Arxiv:2003.13695V3 [Cond-Mat.Mes-Hall] 15 Sep 2020

Quantum Electrodynamic Control of Matter: Cavity-Enhanced Ferroelectric Phase Transition

Yuto Ashida∗ Department of Applied Physics, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan

Atac¸ I˙mamo˘glu and Jer´ omeˆ Faist Institute of Quantum Electronics, ETH Zurich, CH-8093 Zurich,¨ Switzerland

Dieter Jaksch Clarendon Laboratory, University of Oxford, Parks Road, Oxford OX1 3PU, United Kingdom

Andrea Cavalleri Max Planck Institute for the Structure and Dynamics of Matter, 22761 Hamburg, Germany and Clarendon Laboratory, University of Oxford, Parks Road, Oxford OX1 3PU, United Kingdom

Eugene Demler Department of Physics, Harvard University, Cambridge, MA 02138, USA

The light-matter interaction can be utilized to qualitatively alter physical properties of materials. Recent theoretical and experimental studies have explored this possibility of controlling matter by light based on driving many-body systems via strong classical electromagnetic radiation, leading to a time-dependent Hamiltonian for electronic or lattice degrees of freedom. To avoid inevitable heating, pump-probe setups with ultrashort laser pulses have so far been used to study transient light-induced modifications in materials. Here, we pursue yet another direction of controlling quantum matter by modifying quantum fluctuations of its electromagnetic environment. In contrast to earlier proposals on light-enhanced electron-electron interactions, we consider a dipolar quantum many-body system embedded in a cavity composed of metal mirrors, and formulate a theoretical framework to manipulate its equilibrium properties on the basis of quantum light-matter interaction. We analyze hybridization of different types of the fundamental excitations, including dipolar phonons, cavity photons, and plasmons in metal mirrors, arising from the cavity confinement in the regime of strong light-matter interaction. This hybridization qualitatively alters the nature of the collective excitations and can be used to selectively control energy-level structures in a wide range of platforms. Most notably, in quantum paraelectrics, we show that the cavity-induced softening of infrared optical phonons enhances the ferroelectric phase in comparison with the bulk materials. Our findings suggest an intriguing possibility of inducing a superradiant-type transition via the light-matter coupling without external pumping. We also discuss possible applications of the cavity-induced modifications in collective excitations to molecular materials and excitonic devices.

I. INTRODUCTION who observed that coherent microwave radiation leads to an increase of the critical current in superconductors. This phenomenon has been understood from the Eliashberg electron- One of the central goals in both quantum optics and con- photon theory [5–7], in which photons with frequencies be- densed matter physics is to understand macroscopic phenom- low the quasiparticle threshold are shown to modify the dis- ena emerging from the fundamental quantum degrees of free- tribution of quasiparticles in a way that the superconducting dom. Historically, quantum optics strived to study light- order parameter is enhanced. Other seminal works studying matter interactions mainly in few-body regimes [1]. On an- control of matter by light include Floquet engineering of elec- other front, condensed matter physics investigated collective tronic band structures [8–12], and photo-induced pumping by phenomena of many-body systems, and largely focused on op- ultrashort laser pulses for creating metastable nonequilibrium timizing structural and electronic phases of solids. In partic- states [13–28]. Nonlinear optical phenomena can also be un- ular, a number of techniques, including applied strain, elec- derstood from the perspective of light-induced changes in op- tronic doping, and isotope substitution, have been used to en- tical properties of matter [29–31]. For instance, the light- hance a macroscopic order such as ferroelectricity or super- induced transparency can be considered as strong modifica- conductivity. As an interdisciplinary frontier at the intersec- arXiv:2003.13695v3 [cond-mat.mes-hall] 15 Sep 2020 tion of the refractive index and absorption coefficient by a tion of optics and condensed matter physics, there have re- control optical beam [32], and the optical phase conjugation cently been remarkable developments in using classical elec- phenomena arise from the resonant parametric scattering of tromagnetic radiation to control transient states of matter [2]. signal photons and matter excitations [33, 34]. The essential The aim of this paper is to explore yet another route towards element of all these developments is an external pump that controlling the phase of matter by quantum light, namely, by provides classical light field acting on quantum matter. modifying quantum fluctuations of the electromagnetic field in equilibrium – in the absence of an external drive. An alternative approach to controlling matter by light is to Studies of light-induced modifications in material proper- utilize quantum fluctuations of the vacuum electromagnetic ties date back to experiments by Dayem and Wyatt [3,4], field. A prototypical model to understand such quantum light- 2 matter interaction is the Dicke model, which describes an en- semble of two-level atoms coupled to a single electromagnetic mode of a cavity [35–40]. For a sufficiently strong electric dipole, Dicke has shown that one should find instability into the superradiant phase characterized by spontaneous atomic polarization and macroscopic photon occupation [35]. This transition originates from the competition between the decrease of the total energy due to the light-matter coupling and the energy cost of adding photons to the cavity and admixing the excited state of atoms. The Dicke model is exactly solv- able via the Bethe-ansatz equations originally discussed in the central spin model [41–43], and the existence of the superradiant transition has been rigorously proven in the thermodynamic limit [37, 38]. Subsequently to Dicke’s original work, it has been pointed out that the superradiant transition requires such strong interaction that one needs to include the Aˆ2 term in the Hamilto- nian, where Aˆ denotes the vector potential of electromagnetic fields. In turn, the latter term has been shown to prevent the FIG. 1. (a) Schematic phase diagram of a quantum paraelectric ma- transition [44, 45]. Even though ideas have been suggested to terial. The horizontal axis corresponds to a tuning parameter, such circumvent this no-go theorem by using microwave resonators as pressure or isotope substitution, which controls the transition be- [46, 47] or including additional inter-particle interactions [48– tween the ferroelectric and paraelectric phases. The vertical axis is 53], the problem is still a subject of debate [54–58]. An alter- temperature/energy and the red dashed line in the paraelectric phase native proposal using multilevel systems with external pump shows the optical phonon energy. Note that the phonon softens to fields [59, 60] has been experimentally realized [61, 62]. Ana- zero energy at the paraelectric-to-ferroelectric transition point. The logues of the Dicke transition have also been explored in ultra- main goal of this paper is to demonstrate that this transition can also cold atoms, where matter excitations correspond to different be controlled by changing the electromagnetic environment of the system, in particular, by placing a thin film of material in a cavity. momentum states of atoms [63, 64] or hyperfine states in spin- (b,c) Ferroelectric and paraelectric phases in the case of ionic crys- orbit-coupled Bose gases [65]. In all these realizations, exter- tals. Green (blue) circles indicate the ions with the positive (neg- nal driving is essential to obtain the superradiant-type phases. ative) charges. The ferroelectric phase in (b) is characterized by a The present work in this context suggests a promising route nonzero uniform displacement causing an electric polarization (red towards realizing a genuine equilibrium superradiant transi- arrows) while the mean-phonon displacement is absent in the para- tion under no external drive, which has so far remained elu- electric phase in (c). The variables φ and Ω represent the phonon sive. Our consideration is motivated by an observation that field and the bare phonon frequency, respectively (see Sec. III). transition into a superradiant state should be easier to attain in a system naturally close to a phase with spontaneous electric polarization. For instance, materials such as SrTiO3 and varying electromagnetic fields and phonon excitations must KTaO3 remain paraelectric down to zero temperature, but fer- be included. This should be contrasted to most of the previ- roelectric phase can be induced by applying pressure or iso- ous studies of the superradiant transitions, which have focused tope substitution. The primary goal of this paper is to inves- on a simplified case of either many-body systems coupled to tigate cavity-induced changes in a quantum paraelectric at the a single spatially uniform mode in a cavity, or two-level sys- verge of the ferroelectric order (see Fig.1(a-c)). We formu- tems coupled to a single mode obtained after truncating high- late a simple yet general theoretical framework for analyz- lying spatially varying modes [48]. The second key element ing hybridization of matter excitations, such as infrared active of our analysis is including nonlinearity of dipolar phonons phonons, and light modes confined in systems where metallic in materials. In the bulk geometry, the importance of such cladding layers act as cavity mirrors (see Fig.2). Resulting nonlinearities for understanding the quantum phase transition polariton excitations consist of infrared active phonons in the has been discussed in Ref. [66]. In contrast, we focus on the quantum paraelectric, electromagnetic fields in the cavity, and cavity geometry and demonstrate that a combination of non- plasmons in the metallic electrodes. We show that the spec- linearities and cavity confinement leads to substantial phonon trum of such cavity polaritons is qualitatively different from softening. This softening in turn enhances the superradiant the bulk spectrum, and can exhibit significant mode-softening, transition, thus opening a way to controlling quantum matter indicating enhancement of the ferroelectric instability. by modifying electromagnetic vacuum. Two essential elements of our analysis place it outside the Before concluding this section, we put our work in a recently discussed no-go theorems for the superradiant transi- broader context of studies on quantum matter strongly inter- tion [44, 45, 55, 56, 58]. Firstly, we include multiple modes acting with electromagnetic fields. Importantly, our model for both light and matter degrees of freedom. More precisely, bears a close connection with physical systems discussed we assume systems to be large enough in the direction parallel in several areas, including cavity quantum electrodynamics to the interfaces in such a way that a continuum of spatially (QED), plasmonics, and polaritonic chemistry. Firstly, in the 3

