WorksPhysics and Astronomy
Time-Varying Medium from an Ensemble of Laser-Dressed Quantum Emitters
Abstract
We propose a minimal quantum-mechanical model of a frequency dispersive, time-modulated medium that explicitly incorporates the pumping mechanism. It consists of a homogeneous ensemble of quantum emitters dressed by a pump laser. We analyze their dispersive, time-varying response to a perturbative probe field by deriving the macroscopic polarizability of the ensemble, and then solving Maxwell's equations via the Floquet formalism. This yields a rich band structure that departs from conventional analyses in which the modulation source is neglected, arising from the interplay between the dispersive polaritonic and time-varying responses. These results shed light on the role of the modulation source in time-varying media and pave the way toward realizing dispersive time-varying media from driven quantum-mechanical systems.
Photonic time-varying media (TVM), materials whose optical properties are modulated in time [1, 2, 3, 4], have emerged as a promising platform for controlling light, enabling linear frequency conversion, amplification, and non-reciprocity. A hallmark of TVM is the emergence of momentum gaps in the photonic dispersion relation under fast periodic modulation [5, 6, 7, 8, 9, 10, 11, 12, 13]. Such momentum gaps host complex-frequency modes and can temporally amplify light when the modulation is sufficiently strong. However, theoretical analyses typically rely on classical electromagnetism, leaving the quantum-mechanical description of TVM largely unexplored [14, 15, 16, 17, 18, 19], particularly from a microscopic point of view of the medium. Moreover, these analyses often assume a non-dispersive, phenomenological time-dependent permittivity, without explicitly accounting for the pumping mechanism responsible for the modulation. This raises an important question: how are conventional TVM effects modified when the quantum-mechanical nature of the medium, its dispersive response, and the underlying pump are all explicitly accounted for?
In this Letter, we address this question by proposing a minimal quantum-mechanical model of a dispersive time-varying medium that explicitly includes the pumping mechanism. The model consists of a homogeneous ensemble of quantum emitters (QEs) driven into the Mollow regime Figure 1(a), where the pump laser dresses the states of the system [20, 21, 22, 23]. We study the linear response of this driven system to a weak probe field by analytically deriving its dispersive polarizability, following the approach of Ref. [15] but explicitly accounting for the pumping mechanism. This polarizability is inserted into Maxwell’s equations, which are then solved using the Floquet theorem to obtain the electromagnetic modes supported by the driven medium. The resulting photonic band structure, centered at the pump laser frequency, displays flat momentum gaps at the emitter dressed frequencies, with exceptional points at the band edges that we characterize via the phase rigidity [24, 25], as well as dispersive momentum gaps and avoided crossings. These features are a direct consequence of including dispersion and the modulation source in the analysis.

Figure 1. Polarizability for the minimal quantum-mechanical model of TVM proposed in this Letter. (a) Schematic representation of the homogeneous ensemble of QEs driven by a strong pump laser. The dispersive, time-varying response to a weak probe laser is characterized by the two-time polarizability . (b) Behaviour of polarizability in the frequency domain under resonant pumping. is represented in units of . We fix . White arrows denote frequency splittings of value .
We start by analyzing the dynamics of a single QE pumped by a laser, which will serve to build the macroscopic response from the ensemble of driven QEs. The ground and excited states of the QE are and , and the transition frequency is . Within the Rotating-Wave Approximation, the Hamiltonian of the pumped QE is [26, 27, 28, 29]
where the first term on the right-hand side corresponds to the bare QE Hamiltonian and the second term accounts for the interaction of the QE with the pump field. Here, represents the lowering operator of the QE, is the raising operator, and is the Rabi frequency. We neglect any losses because we aim to analyze the time-periodic response of the system, and thus focus on timescales short compared to the decay rates of the system. The unitary time evolution of the system is thus governed by the propagator [30], with the temporal ordering operator.
The evaluation of the time evolution of the pumped system is facilitated by moving to the rotating frame at the laser frequency through the unitary transformation , which yields the time-independent Hamiltonian in the rotating frame [26, 27], where is the laser detuning. As a result, the propagator that governs the time evolution can be decomposed as , with , which does not involve any time-integral. This decomposition can be interpreted as a Floquet decomposition, since the Hamiltonian is time-periodic [30, 31, 32, 33]. Diagonalizing yields the dressed states in the rotating frame,
where is the Stückelberg (mixing) angle [26, 27], defined by and . Here, is the generalized Rabi frequency, which corresponds to the energy splitting between and . In the Mollow regime, where the Rabi frequency dominates over the losses, the populations of the ground and excited bare states oscillate at frequency [34, 35, 36, 37] – known as Rabi oscillations – and the emission spectrum displays the characteristic Mollow triplet: a central peak at the pump frequency and two sidebands at [20, 21, 22, 23].
