The two-mode parametric amplifier of Mollow and Glauber (Phys. Rev. 160, 1097 (1967)) evolves an arbitrary initial state through two-mode squeezing, and the weight function P of the diagonal coherent-state (Glauber-Sudarshan) representation is the Fourier transform of the initial normally ordered characteristic function multiplied by a kernel that grows like exp⁡[(∣η∣2+∣ζ∣2)sinh⁡2κt]\exp[(|\eta|^2+|\zeta|^2)\sinh^2\kappa t]. The source proved that a non-negative P requires the correlation conditions μ(±)(t)≥1\mu^{(\pm)}(t)\ge1 with μ(±)(t)=μ(±)(0)e±2κt\mu^{(\pm)}(t)=\mu^{(\pm)}(0)e^{\pm2\kappa t} (necessary for arbitrary states, and not known to be sufficient), and that thermal and displaced-Gaussian inputs lose their P function beyond finite explicitly computable times; it left open the sufficient conditions for arbitrary initial states, the classification of initial characteristic functions by existence interval, the singularity structure of P beyond breakdown, and extensions to nonresonant or non-quadratic couplings. Later literature resolved the pure-state classification (Cahill 1969), the Gaussian sector (covariance criterion, breakdown fully explicit), the representation question (regularized and generalized P representations), and the distribution-theoretic singularity hierarchy (Sperling 2016). The surviving open core is the necessary-and-sufficient characterization of when PP is a regular (ordinary, non-singular) non-negative function — the strictly regular subclass of P-classical states, excluding the atomic boundary formed by the coherent states and their finite mixtures — and of the P-existence interval and earliest singular time for arbitrary, non-Gaussian initial states of the amplifier, stated directly in terms of the initial characteristic function.

Background

A coherent state ∣α⟩|\alpha\rangle of a bosonic mode with annihilation operator aa is the eigenstate a∣α⟩=α∣α⟩a|\alpha\rangle=\alpha|\alpha\rangle; it is the quantum state whose behavior most closely resembles a classical harmonic oscillation. Glauber and Sudarshan showed that any density operator of one or several modes admits a diagonal expansion ρ=∫P(α,β) ∣α,β⟩⟨α,β∣ d2α d2β\rho=\int P(\alpha,\beta)\,|\alpha,\beta\rangle\langle\alpha,\beta|\,d^2\alpha\,d^2\beta in products of coherent-state projectors, where the weight function PP — the Glauber-Sudarshan or P function — always exists as a generalized function (a distribution), but need not be an ordinary non-negative function. Whenever PP is a non-negative generalized function — a probability measure, possibly atomic — the state is called classical (P-classical): all its statistics are then those of a classical ensemble of coherent fields, the coherent states themselves being classical with PP an atomic delta measure. Otherwise the state is nonclassical, which is the foundational definition of quantum light in quantum optics. Failure comes in two grades: PP may be a regular function taking negative values, or it may be so singular — derivatives of Dirac delta functions, or worse — that it cannot be interpreted as a density at all. Squeezed states, the outputs of parametric amplifiers, are the canonical examples.

The non-degenerate parametric amplifier consists of two modes AA and BB with frequencies ωa\omega_a, ωb\omega_b, coupled by a classical pump at the sum frequency. In the rotating-wave, undepleted-pump approximation (the rotating-wave approximation keeps only the energy-conserving coupling terms, and the undepleted-pump approximation treats the classical pump field as a prescribed constant of unchanged amplitude) it is modeled by the quadratic Hamiltonian

H(t)=ℏωaa†a+ℏωbb†b−ℏκ(a†b†e−iωt+ab eiωt),ω=ωa+ωb,H(t)=\hbar\omega_a a^\dagger a+\hbar\omega_b b^\dagger b-\hbar\kappa\left(a^\dagger b^\dagger e^{-i\omega t}+ab\,e^{i\omega t}\right),\qquad \omega=\omega_a+\omega_b,

where κ\kappa is the coupling constant. The Heisenberg solutions mix each mode with the creation operator of the other, a(t)=a e−iωatcosh⁡κt+i b†e−iωatsinh⁡κta(t)=a\,e^{-i\omega_a t}\cosh\kappa t+i\,b^\dagger e^{-i\omega_a t}\sinh\kappa t and analogously for b(t)b(t); this is two-mode squeezing, the mechanism behind strong quantum correlations between the signal and idler outputs.

