Papers
Topics
Authors
Recent
Search
2000 character limit reached

Input-Output Scattering Approach

Updated 5 July 2026
  • Input-Output Scattering Approach is a formalism that relates ingoing fields, internal system dynamics, and outgoing observables using boundary conditions and Green’s functions.
  • It employs methods such as Keldysh path-integrals, diagrammatic techniques, and Volterra series to model nonlinear interactions and multi-photon processes.
  • The approach underpins quantum-optical experiments, circuit and cavity QED, and macroscopic scattering analyses in lossy or moving-boundary systems.

Searching arXiv for the cited works to ground the article in the current record.
Search query: 2509.07563 input-output scattering approach Keldysh path integral
The input-output scattering approach is a family of formalisms for open quantum systems in which ingoing fields, internal degrees of freedom, and outgoing fields are related by boundary conditions, Green’s functions, scattering matrices, or equivalent generating functionals. In quantum optics it is a well-known tool and is ubiquitous in the description of quantum systems probed by light; owing to the generality of the setup it describes, it finds application in a wide variety of experiments in circuit and cavity QED, as well as in waveguide QED, microwave amplification, nonreciprocal transport, moving-boundary problems, and scattering by lossy macroscopic media [2509.07563][2411.18543].

1. Standard operator structure and boundary relations

In its Markovian form, the approach describes an open (m)-port system by Heisenberg equations of motion supplemented by an input-output boundary condition. A representative formulation is
[
\dot Z = -i[Z,H_S]
+\tfrac12\sum_{j=1}m\Bigl{L_j\dagger[Z,L_j]+[L_j\dagger,Z]L_j\Bigr}
+\sum_{j=1}m\Bigl{b_{j,\rm in}[L_j\dagger,Z]+[Z,L_j]\,b_{j,\rm in}\dagger\Bigr},
]
together with
[
b_{j,\rm out}(t)=b_{j,\rm in}(t)+L_j(t).
]
This is the operator-level core of the Gardiner–Collett formalism and is the starting point for several nonlinear and network generalizations [1407.8108].

For one-dimensional scattering geometries, the same structure appears as boundary relations at the ends of a system coupled to left- and right-propagating waveguides. For a many-body bosonic system coupled at its ends to single-mode waveguides, the Markov and rotating-wave approximations give
[
cL_{\rm out}(t)=cL_{\rm in}(t)-i\sqrt{\gamma_1}\,a_1(t),\qquad
cR_{\rm out}(t)=cR_{\rm in}(t)-i\sqrt{\gamma_N}\,a_N(t),
]
with (\gamma_i=2\pi|\xi_i|2). The same pattern recurs in cavity input-output relations, where the intracavity mode (c_j) obeys a ring-down equation and the output amplitude is the superposition of direct transmission and cavity leakage [1702.01632][1909.04300].

This operator boundary structure is not restricted to bosonic photonic systems. In dissipative quantum-dot circuits, the generalized input-output method defines fermionic input and output fields (d_{n,\rm in/out}), (a_{m,\rm in/out}) and yields
[
\mathbf F_{\rm out}(t)=\mathbf F_{\rm in}(t)-i\sqrt{\tfrac{2}{\pi}\,\mathbf K}\,\mathbf O(t),
]
with (\mathbf K=\mathrm{diag}(\Gamma,\Gamma,\kappa,\kappa)), so that the same formal logic applies to resonant electron transport [2004.05408].

2. Scattering matrices, Green’s functions, and output observables

A central output of the approach is the scattering matrix. In a Keldysh treatment of a system mode (a) coupled to a Markovian bath, the retarded Green’s function is defined by
[
GR(t-t')=-i\,\Theta(t-t')\,\langle[a(t),a\dagger(t')]\rangle,
\qquad
GR[\omega]=\bigl[\omega-\omega_0+i\kappa/2-\SigmaR(\omega)\bigr]{-1},
]
where the bath contributes (\SigmaR_{\rm bath}=-i\kappa/2). The single-photon scattering matrix for coherent elastic scattering is then
[
S(\omega)=1-i\,\kappa\,GR(\omega),
]
which in the simple linear case reduces to
[
S(\omega)=\frac{\omega-\omega_0-i\kappa/2}{\omega-\omega_0+i\kappa/2}.
]
This makes the relation between response theory and scattering explicit [2509.07563].

The same scattering viewpoint can be formulated as a multimode relation. For the dc-SQUID microwave amplifier, the running-state dynamics are linearized around the Josephson oscillation and organized into an admittance matrix
[
Y(\omega)=MI(\omega)\,[MV(\omega)]{-1},
]
from which the multimode scattering matrix is obtained as
[
S(\omega)=[I+Y(\omega)]{-1}[I-Y(\omega)].
]
The complete input and output vectors include common- and differential-mode fields together with sidebands at (\omega+n\omega_J), so the high-frequency dynamics are expressed as a scattering between the participating modes [1206.4706].