Our objective is to understand the electromagnetic response of such driven quantum system under a weak probe field.
For that purpose, we derive the Heisenberg evolution of the dipole operator, which will be used to obtain the polarizability via the Kubo formula [38, 39]. For simplicity, we assume that the pump field and the transition dipole moment are linearly polarized along the same direction, which we fix along the -axis. This allows us to work with the scalar dipole operator , where is the magnitude of the transition dipole moment. Computing , with , one obtains [30]
where we have introduced and the operators , , and , which are the lowering, raising and -Pauli matrices in the basis . Additionally, , , , and .
We are now in a position to derive the two-time polarizability of a single driven emitter in response to a perturbative probe field. For simplicity, we consider that the probe electric field is also polarized in the -direction. Within linear-response theory, the probe-induced polarizability is given by the Kubo formula [15], , where is the Heaviside step function and the expectation value is evaluated with respect to the initial state, which hereafter we fix to be the ground state . Using Eq. (4), we obtain
where
Equation (5) reveals that exhibits oscillatory dependence on both and , with contributions at the pump-laser and generalized Rabi frequencies. To further characterize , we investigate its behaviour in the frequency domain via Fourier transformation of Eq. (5). Figure 1(b) shows the absolute value of , for a resonantly driven QE (), with eV. We choose this value of the Rabi frequency for the sake of visibility, as the effect is scalable to any value of provided it remains much larger than the losses. From Figure 1(b), we identify two distinct spectral features. First, two manifolds of vertical lines, each split into three lines, which arise due to the dispersive response. Second, three manifolds of diagonal lines, separated by , each split into two lines separated by (white arrows). These diagonal lines correspond to fixed values of , revealing that the temporal modulation of the medium couples frequency components separated by some combination of and , a signature of a time-varying response [40, 41]. Analytical expressions for are derived in the Supplemental Material (SM) [42].
The electromagnetic response of a laser-dressed QE can be next extended to the macroscopic response of an homogeneous ensemble by assuming that all QEs are identical and non-interacting. The macroscopic polarization induced by the probe field is then obtained by multiplying the single-emitter polarizability by the emitter number density per unit volume .
In the following, we use Maxwell’s equations and the Floquet theorem to derive the dispersion relation of the electromagnetic modes supported by the pumped medium that can be excited by a perturbative probe field. To this end, we consider that these modes propagate in the -direction, with electric field, electric displacement field, magnetic field and polarization given by , , , and . The time evolution of these fields is governed by the Maxwell equations, together with the constitutive relation. Specifically, the Faraday and Ampère equations, respectively, yield
with the magnetic vacuum permeability. Considering a vacuum background for the QEs, the constitutive relation is given by , where the polarization can be obtained from the polarizability as
From this polarizability and the Maxwell’s equations in Eq. (6), the dispersion relation of these modes can be obtained.
To simplify the analytical treatment and the numerical calculation, we henceforth assume that the laser frequency is an integer multiple of the generalized Rabi frequency, . The exponential appearing in Eq. (5) then reduces to , where , reducing the number of indices from two to one. The polarization can thus be written as
where we have defined
as well as the auxiliary field variables
This auxiliary variable satisfies the differential equation
and plays a key role in constructing the Floquet eigenvalue problem, as they will serve to obtain a closed system of equations for the Floquet harmonics of the and fields in this time-varying, dispersive setup.
According to Floquet theory [1, 3], since the response function is periodic in time, we adopt the following harmonic expansions for the displacement field, the magnetic field, and the auxiliary variables:
where is the Floquet quasifrequency. Substituting these expansions into Eq. (6a), Eq. (6b), and Eq. (13), we obtain
These coupled equations constitute an eigenvalue problem in the Floquet quasifrequency, whose solution yields the allowed modes. In the SM [42], we describe the extension of this formulation to the general case in which is not necessarily an integer multiple of .
In the remainder of this Letter, we numerically solve the eigenvalue problem in Eq. (13) and analyze the resulting band structure. Specifically, we plot in Figure 2 the band structure obtained for QEs per , which is sufficient to observe the emergence of momentum gaps and consistent with experiments [43, 44, 45, 46, 47]. The remaining emitter and laser parameters are as in Figure 1(b) [48]. Black dots denote modes with purely real Floquet frequency, grey dots denote the real frequency part of modes with complex Floquet frequency, and red dots denote the corresponding imaginary part, shifted vertically by its real part for clarity. We focus on the range of Floquet frequencies around the pump frequency , in contrast with conventional analyses of periodically time-varying media, where the first Brillouin zone is centered at zero frequency. The band structure is thus centered precisely within the range where the Rabi splitting of the undriven polaritonic dispersion falls [49, 50, 51]. To aid the interpretation of the driven band structure, we overlay in Figure 2 the dispersion relation in the undriven case (blue dots) [52], which exhibits the characteristic upper and lower polariton branches separated by the Rabi splitting at . This polaritonic structure underlies the dispersive band structure of the driven medium.

