unitary conjugation of ρmeas\rho_{\mathrm{meas}} and shares its spectrum, so ρmeasTB⪰0\rho_{\mathrm{meas}}^{T_B} \succeq0 if and only if ρmeas⪰0\rho_{\mathrm{meas}} \succeq0. Hence ρmeas\rho_{\mathrm{meas}} is separable if and only if ρmeas⪰0\rho_{\mathrm{meas}} \succeq0.

Consequently, if ρmeas\rho_{\mathrm{meas}} has a negative eigenvalue then no separable state reproduces the data DD; equivalently, entanglement is certified exactly by the negativity of ρmeas\rho_{\mathrm{meas}}. Since the argument uses DD alone, no other analysis of the same data can certify more. ■

Note that the above two theorems have been independently proved by Lean Theorem Prover [42, 43].

The two results give a particularly simple interpretation of incomplete QQQT. The relevant question is not whether a particular completion of the density matrix is entangled, but whether every physical state consistent with the measured data must be entangled [44, 45]. Theorem 2 answers this question exactly. Operationally, one constructs ρmeas\rho_{\mathrm{meas}} from the 23 measured moments and diagonalizes it. A negative eigenvalue implies that every state compatible with the data is entangled; if ρmeas\rho_{\mathrm{meas}} is positive semidefinite, at least one compatible separable state exists. The semidefinite optimization required for general incomplete tomography therefore reduces in this system to the single diagonalization of Lemma 1. Figure 2 verifies Theorem 1 numerically using randomly generated states.

Theorem 2 establishes that no analysis of the same 23 observables can outperform the reconstruction. Nevertheless, incomplete tomography can still miss genuine entanglement, and this loss can be quantified.

Proposition 1 (Pure states retain half the negativity). For a Schmidt-rank-2 state ∣ψ⟩=λ1∣1⟩A∣1⟩B+λ2∣2⟩A∣2⟩B|\psi\rangle= \sqrt{\lambda_1}|1\rangle_A|1\rangle_B + \sqrt{\lambda_2}|2\rangle_A|2\rangle_B, in the physically relevant meson Schmidt bases, λmin⁡(ρmeasTB)=−12λ1λ2\lambda_{\min}(\rho_{\mathrm{meas}}^{T_B}) = -\frac{1}{2}\sqrt{\lambda_1\lambda_2}, exactly half the true value −λ1λ2-\sqrt{\lambda_1\lambda_2}.

Proposition 2 (Mixed-state blind windows). Genuine blind windows exist for mixed states. A maximally entangled state diluted with isotropic noise, ρ(p)=p∣e+⟩⟨e+∣+1−p6I6\rho(p) = p|e_+\rangle\langle e_+| + \frac{1-p}{6}\mathbb{I}_6, is entangled for p>1/4p > 1/4 but uncertifiable from the 23 accessible parameters until p>2/5p > 2/5. A mixture of two maximally entangled Bell-like states is entangled for every p∈[0,1]p \in[0,1] yet uncertifiable over a window of width 2−1≈0.41\sqrt{2}-1 \approx0.41 centered at p=12p = \frac{1}{2}.

Thus, pure-state entanglement remains visible with a reduced negativity, whereas mixed states can contain genuine blind regions. By Theorem 2, these are fundamental limitations of the accessible observables rather than shortcomings of the reconstruction: within each blind window, at least one separable state reproduces the same measured data. Verification of Propositions are included in the Appendix.

Remarks: We found the above results striking. Although this might be an coincidence, the 12 missing parameters due to the strong decay of the vector meson is somehow special. For example, discarding just the three dipoles PiP_i alone, or the nine CijC_{ij} alone, or the nine correlations Ki,αK_{i,\alpha}, or both single-particle polarizations {ai,Pi}\{a_i,P_i\}, each yields separable inputs reconstructed with a negative eigenvalue, which are false positives; then, our written theorems would be invalid. We pushed this further by randomly picking 12-element subsets to be zeroed, and many had failed the same way. This finding, only in this particular condition, enables the QQQT of this hadronic systems of Λϕ\Lambda\phi and AK∗0AK^{*0} with incomplete density matrix.

Physics implications. Let p(ΔR)p(\Delta R) denote the fraction of spin coherence a quark retains through hadronization, one factor per hadron formed. In open-quantum-system language, where hadronization acts on the initial ssˉs\bar{s} spin state as a quantum channel [24], pp is the strength of the local map on one side; for ΛΛˉ\Lambda\bar{\Lambda} two such maps act independently, so that Cijmes=p2CijpartonC^{\mathrm{mes}}_{ij} = p^2 C^{\mathrm{parton}}_{ij}. It runs with the pair separation because a wider separation means more string breaking in between. Measuring both systems then gives the following.

  • Constraints to quark-level correlation. Each hadron is dressed locally, and local dressing cannot create entanglement, so whatever spin correlation survives into the observed hadrons is a lower limit on what the quark pair carried at production, independently of how hadronization works. This can be used to distinguish initial-state verse final-state effects [46] that provide spin correlations.

  • Baryon–meson comparison. A ΛΛˉ\Lambda\bar{\Lambda} forms when the ss picks up a udud diquark, whereas a vector meson forms when the sˉ\bar{s} picks up a single partner; if coherence is lost by entangling with the surrounding strings, dragging along two extra quarks need not cost the same as dragging along one. The comparison isolates this, although not through the same observable in the two systems: for ΛΛˉ\Lambda\bar{\Lambda} the coherence is read directly off CijC_{ij}, whereas for A ⁣VA\!V that dipole–dipole sector is precisely what is lost, and the meson factor enters only through Ki,αK_{i,\alpha} and Li,αL_{i,\alpha}. Comparing how fast the two systems degrade, rather than their absolute normalizations, is what separates pmesonp_{\mathrm{meson}} from pbaryonp_{\mathrm{baryon}}, a new physics quantity characterizing hadronization that has never been measured.

  • Quantum-to-classical transitions. Another physics question we can ask is how far out in ΔR\Delta R each system remains quantum entangled. The ΛΛˉ\Lambda\bar{\Lambda} system exposes the full spin state, while ΛV\Lambda V necessarily reports less than it holds: the reconstruction can miss entanglement but, by Theorem 1, can never create it. Consequently, if ΛV\Lambda V is still certified at separations where ΛΛˉ\Lambda\bar{\Lambda} has gone classical, the conclusion follows immediately: spin coherence survives better inside a meson than inside a baryon.

Summary. We have developed a qubit–qutrit quantum tomography (QQQT) technique for spin-12⊗1\frac{1}{2}\otimes1 ΛV\Lambda V systems and shown that, despite the intrinsically incomplete reconstruction, the 23 experimentally accessible spin parameters are sufficient to provide an exact criterion for when entanglement is implied by the data. The resulting test reduces the general incomplete-tomography problem to a single diagonalization, while its loss of sensitivity can be quantified analytically for both pure and mixed states. Applied to Λϕ\Lambda\phi and ΛK∗0\Lambda K^{*0} production, this framework requires only angular-moment measurements in existing data sets and provides a new complement to ΛΛˉ\Lambda\bar{\Lambda} studies [14]: comparing baryon–meson and baryon–antibaryon spin correlations can directly probe how quantum coherence evolves through different hadronization channels [22, 24, 47]. More broadly, the same construction applies to any spin-12⊗1\frac{1}{2}\otimes1 system analyzed through a self-analyzing weak decay and a parity-conserving two-body decay.

Acknowledgments.

The authors thank the Department of Energy’s Summer Undergraduate Laboratory Internship (SULI) for supporting V.Kanachova’s research project at Brookhaven National Lab. We also acknowledge useful informal discussions with members of the local ePIC group at Brookhaven National Laboratory. The authors acknowledge the usage of Artificial Intelligence models, ChatGPT 5 and Claude Opus 5, to polish the language as well as perform numerical verifications. The work of F. Liu and Z. Tu was supported by the U.S. Department of Energy under Award No. DE-SC0012704 and by the Brookhaven National Laboratory Laboratory Directed Research and Development (LDRD) Project 26-029.