curring at the metal-insulator interfaces, i.e., a finite plasma frequency of the metallic cavity mirrors. We demonstrate that the proper consideration of these aspects is important to accurately understand several key features in the strong light- matter coupling regime, which are not readily attainable in the conventional bulk settings. The resulting cavity-induced changes provide a general route towards controlling energy- level structures not only in paraelectric materials, but also in many other systems such as molecular solids, organic light- emitting devices, and excitonic systems. Our analysis can further be extended to include fabricated surface nanostructures [133, 134]. In view of the versatility of the present formulation, we anticipate that our results will be useful for advancing our understanding of a variety of physical systems lying at the intersection of condensed matter physics, quantum optics, and FIG. 2. (Top) Setup of the heterostructure cavity configuration quantum chemistry. considered in the present manuscript. A slab of paraelectric mate- The remainder of the paper is organized as follows. In rial with a finite thickness is sandwiched between two semi-infinite Sec.II, we summarize our main results at a nontechnical level. metallic electrodes. Right panel represents spatial profiles of typical In Sec. III, we present a theoretical framework for describ- energy modes along the spatial direction perpendicular to the metal- ing hybridized collective modes in both bulk and cavity set- insulator interfaces. (Bottom) Schematic figure illustrating three fun- tings. As a first step, we neglect phonon nonlinearities and damental excitations relevant to the theory. A phonon excitation de- determine dispersion relations of elementary excitations. In scribes an optical collective mode induced by, e.g., the deformation Sec.IV, we include phonon nonlinearities based on the vari- of ionic crystals around the equilibrium configuration. More gen- ational approach and analyze their effects on the paraelectric- erally, it can also represent dipole moments of molecular solids or assembled organic molecules. The interaction between phonons and to-ferroelectric phase transition. We show that the mode con- cavity photons is mediated by the dipole coupling, whose strength finement in the cavity geometry induces significant softening is represented by g. A plasmon, a collective excitation of electronic of polariton modes, indicating the enhanced propensity for gases in metal mirrors, leads to the electromagnetic coupling charac- ferroelectricity. In Sec.V, we present a summary of results terized by the plasma frequency ωp. and suggest several interesting directions for future investiga- tions.