The approach also gives direct access to output-field statistics. In the Schwinger–Keldysh path-integral formulation, one introduces source fields (\chi(t)) and (\chi'(t)) coupled to (b_{\rm out}(t)) and (b_{\rm out}\dagger(t)), defines (\Lambda_{\rm out}[\chi,\chi']), and works with the cumulant generating functional (\mathcal S_{\rm out}=\ln\Lambda_{\rm out}). Normal- and time-ordered output cumulants follow by functional differentiation, including
[
G{(1)}(t,t')=\langle b_{\rm out}\dagger(t)\,b_{\rm out}(t')\rangle,
\qquad
G{(2)}(t_1,t_2)=\langle b_{\rm out}\dagger(t_1)b_{\rm out}\dagger(t_2)b_{\rm out}(t_2)b_{\rm out}(t_1)\rangle.
]
This places output coherences, scattering amplitudes, and nonequilibrium field theory within one framework [2509.07563].

3. Path-integral, diagrammatic, and Volterra formulations

One major line of development recasts input-output theory in the Schwinger–Keldysh path-integral language. Starting from a system Hamiltonian (H_S[a\dagger,a]) coupled to bath modes (b_k) through
[
V=\sum_k\bigl(g_k\,a\dagger b_k + g_k*\,b_k\dagger a\bigr),
]
one writes the evolution on a closed-time contour, inserts coherent-state resolutions of the identity, integrates out the bath variables at intermediate times under the Markov approximation with flat spectrum (\rho(\omega)\approx 2\pi\kappa), and performs the Keldysh rotation
[
\psi{\rm cl}=\tfrac1{\sqrt2}(\psi++\psi-),\qquad
\psi{\rm q}=\tfrac1{\sqrt2}(\psi+-\psi-).
]
The resulting effective action involves only system fields coupled to input/output source fields and supports a double-contour diagrammatics in the ((R,A,K)) basis [2509.07563].

Within that diagrammatics, (\phi{\rm cl}) are drawn as full lines and (\phi{\rm q}) as dashed lines. The Gaussian part gives the propagators (GR), (GA), and (GK). Each interaction vertex carries either one or three quantum legs, which enforces causality and ensures that closed loops of (GR) or (GA) alone vanish. Infinite classes of diagrams, including tadpole loops of the Keldysh propagator, can be resummed by shifting (\omega_0\to\omega_0+4U\,n_B) or by solving the Dyson equation
[
\mathbf G=\mathbf G_0+\mathbf G_0\,\mathbf\Sigma\,\mathbf G.
]
This gives a uniform way of obtaining perturbative results for nonlinear systems [2509.07563].

A second line of development uses Green’s functions directly at the S-matrix level. In multiphoton scattering, once the output fields are rewritten using the input-output relations and Wick’s theorem is applied, the nontrivial contributions reduce to time-ordered correlation functions of system operators evolving under an effective non-Hermitian Hamiltonian
[
H_{\rm eff}=H_{\rm sys}-\tfrac{i}{2}\sum_j\gamma_j a_j\dagger a_j-\dots.
]
Connected (n)-photon S-matrix elements are identified with (2n)-point Green’s functions, and a diagrammatic enumeration of time-orderings yields exact closed forms as finite sums of rational functions [1702.01632].

A third line of development replaces explicit internal dynamics by a quantum Volterra series. For a general component, one formally averages out the internal operators (L_j(t)) against the initial state and expands the output as an infinite series of multilinear functionals of the past inputs. The kernels
[
K{(\pm)}_{j;\,j_1\cdots j_n}(\tau_1,\ldots,\tau_n)
]
are determined by (H_S), the coupling operators (L_j), and the initial state (\rho_0). For weak nonlinearity,
[
H=H_\ell+\mu H_{nl},\qquad L_j=L_{j,\ell}+\mu L_{j,nl},\qquad \mu\ll1,
]
one obtains a truncation with a linear-response kernel (K{(1)}), a lowest nonlinear correction (K{(2)}), and complexity that grows only as (O(N)) with network size rather than exponentially [1407.8108].