Figure 2. Dispersion relation for a homogeneous ensemble of laser-dressed QEs. Modes with purely real Floquet frequency are represented by black dots, with the -axis value corresponding to . Modes with a complex Floquet frequency are represented by grey and red dots: the grey dots show , while the red dots show the value of shifted by . Blue dots correspond to the dispersion relation in the absence of a pump. The emitter and laser parameters are the same as in Figure 1(b), and QEs per . Green dashed boxes indicate the regions analyzed in Figure 3.
The driven band structure in Figure 2 organizes into two physically distinct regions in momentum space. In the -region centered around the pump momentum (), the coupling between positive- and negative-frequency Floquet replicas systematically gives rise to momentum gaps. In the outer -region ( and ), the coupling is weaker and momentum gaps do not generally open. We discuss these two regions in the following.
In the -region around the pump momentum, positive- and negative-frequency Floquet replicas interact giving rise to several momentum gaps of different origin. First, flat momentum gaps open at , arising from the parametric coupling of two photonic modes whose frequencies differ by [53] – for example, modes and . These gaps are the hallmark of parametric amplification of the probe assisted by the pump, resembling the conventional momentum gaps [1, 2, 3, 4]. To further characterize these momentum gaps, we compute the phase rigidity for each mode [24, 25], where and are the left and right eigenvectors of the Floquet matrix [Eq. (13)] for this mode. A zoom of one of these flat gaps is shown in Figure 3(a), where (purple in the color bar) marks exceptional points (EPs) at the band edges, consistent with the conventional case.

Figure 3. Phase rigidity around the momentum gaps. The color bar indicates the value of the phase rigidity, with purple indicating values close to 0. Red dots show the value of shifted by . The emitter and laser parameters are the same as in Figure 2.
Second, dispersive momentum gaps appear at and , arising from the interaction between polaritonic and photonic modes of opposite frequency sign [25, 54, 55, 56]. A zoom of one of these regions is shown in Figure 3(b). The smallest dispersive gaps connect the two sides of the central -region around the pump momentum, and stem from the coupling of photonic modes to a next-order replica of the polaritonic mode. The largest dispersive gaps bridge each of these two sides of the central -region with the outer -regions. In addition, broad flat momentum gaps appear at and . We note that gain at Mollow sideband frequencies has also been predicted [57] and observed experimentally [58, 59, 60] in laser-dressed quantum emitters, but in a steady-state regime where amplification manifests as spatial gain rather than temporal gain . EPs appear at the edges of all these gaps, as confirmed by the vanishing phase rigidity in Figure 3(b).
In the outer -region, momentum gaps are generally absent due to the weaker coupling between modes, as already mentioned. However, an exception arises when the polaritonic dispersion brings replicas of opposite frequency sign into resonance. This is illustrated in Figure 3(c), which shows a zoom around : the intersection of the photonic and polaritonic bands gives rise either to a momentum gap with EPs at its edges (around , where the interacting replicas carry frequencies of opposite sign) or to an avoided crossing (around , where they carry frequencies of equal sign). Altogether, these results highlight the rich band structure that emerges within the minimal quantum-mechanical model of an ensemble of laser-dressed quantum emitters.
In summary, we have derived the photonic band structure of a minimal quantum-mechanical model of a dispersive, time-varying medium that explicitly incorporates the source of the temporal modulation. The medium consists of a homogeneous ensemble of laser-dressed QEs, making our results applicable to a broad range of physical implementations, including organic molecules, atoms, quantum dots, and color centers. The resulting band structure exhibits qualitative differences from that predicted by models assuming ad-hoc, non-dispersive modulations of the permittivity, due to the interplay between the dispersive polaritonic and time-periodic responses of the laser-dressed QEs, and could find applications in the engineering of momentum gaps [7, 12, 25, 55, 56, 61, 62, 63]. This framework can serve to investigate light amplification and gain in driven systems beyond classical TVM treatments, shedding light on their potential as sources of quantum light. Next steps include a rigorous incorporation of losses, which may suppress the EPs of dispersive gaps without necessarily inhibiting amplification [25, 64], and going beyond the homogeneous approximation.