field of cavity QED, one is typically interested in small quantum systems coupled to cavity photons [35, 67–72]; promi- II. SUMMARY OF MAIN RESULTS nent examples include individual atoms [73–75], quantum dots [76–81], and light-emitting defects in solids such as In this section, we summarize the main results of the paper nitrogen- and silicon-vacancy centers [82–88]. These systems at a nontechnical level before presenting a detailed theoreti- have been suggested as platforms for realizing single-photon cal formulation in subsequent sections. Our most important transistors [89–91] and sources [92–95], and can thus pro- contribution is a theoretical proposal of a new approach for vide a promising route to enabling photonic quantum infor- controlling many-body states of matter using quantum light mation processing [96–98]. Achieving a strong light-matter that does not rely on external pumping. Physical phenomenon coupling has also been the subject of intense research in the that underlies this proposal is the cavity-induced softening of fields of plasmonics in nanostructures [99–101] and polari- a dipolar phonon mode, which originates from the interplay tonic chemistry [102–107]. In both areas, cavity setups hold between phonon nonlinearities and modifications of the hy- promise for realizing systems with a broad range of inter- brid light-matter collective excitations in the cavity geometry esting physical properties, including superconductivity [108– (see Fig.3(a,b)). Softening of the phonon frequency to zero 110], charge and energy transport [111–118], hybridized ex- indicates a phase transition into the state with spontaneous citations in molecular crystals [119–122] and light-harvesting symmetry breaking and formation of electric dipolar moments complexes [123–126], chemical reactivity of organic com- (red dashed line in Fig.1(a)). Thus, our analysis demonstrates pounds [127–129], and energy transfer via phonon nonlinear- that cavity confinement shifts the transition point in favor of ity [130]. Strong coupling between the zero-point fluctuations the ferroelectric phase, i.e., the paraelectric-to-ferroelectric of the electromagnetic field (i.e., the vacuum fluctuations) and transition occurs at a lower applied pressure/isotope substi- vibrational excitations of individual molecules has been ex- tution in the cavity geometry as compared to the bulk case perimentally realized [131], and its nontrivial influence on the (Fig.3(c)). Said differently, it is possible to start with a ma- superconductivity has recently been reported [132]. terial that is in the symmetry-unbroken paraelectric phase in a In this context, one of the main novelties of the present bulk sample, yet will be in the ferroelectric phase when placed work is a general formalism that allows one to include both inside the cavity (black downarrow in Fig.3(c)). the dispersive nature of polariton excitations in the dielectric The main focus of this paper is on many-body systems slab of finite thickness and the surface plasmonic effects oc- that are naturally close to the superradiant-type transition, i.e., 4 paraelectric materials close to the ferroelectric phase. For in- perature and thermal fluctuations at finite temperature. Plac- stance, SrTiO3 and KTaO3 are paraelectric at ambient con- ing the focus on equilibrium systems makes the present work ditions, but isotope substitution of 18O for 16O or applying distinct from the previous studies on materials subject to clas- pressure can bring the materials into a phase with spontaneous sical light waves, in which the use of strong external drive is electric polarization [135]. The key property of such materi- essential. It is such quantum aspect of light-matter interaction als relevant to our analysis is the existence of a low-energy in- that motivates us to ask the two questions: frared optical phonon mode. To understand collective modes of such systems in resonator settings shown in Fig.2, we need (A) How does the light-matter coupling modify polariton to consider hybridization of the electric dipole moment of modes of paraelectric materials in the cavity geometry? the optical phonon with the quantized electromagnetic modes confined in the cavity and with plasmons in metal mirrors. (B) Is it possible to enhance the ferroelectric instability by modifying quantum electromagnetic environment of a Our analysis of quantum light-matter coupling has implica- many-body system under no external pump? tions for understanding both quantum fluctuations at zero tem- To address these questions, we formulate a simple yet generic theoretical framework for analyzing interaction between three types of excitations: infrared active phonons, cavity photons, and plasmons in metal mirrors. We focus on the heterostructure, in which energies can be confined in vol- umes much smaller than the free-space wavelengths of photons, leading to strong enhancement of light-matter interaction [136]. Strong hybridization between the continuum of electromagnetic fields and matter excitations necessitates deriv- ing new collective modes starting from the microscopic model and not relying on modes defined in the bulk geometry. In our analysis, the size of the cavity in the longitudinal direction is assumed to be large enough for collective eigenmodes to be characterized by a conserved in-plane momentum. Thus, the present model is distinct from most of the previous studies on the superradiant phases, where only one or at most a few modes in a cavity have been included. In the ultrastrong coupling regime, it is essential to accurately take into account the full continuum of modes when analyzing thermodynamic properties of the system. The careful inclusion of all the proper modes becomes particularly important when the phonon nonlinearity is included because, as discussed below, the frequency of every mode is affected by quantum and thermal fluctuations of the other modes (see Sec.IV for further details). In view of the importance of identifying the correct cavity eigenmodes, we begin our analysis with a detailed discussion of the quadratic theory to answer the first question (A). Inter- FIG. 3. Summary of the cavity-induced changes of the collective ex- estingly, even when the system is far from the transition point, citations and the phase diagram. (a) Bulk band dispersions. The hy- the cavity confinement can qualitatively alter the nature of col- bridization between a photon and an optical phonon at the frequency lective excitations (see e.g., Fig.6 in Sec. III). Building upon Ω (black dashed lines) leads to two gapped branches correspond- this, we proceed to address the second question (B) and show ing to the upper and lower transverse phonon polaritons (black solid that, after including phonon nonlinearities in the analysis, the curves). (b) Collective excitations confined in the cavity as a function cavity polaritons can exhibit considerable mode softening in of the in-plane wavevector q. The hybridization leads to softening of comparison to their analogues in bulk materials. This soften- the optical phonon modes (red curves and black solid arrow) and also ing ultimately induces the paraelectric-to-ferroelectric phase supports the coupled surface modes (blue curve). (c) Phase diagram transition when the frequency of the lowest energy mode goes in the cavity geometry. As the tuning parameter r is decreased, the transition from the paraelectric phase (PE) to the ferroelectric phase to zero (see Fig.3(b)). (FE) occurs when the renormalized effective phonon frequency goes We note that the phenomenon of cavity enhancement of the down to zero. The FE is enhanced in the cavity setting in comparison ferroelectric phase is not present at the level of a quadratic with the bulk geometry, for which the transition point is indicated by model, and can only be found when phonon nonlinearities are the black dashed line. The inset shows the dependence on the plasma included in the analysis. The phonon nonlinearities in gen- frequency ωp at the insulator thickness d = 1. The parameters are eral render the formation of a spontaneous dipole moment c = ~ = β = 1, ωp = 5, U = 1, g = 3.5, and the UV momentum energetically unfavorable. We show that the cavity confine- 2 cutoff is Λc = 10 . ment alleviates this adverse effect of phonon-phonon inter- 5 actions, thus allowing one to induce ferroelectricity in quan- where the first (second) term describes the dynamics of the tum paraelectrics that otherwise exhibit no spontaneous po- electromagnetic field (dipolar matter field), and the last term larizations even at zero temperature. We demonstrate that represents the light-matter coupling between these two fields. the phenomenon of cavity-enhanced ferroelectricity is more We specify each of those terms below. pronounced when confinement effects are enhanced by using a cavity with a thinner slab of paraelectric material and/or metallic mirrors with a higher plasma frequency (see 1. Electromagnetic field Fig.3(c)). A few remarks are in order. While our theory is formu- The free evolution of the electromagnetic field is described lated for a heterostructure based on a paraelectric sandwiched by the Hamiltonian between metallic mirrors, the present approach of controlling " # material excitations via modifying quantum electromagnetic Z ˆ 2 2 2 ˆ 3 Π 0c ˆ environment can readily be generalized to a wide range of Hlight = d r + ∇ × A , (2) 20 2 platforms. To demonstrate this point, we discuss several other potential applications of such cavity-based control of polari- where 0 is the vacuum permittivity, c is the vacuum light ton modes, including molecular materials and excitonic de- speed, Aˆ (r) is the vector potential of photons, and Πˆ (r) is vices. Thus, our analysis may find practical applications in its canonically conjugate variable. The integral is taken over a improving the efficiency of organic light-emitting devices and cubic whose volume is V = L3 and we choose the Coulomb realizing photon-exciton converters (see Sec.V and App.F). gauge We also consider the role of dissipation on the polariton softening and phase transition using a standard Drude-type model ∇ · Aˆ = 0. (3) of conductivity in metallic mirrors. Our formalism can be extended to include other mechanisms of dissipation. Assuming the periodic boundary conditions and introducing discrete wavenumbers ki = 2πni/L with i = x, y, z and ni being integer numbers, we represent the vector fields in terms of their Fourier components as III. PHONON-PHOTON-PLASMON HYBRIDIZATION r X ~ ik·r Aˆ(r) = aˆkλe kλ + H.c. , (4) Our main goal is to understand the qualitative physics of a 20V ωk kλ dipolar insulator coupled to quantized electromagnetic fields r in a cavity rather than to make predictions specific to partic- X 0ωk Πˆ (r) = −i ~ aˆ eik·r − H.c. , (5) ular experimental setups. Thus, we consider a generic micro- 2V kλ kλ scopic model and will provide order of magnitude estimates kλ for a concrete system at the end of our analysis in Sec.IV. where ωk = c |k| is the vacuum photon dispersion, and aˆkλ We here begin by introducing a general theory of the light- † (aˆkλ) is the annihilation (creation) operator of photons with matter interaction and formulate a framework for obtaining momentum k and polarization λ. The Coulomb gauge (3) hybridized eigenmodes in a way that allows us to quantize the leads to the transversality condition of the orthonormal polar- fields. Throughout this paper, we will assume that phonon ization vectors kλ: frequencies are the same for all phonon polarizations, while generalization of our analysis to anisotropic systems may in- k · kν = 0, (6) troduce additional interesting features (see e.g., Ref. [137]). † kν = δλν . (7) In the next section, we will apply the framework to an ionic kλ crystal confined in the cavity as a concrete case and analyze The Hamiltonian can be diagonalized in the basis of momen- its paraelectric-to-ferroelectric phase transition. tum k and polarization λ as ˆ X † Hlight = ~ωkaˆkλaˆkλ, (8) A. Bulk kλ where we omit the vacuum energy. Before moving to the case of the cavity setting, we first discuss the three-dimensional bulk setup with no boundary effects and determine the bulk dispersion relations for the 2. Matter field phonon-photon hybridized modes, i.e., the phonon polaritons. Later, we generalize the theory to the cavity configuration by We consider a dipolar insulator whose excitation is repre- taking into account confinement effects and also contributions sented by a real-valued vector field φˆ(r). For instance, it can from plasmons in metal mirrors. originate from infrared optical phonon modes corresponding The Hamiltonian of the bulk system is given by to vibrations of ions around the equilibrium configuration in ionic crystals such as MgO, SiC and NaCl [138]. In other se- Hˆtot = Hˆlight + Hˆmatter + Hˆl−m, (1) tups, it can represent collective dipolar modes associated with 6 assembled organic molecules, molecular crystals, and exci- where Z∗e is the effective charge of a dipolar mode, which tonic systems. For the sake of simplicity, we neglect the bare corresponds to the charge of ions in the case of ionic crystals, dispersion of the dipolar mode, which is a common assump- and VC is the Coulomb potential term whose explicit form is tion in the analysis of optical phonons, and represent its fre- specified below. We note that, aside a constant factor, the con- quency by Ω. The effective Hamiltonian of the matter field is jugate field πˆ physically corresponds to the current operator ˆ thus given by j ∝ ∂tφ associated with the dipolar field φ. We here invoke neither the dipole approximation nor the ro- Hˆ = Hˆ + Hˆ , (9) matter 0 int tating wave approximation, which are often used in literature, where Hˆ0 is the quadratic part but can break down especially in strong-coupling regimes Z 3 2 [89, 140]. The light-matter interaction (13) including both d r πˆ 1 2 ˆ2 ˆ 2 Hˆ0 = + MΩ φ , (10) πˆ · A and A terms allows one to explicitly retain the gauge v 2M 2 invariance of the theory. It is also noteworthy that the light- matter coupling (13) is nonvanishing even in the absence of and Hînt is the most relevant interaction photons. This interaction with the vacuum electromagnetic Z d3r 2 Hˆ = U φˆ · φˆ . (11) field originates from the zero-point fluctuations of the electric int v field and can be viewed as the light-matter interaction mediated by virtual photons. In this paper, we focus on such a low ˆ ˆ Here, π(r) is a conjugate field of the matter field φ(r), v is photon, quantum regime with no external pump. the unit-cell volume in the insulator, M is the effective mass, Our choice of the Coulomb gauge leads to the nondynam- which corresponds to the reduced mass of two ions in the ical contribution VC resulting from the constraint relating the case of ionic crystals, and U ≥ 0 characterizes the interaction longitudinal component of the electric field to that of the mat- strength. Physically, the interaction term can originate from, ter field. To see this explicitly, we first decompose the electric e.g., phonon anharmonicity. When we will discuss in SecIV field in terms of the transverse and longitudinal components the transition between the paraelectric phase with hφi = 0 as and the ferroelectric phase with hφi= 6 0, the phonon frequency Ω2 should be considered as the tuning parameter gov- Eˆ = Eˆ⊥ + Eˆk, (14) erning the transition in the effective theory (cf. Fig.1). We note that the present model neglects the coupling between the which satisfy ∇ · Eˆ⊥ = 0 and ∇ × Eˆk = 0. On one hand, the gapped electron-hole excitations and soft phonons, which is transverse part can be directly related to the vector potential typically considered as weak [139]. as φˆ(r) πˆ(r) The matter field and its conjugate variable ∂Aˆ can be quantized through the canonical quantization. In the Eˆ⊥ = − . (15) Coulomb gauge, it is useful to decompose them in terms ∂t of the transverse and longitudinal components as φˆ(r) = On the other hand, the longitudinal part is subject to the con- φˆk(r) + φˆ⊥(r) and πˆ(r) = πˆ k(r) + πˆ ⊥(r). The quadratic straint resulting from the Coulomb gauge ˆ ˆ part H0 of the matter Hamiltonian Hmatter can readily be di- Z∗e ˆk ˆk agonalized as (see AppendixA) E = − φ , (16) 0v " # ˆ X ˆ⊥† ˆ⊥ X ˆk† ˆk where we recall that φˆk is the longitudinal part of the matter H0 = ~Ω φ φkλ + φ φ , (12) kλ k k ˆk kλ k field satisfying ∇ × φ = 0. This longitudinal component leads to the expression of the Coulomb potential term VC as ˆ⊥ ˆ⊥† where φkλ (φkλ) represents an annihilation (creation) opera- follows [141]: tor of a transverse phonon with momentum k and polarization k k† Z 2 ∗ 2 Z 3 2 λ, while φˆ (φˆ ) is its longitudinal counterpart. 0 3 k (Z e) d r ˆk k k VC = d r Eˆ = φ . (17) 2 20v v