Appendix

The Appendix is organized as follows. Sections A–B fix the state and the frame. Sections C–E determine experimental accessibility: we construct the two decay analyzers, evaluate the master angular distribution W(θ1,ϕ1,θ2,ϕ2)W(\theta_1,\phi_1,\theta_2,\phi_2) in closed form, read the 23/12 accessibility split directly from it, give an estimator for each accessible parameter, and classify the parameters by parity. Section F provides a comparison with the ΛΛˉ\Lambda\bar{\Lambda} system, where all 15 parameters are measured. Section G constructs ρmes\rho_{\mathrm{mes}} and gives the proof of Lemma 1 and the verification of Propositions 1 and 2.

Conventions: frame, operators, and the joint state

All spin components are defined with respect to the pair-rest-frame triad {k^,r^,n^}\{\hat{k},\hat{r},\hat{n}\} of Figure 1: k^\hat{k} along the Λ\Lambda momentum, n^∝p^beam×k^\hat{n}\propto\hat{p}_{\mathrm{beam}}\times\hat{k}, and r^=n^×k^\hat{r}=\hat{n}\times\hat{k}, so that k^×r^=n^\hat{k}\times\hat{r}=\hat{n}. Throughout we identify

(x^,y^,z^)≡(r^,n^,k^),(5)(\hat{x},\hat{y},\hat{z})\equiv(\hat{r},\hat{n},\hat{k}), \tag*{(5)}

which makes (x^,y^,z^)(\hat{x},\hat{y},\hat{z}) right-handed and fixes the index convention throughout.

The Pauli matrices are σx=(0110)\sigma_x=\begin{pmatrix}0&1\\1&0\end{pmatrix}, σy=(0−ii0)\sigma_y=\begin{pmatrix}0&-i\\i&0\end{pmatrix}, σz=(100−1)\sigma_z=\begin{pmatrix}1&0\\0&-1\end{pmatrix}, and the spin-1 generators in the helicity basis {∣+1⟩,∣0⟩,∣−1⟩}\{\lvert+1\rangle,\lvert0\rangle,\lvert-1\rangle\} are

Sz=(10000000−1),Sx=12(010101010),S_z=\begin{pmatrix}1&0&0\\0&0&0\\0&0&-1\end{pmatrix},\qquad S_x=\frac{1}{\sqrt{2}}\begin{pmatrix}0&1&0\\1&0&1\\0&1&0\end{pmatrix},
Sy=12(0−i0i0−i0i0).S_y=\frac{1}{\sqrt{2}}\begin{pmatrix}0&-i&0\\i&0&-i\\0&i&0\end{pmatrix}.

The quadrupole tensor is Tjk≡32(SjSk+SkSj)−2δjk13T_{jk}\equiv\frac{3}{2}(S_jS_k+S_kS_j)-2\delta_{jk}\mathbb{1}_3, with five independent components Tzz,Txx−yy,Tyz,TxzT_{zz},T_{xx-yy},T_{yz},T_{xz}, TxyT_{xy}; e.g. Tzz=diag⁡(1,−2,1)T_{zz}=\operatorname{diag}(1,-2,1), and all satisfy Tr⁡Tjk=0\operatorname{Tr}T_{jk}=0. A fact used repeatedly below is that Sx,Sz,Tzz,Txz,Txx−yyS_x,S_z,T_{zz},T_{xz},T_{xx-yy} are real, whereas Sy,Tyz,TxyS_y,T_{yz},T_{xy} are purely imaginary; this underlies the sign-flip rule of Sec. G.

In the operator basis {12,σi}⊗{13,Si,Tjk}\{\mathbb{1}_2,\sigma_i\}\otimes\{\mathbb{1}_3,S_i,T_{jk}\} the joint state is Eq. (2) of the main text,

ρ=16[12⊗13+ai(σi⊗13)+12⊗(ρ00Tzz+R10Txz+R1mTxx−yy)+12⊗(PxSx+PySy+PzSz)+12⊗(I10Tyz+I1mTxy)+Cij(σi⊗Sj)+Ki,α(σi⊗Tαeven)+Li,α(σi⊗Tαodd)],(6)\begin{aligned} \rho={}&\frac{1}{6}\bigl[\mathbb{1}_2\otimes\mathbb{1}_3+a_i(\sigma_i\otimes\mathbb{1}_3)+\mathbb{1}_2\otimes(\rho_{00}T_{zz}+R_{10}T_{xz}\\ &\quad+R_{1m}T_{xx-yy})+\mathbb{1}_2\otimes(P_xS_x+P_yS_y+P_zS_z)\\ &\quad+\mathbb{1}_2\otimes(I_{10}T_{yz}+I_{1m}T_{xy})+C_{ij}(\sigma_i\otimes S_j)\\ &\quad+K_{i,\alpha}(\sigma_i\otimes T_\alpha^{\mathrm{even}})+L_{i,\alpha}(\sigma_i\otimes T_\alpha^{\mathrm{odd}})\bigr], \tag*{(6)} \end{aligned}

with αeven∈{zz,xz,xx−yy}\alpha^{\mathrm{even}}\in\{zz,xz,xx-yy\}, αodd∈{yz,xy}\alpha^{\mathrm{odd}}\in\{yz,xy\} and the count 3+3+3+2+2+9+9+6=353+3+3+2+2+9+9+6=35.

One convention matters when comparing with the spin-alignment literature: tracing out the meson reproduces ρΛ\rho_\Lambda exactly, but tracing out the Λ\Lambda returns the operator content of ρB\rho_B with rescaled coefficients, Pi(3)=23PiP_i^{(3)}=\frac{2}{3}P_i and quadrupole(3)=13×^{(3)}=\frac{1}{3}\timesquadrupole relative to the single-particle convention of Eq. (1). All values below use the joint convention of Eq. (6).

Basis states

Two bases recur below: the Bell-like states label the mixed-state families, while the Clebsch–Gordan states provide the physically motivated Schmidt bases used in Sec. G.

Product basis — 6 states ∣a,b⟩|a,b\rangle, a∈{↑,↓}a \in\{\uparrow,\downarrow\}, b∈{+1,0,−1}b \in\{+1,0,-1\}.

Generalized Bell basis — since min⁡(dA,dB)=2\min(d_A,d_B)=2, entanglement is capped at 1 ebit (Schmidt rank ≤2\le2); a complete orthonormal basis of maximally entangled states is

∣eb±⟩=12(∣↑,b⟩±∣↓,b+1⟩),b=+1,0,−1(mod3),(7)|e_b^\pm\rangle= \frac{1}{\sqrt{2}}\left(|\uparrow,b\rangle\pm|\downarrow,b+1\rangle\right),\quad b=+1,0,-1\pmod{3}, \tag*{(7)}

indexed e0,e1,e2e_0,e_1,e_2 for b=+1,0,−1b=+1,0,-1, so ∣e0+⟩=12(∣↑,+1⟩+∣↓,0⟩)|e_0^+\rangle=\frac{1}{\sqrt{2}}(|\uparrow,+1\rangle+|\downarrow,0\rangle) and ∣e1+⟩=12(∣↑,0⟩+∣↓,−1⟩)|e_1^+\rangle=\frac{1}{\sqrt{2}}(|\uparrow,0\rangle+|\downarrow,-1\rangle).

Clebsch–Gordan basis — from 12⊗1→J=32⊕12\frac{1}{2}\otimes1\to J=\frac{3}{2}\oplus\frac{1}{2}:

∣32,12⟩=23∣↑,0⟩+13∣↓,+1⟩,(8)\left|\frac{3}{2},\frac{1}{2}\right\rangle=\sqrt{\frac{2}{3}}|\uparrow,0\rangle+\sqrt{\frac{1}{3}}|\downarrow,+1\rangle, \tag*{(8)}
∣12,12⟩=13∣↑,0⟩−23∣↓,+1⟩,(9)\left|\frac{1}{2},\frac{1}{2}\right\rangle=\sqrt{\frac{1}{3}}|\uparrow,0\rangle-\sqrt{\frac{2}{3}}|\downarrow,+1\rangle, \tag*{(9)}

(and the M→−MM\to-M images), entangled with Schmidt coefficients (13,23)(\frac{1}{3},\frac{2}{3}), not maximal; the stretched states ∣32,±32⟩\left|\frac{3}{2},\pm\frac{3}{2}\right\rangle are product states.