Acknowledgments
We thank Thomas F. Allard, Miguel Verde, and Jaime E. Sustaita-Osuna for helpful discussions. We have used AI tools (Claude) to optimize the numerical code used to solve the Floquet eigenvalue problem. A.J.D and P.A.H acknowledge funding from the European Union through the ERC Grant TIMELIGHT under GA101115792 and from the Spanish Ministry for Science, Innovation, and Universities - Agencia Estatal de Investigación (AEI) through Grant No. RYC2021-031568-I (Ramón y Cajal program). D.M.C. acknowledges support from MCINN projects PID2021-126964OB-I00 (QENIGMA), TED2021-130552B-C21 (ADIQUNANO), and PID2024-156077OB-I00 (DQU0TE), and the support by the project NPhoQs funded by Programa Fundamentos 2024 from the BBVA Foundation through the grant EIC24-1-17304. We acknowledge funding from the Spanish Ministry for Science, Innovation, and Universities - Agencia Estatal de Investigación (AEI) through Grants No. CEX2023001316-M (through the María de Maeztu program for Units of Excellence in Research and Developments)
Data availability. The data that support the findings of this manuscript will be available online.
Supplemental Material: Time-Varying Medium from an Ensemble of Laser-Dressed Quantum Emitters
Adrián Juan-Delgado; Diego Martín-Cano; Paloma A. Huidobro
Complete derivation of the polarizability
In this section, we provide further details on the analytical derivation of the expression presented in the main text for the time-domain polarizability [Eq. (4)] of the main text. Additionally, we derive the analytical expression for the frequency-domain polarizability .
Within linear-response theory, the time-domain polarizability is obtained via the Kubo formula [1]
where is the Heaviside step function, and is the transition dipole moment operator. The expectation value is taken with respect to the initial state and the operators are evolved in the Heisenberg picture with respect to the bare Hamiltonian including the pump field (equivalently, in the interaction picture with respect to a probe field). To compute this expression, we begin from the Heisenberg equation of motion for the dipole moment operator [Eq. (3)] of the main text, which is derived in Ref. [2],
where , is the pump-laser frequency, and is the generalized Rabi frequency, with the detuning between the emitter transition frequency and the pump-laser frequency. The operators , , and , are, respectively, the lowering, raising and -Pauli matrices in the basis [see Eq. (2)] of the main text. The coefficients are given by , , , and , where , and . For and positive (i.e., ), these coefficients reduce to , , and , while the dressed states (in the rotating frame) are simply , .
For an initial ground state , the two-time correlation function of the dipole moment operator becomes
The commutator in Eq. (20) is then given by
Substituting into Eq. (20), the resulting expression of the time-domain polarizability is
which can be written in the form of Eq. (5) of the main text by introducing
We next calculate the frequency-domain polarizability by computing the Fourier transform of , which is given by
Using the integral representation of the Heaviside step function [1], , where is a positive infinitesimal, we obtain that Eq. (26) becomes
We next group the terms oscillating at and , which yields
Applying the identity , we obtain
Evaluating next the integral over via the Dirac delta functions leads to
which can be written in the compact form
where we have defined
In the main text, Figure 1(b) is obtained by using Eq. (31), with .
Interestingly, we can reduce the number of indices in Eq. (31) from four to two, which facilitates the alternative derivation of the Floquet eigenvalue problem with as eigenvalue (Section SII of this Supplemental Material). To this end, we introduce and , so that
We have introduced and we have defined
The summation over and in Eq. (34) is restricted to values for which and both belong to . More specifically, if then runs only over ; if , over ; if , over ; if , over ; and if , only over . The same applies to and .
To further characterize the behaviour of the polarizability, we plot in Figure S1 the non-zero coefficients, normalized by . Specifically, we find that is only non-zero for (solid red line) and (dashed blue line), and for . Figure S1 and Figure S1 show the real part of for , associated with the antidiagonal central lines in Figure 1(b) of the main text, plotted as a function of and , respectively, so as to center the two relevant resonance pairs in each panel. In both cases, each curve displays two dispersive lineshapes with sharp sign transitions of width . For (solid red line), exhibits these transitions at and in panel (a), and at and in panel (b). For (dashed blue line), the transitions occur at and in panel (a), and at and in panel (b). Notably, the two curves for share one common resonance at , while the second resonance of each is displaced by , reflecting the sideband structure induced by the drive.

Figure S1. Normalized coefficients under resonant pumping. The index takes values (a,b) , (c) , and (d) . Red solid lines correspond to and dashed blue lines to . All other combinations of combinations give . We fix , eV, and eV.
Figure S1 shows for , plotted as a function of . For (solid red line), the sign transitions occur at and , and for (dashed blue line) they appear at and . In contrast to the case, the two dispersive features within each curve now have equal amplitudes, so that the curve is symmetric about and the curve is symmetric about .