Mollow and Glauber (Paper II, 1967) derived the exact joint evolution of the two-mode density operator and of its phase-space representatives. The Wigner function W(α,β,t)W(\alpha,\beta,t) — the symmetric quasiprobability obtained as the Fourier transform of the ordinary characteristic function — is transported exactly along the classical trajectories (Liouville flow) for arbitrary initial states, at all times. The P function, by contrast, is tied to the normally ordered characteristic function χN(η,ζ,t)\chi_N(\eta,\zeta,t), the expectation value of the normally ordered displacement exponential (normal ordering places all creation operators to the left of annihilation operators; χN\chi_N generates all normally ordered moments, hence all photon-counting statistics). Their Eq. (6.15) expresses P(α,β,t)P(\alpha,\beta,t) as the Fourier integral of the initial χN(η,ζ,0)\chi_N(\eta,\zeta,0) with a kernel containing the factor exp⁡[(∣η∣2+∣ζ∣2)sinh⁡2κt]\exp[(|\eta|^2+|\zeta|^2)\sinh^2\kappa t], which grows exponentially in time; equivalently, in decoupled variables one argument of χN\chi_N is rescaled by eκte^{\kappa t} and the other by e−κte^{-\kappa t}.

Because of this growing kernel the Fourier integral can diverge beyond a finite time. For two independent thermal (chaotic) inputs with equal mean occupation ⟨n⟩\langle n\rangle the amplified state keeps a Gaussian, non-negative P only for ∣t∣<τ|t|<\tau with e2κτ=1+2⟨n⟩e^{2\kappa\tau}=1+2\langle n\rangle; for displaced thermal inputs with occupations ⟨n⟩,⟨m⟩\langle n\rangle,\langle m\rangle the condition is M(t)≥0M(t)\ge0, i.e. cosh⁡2κt≤1+2⟨n⟩⟨m⟩/(1+⟨n⟩+⟨m⟩)\cosh 2\kappa t\le1+2\langle n\rangle\langle m\rangle/(1+\langle n\rangle+\langle m\rangle); and if either input mode is the vacuum, a P representation exists only at t=0t=0. Introducing the decoupled fluctuation operators c(±)(t)c^{(\pm)}(t) — linear combinations of a(t)a(t) and b†(t)b^\dagger(t) that evolve as e±κte^{\pm\kappa t} — with correlation measures μ(±)(t)=2 tr{ρ c^(±)†(t)c^(±)(t)}=μ(±)(0)e±2κt\mu^{(\pm)}(t)=2\,\mathrm{tr}\{\rho\,\hat c^{(\pm)\dagger}(t)\hat c^{(\pm)}(t)\}=\mu^{(\pm)}(0)e^{\pm2\kappa t}, the authors proved a model-independent necessary condition: a non-negative P at time tt requires μ(±)(t)≥1\mu^{(\pm)}(t)\ge1, so classicality can survive at most within the time interval 1/μ(+)(0)≤e2κt≤μ(−)(0)1/\mu^{(+)}(0)\le e^{2\kappa t}\le\mu^{(-)}(0). Within the Gaussian family this condition is also sufficient, but for general initial states it is not known to be so — the absence of any general sufficiency criterion is confirmed by the modern literature (Sperling and Vogel 2020; Alexanian 2016; Ryl et al. 2015) — and the authors left open: the sufficient conditions for arbitrary initial states, the classification of initial characteristic functions by their P-existence interval, the detailed singularity structure of P beyond breakdown and its relation to observables, and the fate of these questions for nonresonant couplings and non-quadratic Hamiltonians (where the Wigner function no longer follows Liouville flow and the decoupled exponential modes do not exist).