Decay analyzers and the master angular distribution

All experimentally accessible information in Eq. (6) is contained in a single function of four angles. Let q^1\hat{q}_1 be the proton direction from Λ→pπ−\Lambda\to p\pi^-, evaluated in the Λ\Lambda rest frame, and q^2\hat{q}_2 the K+K^+ direction from V→PPV\to PP, evaluated in the VV rest frame, both resolved on the common axes of Eq. (5), with polar and azimuthal angles (θ1,ϕ1)(\theta_1,\phi_1) and (θ2,ϕ2)(\theta_2,\phi_2) as in Figure 1. Write

n^(θ1,ϕ1)=(sin⁡θ1cos⁡ϕ1,sin⁡θ1sin⁡ϕ1,cos⁡θ1)(10)\hat{n}(\theta_1,\phi_1)=(\sin\theta_1\cos\phi_1,\sin\theta_1\sin\phi_1,\cos\theta_1) \tag*{(10)}

for the proton unit vector.

The two decays have very different analyzers. The weak, parity-violating Λ→pπ−\Lambda\to p\pi^- has the fully informative analyzer

FΛ(θ1,ϕ1)=12+αΛn^(θ1,ϕ1)⋅σ,αΛ≈0.75,(11)F^\Lambda(\theta_1,\phi_1)=\mathbb{1}_2+\alpha_\Lambda\hat{n}(\theta_1,\phi_1)\cdot\boldsymbol{\sigma},\qquad\alpha_\Lambda\approx0.75, \tag*{(11)}

which contains the rank-0 and full rank-1 (dipole) content of the Λ\Lambda spin. The strong/electromagnetic, parity-conserving V→PPV\to PP proceeds through the single helicity amplitude V(0,0)V(0,0), so its analyzer GV(θ2,ϕ2)G^V(\theta_2,\phi_2) is built from rank-0 and rank-2 multipoles only [41, 48]: a two-body decay into two spinless particles couples to L=0,2L=0,2 and never to L=1L=1, irrespective of how the meson was produced. This structural asymmetry between FΛF^\Lambda and GVG^V determines the accessibility results below.

Evaluating W∝Tr⁡[ρ(FΛ⊗GV)]W\propto\operatorname{Tr}[\rho(F^\Lambda\otimes G^V)] for the full 35-parameter ρ\rho gives a factorized result in which the correlation sector uses the same five meson-side angular functions as the single-particle sector:

Accessibility: reading the 23/12 split off W

\begin{aligned} W(\theta_1,\phi_1,\theta_2,\phi_2)&\propto1\\ &\quad+3[\rho_{00}g_{00}+R_{10}g_{10}+R_{1m}g_{1m}\\ &\qquad+I_{10}h_{10}+I_{1m}h_{1m}]\\ &\quad+\alpha_\Lambda\hat{n}(\theta_1,\phi_1)\cdot a\\ &\quad+3\alpha_\Lambda\sum_{i=x,y,z}\hat{n}_i(\theta_1,\phi_1)\left[K_{i,zz}g_{00}\\ &\qquad+K_{i,xz}g_{10}+K_{i,xx-yy}g_{1m}\\ &\qquad+L_{i,yz}h_{10}+L_{i,xy}h_{1m}\right], \tag*{(12)} \end{aligned}

The four blocks are the isotropic term, meson alignment (5 terms, VV angles only), Λ\Lambda polarization (3 terms, Λ\Lambda angles only), and correlations (3×5=153\times5=15 genuine four-angle terms). Here every gα,hαg_\alpha,h_\alpha is evaluated at (θ2,ϕ2)(\theta_2,\phi_2). The five meson-side functions are

g00(θ2,ϕ2)=sin⁡2θ2−23,g_{00}(\theta_2,\phi_2)=\sin^2\theta_2-\frac{2}{3},
g10(θ2,ϕ2)=−12sin⁡2θ2cos⁡ϕ2,g_{10}(\theta_2,\phi_2)=-\frac{1}{2}\sin2\theta_2\cos\phi_2,
g1m(θ2,ϕ2)=−sin⁡2θ2cos⁡2ϕ2,(13)g_{1m}(\theta_2,\phi_2)=-\sin^2\theta_2\cos2\phi_2, \tag*{(13)}
h10(θ2,ϕ2)=12sin⁡2θ2sin⁡ϕ2,h_{10}(\theta_2,\phi_2)=\frac{1}{2}\sin2\theta_2\sin\phi_2,
h1m(θ2,ϕ2)=12sin⁡2θ2sin⁡2ϕ2.h_{1m}(\theta_2,\phi_2)=\frac{1}{2}\sin^2\theta_2\sin2\phi_2.

The relative block factors (1,3,αΛ,3αΛ)(1,3,\alpha_\Lambda,3\alpha_\Lambda) follow from the exact trace evaluation and have been verified numerically. Setting I10=I1m=0I_{10}=I_{1m}=0 and dropping the correlation block reduces Eq. (12) to the conventional three-parameter Schilling–Wolf distribution in ρ00,Re⁡ρ10,Re⁡ρ1,−1\rho_{00},\operatorname{Re}\rho_{10},\operatorname{Re}\rho_{1,-1} used in standard spin-alignment analyses [31, 32]; the hαh_\alpha terms are present in the decay but are not fitted there, as discussed in Sec. E.

Accessibility: reading the 23/12 split off W

Equation (12) determines accessibility by inspection, without further dynamical input. Counting the independent angular structures it contains,

3⏟n^i+5⏟gα,hα+15⏟n^i×(gα,hα)=23,(14)\underbrace{3}_{\hat{n}_i}+\underbrace{5}_{g_\alpha,h_\alpha}+\underbrace{15}_{\hat{n}_i\times(g_\alpha,h_\alpha)}=23, \tag*{(14)}

one for each accessible parameter: the three aia_i, the five meson multipoles ρ00,R10,R1m,I10,I1m\rho_{00},R_{10},R_{1m},I_{10},I_{1m}, and the fifteen correlations KiαK_{i\alpha} (nine) and LiαL_{i\alpha} (six).

The remaining twelve parameters in Eq. (6) — the meson dipole PiP_i and the nine dipole-dipole correlations CijC_{ij} — are identically absent from Eq. (12), for all (θ1,ϕ1,θ2,ϕ2)(\theta_1,\phi_1,\theta_2,\phi_2). The reason is structural: detecting PiP_i would require a rank-1 harmonic of (θ2,ϕ2)(\theta_2,\phi_2) in WW, and detecting CijC_{ij} would require the product of n^i\hat{n}_i with such a harmonic. No rank-1 function of the meson angles occurs anywhere in Eq. (12), because GVG^V contains no L=1L=1 multipole. The inaccessible sector is therefore exactly

3⏟Pi+9⏟Cij=12,(15)\underbrace{3}_{P_i}+\underbrace{9}_{C_{ij}}=12, \tag*{(15)}

reproducing Table I of the main text. Equivalently, one may add a multiple of the meson dipole operator to ρ\rho, varying PiP_i and CijC_{ij} while leaving every moment of Eq. (12) unchanged. The incompleteness is therefore intrinsic, not a fit artifact.

Method of moments: the estimator for each parameter

All 23 accessible parameters are angular moments of the single-particle or joint decay distributions of the Λϕ\Lambda\phi and ΛK∗0\Lambda K^{*0} systems and can therefore be extracted from existing data sets. The main experimental complication is that the pair acceptance in Ω≡(θ1,ϕ1,θ2,ϕ2)\Omega\equiv(\theta_1,\phi_1,\theta_2,\phi_2) does not factorize between the two decays. This can be corrected by event mixing: a Λ\Lambda and a VV from different events in the same event class are paired to reproduce the detector acceptance while removing the physical two-particle correlation,

W(Ω)∝Wsame(Ω)Wmix(Ω),(16)W(\Omega) \propto\frac{W_{\mathrm{same}}(\Omega)}{W_{\mathrm{mix}}(\Omega)}, \tag*{(16)}

after which the 15 angular correlations are extracted. Here event mixing serves only as an acceptance correction. Operationally, one reconstructs the Λ\Lambda and VV candidates, corrects the joint and single-particle angular distributions, extracts the 23 moments below, constructs ρmeas\rho_{\mathrm{meas}}, and diagonalizes it. Carrying this out differentially in the pair separation ΔR\Delta R turns the entanglement measurement into a probe of hadronization, allowing the evolution of the signal to be compared among Λϕ\Lambda\phi, ΛK∗0\Lambda K^{*0}, and control channels such as Λρ0\Lambda\rho^0, which shares no valence quark with the Λ\Lambda [49].