Figure S1 shows for , plotted as a function of . For (solid red line), the sign transitions occur at and , and for (dashed blue line) they appear at and . As in panel (c), the two dispersive features within each curve have equal amplitudes, so that the curve is symmetric about and the curve is symmetric about .
Floquet eigenvalue problem with Floquet quasifrequency as eigenvalue
In this section, we derive in detail the Floquet eigenvalue problem with the Floquet quasifrequency as eigenvalue. We do so first by considering that the laser frequency is an integer multiple of the generalized Rabi frequency, as assumed in the main text, and then extend the formulation to the general case.
The laser frequency is an integer multiple of the generalized Rabi frequency
In this case, we start from the set of equations obtained in the main text for the Floquet harmonics , , and [see Eqs. (13a), (13b), and (13c) of the main text]. To cast the eigenvalue problem in matrix form, we build state vector with these harmonic coefficients, normalizing them so that all components share the same units and all matrix elements have units of frequency. To this end, we define
which have the same units as . With these definitions, the set of equations in Eq. (13) of the main text can be rewritten as
where is a dimensionless quantity.
The eigenvalue problem can then be written in matrix form as
where the vector is given by
with the column vector of normalized harmonic coefficients , the vector of harmonic coefficients , and , the vectors of harmonic coefficients and , respectively; the ellipses indicate the concatenation over all auxiliary-variable indices. The matrix takes the form
where we have introduced
Figure 3 of the main text is obtained by solving numerically the eigenvalues of . We note that the state vector has components, where is the number of Floquet harmonics retained in the truncated expansions of , and . Of these, only correspond to physical fields ( and ); the remaining components arise from the auxiliary variables introduced to close the system of equations, and give rise to spurious eigenmodes with no physical meaning. To filter them out, we compute for each eigenmode the weight of its eigenvector in the subspace:
Physical modes concentrate their weight in the subspace, while spurious modes have close to zero. We retain only modes exceeding a threshold of relative to the maximum value of across all modes.
General case: incommensurate laser and Rabi frequencies
In this section we derive the general eigenvalue problem without assuming that the laser frequency is an integer multiple of the generalized Rabi frequency. We start by substituting the time-domain polarizability Eq. (5) of the main text into the polarization in Eq. (9) of the main text, which yields
where we have defined , and the generalized auxiliary variables
which satisfy
We next adopt Floquet ansatzes for the displacement field, the magnetic field, and the auxiliary variables. Since both modulation frequencies appear explicitly, these read
Substituting these expressions into the Faraday equation [Eq. (6a) of the main text], the Ampère equation [Eq. (6b) of the main text], and Eq. (50), we obtain
We can then define the normalized quantities and , which have the same units as the magnetic field harmonic , and solve the eigenvalue problem. Notably, Eqs. (S27) show that, in the general incommensurate case, the matrix elements couple harmonics through two independent indices, making the construction and diagonalization of the matrix significantly more complex than in the commensurate case (Section SII.A.).
Alternative Floquet eigenvalue problem with momentum as eigenvalue
In this section, we describe an alternative derivation of the Floquet eigenvalue problem, casting in this case the squared momentum as eigenvalue. To this end, we extend the formulation presented in Ref. [1] to the case with two modulation frequencies in the response function, which arise as we have explicitly included the modulation source. In this case, we start from the wave equation for the electric field in the frequency domain, which can be written as
where we have assumed again a vacuum background for the emitters. Additionally, we have directly adopted a scalar description, since the transition dipole moment, pump field, and probe field are assumed to be linearly polarized along the same direction (e.g., in the -direction).
We next propose the following Floquet ansatz for the electric field in the time domain,
whose Fourier transform yields
Substituting into Eq. (57), the wave equation becomes
Performing the index shifts and , we obtain
This equation can be rearranged as
By reindexing through a one-to-one mapping, the above equation can be cast as a matrix eigenvalue problem. Specifically, we truncate the -manifold to dimension , so that , and similarly the -manifold to dimension , so that . The combined -manifold then has dimension , with , and the index runs over . The mapping from to is explicitly given by
The eigenvalue problem then takes the matrix form
where the matrix elements are
This formulation yields a smaller matrix problem than that obtained in Section SII, where the eigenvalue corresponds to , since it only requires solving for the harmonics of the electric field rather than those of the displacement field, the magnetic field, and all auxiliary variables, resulting in significantly shorter computation times. However, the electromagnetic modes obtained within the formulation with as eigenvalue include both evanescent and propagating solutions, making it numerically difficult to isolate the latter. This complicates the physical analysis of the band structure and, in particular, that of the dispersive momentum bandgaps. Nevertheless, we have verified that the results obtained with this formulation are consistent with those presented in the main text. Moreover, the calculation of the phase rigidity and the identification of the exceptional points (EPs) are naturally obtained by constructing the eigenvalue problem in rather than in . Indeed, EPs correspond to values of at which two eigenvalues coalesce together with their associated eigenvectors, so that the phase rigidity vanishes. When the eigenvalue problem is instead formulated in for fixed , such coalescence does not appear directly as a degeneracy of the eigenvalues and requires additional analysis to identify.