4. Nonlinearity, many-body response, and output statistics

The input-output scattering approach is especially useful when the observable of interest is not only transmission or reflection, but the full output-field statistics of a nonlinear device. For the Kerr oscillator with
[
H_S=\omega_0\,a\dagger a+\tfrac U2\,a\dagger a\dagger a a,
]
the Keldysh action contains a quartic interaction
[
S_{\rm Kerr}
=S_0[\phi]
-U\int dt\,
\bigl(\bar\phi{\rm cl}\phi{\rm cl}+\bar\phi{\rm q}\phi{\rm q}\bigr)\,
\bigl(\bar\phi{\rm q}\phi{\rm cl}+\bar\phi{\rm cl}\phi{\rm q}\bigr).
]
Expanding perturbatively in (U) and using Wick’s theorem yields corrections to the average output field (\langle b_{\rm out}(t)\rangle). To second order in (U), these corrections reduce the magnitude of (\langle b_{\rm out}\rangle), which is a reduction of the coherent reflection amplitude (S(0)=\langle b_{\rm out}\rangle/\langle b_{\rm in}\rangle) [2509.07563].

The same example also illustrates a common interpretive point. The net result can be a reflection probability
[
R=|S(0)|2<1
]
even though no photons are truly absorbed; rather the output field becomes squeezed so that its displacement amplitude is reduced. The same diagrammatic expansion gives access to (G{(1)}(\tau)), the squeezing spectrum, and the antibunching/bunching function (g{(2)}(\tau)) [2509.07563].

In weakly nonlinear quantum networks, the Volterra formulation provides an alternative reduction. For a Kerr cavity, an optomechanical transducer, and a nonlinear coherent-feedback loop containing a quantum amplifier, the method keeps a small set of causal kernels up to the desired order. The stated benefit is that it can model weakly nonlinear quantum networks and can formulate quantum networks with both nonlinear and nonconservative components, including quantum amplifiers, which cannot be modelled by the Hudson–Parthasarathy model and the quantum transfer function model [1407.8108].

At the few-photon level, the diagrammatic Green-function approach makes the nonlinear response visually explicit. For a single two-level emitter, two collocated two-level atoms, and Bose–Hubbard arrays of Kerr resonators, the connected amplitudes are obtained by summing over eigenstates of (H_{\rm eff}) and over the allowed absorption and emission sequences. This separates purely elastic contributions from genuinely nonlinear inelastic processes through cluster decomposition of the full S-matrix [1702.01632].

5. Interference, nonreciprocity, and modified scattering channels

A recurring theme across implementations is that the scattering approach exposes interference between distinct physical channels. For a giant atom beyond the electric-dipole approximation, the modified input-output approach introduces an additional low-(Q) cavity mode (b) representing the quasi-direct non-dipolar channel. Under the Markov and rotating-wave approximations, the output fields satisfy
[
a_{out}R(t)=a_{in}R(t)-\sum_j\sqrt{\gamma_j}\,e{+ikx_j}\,\sigma_-(t)+\sqrt{\gamma_bR}\,b(t),
]
[
a_{out}L(t)=a_{in}L(t)-\sum_j\sqrt{\gamma_j}\,e{-ikx_j}\,\sigma_-(t)+\sqrt{\gamma_bL}\,b(t).
]
For one-port drive, the right-going output amplitude takes the form
[
a_{out}R(\omega)=t(\omega)\,a_{in}L(\omega),\qquad t(\omega)=t_{\rm res}(\omega)+t_{\rm bg}(\omega),
]
with a resonant channel and a background quasi-direct channel (t_{\rm bg}(\omega)\approx \mu e{i\phi}). The total transmission probability (T(\omega)=|t(\omega)|2) exhibits the characteristic asymmetric Fano profile, and fits allow extraction of (\gamma_j), (\gamma_c), (\Gamma_0), (\mu), (\phi), (q), and (\Gamma_{\rm tot}) [2605.11041].

An analogous revision occurs in cavity input-output relations for whispering-gallery resonator–waveguide systems. In the weak-scattering regime, (t_0\approx 1), the standard relation
[
a_j{\rm out}=a_j{\rm in}-\sqrt{2\,\kappa_je}\,c_j
]
yields band-stop transmission with off-resonant transmission approaching unity. In the strong-scattering regime, (|t_0|\approx 0), the off-resonant field approaches to zero, but more than (90\%) coupling efficiency can still be achieved due to the Purcell-enhanced channeling. The cavity-impact factor is
[
G_j(\omega)=\frac{2\,\kappa_j}{[\kappa_j2+(\omega-\omega_j)2]\,\tau_j},
]
and the corresponding output relation becomes
[
a_j{\rm out}
=\sqrt{\frac{\Gamma_j(\omega)\,\kappa_je}{\kappa_j0+\kappa_je}}\;a_j{\rm in}.
]
The CIOR is therefore essentially different in the strong-scattering regime, and polarization control selects either band-stop or band-pass behavior [1909.04300].