The four Λ\Lambda-side functions 1,n^x,n^y,n^z1,\hat n_x,\hat n_y,\hat n_z are mutually orthogonal over the Λ\Lambda solid angle, and the six meson-side functions 1,g00,g10,g1m,h10,h1m1,g_{00},g_{10},g_{1m},h_{10},h_{1m} are mutually orthogonal over the VV solid angle — all cross-integrals vanish by direct integration. Thus the 23 structures of Eq. (12) do not mix, and each parameter follows from a single angular moment, without fitting. Writing ⟨f⟩\langle f\rangle for the event average of f(θ1,ϕ1,θ2,ϕ2)f(\theta_1,\phi_1,\theta_2,\phi_2) over the acceptance-corrected sample of Eq. (16),

ai=3αΛ⟨n^i⟩,a_i=\frac{3}{\alpha_\Lambda}\langle\hat n_i\rangle,
ρ00=154⟨g00⟩,R10=5⟨g10⟩,R1m=54⟨g1m⟩,\rho_{00}=\frac{15}{4}\langle g_{00}\rangle,\quad R_{10}=5\langle g_{10}\rangle,\quad R_{1m}=\frac{5}{4}\langle g_{1m}\rangle,
I10=5⟨h10⟩,I1m=5⟨h1m⟩,I_{10}=5\langle h_{10}\rangle,\quad I_{1m}=5\langle h_{1m}\rangle,
Ki,zz=454αΛ⟨n^ig00⟩,Ki,xz=15αΛ⟨n^ig10⟩,K_{i,zz}=\frac{45}{4\alpha_\Lambda}\langle\hat n_i g_{00}\rangle,\quad K_{i,xz}=\frac{15}{\alpha_\Lambda}\langle\hat n_i g_{10}\rangle,
Ki,xx−yy=154αΛ⟨n^ig1m⟩,K_{i,xx-yy}=\frac{15}{4\alpha_\Lambda}\langle\hat n_i g_{1m}\rangle,
Li,yz=15αΛ⟨n^ih10⟩,Li,xy=15αΛ⟨n^ih1m⟩.(17)L_{i,yz}=\frac{15}{\alpha_\Lambda}\langle\hat n_i h_{10}\rangle,\quad L_{i,xy}=\frac{15}{\alpha_\Lambda}\langle\hat n_i h_{1m}\rangle. \tag*{(17)}

The prefactors combine the normalization integrals ⟨n^i2⟩=13\langle\hat n_i^2\rangle=\frac{1}{3}, ⟨g002⟩=445\langle g_{00}^2\rangle=\frac{4}{45}, ⟨g102⟩=⟨h102⟩=⟨h1m2⟩=115\langle g_{10}^2\rangle=\langle h_{10}^2\rangle=\langle h_{1m}^2\rangle=\frac{1}{15} and ⟨g1m2⟩=415\langle g_{1m}^2\rangle=\frac{4}{15} with the block factors of Eq. (12). Table S1 summarizes the correspondence parameter ↔\leftrightarrow angular function ↔\leftrightarrow estimator, written explicitly in (θ1,ϕ1,θ2,ϕ2)(\theta_1,\phi_1,\theta_2,\phi_2).

SectorParameterAngular function of (θ1,ϕ1,θ2,ϕ2)(\theta_1,\phi_1,\theta_2,\phi_2)Estimator
Λ\Lambda polarization — 3 parameters, Λ\Lambda angles onlyaxa_xn^x=sin⁡θ1cos⁡ϕ1\hat n_x=\sin\theta_1\cos\phi_13αΛ⟨n^x⟩\frac{3}{\alpha_\Lambda}\langle\hat n_x\rangle
aya_yn^y=sin⁡θ1sin⁡ϕ1\hat n_y=\sin\theta_1\sin\phi_13αΛ⟨n^y⟩\frac{3}{\alpha_\Lambda}\langle\hat n_y\rangle
aza_zn^z=cos⁡θ1\hat n_z=\cos\theta_13αΛ⟨n^z⟩\frac{3}{\alpha_\Lambda}\langle\hat n_z\rangle
meson alignment — 5 parameters, VV angles onlyρ00\rho_{00}g00=sin⁡2θ2−23g_{00}=\sin^2\theta_2-\frac{2}{3}154⟨g00⟩\frac{15}{4}\langle g_{00}\rangle
R10R_{10}g10=−12sin⁡2θ2cos⁡ϕ2g_{10}=-\frac{1}{2}\sin2\theta_2\cos\phi_252⟨g10⟩\frac{5}{2}\langle g_{10}\rangle
R1mR_{1m}g1m=−sin⁡2θ2cos⁡2ϕ2g_{1m}=-\sin^2\theta_2\cos2\phi_254⟨g1m⟩\frac{5}{4}\langle g_{1m}\rangle
I10I_{10}h10=12sin⁡2θ2sin⁡ϕ2h_{10}=\frac{1}{2}\sin2\theta_2\sin\phi_25⟨h10⟩5\langle h_{10}\rangle
I1mI_{1m}h1m=12sin⁡2θ2sin⁡2ϕ2h_{1m}=\frac{1}{2}\sin^2\theta_2\sin2\phi_25⟨h1m⟩5\langle h_{1m}\rangle
correlations — 15 parameters, genuine four-angle moments (i=x,y,z)(i=x,y,z)Ki,zzK_{i,zz}n^i(θ1,ϕ1)(sin⁡2θ2−23)\hat n_i(\theta_1,\phi_1)\left(\sin^2\theta_2-\frac{2}{3}\right)454αΛ⟨n^ig00⟩\frac{45}{4\alpha_\Lambda}\langle\hat n_i g_{00}\rangle
Ki,xzK_{i,xz}−12n^i(θ1,ϕ1)sin⁡2θ2cos⁡ϕ2-\frac{1}{2}\hat n_i(\theta_1,\phi_1)\sin2\theta_2\cos\phi_215αΛ⟨n^ig10⟩\frac{15}{\alpha_\Lambda}\langle\hat n_i g_{10}\rangle
Ki,xx−yyK_{i,xx-yy}−n^i(θ1,ϕ1)sin⁡2θ2cos⁡2ϕ2-\hat n_i(\theta_1,\phi_1)\sin^2\theta_2\cos2\phi_215αΛ⟨n^ig1m⟩\frac{15}{\alpha_\Lambda}\langle\hat n_i g_{1m}\rangle
Li,yzL_{i,yz}12n^i(θ1,ϕ1)sin⁡2θ2sin⁡ϕ2\frac{1}{2}\hat n_i(\theta_1,\phi_1)\sin2\theta_2\sin\phi_215αΛ⟨n^ih10⟩\frac{15}{\alpha_\Lambda}\langle\hat n_i h_{10}\rangle
Li,xyL_{i,xy}12n^i(θ1,ϕ1)sin⁡2θ2sin⁡2ϕ2\frac{1}{2}\hat n_i(\theta_1,\phi_1)\sin^2\theta_2\sin2\phi_215αΛ⟨n^ih1m⟩\frac{15}{\alpha_\Lambda}\langle\hat n_i h_{1m}\rangle
inaccessible — 12 parameters: no rank-1 harmonic of (θ2,ϕ2)(\theta_2,\phi_2) exists in WWPiP_iwould require Y1m(θ2,ϕ2)Y_{1m}(\theta_2,\phi_2)absent from WW
CijC_{ij}would require n^i(θ1,ϕ1)Y1m(θ2,ϕ2)\hat n_i(\theta_1,\phi_1)Y_{1m}(\theta_2,\phi_2)absent from WW

Table S1. Correspondence between the parameters of Eq. (6), the orthogonal angular structures of Eq. (12), and their moment estimators. The 23 accessible parameters split as 3+5+15; the 12 inaccessible ones are exactly those that would need a rank-1 (dipole) harmonic of the meson decay angles, which V→PPV\to PP does not provide. The eight parameters I10I_{10}, I1mI_{1m}, Li,yzL_{i,yz}, Li,xyL_{i,xy} are carried entirely by the sin⁡ϕ2\sin\phi_2 and sin⁡2ϕ2\sin2\phi_2 structures omitted by the conventional Schilling–Wolf fit.