Band structure for different emitter densities
In this section, we analyze the emergence of the band structure in Figure 2 and Figure 3 of the main text as a function of the modulation strength. To this end, we investigate the band structure for different values of emitter density, from emitters per to the value emitters per in Figure 3 of the main text. The emitter and laser parameters are the same as those considered in the main text, i.e., , and . Additionally, to solve the eigenvalue problem numerically, we use the formulation discussed in Section SII.A of this Supplemental Material, which is the same formulation used in the calculation of the band structure in Figure 2 and Figure 3 of the main text. The color convention for black, grey, and red dots in Figure S2 is the same as that adopted in Figure 2 of the main text.

Figure S2. Dispersion relation for an ensemble of quantum emitters in the Mollow regime and different values of emitter density. We fix , and , as in the main text. The emitter density is fixed at (a) emitters per , (b) emitters per , (c) emitters per , and (d) emitters per , the same value as in Figure 3 of the main text. Modes with purely real Floquet frequency are represented by black dots, with the -axis value corresponding to . Modes with a complex Floquet frequency are represented by grey and red dots: the grey dots show , while the red dots show the value of shifted by . Cyan dashed line in (a) represent the light line dispersion relation. Cyan dots in [(b)-(d)] correspond to the polaritonic dispersion relation in the absence of a pump.
We start by setting emitters per in Figure S2, which is equivalent to setting a vanishing modulation strength. The permittivity thus reduces to the vacuum background permittivity and the dispersion relation to the vacuum light line (dashed cyan line). Within the periodic lattice approximation, this relation is replicated at the harmonic modulation frequencies [3, 4], leading to the dispersion relation shown by the black dots in Figure S2, obtained by solving the Floquet eigenvalue problem in Eq. (13) of the main text.
We next increase the emitter density to a value where small momentum bandgaps emerge. Specifically, we fix emitters per in Figure S2, a value ten times smaller than that in Figure 2 and Figure 3 of the main text. In Figure S2 we additionally plot the polaritonic dispersion relation obtained in the absence of pumping (cyan dots), which is given by , with , and which reduces to the light line in Figure S2 when . We observe that at emitters per a small polaritonic Rabi splitting already emerges in Figure S2, leading to the flat bands shown by the black dots in the outer region ( and ). Additionally, we observe flat momentum bandgaps at the frequencies of the Mollow triplet peaks, i.e., at and . However, this value of emitter density (equivalently, modulation strength) is not sufficient to clearly observe the flat momentum bandgaps at , nor the dispersive momentum bandgaps discussed in the main text.
We next fix emitters per m in Figure S2. The polaritonic Rabi splitting obtained in the absence of the pump (cyan dots) increases, and the slope of the bands obtained in the pumped case in the outer -region slightly increases with respect to Figure S2. Additionally, the momentum bandgaps at and the dispersive momentum bandgaps start to emerge. However, the flat momentum bandgaps at the frequencies of the Mollow triplet peaks again exhibit a larger imaginary part of the Floquet frequency (red dots).
Finally, for completeness, we plot in Figure S2 the dispersion relation at the same emitter density emitters per m as in Figure 2 and Figure 3 of the main text, where the momentum bandgaps are clearly visible. Thus, Figure S2 corresponds to Figure 2 of the main text.
References
References
- [1]J. R. Zurita-Sánchez, P. Halevi, and J. C. Cervantes-González, Physical Review A—Atomic, Molecular, and Optical Physics 79, 053821 (2009).
- [2]E. Galiffi, R. Tirole, S. Yin, H. Li, S. Vezzoli, P. A. Huidobro, M. G. Silveirinha, R. Sapienza, A. Alù, and J. B. Pendry, Advanced Photonics 4, 014002 (2022).
- [3]M. Asgari, P. Garg, X. Wang, M. Mirmoosa, C. Rockstuhl, and V. Asadchy, Advances in Optics and Photonics 16, 958 (2024).
- [4]A. Boltasseva, V. Shalaev, and M. Segev, Optical Materials Express 14, 592 (2024).
- [5]J. Reyes-Ayona and P. Halevi, Applied Physics Letters 107 (2015).