Nonreciprocity is another setting in which the scattering representation is structurally informative. In the dc-SQUID amplifier, inclusion of at least two Josephson harmonics yields a ratchet-like pump and unequal forward and reverse gains,
[
G_{\rm fwd}(\omega_m)=|S_{CD}(\omega_m)|2,\qquad
G_{\rm rev}(\omega_m)=|S_{DC}(\omega_m)|2,
]
with directionality
[
D(\omega_m)=10\log_{10}[G_{\rm fwd}/G_{\rm rev}]
]
that peaks at a bias (\epsilon\approx 0.45). In the quantum-dot realization, directionality is enforced by (\Phi=0), impedance matching by (\Sigma=2\Gamma), and under the two optimal conditions the on-resonance scattering matrix becomes
[
\widetilde S(\epsilon_d)=
\begin{pmatrix}
0&0\[1ex]
\widetilde S_{21}(\epsilon_d)&0
\end{pmatrix},
]
so that only one transmission direction remains and the device acts as a quantum diode in the resonant transport regime [1206.4706][2004.05408].

6. Moving boundaries, loss, and macroscopic quantum scattering

The scattering formulation also extends to cases where the scatterer itself is dynamical or lossy. For a quantized moving mirror interacting with left- and right-propagating optical fields, the Hudson–Parthasarathy formalism writes the joint unitary (U(t)) as a QSDE driven by gauge or scattering processes (d\Lambda_{ab}(t)=b_a\dagger(t)b_b(t)\,dt). In ((S,L,H)) language one has (L=0), (H=H_{\rm mirror}), and a (2\times2) operator-valued scattering matrix (S(Q)). The output fields satisfy
[
b_{{\rm out},a}(t)=S_{ab}(Q(t))\,b_{{\rm in},b}(t),
]
and in the single-channel case,
[
b_{\rm out}(t)=e{i\theta(Q(t))}\,b_{\rm in}(t).
]
For coherent drive from one side, the momentum obeys a radiation-pressure equation in which the force is proportional to (\theta'(Q_t)\,b\dagger(t)b(t)) [1409.3527].

This moving-boundary problem also makes explicit a mathematical subtlety. Two singular approximation schemes lead to different stochastic limits. In the Holevo time-ordering scheme,
[
S=\exp[-iE_{\ell\ell}],
]
whereas the Stratonovich or midpoint scheme gives
[
S=(I-\tfrac12 iE_{\ell\ell})(I+\tfrac12 iE_{\ell\ell}){-1}.
]
The limit model is therefore highly sensitive to how the approximation scheme is interpreted mathematically. This is illustrated already at the level of one-particle scattering, where two regularizations of a localized potential give different boundary conditions and different phase factors [1409.3527].

For lossy macroscopic objects of arbitrary shape, size, and dispersive optical response, the modified Langevin noise formalism separates the field into three noninteracting bosonic subsystems: (s)-, (e)-, and (m)-polaritons. In the lossless limit, (s)-polaritons reduce to standard photons whereas (e)- and (m)-polaritons disappear. The far-field input-output relation is a block-unitary transformation between ingoing and outgoing operators, with transmission dyadic (T_{ss}), emission dyadics (E_{s\nu}), absorption dyadics (A_{\nu s}), and internal redistribution dyadics (Q_{\nu\nu'}). The radiation-balance constraint is
[
T_{ss}T_{ss}\dagger+E_{se}E_{se}\dagger+E_{sm}E_{sm}\dagger=I_{ss},
]
which reduces in the lossless limit to (T_{ss}T_{ss}\dagger=I_{ss}) [2411.18543].

Because the scattered radiation is collected in the far field while the object is usually left unmeasured, the formalism also yields the reduced density operator of the outgoing (s)-polaritons. In the single-photon case, the outgoing field takes the form
[
\hat\rho_s{(\rm out)}
=P_s\,|\Phi_{1s}\rangle\langle\Phi_{1s}|
+(P_e+P_m)\,|0\rangle\langle0|,
]
so that absorption by the object appears as loss of purity in the reduced photonic state. In the two-photon case, the scattered field becomes a mixture of sectors corresponding to two transmitted photons, one transmitted and one absorbed photon, or both photons absorbed, and the one-photon reduced block generally has rank two when the input is entangled [2411.18543].

Taken together, these constructions show that the input-output scattering approach is not a single formalism but a technically coherent family of representations. Across operator, path-integral, diagrammatic, Volterra, QSDE, and macroscopic-electrodynamic realizations, the common structure is the replacement of explicit bath dynamics by causal relations between ingoing fields, internal response, and outgoing observables, with the scattering matrix or its generalizations serving as the primary observable interface.

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

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

Follow Topic

Get notified by email when new papers are published related to Input-Output Scattering Approach.