The subsequent literature resolved substantial parts of this program. Sudarshan and Klauder, McKenna and Currie (1963-1965) had shown that the diagonal representation exists for every state as a distribution, so the amplifier question concerns regularity and positivity, not bare existence. Cahill (1969) gave a complete classification for pure states — P is a tempered distribution (a distribution acting on the Schwartz space of smooth, rapidly decreasing test functions) exactly for photon-added coherent states [c0+c1a†+⋯+cn(a†)n]∣α⟩[c_0+c_1a^\dagger+\cdots+c_n(a^\dagger)^n]|\alpha\rangle, every other pure state having singularities incompatible with the P form — and introduced a regularized P representation valid for all density operators; Cahill and Glauber (1969) unified the quasiprobabilities through s-ordering, a one-parameter family of quasiprobabilities interpolating between P, Wigner, and Husimi functions (the Husimi function is the everywhere non-negative quasiprobability obtained by Gaussian-smoothing the Wigner function, equivalently the coherent-state expectation value of the density operator). Drummond and Gardiner (1980) introduced generalized P representations — the complex-P on integration contours and the positive P on a doubled phase space — that exist for all states and yield a practical master-equation technology for nonlinear (non-quadratic) Hamiltonians; Chaturvedi, Agarwal and Srinivasan (1994) developed related positivization and regularization schemes. For Gaussian states, which include all amplified thermal and displaced inputs, P-classicality is completely decided by positive semidefiniteness of the normally ordered covariance matrix (equivalent to μ(±)≥1\mu^{(\pm)}\ge1); Alexanian (2016) works the parametric-amplifier case out explicitly, including the inevitability of breakdown at finite time. Sperling (2016) characterized the singularities of P within distribution theory: the maximal singularity is of exponential order in delta-derivatives, the representation is non-unique, test functions must lie beyond the Schwartz class, and finite-order delta-derivative singularities certify nonclassicality while infinite-order ones do not (a thermal state admits an infinitely singular representation and is nevertheless classical). Observable criteria are supplied by moment hierarchies (Richter and Vogel 2002; Shchukin, Richter and Vogel 2005) and by nonclassicality filters and regularized quasiprobabilities (Kiesel and Vogel 2010), with which the first experimental determination of a genuinely nonclassical P function was achieved for photon-added thermal states (Kiesel et al. 2008). Hillery (1985) proved that the only pure classical states are the coherent states themselves, and Wünsche (2004) showed that every classical state has nonclassical states arbitrarily close, so classicality cannot be certified within any finite error margin. What no later work supplies is a necessary-and-sufficient criterion applicable to arbitrary non-Gaussian states, or a classification of initial states by the time interval on which their amplified P function exists: the surviving core stated below.

Problem Statement

For the resonant two-mode parametric amplifier with Hamiltonian

H(t)=ℏωaa†a+ℏωbb†b−ℏκ(a†b†e−iωt+ab eiωt),ω=ωa+ωb,H(t)=\hbar\omega_a a^\dagger a+\hbar\omega_b b^\dagger b-\hbar\kappa\left(a^\dagger b^\dagger e^{-i\omega t}+ab\,e^{i\omega t}\right),\qquad \omega=\omega_a+\omega_b,

acting on an arbitrary initial two-mode state with normally ordered characteristic function χN(η,ζ,0)=⟨eηa†+ζb†e−η∗a−ζ∗b⟩\chi_N(\eta,\zeta,0)=\langle e^{\eta a^\dagger+\zeta b^\dagger}e^{-\eta^*a-\zeta^*b}\rangle (the expectation value of the normally ordered displacement operator, which generates all normally ordered moments), the weight function P(α,β,t)P(\alpha,\beta,t) of the diagonal coherent-state representation is, when it exists as an ordinary function, the Fourier transform of χN(η,ζ,0)\chi_N(\eta,\zeta,0) with a kernel containing the time-growing factor exp⁡[(∣η∣2+∣ζ∣2)sinh⁡2κt]\exp[(|\eta|^2+|\zeta|^2)\sinh^2\kappa t] (source Eq. (6.15)); under the amplifier evolution χN(⋅,t)\chi_N(\cdot,t) is determined exactly from χN(⋅,0)\chi_N(\cdot,0) by the two-mode squeezing transformation, rescaling the decoupled-variable arguments by e±κte^{\pm\kappa t}. Determine, with proofs:

(1) necessary and sufficient conditions, stated in terms of the initial state (equivalently χN(η,ζ,0)\chi_N(\eta,\zeta,0)) and tt, under which P(α,β,t)P(\alpha,\beta,t) exists as a non-negative, regular (ordinary, non-singular) function. The conditions must establish that P(α,β,t)P(\alpha,\beta,t) is a regular non-negative function and not merely a non-negative measure or distribution: a bare restatement of the convergence of the defining Fourier integral at time tt, or positivity established only in the distributional sense, does not qualify. Within the Gaussian family the conditions must reduce to positive-definiteness of the normally ordered covariance matrix, equivalently the strict correlation bounds μ(±)(t)>1\mu^{(\pm)}(t)>1 (the boundary consists of the marginal cases in which a correlation bound equals one — at t=0t=0, the coherent and vacuum inputs — for which PP degenerates to a singular delta measure rather than a regular function), where μ(±)(t)=2 tr{ρ c^(±)†(t)c^(±)(t)}=μ(±)(0)e±2κt\mu^{(\pm)}(t)=2\,\mathrm{tr}\{\rho\,\hat c^{(\pm)\dagger}(t)\hat c^{(\pm)}(t)\}=\mu^{(\pm)}(0)e^{\pm2\kappa t} are the correlation measures of the decoupled fluctuation operators c^(±)(t)=c(±)(t)−⟨c(±)(t)⟩\hat c^{(\pm)}(t)=c^{(\pm)}(t)-\langle c^{(\pm)}(t)\rangle, built from the linear combinations c(±)(t)=2−1/2[a(t)eiωat±i b†(t)e−iωbt]c^{(\pm)}(t)=2^{-1/2}[a(t)e^{i\omega_a t}\pm i\,b^\dagger(t)e^{-i\omega_b t}] that evolve as c(±)(t)=c(±)(0)e±κtc^{(\pm)}(t)=c^{(\pm)}(0)e^{\pm\kappa t}; the non-strict bounds μ(±)(t)≥1\mu^{(\pm)}(t)\ge1 are necessary for arbitrary states and not known to be sufficient;

(2) a complete classification of initial characteristic functions by the maximal time interval on which P(α,β,t)P(\alpha,\beta,t) exists as an ordinary function, including the earliest singular time τ∗(χN)\tau_*(\chi_N) of each class, with the classes characterized by an explicit criterion on χN(η,ζ,0)\chi_N(\eta,\zeta,0) rather than by relabeling each initial characteristic function with its own existence interval (the source's thermal and displaced-Gaussian formulas give τ∗\tau_* within the Gaussian family; the classification must extend beyond the Gaussian family). Deliverables (1) and (2) are jointly required and target different properties of PP: (1) concerns a regular non-negative function, whereas (2) classifies by the maximal interval on which P(α,β,t)P(\alpha,\beta,t) exists as an ordinary function of unrestricted sign, τ∗(χN)\tau_*(\chi_N) being that interval's endpoint, the earliest time at which PP ceases to be an ordinary function. Since non-negativity is an additional requirement, the interval of (1) is contained in that of (2), so (1) does not by itself determine (2); (2) requires this interval and τ∗\tau_* as an explicit rule on χN(η,ζ,0)\chi_N(\eta,\zeta,0) covering every initial characteristic function, completeness of the classification meaning exactly that coverage.

Items (1)-(2) for arbitrary, non-Gaussian initial states are the surviving open core. The singularity structure of PP beyond breakdown — which classes of generalized functions can represent the amplified two-mode state, tied to measurable correlation quantities of which the source's μ(±)\mu^{(\pm)} bounds are the prototype — and the fate of these questions for nonresonant couplings or non-quadratic Hamiltonian terms, where the Wigner function no longer follows classical Liouville flow and the decoupled exponential operator modes do not exist, are documented background: they have been substantially resolved or absorbed by later work as recorded in Background, and are not proof deliverables of this problem.