3. Light-matter coupling 4. Bulk dispersions The transverse dipolar matter field φˆ⊥ can directly couple to the electromagnetic field in a bilinear form. For instance, One can diagonalize the quadratic part of the total Hamil- we recall that the field φˆ⊥ can represent transverse displace- tonian, ments of ionic charges with opposite signs and thus can cou- Hˆtot,0 = Hˆlight + Hˆ0 + Hˆl−m, (18) ple to the electric field. In terms of the vector potential Aˆ, the light-matter interaction is thus given by which allows one to obtain an analytical expression of the bulk dispersions. The presence of light-matter interaction leads to Z 3 ∗ ∗ 2 d r Z e ⊥ (Z e) 2 Hˆl−m = − πˆ · Aˆ + Aˆ + VC, the formation of the hybridized modes consisting of photons v M 2M and phonons, known as the phonon polaritons. This hybridiza- (13) tion strongly modifies the underlying dispersion relations such 7 that the avoided crossing occurs between two polariton disper- The solid curves in Fig.4 show the two branches of the bulk sions. At the quadratic level, we can decompose the quadratic phonon polaritons. The splitting between the upper and lower Hamiltonian Hˆtot,0 in Eq. (18) into two parts that solely in- dispersive modes is characterized by the coupling strength g. clude either transverse or longitudinal components of the field The upper branch exhibits a quadratic dispersion at low k = operators as follows: |k| and is gapped, i.e., it takes a finite value pΩ2 + g2 at k = 0, while the lower one linearly grows from zero and saturates ˆ ˆ ⊥ ˆ k Htot,0 = Htot,0 + Htot,0. (19) at Ω even in the limit of k → ∞. Meanwhile, the longitudinal part of the total quadratic The dispersion relations of the transverse modes can now be k ˆ ⊥ Hamiltonian Hˆ only contains the longitudinal matter field obtained by diagonalizing Htot,0 (see AppendixA). The re- tot,0 sulting Hamiltonian is φˆk and can be readily diagonalized as X Hˆ ⊥ = ω γˆ⊥† γˆ⊥ , (20) ˆ k k X ˆk† ˆk tot,0 ~ ks ksλ ksλ Htot,0 = ~Ω φk φk, (23) ksλ k where s = ± denotes an upper or a lower bulk dispersion with where Ωk = pΩ2 + g2 is the longitudinal phonon frequency eigenfrequencies [142]. We note that the longitudinal mode is independent of v u q the momentum k and is thus nondispersive as shown in the u 2 2 2 2 2 2 2 2 2 ωk + Ω + g ± (ωk + Ω + g ) − 4ωkΩ solid horizontal line in Fig.4. ω = t , k± 2 (21) B. Cavity configuration ⊥ ⊥† and γˆksλ (γˆksλ) represents an annihilation (creation) operator of a transverse phonon polariton with momentum k, polariza- We next consider the cavity setting and take into account tion λ, and mode s. We here introduce the coupling strength confinement effects as well as the hybridization of electro- g as magnetic fields with plasmons in metal mirrors in addition s to collective dipolar modes. We focus on the prototypical (Z∗e)2 g = . (22) heterostructure configuration consisting of a dipolar insula- 0Mv tor sandwiched between two semi-infinite ideal metals (see Fig.5). We consider eigenmodes propagating with the two- dimensional in-plane momentum q while the formalism de-

] veloped in this section can be extended to deal with more

Ω complex setups such as fabricated metal surface structures. [ Thin-ﬁlm conﬁgurations considered here have been previously applied to light-emitting devices [143, 144], the electron-tunneling emission [145], and the transmitting waveguide [146]. One of the main novelties introduced by this paper is to reveal the great potential of heterostructure

ω conﬁgurations as a quantum electrodynamic setting towards controlling collective matter excitations and, in particular, to point out a possibility of inducing the superradiant-type quan-

k [Ω/c]