Parity classification. Parity conservation in production predicts that roughly half of the 23 accessible parameters vanish, providing internal null tests. In the pair rest frame momenta reverse under parity while n^∝p^Λ×k^\hat n\propto\hat p_{\Lambda}\times\hat k, a cross product of two polar vectors, is invariant; spin is axial, so a spin component picks up the sign of its axis, sr=−1s_r=-1, sn=+1s_n=+1, sk=−1s_k=-1. A term survives iff the product of ss over its indices is +1+1, equivalently, if it carries an odd number of n^\hat n labels. Applying this rule, 11 of the 23 are parity-allowed, namely ana_n; ρ00,R10,R1m\rho_{00},R_{10},R_{1m}; the three Kn,αK_{n,\alpha}; and the four Lr,α,Lk,αL_{r,\alpha},L_{k,\alpha}; while the remaining 12, ar,ak;I10,I1ma_r,a_k;I_{10},I_{1m}; the six Kr,α,Kk,αK_{r,\alpha},K_{k,\alpha}; and Ln,yz,Ln,xyL_{n,yz},L_{n,xy}, vanish if production conserves parity. The first group includes the familiar results that global hyperon polarization lies along the production-plane normal and that only ρ00,Re⁡ρ1,0,Re⁡ρ1,−1\rho_{00},\operatorname{Re}\rho_{1,0},\operatorname{Re}\rho_{1,-1} are conventionally fitted.

The constraint applies to production, not to the decays: the parity-violating Λ→pπ−\Lambda\to p\pi^- is what makes the hyperon self-analyzing and is already carried by FΛF^\Lambda in Eq. (11). Since production is strong and electromagnetic, the 12 parity-odd parameters are expected to vanish and therefore provide a cross-check of the acceptance correction in Eq. (16): detector holes do not respect the mirror symmetry, so an imperfectly corrected acceptance leaks into precisely these terms while leaving the 11 parity-allowed ones comparatively undisturbed. A significant nonzero value therefore points first to the acceptance or the orientation convention for n^\hat n, and only then to background or genuinely parity-violating production. This does not affect the entanglement analysis, which uses all 23 measured values.

Complete tomography of the system

For comparison with the ΛV\Lambda V case, the same programme is complete for the ΛΛˉ\Lambda\bar\Lambda system. Both members of the pair are spin-12\frac{1}{2}, so the joint Hilbert space is 2×22\times2 and, in the operator basis {12,σi}⊗2\{1_2,\sigma_i\}^{\otimes2}, a Hermitian unit-trace density matrix carries 16−1=1516-1=15 independent real parameters,

ρΛΛˉ=14[12⊗12+Bi+(σi⊗12)+Bj−(12⊗σj)+Cij(σi⊗σj)],(18)\rho_{\Lambda\bar\Lambda}=\frac{1}{4}\left[1_2\otimes1_2+B_i^+(\sigma_i\otimes1_2)+B_j^-(1_2\otimes\sigma_j)+C_{ij}(\sigma_i\otimes\sigma_j)\right], \tag*{(18)}

counted as 3⏟Bi++3⏟Bj−+9⏟Cij=15\underbrace{3}_{B_i^+}+\underbrace{3}_{B_j^-}+\underbrace{9}_{C_{ij}}=15, with B±B^\pm the single-hyperon polarization vectors and Cij=⟨σi⊗σj⟩C_{ij}=\langle\sigma_i\otimes\sigma_j\rangle the spin-spin correlation matrix. All components are referred to the same triad (x^,y^,z^)≡(r^,n^,k^)(\hat x,\hat y,\hat z)\equiv(\hat r,\hat n,\hat k) of Sec. A, with k^\hat k now along the Λ\Lambda momentum in the ΛΛˉ\Lambda\bar\Lambda rest frame. The key difference from Sec. C is that both analyzers are fully informative. Each hyperon decays weakly, Λ→pπ−\Lambda\to p\pi^- and Λˉ→pˉπ+\bar\Lambda\to\bar p\pi^+, so each carries a nonvanishing rank-1 analyzing power along all three axes,

FΛ=1212+αΛ q^1⋅σ,FΛˉ=1212+αΛˉ q^2⋅σ,(19)F^\Lambda=\frac{1}{2}\mathbb{1}_2+\alpha_\Lambda\,\hat q_1\cdot\boldsymbol{\sigma},\qquad F^{\bar\Lambda}=\frac{1}{2}\mathbb{1}_2+\alpha_{\bar\Lambda}\,\hat q_2\cdot\boldsymbol{\sigma}, \tag*{(19)}

with αΛˉ≃−αΛ\alpha_{\bar\Lambda}\simeq-\alpha_\Lambda if CPCP is conserved. Evaluating W∝Tr⁡[ρΛΛˉ(FΛ⊗FΛˉ)]W\propto\operatorname{Tr}[\rho_{\Lambda\bar\Lambda}(F^\Lambda\otimes F^{\bar\Lambda})] gives the joint decay distribution

W(q^1,q^2)∝1+αΛq^1⋅B++αΛˉq^2⋅B−+αΛαΛˉCijq^1iq^2j,(20)W(\hat q_1,\hat q_2)\propto1+\alpha_\Lambda\hat q_1\cdot\mathbf B^++\alpha_{\bar\Lambda}\hat q_2\cdot\mathbf B^-+\alpha_\Lambda\alpha_{\bar\Lambda}C_{ij}\hat q_{1i}\hat q_{2j}, \tag*{(20)}

where q^1\hat q_1 is the proton direction evaluated in the Λ\Lambda rest frame and q^2\hat q_2 the antiproton direction in the Λˉ\bar\Lambda rest frame, both resolved on the common axes and parametrized by (θ1,ϕ1)(\theta_1,\phi_1) and (θ2,ϕ2)(\theta_2,\phi_2) exactly as in Figure 1.

Every one of the 15 parameters appears in Eq. (20), and the angular structures {q^1i}\{\hat q_{1i}\}, {q^2j}\{\hat q_{2j}\} and {q^1iq^2j}\{\hat q_{1i}\hat q_{2j}\} are mutually orthogonal over the two solid angles. Each parameter is therefore isolated by a single moment, without fitting:

Bi+=3αΛ⟨q^1i⟩,Bj−=3αΛˉ⟨q^2j⟩,Cij=9αΛαΛˉ⟨q^1iq^2j⟩,(21)B_i^+=\frac{3}{\alpha_\Lambda}\langle\hat q_{1i}\rangle,\qquad B_j^-=\frac{3}{\alpha_{\bar\Lambda}}\langle\hat q_{2j}\rangle,\qquad C_{ij}=\frac{9}{\alpha_\Lambda\alpha_{\bar\Lambda}}\langle\hat q_{1i}\hat q_{2j}\rangle, \tag*{(21)}

the prefactors following from ⟨q^1i2⟩=⟨q^2j2⟩=13\langle\hat q_{1i}^2\rangle=\langle\hat q_{2j}^2\rangle=\frac{1}{3}. Written out in the four angles, q^1=(sin⁡θ1cos⁡ϕ1,sin⁡θ1sin⁡ϕ1,cos⁡θ1)\hat q_1=(\sin\theta_1\cos\phi_1,\sin\theta_1\sin\phi_1,\cos\theta_1) and likewise for q^2\hat q_2, so that for example Crn=9⟨sin⁡θ1cos⁡ϕ1sin⁡θ2sin⁡ϕ2⟩/(αΛαΛˉ)C_{rn}=9\langle\sin\theta_1\cos\phi_1\sin\theta_2\sin\phi_2\rangle/(\alpha_\Lambda\alpha_{\bar\Lambda}); Table S2 summarizes the full correspondence. We verified Eq. (21)

SectorParameterAngular functionEstimator
Λ\Lambda polarizationBi+B_i^+q^1i(θ1,ϕ1)\hat q_{1i}(\theta_1,\phi_1)3αΛ⟨q^1i⟩\frac{3}{\alpha_\Lambda}\langle\hat q_{1i}\rangle
Λˉ\bar\Lambda polarizationBj−B_j^-q^2j(θ2,ϕ2)\hat q_{2j}(\theta_2,\phi_2)3αΛˉ⟨q^2j⟩\frac{3}{\alpha_{\bar\Lambda}}\langle\hat q_{2j}\rangle
spin–spin correlationCijC_{ij}q^1iq^2j\hat q_{1i}\hat q_{2j}9αΛαΛˉ⟨q^1iq^2j⟩\frac{9}{\alpha_\Lambda\alpha_{\bar\Lambda}}\langle\hat q_{1i}\hat q_{2j}\rangle
total 3 + 3 + 9 = 15, all accessible (i,j=r,n,k)(i,j=r,n,k)

Table S2. Correspondence between the 15 parameters of the ΛΛˉ\Lambda\bar\Lambda density matrix Eq. (18), the orthogonal angular structures of Eq. (20), and their moment estimators. Unlike the ΛV\Lambda V case of Table S1, no sector is kinematically inaccessible: both decays are weak and self-analyzing, so each supplies a full rank-1 analyzing power.

with a Monte Carlo closure test: events generated from Eq. (20) for a random physical ρΛΛˉ\rho_{\Lambda\bar\Lambda} recover all 15 input parameters.