References

  1. B. R. Mollow and R. J. Glauber, Quantum Theory of Parametric Amplification. II, Physical Review 160, 1097-1108 (1967). DOI: 10.1103/PhysRev.160.1097, https://doi.org/10.1103/PhysRev.160.1097
  2. B. R. Mollow and R. J. Glauber, Quantum Theory of Parametric Amplification. I, Physical Review 160, 1076-1096 (1967). DOI: 10.1103/PhysRev.160.1076, https://doi.org/10.1103/PhysRev.160.1076
  3. E. C. G. Sudarshan, Equivalence of Semiclassical and Quantum Mechanical Descriptions of Statistical Light Beams, Physical Review Letters 10, 277-279 (1963). DOI: 10.1103/PhysRevLett.10.277, https://doi.org/10.1103/PhysRevLett.10.277
  4. J. R. Klauder, J. McKenna and D. G. Currie, On 'Diagonal' Coherent-State Representations for Quantum-Mechanical Density Matrices, Journal of Mathematical Physics 6, 734-739 (1965). DOI: 10.1063/1.1704330, https://doi.org/10.1063/1.1704330
  5. K. E. Cahill, Pure States and the P Representation, Physical Review 180, 1239-1243 (1969). DOI: 10.1103/PhysRev.180.1239, https://doi.org/10.1103/PhysRev.180.1239
  6. K. E. Cahill, Regularization of the P Representation, Physical Review 180, 1244-1255 (1969). DOI: 10.1103/PhysRev.180.1244, https://doi.org/10.1103/PhysRev.180.1244
  7. K. E. Cahill and R. J. Glauber, Density Operators and Quasiprobability Distributions, Physical Review 177, 1882-1902 (1969). DOI: 10.1103/PhysRev.177.1882, https://doi.org/10.1103/PhysRev.177.1882
  8. P. D. Drummond and C. W. Gardiner, Generalised P-representations in quantum optics, Journal of Physics A: Mathematical and General 13, 2353-2368 (1980). DOI: 10.1088/0305-4470/13/7/018, https://doi.org/10.1088/0305-4470/13/7/018
  9. S. Chaturvedi, G. S. Agarwal and V. Srinivasan, Positivization and regularization of quantum phase space distributions, Journal of Physics A: Mathematical and General 27, L39-L45 (1994). DOI: 10.1088/0305-4470/27/2/007, https://doi.org/10.1088/0305-4470/27/2/007
  10. M. Hillery, Classical pure states are coherent states, Physics Letters A 111, 409-411 (1985). DOI: 10.1016/0375-9601(85)90483-9, https://doi.org/10.1016/0375-9601(85)90483-9
  11. A. Wünsche, About the nonclassicality of states defined by nonpositivity of the P-quasiprobability, Journal of Optics B: Quantum and Semiclassical Optics 6, 159-171 (2004). DOI: 10.1088/1464-4266/6/2/006, https://doi.org/10.1088/1464-4266/6/2/006
  12. G. S. Agarwal and K. Tara, Nonclassical character of states exhibiting no squeezing or sub-Poissonian statistics, Physical Review A 46, 485-488 (1992). DOI: 10.1103/PhysRevA.46.485, https://doi.org/10.1103/PhysRevA.46.485
  13. Th. Richter and W. Vogel, Nonclassicality of Quantum States: A Hierarchy of Observable Conditions, Physical Review Letters 89, 283601 (2002). DOI: 10.1103/PhysRevLett.89.283601, https://doi.org/10.1103/PhysRevLett.89.283601
  14. E. Shchukin, Th. Richter and W. Vogel, Nonclassicality criteria in terms of moments, Physical Review A 71, 011802 (2005). DOI: 10.1103/PhysRevA.71.011802, https://doi.org/10.1103/PhysRevA.71.011802
  15. T. Kiesel, W. Vogel, V. Parigi, A. Zavatta and M. Bellini, Experimental determination of a nonclassical Glauber-Sudarshan P function, Physical Review A 78, 021804(R) (2008). DOI: 10.1103/PhysRevA.78.021804, https://doi.org/10.1103/PhysRevA.78.021804
  16. T. Kiesel and W. Vogel, Nonclassicality filters and quasiprobabilities, Physical Review A 82, 032107 (2010). DOI: 10.1103/PhysRevA.82.032107, https://doi.org/10.1103/PhysRevA.82.032107
  17. M. Alexanian, Nonclassicality criteria: Quasiprobability distributions and correlation functions, Physical Review A 94, 043837 (2016). DOI: 10.1103/PhysRevA.94.043837, https://doi.org/10.1103/PhysRevA.94.043837
  18. J. Sperling, Characterizing maximally singular phase-space distributions, Physical Review A 94, 013814 (2016). DOI: 10.1103/PhysRevA.94.013814, https://doi.org/10.1103/PhysRevA.94.013814
  19. J. Sperling and W. Vogel, Quasiprobability distributions for quantum-optical coherence and beyond, Physica Scripta 95, 034007 (2020). DOI: 10.1088/1402-4896/ab5501, https://doi.org/10.1088/1402-4896/ab5501
  20. S. Ryl, J. Sperling, E. Agudelo, M. Mraz, S. Köhnke, B. Hage and W. Vogel, Unified nonclassicality criteria, Physical Review A 92, 011801 (2015). DOI: 10.1103/PhysRevA.92.011801, https://doi.org/10.1103/PhysRevA.92.011801