FIG. 4. Bulk dispersions of the phonon polariton plotted against the norm of the three-dimensional wavevector k = |k|. The solid curves represent the eigenvalues of the transverse modes (cf. Eq. (21)) for the upper (s = +) and the lower (s = −) phonon polaritons, ˆ ⊥ which are obtained by diagonalizing the transverse part Htot,0 of the quadratic light-matter Hamiltonian Hˆtot,0. The solid horizontal line represents the nondispersive longitudinal phonon mode (cf. Eq. (23)) FIG. 5. Schematic figure illustrating the cavity configuration con- ˆ k obtained by diagonalizing the longitudinal part Htot,0. The dotted sidered in this paper. A dipolar insulator whose center is positioned lines show the photon dispersion ω = ck and the nondispersive bare at z = 0 is sandwiched between two semi-infinite metals with inter- phonon frequency at ω = Ω. The parameter is chosen as g = 3.5 Ω. faces at z = ±d/2. 8 tum phase transition via the vacuum electromagnetic environ- Summarizing, the present cavity light-matter system con- ment. More specifically, analyzing the hybridization of cav- sists of two canonically conjugate pairs of variables (Πˆ , Aˆ) ity electromagnetic fields and matter excitations, we demon- and (πˆ, φˆ); the former describes the dynamical electromag- strate that the present cavity architecture enriches the under- netic degrees of freedom and only have transverse compo- lying dispersion relations in a way qualitatively different from nents while the latter describes matter excitations and have the corresponding bulk case. Including phonon nonlinearities, both transverse and longitudinal parts. Their time evolution is we will further show that the heterostructure configuration en- governed by the Hamiltonian consisting of Eqs. (2), (24) and ables one to significantly soften the phonon modes, which ul- (27), and is subject to the constraint (28). timately induces the structural phase transition. We note that the use of the Coulomb gauge remains to be a convenient choice even in the present case of the inhomogeneous geometry. An alternative naive choice would be the 1. Hamiltonian condition ∇ · [(r, ω)Aˆ(r, ω)] = 0 with being the relative permittivity in the medium, which might be viewed as a quan- To reveal essential features of the present cavity setting, we tum counterpart of the gauge constraint in the macroscopic consider an ideal metal with the Drude property whose plasma Maxwell equations. However, it has been pointed out that in p 2 frequency is denoted as ωp = nee /(me0), where e is inhomogeneous media this gauge choice suffers from the diffi- the elementary electric charge, ne is the electron density in culty of performing the canonical quantization of the variables the metal, and me is the electronic mass. We assume that all due to the frequency-domain formulation of the gauge con- the materials are nonmagnetic such that the relative magnetic straint [141]. In contrast, the Coulomb gauge chosen here can permeability µ is always taken to be µ = 1. As illustrated be still imposed on a general inhomogeneous system without in Fig.5, we consider a setup with the center of a dipolar such difficulties; we thus employ it throughout this paper. insulator being positioned at z = 0 while the insulator-metal interfaces being positioned at z = ±d/2. We again choose the Coulomb gauge ∇·Aˆ = 0 and thus the Hamiltonian Hˆlight of 2. Elementary excitations the free electromagnetic field is the same as in the bulk case (cf. Eq. (2)) while the matter Hamiltonian is modified as The quadratic part of the total cavity Hamiltonian can still ˆ ˆ ˆ be diagonalized, which allows us to identify the hybridized Hmatter = H0 + Hint, (24) eigenmodes and the corresponding elementary excitation en- Z 3 2 d r πˆ 1 2 ˆ2 ergies. To this end, we first expand the electromagnetic vector Hˆ0 = + MΩ φ , (25) I v 2M 2 potential in terms of the transverse elementary modes. Owing Z 3 2 to the spatial invariance on the xy plane, we obtain ˆ d r ˆ ˆ Hint = U φ · φ , (26) s I v ˆ X ~ ⊥ A(r)= aˆqnλUqnλ(r)+H.c. , (30) where we define the integral over the insulator region as 20Aωqnλ qnλ R d3r ≡ R d2r R d/2 dz. The light-matter coupling term is I −d/2 U ⊥ (r) = eiq·ρu⊥ (z), (31) also modified as qnλ qnλ Z 2 where A = L2 is the area of interest on the xy plane, q is the 0ωp Hˆ = d3r Aˆ2 two-dimensional in-plane wavevector q = (q , q )T, and ρ = l−m 2 x y M (x, y)T is the position vector on the xy plane. The transverse Z 3 ∗ ∗ 2 ⊥ ⊥ d r Z e ⊥ (Z e) 2 mode function U satisfies ∇ · U = 0. + − πˆ · Aˆ+ Aˆ +VC, (27) qnλ qnλ I v M 2M Because the spatial invariance along the z direction is lost in the heterostructure configuration, an elementary excitation where the first term describes the plasmon coupling in the is now labeled by a discrete number n ∈ (instead of kz in R 3 R 2 R −d/2 R ∞ N metal regions defined as M d r ≡ d r −∞ + d/2 dz, the bulk case) as well as the two-dimensional vector q. We and the second term is the coupling between the insulator and also introduce a discrete variable λ = (Λ,P ) that includes the the cavity electromagnetic field. We recall that the Coulomb polarization label Λ ∈ {TM, TE} and the parity label P = gauge leads to the additional constraint on the longitudinal ±. Here, the polarization label Λ in the discrete subscript λ components of the electric and matter fields in the insulator indicates either the transverse magnetic (TM) or electric (TE) region |z| < d/2: polarization associated with the zero magnetic or electric field in the z direction: Z∗e Eˆk = − φˆk. (28) u⊥ (z) = 0 for Λ = TM, (32) 0v qnλ z ∇ × U ⊥ (r) = 0 for Λ = TE. (33) As in the bulk case, this constraint results in the Coulomb qnλ z potential term V originating from the longitudinal dipolar C Meanwhile, the parity label P indicates the symmetry of the modes as follows: mode function with respect to the spatial inversion along the Z 2 ∗ 2 Z 3 2 0 3 k (Z e) d r ˆk z direction. From the parity symmetry of the Maxwell equa- VC = d r Eˆ = φ . (29) 2 I 20v I v tions, it follows that the z component of the mode function 9

(a) (TM, +), (L, ±) (b) (TM, ー) (c) (TE, ±) ] ] ] Ω Ω Ω [ [ [ TM TM1 2 TE 1

SS (=TM 0 ) AS2 TE 0 ω ω L ± ω

RTE RTM n ∈even RTMn ∈odd n

AS1 q [Ω/c] q [Ω/c] q [Ω/c]

(d) (TM, +), RTM n ∈even (e) (TM, ー), RTM n ∈odd (f) (TE, ±), RTE n … ] … ] ] … Ω Ω Ω [ RTM4 [ [ RTM5 … RTM2 RTM3 RTE2

ω RTE

ω ω 1

RTM1 RTM0

RTE0 q [Ω/c] q [Ω/c] q [Ω/c]

FIG. 6. Elementary excitations of the cavity setting in different sectors (Λ,P ) with Λ =TM, TE, L denoting the transverse magnetic (TM), transverse electric (TE), or longitudinal (L) polarization and P = ± being the parity of transverse components of the electric field against the inversion of the direction perpendicular to the plane. Black dotted curves indicate the bulk dispersions. The top panels plot excitation energies ω against in-plane momentum q in the sectors (a) (TM,+) and (L,±), (b) (TM,−), and (c) (TE,±), respectively. The bottom panels (d), (e), and (f) show the corresponding magnified views below the phonon frequency Ω. In (a) and (b), the TM1,2 modes represent the two highest modes in the TM sector, the SS (=TM0) mode represents the symmetric surface (SS) mode, the AS1,2 modes represent the antisymmetric surface (AS) modes whose dipolar moments point out of the plane at low q. The longitudinal L± modes have the same nondispersive energy independent of the momentum q and the parity P = ±. The assignment of the label TM0 to the SS mode is in accordance with the convention used in studies of plasmonics [147]. In (d) and (e), the RTMn modes with n = 0, 1, 2 ... represent the resonant TM modes close to the dipolar-phonon frequency Ω. The RTMn≥1 modes extend over the insulator region while the most-softened mode RTM0 is essentially the surface mode localized at the interfaces analogous to the SS mode. In (c), the TE0,1 modes represent the two highest modes in the TE sector. In (f), the RTEn modes with n = 0, 1, 2 ... represent the resonant TE modes; all of them extend over the insulator region. The parameters are d = c/Ω, g = 3.5 Ω, and ωp = 5 Ω. For the sake of visibility, the plasma frequency is set to be a rather low value while its specific choice will not qualitatively affect the thermodynamic phase. must have the opposite parity to that of the x, y components manner as in the electromagnetic field. Because the matter (see AppendixC). We thus assign the parity label P to each field resides only in the insulator region, it is natural to expand mode function according to the transverse components of the dipolar matter field as u⊥ (−z) = P u⊥ (z) , (34) r qnλ x,y qnλ x,y ˆ⊥ ~ X ˆ⊥ ⊥ iq·ρ φ (r)= φqnλfqnλ(z)e + H.c. , u⊥ (−z) = −P u⊥ (z) . (35) 2MNdΩ qnλ z qnλ z qnλ Finally, we can impose the orthonormal conditions on the (37) mode functions as where Nd = A/v = N/d is the number of modes per thick- Z ⊥ ⊥† ness, and {f } is a complete set of transverse orthonormal dz u u⊥ = δ δ . (36) qnλ qnλ qmν nm λν ⊥ iq·ρ functions in the insulator region satisfying ∇ · (fqnλe ) = We can expand the matter degrees of freedom in the similar 0. We note that a position variable r in the matter field φˆ 10