- [6]E. Galiffi, Y.-T. Wang, Z. Lim, J. B. Pendry, A. Alù, and P. A. Huidobro, Physical Review Letters 125, 127403 (2020).
- [7]J. Park, H. Cho, S. Lee, K. Lee, K. Lee, H. C. Park, J.-W. Ryu, N. Park, S. Jeon, and B. Min, Science advances 8, eabq6220 (2022).
- [8]J. Glazola-Luna and P. Halevi, Applied Physics Letters 122 (2023).
- [9]X. Wang, M. S. Mirmoosa, V. S. Asadchy, C. Rockstuhl, S. Fan, and S. A. Tretyakov, Science advances 9, eadg7541 (2023).
- [10]S. Horsley, E. Galiffi, and Y.-T. Wang, Physical Review Letters 130, 203803 (2023).
- [11]E. Galiffi, A. C. Harwood, S. Vezzoli, R. Tirole, A. Alù, and R. Sapienza, Nature Photonics 20, 163 (2026).
- [12]B. Huang, Z. Zhu, G. Yu, and Z. Gao, arXiv preprint arXiv:2604.17408 (2026).arxiv.org/abs/2604.17408
- [13]Z. Zhu, B. Huang, S. Xu, J. Chen, Y. Meng, Z. Zhu, X. Xi, and Z. Gao, arXiv preprint arXiv:2512.16516 (2025).arxiv.org/abs/2512.16516
- [14]M. Lyubarov, Y. Lumer, A. D ikopoltsev, E. Lustig, Y. Sharabi, and M. Segev, Science 377, 425 (2022).
- [15]J. Sloan, N. Rivera, J. D. Joannopoulos, and M. Soljačić, ACS Photonics 11, 950 (2024).
- [16]J. Bae, K. Lee, B. Min, and K. W. Kim, Nature com- munications 17, 858 (2025).
- [17]J. Park, K. Lee, R.-Y. Zhang, H.-C. Park, J.-W. Ryu, G. Y. Cho, M. Y. Lee, Z. Zhang, N. Park, W. Jeon, et al., Physical Review Letters 135, 133801 (2025).
- [18]J. E. Sustaeta-Osuna, T. F. Allard, F. J. García-Vidal, and P. A. Huidobro, Physical Review Letters 136, 136903 (2026).
- [19]Y. Kim, K. Lee, K. W. Kim, and B. Min, Current Optics and Photonics 10, 245 (2026).
- [20]B. Mollow, Physical Review 188 (1969).
- [21]F. Schuda, C. Stroud Jr, and M. Hercher, Journal of Physics B: Atomic and Molecular Physics 7, L198 (1974).
- [22]H. Walther, in Laser Spectroscopy: Proceedings of the Second International Conference. Megève, June 23–27, 1975 (Springer, 1975) pp. 358–369.
- [23]F. Wu, R. Grove, and S. Ezekiel, Physical Review Letters 35, 1426 (1975).
- [24]J. Wiersig, Physical Review Research 5, 033042 (2023).
- [25]T. F. Allard, J. E. Sustaeta-Osuna, F. J. García-Vidal, and P. A. Huidobro, Physical Review Letters 136, 106903 (2026).
- [26]D. Steck, Quantum and atom optics (2007).
- [27]A. Juan-Delgado, Photon statistics and entanglement from two interacting quantum emitters, Ph.D. thesis, Universidad del País Vasco-Euskal Herriko Unibertsitatea (2025).
- [28]C. Gerry, P. Knight, and P. Knight, Introductory quantum optics (Cambridge university press, 2005).
- [29]R. Loudon, The quantum theory of light (OUP Oxford, 2000).
- [30]A. Juan-Delgado and A. Chenu, Physical Review A 104, 022219 (2021).
- [31]J. H. Shirley, Physical Review 138, B979 (1965).
- [32]Y. B. Zel’dovich, JETP 24, 1006 (1967).
- [33]K. Szczygielski, Linear Algebra and its Applications 609, 176 (2021).
- [34]C. Cohen-Tannoudji and S. Reynaud, Journal of Physics B: Atomic and Molecular Physics 10, 345 (1977).
- [35]J. Dalibard and C. Cohen-Tannoudji, JOSA B 2, 1707 (1985).
- [36]E. Del Valle and F. Laussy, Physical review letters 105, 233601 (2010).
- [37]A. Juan-Delgado, Óptica Pura y Aplicada 59, 1 (2026).
- [38]R. Kubo, Journal of the physical society of Japan 12, 570 (1957).
- [39]R. Kubo, M. Yokota, and S. Nakajima, Journal of the Physical Society of Japan 12, 1203 (1957).