The contrast with Sec. D is immediate. For ΛΛˉ\Lambda\bar\Lambda the tomography is complete: 15 of 15 parameters are measured, ρΛΛˉ\rho_{\Lambda\bar\Lambda} is reconstructed in full, and the Peres–Horodecki criterion — exact in 2×22\times2 as it is in 2×32\times3 — can be applied directly to the measured matrix. No analogue of the reconstruction in Sec. G is required, and no blind windows arise. Thus the incompleteness arises specifically when one weakly decaying hyperon is replaced by a strongly decaying vector meson: it is the loss of the meson rank-1 analyzing power, not the increase in Hilbert-space dimension, that removes 12 of the 35 parameters.

Reconstruction of meas and proofs

The reconstruction as a partial time reversal. Having identified the measured sector, we turn to the entanglement test. Throughout we use the partial transpose on the meson, ρ(a,b),(a′,b′)TB=ρ(a,b′),(a′,b)\rho^{T_B}_{(a,b),(a',b')}=\rho_{(a,b'),(a',b)}, for which the Horodecki theorem makes PPT necessary and sufficient for separability in 2×32 \times3 [36, 37, 50], and the labels TT-even for {Sx,Sz,Tzz,Txz,Txx−yy}\{S_x,S_z,T_{zz},T_{xz},T_{xx-yy}\} and TT-odd for {Sy,Tyz,Txy}\{S_y,T_{yz},T_{xy}\}, following the reality properties of Sec. A.

Setting a parameter to zero can be viewed as an averaging operation: if a linear map G\mathcal{G} flips its sign while preserving the others, then 12[ρ+G(ρ)]\frac{1}{2}[\rho+\mathcal{G}(\rho)] sets it to zero. By Sec. D, the required map flips every SiS_i while preserving every TjkT_{jk} — precisely time reversal on the meson factor. The textbook operator is Θ=UK\Theta=UK, with KK complex conjugation in the helicity basis (Θ∣m⟩=(−1)m∣−m⟩\Theta\lvert m\rangle=(-1)^m\lvert-m\rangle) and

U=(00−1010−100)U=\begin{pmatrix} 0 & 0 & -1 \\ 0 & 1 & 0 \\ -1 & 0 & 0 \end{pmatrix}

(basis order ∣+1⟩,∣0⟩,∣−1⟩\lvert+1\rangle,\lvert0\rangle,\lvert-1\rangle). (S18)

It acts as required, ΘSiΘ−1=−Si\Theta S_i\Theta^{-1}=-S_i and ΘTjkΘ−1=Tjk\Theta T_{jk}\Theta^{-1}=T_{jk}. Because it is antiunitary, it admits no one-sided action on a tensor product; however, for Hermitian OO one has OT=O∗O^T=O^* and hence ΘOΘ−1=UOTU†\Theta O\Theta^{-1}=UO^TU^\dagger: time reversal is transposition followed by a unitary rotation. Partial transposition supplies the first step; the second is

UB(ρ)≡(12⊗U)ρ(12⊗U)†,(22)\mathcal{U}_B(\rho)\equiv(\mathbb{1}_2\otimes U)\rho(\mathbb{1}_2\otimes U)^\dagger, \tag*{(22)}

a rotation by π\pi about y^\hat{y} of the meson spin (U=−e−iπSyU=-e^{-i\pi S_y}). Its action, together with that of transposition, on the operator basis is

O↦OT:Sx  Sy  Sz  Tzz  Txz  Txx−yy  Tyz  Txy+  −  +  +  +  +  −  −O↦UOU†:−  +  −  +  +  +  −  −O↦UOTU†:−  −  −  +  +  +  +  +(23)\begin{aligned} O\mapsto O^T:\quad&S_x\ \ S_y\ \ S_z\ \ T_{zz}\ \ T_{xz}\ \ T_{xx-yy}\ \ T_{yz}\ \ T_{xy} \\ &+\ \ -\ \ +\ \ +\ \ +\ \ +\ \ -\ \ - \\ O\mapsto UOU^\dagger:\quad&-\ \ +\ \ -\ \ +\ \ +\ \ +\ \ -\ \ - \\ O\mapsto UO^TU^\dagger:\quad&-\ \ -\ \ -\ \ +\ \ +\ \ +\ \ +\ \ + \tag*{(23)} \end{aligned}

The last row gives the time-reversal pattern Si→−SiS_i\to-S_i, Tjk→TjkT_{jk}\to T_{jk}: on the BB factor of any product state UB∘TB\mathcal{U}_B\circ T_B reproduces ΘρBΘ−1\Theta\rho_B\Theta^{-1} exactly, which is why we call it partial time reversal, the standard reading of partial transposition [51]. It is linear but, like TBT_B itself, positive rather than completely positive, so it is not a physical channel — precisely what enables entanglement detection. Since the Λ\Lambda-side operators are untouched, the composition flips exactly PiP_i and CijC_{ij} — the inaccessible sector of Sec. D — and fixes all 23 accessible parameters. Averaging over Z2={1,UB∘TB}\mathbb{Z}_2=\{1,\mathcal{U}_B\circ T_B\} therefore yields Eq. (3) of the main text,

ρmeas=12[ρtrue+UB(ρtrueTB)],\rho_{\mathrm{meas}}=\frac{1}{2}\left[\rho_{\mathrm{true}}+\mathcal{U}_B\left(\rho_{\mathrm{true}}^{T_B}\right)\right],

as follows directly by applying the flip pattern of Eq. (23) to all 35 parameters.

Proof of Lemma 1. Every parameter surviving in ρmeas\rho_{\mathrm{meas}} is either invariant under both TBT_B and UB\mathcal{U}_B (the 15 routinely measured ones) or flipped by both (I10I_{10}, I1mI_{1m}, Li,αL_{i,\alpha}), as seen from Eq. (23). In either case TBT_B and UB\mathcal{U}_B act identically on every operator present in ρmeas\rho_{\mathrm{meas}}, so ρmeasTB=UB(ρmeas)\rho_{\mathrm{meas}}^{T_B}=\mathcal{U}_B(\rho_{\mathrm{meas}}). Since UB\mathcal{U}_B is a unitary conjugation, they share a spectrum.

Theorems 1 and 2 are proved in the main text. One consequence for the two sections below is worth stating: when an entangled state escapes detection, this is not a reconstruction artifact: some separable state reproduces every measured number, so no method can certify entanglement from those data.