TABLE I. Summary of physical properties for each of elementary excitations. The labels for each band in the first column are consistent with those indicated in Fig.6. The second and third columns represent the parity P = ± and the polarization Λ ∈ {TM, TE, L} of each mode, respectively. The fourth column indicates the direction of the dipole moments in the insulator at low in-plane wavevector q with respect to the plane parallel to the interfaces. The fifth column shows whether or not the divergence of the electric field is vanishing. The sixth and seventh columns represent whether the longitudinal wavenumber κη takes a real value or a purely imaginary value at low and high q, respectively. The subscript η ∈ {q, n, λ} with λ = (Λ,P ) represents a set of all the variables specifying each eigenmode. The final two columns indicate the physical nature of each hybridized band at low and high q, where we abbreviate elementary excitations as plasmons (PL), photons (PT), phonons (PN), polaritons (P), and specify surface modes by S. κ Mode Band P Λ Dipole at low q Divergence η low q high q low q high q

TM2 + TM In-plane zero Real Real PL PT

TM1 − TM In-plane zero Real Real PL-P PT

RTMn≥1 sgn(n) TM In-plane zero Real Real PN-P PN

RTM0 + TM In-plane zero Real Imaginary PN-P S-PN-PL-P

L± ± L Out-of-plane nonzero Real Real PN-P PN-P

SS (=TM0) + TM In-plane zero Real Imaginary PN-P S-PN-PL-P

AS2 − TM Out-of-plane zero Imaginary Imaginary PN-P S-PN-PL-P

AS1 − TM Out-of-plane zero Imaginary Imaginary PT S-PN-PL-P

TE1 − TE In-plane zero Real Real PL PT

TE0 + TE In-plane zero Real Real PL-P PT

RTEn≥0 sgn(n) TE In-plane zero Real Real PN-P PN

should be interpreted as the parameters restricted to the region ωqnλ of an elementary eigenmode now explicitly depends on |z| < d/2. In the discrete variable λ = (Λ,P ), the polariza- the polarization Λ ∈ {TM, TE, L} in contrast to the bulk tion label takes either TM or TE modes, Λ ∈ {TM, TE}, and case in Eq. (21). In terms of the obtained elementary eigen- the parity label P = ± is defined in the same manner as in modes, the quadratic part of the total light-matter Hamilto- Eqs. (34) and (35). Another possible matter excitations are nian, Hˆtot,0 = Hˆlight + Hˆ0 + Hˆl−m, can be diagonalized as the longitudinal modes, for which we can expand the matter field as ˆ X † Htot,0 = ~ωqnλγˆqnλγˆqnλ, (39) r qnλ ˆk ~ X ˆk k iq·ρ φ (r) = φqnλfqnλ(z)e + H.c. , 2MNdΩ † qnλ where γˆqnλ (γˆqnλ) represents an annihilation (creation) op- (38) erator of a hybridized elementary excitation with in-plane wavevector q and discrete labels n and λ. The spatial profile k of each excitation is characterized by an eigenmode function where {fqnλ} is a complete set of longitudinal orthonor- ⊥ k mal functions in the insulator region, which satisfies ∇ × uqnλ or fqnλ depending on the polarization, whose explicit k iq·ρ functional forms are given in AppendixC. (fqnλe ) = 0. We denote this mode by the label λ = (L,P ), where the polarization label Λ is fixed to be the longi- Figure6 shows the elementary excitations in the cavity set- tudinal one Λ = L while the parity P = ± of the mode func- ting for each sector of λ = (Λ,P ), and TableI summarizes tion is defined in the same way as in the transverse case. The the corresponding physical properties. Figures6(a) and (b) longitudinal mode associates with an excitation of the longi- plot the in-plane dispersions in the sectors λ = (TM, +) and tudinal electric field via Eq. (28). (L, ±), and λ = (TM, −), respectively. Their magnified plots An explicit form of the transverse mode functions {u⊥ } around the phonon frequency Ω are also given in Figs.6(d) qnλ and (e). These rich band structures in the TM polarization ⊥ k and {fqnλ}, the longitudinal mode functions {fqnλ}, and have different physical origins as described below. the corresponding excitation energies {ωqnλ} can be deter- First of all, a salient feature of the obtained cavity polariton mined from solving the eigenvalue problem with imposing is the emergence of a large number of soft-phonon modes ly- suitable boundary conditions at the interfaces as detailed in ing below the bare phonon frequency Ω, which we term as the AppendixC. In the transverse modes, we note that both mat- resonant TM (RTM) modes (see Fig.6(d,e)). The RTM modes ter and light fields as well as the hybridized eigenmodes share are labeled by an integer number n = 0, 1, 2 ... as RTMn. the same spatial profile represented by u⊥ owing to the ⊥ qnλ The parity P of the corresponding mode function uqnλ co- bilinear form of the light-matter coupling. Meanwhile, the incides with the parity of n. The RTMn modes with n ≥ 1 longitudinal modes consist of only the matter field and their lie between Ω and the lower bulk dispersion (plotted as the spatial profiles are characterized in terms of the mode func- lower black dotted curve in Fig.6), and share the common k tions fqnλ. It is also noteworthy that an excitation energy structures. For instance, all of the RTMn≥1 modes start from 11 frequencies slightly below Ω and saturate to it at high q ≡ |q|. as the emergence of roton-type excitations (cf. Fig. 12 in Ap- Thus, these modes predominantly consist of the hybridization pendixE). Importantly, the correct treatment of this RTM 0 of dipolar phonons and cavity electromagnetic fields at low q mode turns out to be crucial especially when we discuss the while the phonon contribution eventually becomes dominant superradiant-type quantum phase transition as detailed in the at high q. The field amplitudes of these RTMn≥1 modes ex- next section. Meanwhile, we remark that the reflectivity does tend over the insulator region owing to the real-valuedness not qualitatively change physical properties of the TE modes, of the longitudinal wavenumber κη over the entire plane of which are analogous to the modes supported in Fabry-Perot the in-plane wavevector q. Here, the subscript η is used to cavities [149]. summarize all the indices as η = {q, n, λ}. The origin of the emergence of many RTM modes below the bare phonon Secondly, we remark that the rich structures of the elemen- frequency can be understood from simple physical arguments tary excitations demonstrated above can further be tuned by based on the frequency-dependent dielectric functions as dis- varying the thickness d of the dipolar insulator. The increase cussed in Sec. III B 3. of d generally leads to a generation of more delocalized bulk Remarkably, the lowest resonant mode RTM0, which is the modes extending over the insulator. Thus, aside from the lo- most-softened phonon mode among all the elementary exci- calized surface modes, the eigenmodes become denser and tations in all the sectors, has a qualitatively different physical eventually converge to the corresponding bulk dispersions in nature compared to the RTMn≥1 modes discussed above. Its the limit of d → ∞. In contrast, the decrease of d leads to the longitudinal wavenumber κη changes from a real value to a filtering of the delocalized modes; for instance, a degenerate purely imaginary one as the dispersion crosses the lower bulk frequency of the TE0 and SS modes at zero-in-plane momen- dispersion in the bulk (see Fig.6(d)). Moreover, the saturated tum shifts up for a smaller d. Thus, with decreasing d, the frequency value of the RTM0 mode at q → ∞ remains sub- localized nature of the modes plays more and more crucial stantially below Ω with a nonvanishing gap. Accordingly, the roles in determining the physical properties of the present het- RTM0 mode can be interpreted as the low-energy analogue of erostructure configuration. We assess the effects of changing the surface polariton mode labeled as the SS mode in Fig6(a); the thickness in more detail in the next section by including both modes result from the intrinsic hybridization of dipolar the most relevant nonlinear effect in a self-consistent manner. phonons, cavity electromagnetic fields, and plasmons while The nonlinearity renormalizes the value of the bare phonon the major contribution at high q comes from the phonon (plas- frequency Ω and thus effectively modifies the underlying el- mon) part in the case of the RTM0 (SS) mode. ementary excitations. We will see that a thinner insulator is We now proceed to discuss the symmetric and antisymmet- more preferable for realizing the cavity-enhanced ferroelec- ric surface modes lying in the frequency regime pΩ2 + g2 < tricity. The reason is that a thin heterostructure configuration ω < ωp, which are denoted as SS(=TM0) and AS2 in can alleviate the adverse effect of the interaction among the Figs.6(a) and (b), respectively. They essentially arise from energetically dense phonon excitations and is thus advanta- the splitting of the coupled surface phonon polaritons at two geous for inducing the ultimate softening of the RTM0 mode, metal-insulator interfaces [148]. These surface modes result which causes the ferroelectric instability. from the intrinsic hybridization of the collective dipolar mode, electromagnetic fields in the cavity, and plasmons in metal Finally, we comment on the importance of considering all mirrors, and can emerge only in the case of the TM polariza- the above-mentioned hybridized modes in the cavity geome- tion, in which the continuity of the z component of the elec- try, thus keeping track of the continuum of excitations for ev- tric field can be mitigated. Thus, the surface modes including ery dispersion. In general, nonresonant modes can give rise to the SS(=TM0), AS1,2, and RTM0 modes can be excited only substantial modifications of dipolar moments in the confined by p-polarized light. Additional discussion of the hybridized geometry as compared to their counterparts in free space, eigenmodes in our cavity system can be found in AppendixE. especially in the regime of ultrastrong coupling. This phe- Several remarks are in order. Firstly, we here emphasize nomenon has been discussed for systems with dipole-dipole the importance of explicitly including plasmon contributions interaction and demonstrated to have a qualitative effect on into the analysis. This treatment ensures that the tails of the thermodynamic properties of a system [150, 151]; we here evanescent fields properly extend into the surrounding metal recall that dipolar interactions can be understood as arising mirrors. In many of the previous studies, the mirrors are as- from the adiabatic elimination of nonresonant modes under sumed to be perfectly reflective such that the electric fields certain conditions [51]. In our analysis, all the confinement must exactly vanish at the interfaces and are absent in the effects due to the cavity, including dipole-dipole interactions, metal regions. In such analyses, the physical nature of the are consistently included into the theory without relying on localized surface modes discussed here cannot be addressed approximations such as rotating-wave approximations or adi- appropriately. For instance, the most-softened phonon mode abatic eliminations, which can fail in the presence of closely RTM0 is one important example of such localized modes; degenerate low-energy modes such as the RTM/RTE modes as an analysis assuming the perfect reflectivity cannot capture discussed here. The inclusion of all the continuum modes be- its unique features, including the qualitative changes associ- comes further important when we will study the phonon non- ated with increasing in-plane momentum q (cf. Fig. 11 in Ap- linearity below. There, due to mode couplings mediated by pendixE), the nonvanishing softening from the bare phonon the nonlinearity, an energy of individual modes can depend frequency Ω even in the limit q → ∞ (cf. Fig.6(d)), as well on fluctuations of the other modes. 12