- [40]M. S. Mirmoosa, T. Koutserimpas, G. Ptitsyn, S. Tretyakov, and R. Fleury, New Journal of Physics 24, 063004 (2022).
- [41]M. Verde and P. A. Huidobro, Physical Review Research 8, 023073 (2026).
- [42]See Supplemental Material, which includes Refs. [2, 3, 15, 30], for the complete derivation of the analytical expression of the polarizability; the complete derivation of the Floquet eigenvalue problem with ωF as eigenvalue, for the case in which ωL is an integer multiple of Ωr; the extension of this eigenvalue problem to the general case in which ωL is not an integer multiple of Ωr; an alternative derivation of the Floquet eigenvalue problem with k² as eigenvalue; and the analysis of the band structure for different values of Nd.
- [43]V. M. Acosta, E. Bauch, M. P. Ledbetter, C. Santori, K.-M. Fu, P. E. Barclay, R. G. Beausoleil, H. Linget, J. F. Roch, F. Treussart, et al., Physical Review B—Condensed Matter and Materials Physics 80, 115202 (2009).
- [44]N. B. Manson, M. Hedges, M. S. Barson, R. Ahlefeldt, M. W. Doherty, H. Abe, T. Oshima, and M. J. Sellars, New Journal of Physics 20, 113037 (2018).
- [45]M. Auzinsh, A. Berzins, D. Budker, L. Busaitė, R. Ferber, F. Gahbauer, R. Lazda, A. Wickenbrock, and H. Zheng, Physical Review B 100, 075204 (2019).
- [46]S. Skoff, R. Hendricks, C. Sinclair, J. Hudson, D. Segal, B. Sauer, E. Hinds, and M. Tarbutt, Physical Review A—Atomic, Molecular, and Optical Physics 83, 023418 (2011).
- [47]T. Asano, Z. Fang, and A. Madhukar, Journal of Applied Physics 107 (2010).
- [48]To determine the value of the dipole moment, we use the spontaneous emission formula d = √(3πγε₀ℏc³/ω₀³), which connects the dipole moment d of a two-level QE to its transition frequency ω₀ and decay rate γ. We set ℏγ = 10⁻⁷ eV, a value representative of the losses in organic molecules at cryogenic temperatures, which also justifies neglecting dissipation in the derivation of the Floquet eigenvalue problem.
- [49]H. Kramers and W. Heisenberg, Master of Modern Physics, Princeton Series in Physics (Princeton University Press, Princeton, 1998) 121, 9780691219523 (1925).
- [50]R. Loudon and S. M. Barnett, Journal of Physics B: Atomic, Molecular, and Optical Physics 39, S555 (2006).
- [51]P. R. Berman, R. W. Boyd, and P. W. Milonni, Physical Review A—Atomic, Molecular, and Optical Physics 74, 053816 (2006).
- [52]This dispersion relation is given by k = (ωF/c)√(1 + Nd/ε₀ α₀(ω)), with α₀(ω) = d²/ℏ [1/(ω+ω₀) − 1/(ω−ω₀)].
- [53]D. Holberg and K. Kunz, IEEE Transactions on Antennas and Propagation 14, 183 (1966).
- [54]F. Feng, N. Wang, and G. P. Wang, Applied Physics Letters 124 (2024).
- [55]M. G. Ozlu, V. Mkhitaryan, C. B. Fruhling, A. Boltasseva, and V. M. Shalaev, Physical Review Research 7, 023214 (2025).
- [56]Y.-T. Wang and Y.-H. Chen, Applied Physics Letters 126 (2025).
- [57]B. Mollow, Physical Review A 5, 2217 (1972).
- [58]F. Y. Wu, S. Ezekiel, M. Ducloy, and B. Mollow, Physical Review Letters 38, 1077 (1977).
- [59]M. T. Grunaisen, K. R. MacDonald, A. L. Gaeta, R. W. Boyd, and D. J. Harter, Physical Review A 40, 3464 (1989).
- [60]A. Maser, G. Gmeiner, T. Utikal, S. Götzinger, and V. Sandoghdar, Nature Photonics (2016).
- [61]X. Wang, P. Garg, M. S. Mirmoosa, A. G. Lampianidis, C. Rockstuhl, and V. S. Asadchy, Nature Photonics 19, 149 (2025).
- [62]A. Caballero, T. F. Allard, and P. A. Huidobro, ACS Photonics 13, 3012 (2026).
- [63]M. Salehi, M. Ciabattoni, and F. Monticone, arXiv preprint arXiv:2604.13444 (2026).arxiv.org/abs/2604.13444
- [64]N. Bernier, L. Tóth, A. Feofanov, and T. Kippenberg, Physical Review A 98, 023841 (2018).