Verification of Proposition 1 (pure entangled states). Any pure state of a 2×32\times3 system has Schmidt rank ≤min⁡(2,3)=2\le\min(2,3)=2 and is entangled iff the rank is exactly 2: ∣ψ⟩=λ1∣1⟩A∣1⟩B+λ2∣2⟩A∣2⟩B\lvert\psi\rangle=\sqrt{\lambda_1}\lvert1\rangle_A\lvert1\rangle_B+\sqrt{\lambda_2}\lvert2\rangle_A\lvert2\rangle_B. For ∣1⟩B=∣+1⟩\lvert1\rangle_B=\lvert+1\rangle, ∣2⟩B=∣0⟩\lvert2\rangle_B=\lvert0\rangle the Clebsch–Gordan-motivated basis of Sec. B — direct computation gives λmin⁡(ρmeasTB)=−12λ1λ2\lambda_{\min}(\rho_{\mathrm{meas}}^{T_B})=-\frac{1}{2}\sqrt{\lambda_1\lambda_2} for all λ1,λ2>0\lambda_1,\lambda_2>0, exactly half of λmin⁡(ρtrueTB)=−λ1λ2\lambda_{\min}(\rho_{\mathrm{true}}^{T_B})=-\sqrt{\lambda_1\lambda_2}. The criterion is simple: since UU is real and symmetric with U2=1U^2=1, one has ⟨ψ∣U∣ψ⟩=⟨ψ∣U∣ψ⟩∗\langle\psi|U|\psi\rangle=\langle\psi|U|\psi\rangle^*, and the factor is exactly 12\frac{1}{2} whenever α≡⟨2∣BU∣1⟩B\alpha\equiv\langle2|_B U|1\rangle_B vanishes, i.e. whenever the two Schmidt vectors are not connected by the π\pi rotation UU. This includes the Clebsch–Gordan basis and the full rotated real family ∣1⟩B=cos⁡s∣+1⟩+sin⁡s∣−1⟩\lvert1\rangle_B=\cos s\lvert+1\rangle+\sin s\lvert-1\rangle, ∣2⟩B=∣0⟩\lvert2\rangle_B=\lvert0\rangle, independently of ss.

Remark: numerical evidence on pure-state detection. The factor 12\frac{1}{2} is not universal: over randomly generated Schmidt bases the ratio λmin⁡(ρmeasTB)/(−λ1λ2)\lambda_{\min}(\rho_{\mathrm{meas}}^{T_B})/(-\sqrt{\lambda_1\lambda_2}) varies, dipping to ≈0.11\approx0.11 near the separable boundary. λmin⁡\lambda_{\min} was nevertheless strictly negative in all 3×1043\times10^4 random pure entangled states tested (random complex Schmidt vectors on both sides, λ1\lambda_1 uniform on (0.001,0.999)(0.001,0.999)), the closest approach to zero being −4.7×10−3-4.7\times10^{-3} at λ1=0.999\lambda_1=0.999. A global numerical maximization of λmin⁡(ρmeas)\lambda_{\min}(\rho_{\mathrm{meas}}) over pure entangled states, with the Schmidt weight and both Schmidt bases free, likewise found no non-negative value. We therefore conjecture, without proof, that ρmeas\rho_{\mathrm{meas}} detects every pure entangled state of this system. The natural single witness — the negative eigenvector of ρtrueTB\rho_{\mathrm{true}}^{T_B} mapped through UB\mathcal{U}_B — does not suffice: its expectation value depends on α\alpha, and it ceases to certify for Schmidt bases with ∣α∣\lvert\alpha\rvert near unity, where detection is instead supplied by a different eigenvector. By Theorem 2, the conjecture is equivalent to stating that no pure entangled state of this system shares its 23 accessible parameters with any separable state. Note that the ratio above tends to zero as λ1→0\lambda_1\to0, as it must: the state becomes separable in that limit.

Verification of Proposition 2 (mixed-state blind windows). Mixed states exhibit genuine blind windows. We consider two families: dilution by noise and, more strikingly, a mixture of two entangled states.

Dilution by noise. Dilute the maximally entangled state ∣e0+⟩\lvert e_0^+\rangle of Eq. (7) with isotropic noise, ρtrue(p)=p∣e0+⟩⟨e0+∣+16(1−p)16\rho_{\mathrm{true}}(p)=p\lvert e_0^+\rangle\langle e_0^+\rvert+\frac{1}{6}(1-p)\mathbb{1}_6, p∈[0,1]p\in[0,1]. Direct symbolic evaluation gives

spec⁡(ρtrue(p)TB)={16+p3 (×3), 16−2p3, 16−p6 (×2)}\operatorname{spec}\left(\rho_{\mathrm{true}}(p)^{T_B}\right)=\left\{\frac{1}{6}+\frac{p}{3}\ (\times3),\ \frac{1}{6}-\frac{2p}{3},\ \frac{1}{6}-\frac{p}{6}\ (\times2)\right\}

so ρtrue(p)\rho_{\mathrm{true}}(p) is entangled iff p>1/4p>1/4 (only 16−2p3\frac{1}{6}-\frac{2p}{3} can go negative on [0,1][0,1]). This family populates only the four inaccessible parameters Pz,Cxx,Cyy,CzzP_z,C_{xx},C_{yy},C_{zz}, and

spec⁡(ρmeas(p)TB)={16+p12 (×2), 16−5p12, 16−p6, 16+5p24±5p8}.\operatorname{spec}\left(\rho_{\mathrm{meas}}(p)^{T_B}\right)=\left\{\frac{1}{6}+\frac{p}{12}\ (\times2),\ \frac{1}{6}-\frac{5p}{12},\ \frac{1}{6}-\frac{p}{6},\ \frac{1}{6}+\frac{5p}{24}\pm\frac{\sqrt{5}p}{8}\right\}.

whose smallest member 16−5p12\frac{1}{6}-\frac{5p}{12} turns negative only for p>2/5p>2/5 (the remaining five entries are non-negative on [0,1][0,1] by inspection, including the lower ±\pm branch, which equals 16−35−524p>9−3524>0\frac{1}{6}-\frac{3\sqrt{5}-5}{24}p>\frac{9-3\sqrt{5}}{24}>0 throughout). Thus for p∈(14,25)p\in(\frac{1}{4},\frac{2}{5}) the true state is entangled while ρmeas(p)⪰0\rho_{\mathrm{meas}}(p)\succeq0; by Lemma 1 the reconstruction is then PPT, hence separable, and reproduces the data by construction, so Theorem 2 establishes the claim. For example, at p=0.3p=0.3: λmin⁡(ρtrueTB)=−130<0\lambda_{\min}(\rho_{\mathrm{true}}^{T_B})=-\frac{1}{30}<0 (entangled) while λmin⁡(σTB)=+124>0\lambda_{\min}(\sigma^{T_B})=+\frac{1}{24}>0 for σ≡ρmeas(0.3)\sigma\equiv\rho_{\mathrm{meas}}(0.3) (separable). The two states agree on all 23 accessible parameters and differ only in Pz=940P_z=\frac{9}{40}, Cxx=9240C_{xx}=\frac{9\sqrt{2}}{40}, Cyy=−9240C_{yy}=-\frac{9\sqrt{2}}{40}, Czz=940C_{zz}=\frac{9}{40}, all zero in σ\sigma.

Mixtures of two entangled states. Take ρtrue(p)=p∣e0+⟩⟨e0+∣+(1−p)∣e1+⟩⟨e1+∣\rho_{\mathrm{true}}(p)=p|e_0^+\rangle\langle e_0^+|+(1-p)|e_1^+\rangle\langle e_1^+|, a mixture of two maximally entangled states with no separable component. Its partial transpose carries the eigenvalue p4−145p2−8p+4\frac{p}{4}-\frac{1}{4}\sqrt{5p^2-8p+4}, negative for every p∈[0,1)p\in[0,1) since 5p2−8p+4−p2=4(1−p)2>05p^2-8p+4-p^2=4(1-p)^2>0, and −12-\frac{1}{2} at p=1p=1: so the state is entangled throughout. The reconstruction has smallest eigenvalue

min⁡{3p8−1817p2−16p+4, 38−3p8−1817p2−18p+5}\min\left\{\frac{3p}{8}-\frac{1}{8}\sqrt{17p^2-16p+4},\ \frac{3}{8}-\frac{3p}{8}-\frac{1}{8}\sqrt{17p^2-18p+5}\right\}

the first branch negative only for p<1−12p<1-\frac{1}{\sqrt{2}} and the second only for p>12p>\frac{1}{\sqrt{2}} (both radicands are positive throughout, discriminants −16<0-16<0). Thus entanglement is uncertifiable for p∈[1−12,12]p\in[1-\frac{1}{\sqrt{2}},\frac{1}{\sqrt{2}}], a window of width 2−1≈0.414\sqrt{2}-1\approx0.414; at the midpoint p=12p=\frac{1}{2}, λmin⁡(ρtrueTB)=1−58≈−0.155\lambda_{\min}(\rho_{\mathrm{true}}^{T_B})=\frac{1-\sqrt{5}}{8}\approx-0.155 while λmin⁡(ρmeasTB)=+18\lambda_{\min}(\rho_{\mathrm{meas}}^{T_B})=+\frac{1}{8}. The two endpoints, pure Bell states, are correctly detected with λmin⁡=−14\lambda_{\min}=-\frac{1}{4} each (Proposition 1); only their mixture loses the signal in the accessible sector. A scan of 300 random mixtures of two random entangled pure states produced no blind window, indicating that this is a structured feature of this pair rather than generic behaviour, and showing that no bound of the form “entangled enough implies certifiable” is available.