3. Origin of the resonant soft-phonon modes the mode frequency deviates slightly from the corresponding value in the loss-less limit and now has a finite lifetime. How- We here provide simple explanations to elucidate the origin ever, when the bare phonon frequency approaches zero (i.e., of the resonant softened phonon modes that are denoted by the Ω → 0), we find that not only the real part of the RTM0 mode labels RTM and RTE in Fig.6. To this end, we consider a fre- becomes zero, but also its imaginary part vanishes. This in- quency regime in which the metallic permittivity is negative dicates that the lifetime of the softened mode diverges at the transition point. We emphasize that this conclusion holds true p(ω) < 0 while the insulator dielectric function is positive ξ(ω) > 0 (see also Eqs. (C10) and (C11) in App.C). To be even for large γ. Thus, we conclude that the Ohmic losses specific, we focus on the RTM modes at the zero in-plane mo- in the cavity should not affect our main results, namely, the mentum q = 0 while similar arguments can be given for the prediction of the cavity-induced softening and the resulting corresponding RTE modes as well. enhancement of ferroelectricity. Inside the insulator region, the RTM modes with q = 0 To understand the observed weak sensitivity to Ohmic have the spatial profiles of the magnetic field either sin(κz) or losses in metals, we recall that electric current in a metal is de- cos(κz), which we denote by the labels P = + or P = −, termined by the gradient of the electrochemical potential that respectively. We here recall that the parity of the eigenmodes includes both the electric field and the gradient of the chemical is defined by that of the transverse components of the electric potential. A simple example of this is a problem of screening field. In these cases, the longitudinal wavevector κ should of a static charge inside a disordered metal; in this case, one satisfy the following equations: finds finite electric field on the scale of a screening length, but no electric currents. This is also the reason why in our setup s ( κd the system can develop static ferroelectric order without elec- ξ(ω) cot 2 (P = +), = κd (40) tric currents, i.e., no Ohmic losses. |p(ω)| − tan (P = −). 2 It is useful to provide another perspective on why losses The Maxwell equations also lead to the additional constraint do not affect the phase transition discussed here. In general, static impurities in metals, which cause Ohmic losses, p ω κ = ξ(ω) . (41) should give rise to scattering between modes at the same en- c ergy, but different propagation directions and polarizations. If In the case of the frequency regime slightly below the bare one launches a mode with a well defined momentum, scat- phonon frequency Ω of paraelectric materials, the dielectric tering will lead to depletion in the population of this mode, function ξ(ω) takes a singularly large value compared with which is understood as loss. However, such scattering should |p(ω)|. Thus, the conditions (40) and (41) lead to the approx- not affect equilibrium thermodynamic properties of the sys- imate eigenmode equation tem because it does not associate with actual dissipation of energy. More specifically, as the scattering modes have the p πc ξ(ω)ω ' n (n = 1, 2,...). (42) same energy and thus equal populations in equilibrium, the d detailed balance condition ensures equal rates for scattering Owing to the singular nature of ξ(ω) in the limit of ω → Ω−0, processes going in the opposite directions, leading to no net this condition can in principle be attained for an arbitrary integer n. It is this resonant structure that allows for generating a large number of soft-phonon modes in the heterostructure (a) RTM0 (b) RTM0 configuration. In practice, when the damping of the phonon mode is included, there should exist an effective cutoff on possible values of n, but still many eigensolutions should be al- Ω decrease lowed as long as the damping rate is small enough in compar-