References

References

  1. [1]A. Einstein, B. Podolsky, and N. Rosen, Phys. Rev. 47, 777 (1935).
  2. [2]J. S. Bell, Physics Physique Fizika 1, 195 (1964).
  3. [3]J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, Phys. Rev. Lett. 23, 880 (1969).
  4. [4]A. J. Barr, M. Fabbrichesi, R. Floreani, E. Gabrielli, and L. Marzola, Prog. Part. Nucl. Phys. 139, 104134 (2024), arXiv:2402.07972 [hep-ph].arxiv.org/abs/2402.07972
  5. [5]Y. Afik et al., Eur. Phys. J. Plus 140, 855 (2025), arXiv:2504.00086.DOI
  6. [6]D. E. Eck, J. Phil. Trans. Roy. Soc. A 380, 20210063 (2021), arXiv:2108.08792.
  7. [7]ATLAS Collaboration, Nature 633, 542 (2024), arXiv:2311.07288 [hep-ex].DOI
  8. [8]Y. Afik and J. R. Muñoz de Nova, Eur. Phys. J. Plus 136, 907 (2021), arXiv:2003.02280 [quant-ph].DOI
  9. [9]CMS Collaboration, Phys. Rev. D 100, 072002 (2019), arXiv:1907.03729 [hep-ex].DOI
  10. [10]CMS Collaboration, Rep. Prog. Phys. 87, 117801 (2024), arXiv:2406.03976 [hep-ex].DOI
  11. [11]CMS Collaboration, Phys. Rev. D 110, 112016 (2024), arXiv:2409.11067.DOI
  12. [12]Y. Afik and J. R. Muñoz de Nova, Quantum 6, 820 (2022), arXiv:2203.05582.DOI
  13. [13]K. Cheng, T. Han, and M. Low, Phys. Rev. D 109, 116005 (2024), arXiv:2311.09166.DOI
  14. [14]STAR Collaboration, Nature 650, 65 (2026), arXiv:2506.05499 [nucl-ex].
  15. [15]N. A. Törnqvist, Found. Phys. 11, 171 (1981).
  16. [16]W. Gong, G. Parida, Z. Tu, and R. Venugopalan, Phys.
  17. [17]Y. Shi and J.-C. Yang, Eur. Phys. J. C 80, 116 (2020), arXiv:1912.04111.DOI
  18. [18]BESIII Collaboration, Nature Phys. 15, 631 (2019), arXiv:1808.08917.
  19. [19]BESIII Collaboration, Nature Commun. 16, 4948 (2025), arXiv:2505.14988.
  20. [20]X.-B. Wu et al., JHEP 11, 082, arXiv:2507.15387.
  21. [21]Y. Afik, Y. Kats, J. R. Muñoz de Nova, A. Sofer, and D. Uzan, Phys. Rev. D 111, L111902 (2025), arXiv:2406.04402.DOI
  22. [22]F. Liu and Z. Tu, arXiv e-prints (2026), arXiv:2606.17240 [hep-ph].arxiv.org/abs/2606.17240
  23. [23]A.-I. Lv et al., arXiv e-prints (2025), arXiv:2508.03496.
  24. [24]J. Barata, I. Cuntín, E. Rico, and B. Wu, arXiv e-prints (2026), arXiv:2606.30737 [hep-ph].DOI
  25. [25]B. Andersson, G. Gustafson, G. Ingelman, and T. Sjöstrand, Phys. Rept. 97, 31 (1983).
  26. [26]B. R. Webber, Nucl. Phys. B 238, 492 (1984).
  27. [27]C. Bierlich et al., SciPost Phys. Codebases 8 (2022), arXiv:2302.11601.
  28. [28]S. von Kuk, K. Lee, J. K. L. Michel, and X. Sun, arXiv e-prints (2025), arXiv:2503.22607.arxiv.org/abs/2503.22607
  29. [29]K.-B. Zhang, S.-Y. Wei, et al., Phys. Lett. B 839, 137821 (2023), arXiv:2301.04096.DOI
  30. [30]K.-B. Chen, Z.-T. Liang, Y.-K. Song, and S.-Y. Wei, Phys. Rev. D 105, 034027 (2022).
  31. [31]K. Schilling and G. Wolf, Nucl. Phys. B15, 397 (1970).
  32. [32]STAR Collaboration, Nature 614, 244 (2023), arXiv:2204.02302 [nucl-ex]. X.-L. Sheng, L. Oliva, Z.-T. Liang, Q. Wang, and X.-N. Wang, Phys. Rev. Lett. 131, 042304 (2023), arXiv:2205.15689.
  33. [34]Z.-F. Xu and X.-G. Huang, arXiv e-prints (2024), arXiv:2408.06581.arxiv.org/abs/2408.06581
  34. [35]X.-L. Sheng, Z.-T. Liang, and Q. Wang, arXiv e-prints (2024), arXiv:2402.04540.
  35. [36]A. Peres, Phys. Rev. Lett. 77, 1413 (1996), arXiv:quant-ph/9604005.DOI
  36. [37]M. Horodecki, P. Horodecki, and R. Horodecki, Phys. Lett. A 223, 1 (1996), arXiv:quant-ph/9605038.DOI
  37. [38]M. Banacki, M. Ohta, A. S. Shaleena, P. Caban, M. Eckstein, K. Sakurai, M. Spannowsky, and P. Horodecki, arXiv e-prints (2026), arXiv:2607.26122 [quant-ph].arxiv.org/abs/2607.26122
  38. [39]A. J. Barr, Phys. Lett. B 825, 136866 (2022), arXiv:2106.01377.DOI
  39. [40]R. Ashby-Pickering, A. J. Barr, and A. Wierczucka, JHEP 05, 020, arXiv:2209.13990.
  40. [41]W. M. Kloet and F. Tabakin, arXiv e-prints (1999), arXiv:nucl-th/9908068.DOI
  41. [42]L. de Moura, S. Kong, J. Avigad, F. van Doorn, and J. von Raumer, in Automated Deduction - CADE-25, edited by A. P. Felty and A. Middeldorp (Springer International Publishing, Cham, 2015) pp. 378–388.
  42. [43]The mathlib Community, in Proceedings of the 9th ACM SIGPLAN International Conference on Certified Programs and Proofs, CPP 2020 (ACM, New Orleans, LA, USA, 2020).
  43. [44]K. M. R. Audenaert and M. B. Plenio, New J. Phys. 8, 266 (2006), arXiv:quant-ph/0608067.
  44. [45]O. Gühne and G. Tóth, Phys. Rept. 474, 1 (2009), arXiv:0811.2803.DOI
  45. [46]K. Zhang, X. Wang, and X. Luo, (2026), arXiv:2607.00413 [nucl-th].arxiv.org/abs/2607.00413
  46. [47]J. Datta, A. Deshpande, D. E. Kharzeev, C. J. Naim, and Z. Tu, Phys. Rev. Lett. 134, 111902 (2025), arXiv:2410.22331 [hep-ph].DOI
  47. [48]Z. J. Ajaltouni and E. Conte, arXiv e-prints (2004), arXiv:hep-ph/0409262.DOI
  48. [49]S. Navas et al., Phys. Rev. D 110, 030001 (2024).
  49. [50]P. Horodecki, Phys. Lett. A 232, 333 (1997), arXiv:quant-ph/9703004.DOI
  50. [51]R. Horodecki, P. Horodecki, M. Horodecki, and K. Horodecki, Rev. Mod. Phys. 81, 865 (2009), arXiv:quant-ph/0702225.DOI

Paper details

vlnc:W_9wuxc6cd
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 4

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

ChatGPT 5, Claude Opus 5 (bare)

View statement
Looking for related papers…
Contents