ison with the phonon frequency. =0)] =0)] q q ( ( ω ω Ω decrease

4. Effects of cavity losses Re[ Im[

The analyses presented in the previous sections can be generalized to include additional complexities of experimental systems. In particular, in this section we consider the effect of scattering of electrons in metals by disorder, which can Loss rate γ Loss rate γ manifest itself as Ohmic losses. We use a phenomenological approach of adding an imaginary part to the permittivity of FIG. 7. (a) Real and (b) imaginary parts of the complex eigenvalues metallic mirrors. This corresponds to using the Drude model for the most softened mode (i.e., RTM0 mode) at q = 0 in the pres- of electron dynamics, where damping effect is characterized ence of nonzero cavity-loss rate γ. From top to bottom (resp. bottom by nonzero loss rate γ (see AppendixD). We use this model to to top) curves in (a) (resp. (b)), the bare phonon frequency Ω is set analyze the complex eigenvalues of the mode that exhibits the to be 1, 0.5, 0.1, and 0.01. The parameters are d = 1, ωp = 5, and g = 3.5. strongest softening, i.e., the RTM0 mode. As shown in Fig.7, 13 dissipated energy. by Eq. (27). The inclusion of such nonlinear terms is partic- One may be concerned that this argument fails when in- ularly important close to the phase transition, where phonon elastic scattering processes are included. However, as the fluctuations become significant. RTM0 mode frequency approaches zero, inelastic scattering We use variational approach to analyze the phase diagram processes become strongly suppressed because of the phase- of the nonlinear model (43). It is convenient to change nota- space argument (e.g., the density of states of acoustic phonons tions to bring our model into the form commonly used in the scales as ∝ ω2 at small frequencies). A more rigorous analy- study of second-order phase transitions. We replace the bare 2 sis of effects of dissipation can be performed using an effec- phonon frequency Ω in Hˆ0 (cf. Eq. (25)) with a parameter r, tive model of coupling to a bosonic reservoir, or introducing which corresponds to the tuning parameter of a physical sys- the Green’s function formalism, or the microscopic Hopfield tem such as pressure, isotope concentration, or chemical com- approach. However, we expect that the conclusion of insensi- position. In the quadratic (i.e., noninteracting) theory, para- tivity of the phase transition to dissipation will not be affected. electric and ferroelectric phases correpond to r ≥ 0 and r < 0 respectively, while conditions for the interacting system will be discussed below. We separate the quadratic part of the total IV. CAVITY-ENHANCED FERROELECTRIC PHASE Hamiltonian as Hˆ (r). Then, the full interacting Hamil- TRANSITION IN AN INTERACTING SYSTEM tot,0 tonian can be written as Hˆtot(r) = Hˆtot,0(r) + Hînt. The thermal equilibrium of the system is described by the usual Building upon analysis of hybridized light-matter excita- Gibbs state tions in the cavity geometry, we now discuss the paraelectric- ˆ ˆ ˆ to-ferroelectric phase transition, which is the structural quan- e−βHtot(r) e−β(Htot,0(r)+Hint) ρˆ(r) = = , (44) tum phase transition commonly observed in ionic crystals. Z(r) Z(r) Our method of determining the phase transition point is based ˆ on identifying the softening of the dipolar phonon mode, where Z(r) ≡ Tr[e−βHtot(r)]. Within variational approach, which is a standard indication of the continuous second or- we approximate the Hamiltonian in Eq. (44) by the quadratic der phase transition into the broken symmetry phase. Interest- Hamiltonian with an effective parameter reff that should be ingly, we find that in the cavity setting the transition point is determined from the variational principle: shifted in favor of the ferroelectric phase. This softening results from the interplay between the modified dispersions and min DKL [ˆρ0(reff )|ρˆ(r)] ≥ 0, (45) feedback effects through nonlinear interactions between the reff energetically dense elementary excitations. where DKL[·|·] is the Kullback-Leibler divergence [152], and an effective Gaussian state is defined as ˆ A. Variational principle e−βHtot,0(reff ) ρˆ0(reff ) = (46) Z0(reff ) Our primary objective in this section is to understand how −βHˆtot,0(r ) the interaction term between the phonon modes (cf. Eq. (26)) with Z0(reff ) ≡ Tr[e eff ]. modifies physical properties of quantum paraelectrics, espe- Our goal is to find reff that provides the best approximation cially in the vicinity of the transition into the ferroelectric to the interacting theory. To this end, we consider phase. The paraelectric-to-ferroelectric phase transition is es- D E sentially a structural transition in an ionic crystal, where the Fv ≡ F0(reff ) + Hˆtot(r) − Hˆtot,0(reff ) , (47) phonon field φˆ plays the role of the order parameter. When eff it develops an expectation value, we find the ferroelectric where we denote an expectation value with respect to the phase characterized by spontaneous symmetry breaking. This effective density matrix as h· · · ieff = Tr[ˆρ0(reff ) ··· ], and corresponds to asymmetric distortion of charges within the 1 F0(reff ) = − β ln Z0(reff ) is the Gaussian variational free en- unit cell, giving rise to a uniform electric polarization (cf. ergy. The variational condition (45) can then be simplified as Fig.1(b)). In the bulk system, this transition is associated [153] with softening of the upper phonon-polariton mode, while the acoustic lower polariton mode is irrelevant as it reduces to a min Fv ≥ F (r), (48) r pure photon mode in the q → 0 limit. In the cavity geometry eff of our interest, we thus need to analyze softening of the lowest 1 where F (r) = − β ln Z(r) is the exact free energy. The opti- optical mode, i.e., the RTM0 mode. mal effective tuning parameter r∗ , which is defined by We consider a system described by the Hamiltonian eff r∗ = arg min F , (49) Hˆ = Hˆ + Hˆ + Hˆ , (43) eff v tot matter light l−m reff ˆ where Hmatter now includes both the quadratic terms (25) and gives the best approximation of the interacting model in terms ˆ the most relevant interaction term (26), Hlight is the Hamil- of the linear theory. Within the present variational formal- tonian for the free electromagnetic field (2), and Hˆl−m de- ism, the phase transition point is determined from the condi- ∗ scribes the light-matter coupling in the cavity setting as given tion reff = 0 that is achieved for a certain value of the bare 14 tuning parameter r. When phonon nonlinearities are repul- values of parameters and should be a generic feature of dipolar sive, the transition takes place for a negative value of the bare insulators confined in the cavities. We find that this effect be- parameter r. In other words, the interaction feedback gives comes more pronounced for insulating films of smaller thick- rise to the renormalization of the effective phonon frequency ness d. We also observe that increasing the plasma frequency ∗ 2 ˆ ∗ reff = Ωeff in the variational theory Htot,0(reff ), and the tran- has qualitatively similar effect to decreasing the paraelectric sition occurs when this renormalized frequency of paraelectric slab thickness (cf. the inset in Fig.3(c)). This can be under- materials goes down to zero. stood as a result of tighter mode confinement by metals with We note that the variational method presented here is closely related to the standard mean-field approach, since we choose a simple Gaussian family of variational states. We ap- (a) 0 r =-20 proximate the free energy of an interacting system by the free -0.2 energy of a noninteracting system with an optimized varia- 2 -0.4

[ × 10 ] d decrease tional parameter reﬀ . For this class of variational states, inter- -0.6 action terms in Eq. (47) can be calculated by factorizing them -0.8 into the product of the expectation values of the quadratic op- -1 erators with respect to the noninteracting theory, in the same r =-40 manner as it is done in a standard mean-ﬁeld theory [154] (see -2.5 v

AppendixG for technical details). We note that this theoret- F -3 ical approach is exact for the spherical model, in which the -3.5 number of components in the order parameter is effectively inﬁnite. -4 -4 r =-60

B. Results -6 -8 We show the obtained variational free energy Fv against 0 2 4 6 8 10 the variational parameter reff at different insulator thicknesses r d in Fig.8(a). From top to bottom panels, we decrease the eff bare tuning parameter r. The thermal equilibrium value r∗ (b) 5 U=1 eff 4 of the effective phonon frequency is determined from identi- 3 FE PE fying the minimum of each variational free-energy landscape. 2 1 In all the panels, a thinner insulator thickness d in the cav- 0 ity setting leads to a smaller equilibrium phonon frequency 5 U=2 ∗ 4 reff , indicating the phonon softening owing to the coupling d 3 FE PE with light modes strongly confined in the cavity. With further 2 decreasing the thickness d or the tuning parameter r, the equi- 1 0 librium phonon frequency eventually goes down to zero from 5 ∗ U=3 above, reff → 0+, at which the paraelectric-to-ferroelectric 4 phase transition occurs (see e.g., the red solid curve in the 3 FE PE 2 bottom panel of Fig.8(a)). 1 The corresponding phase diagram of the cavity setting with 0 -400 -300 -200 -100 0 varying thickness d is plotted in Fig.8(b) for different values r of the phonon nonlinearity U. Here, the horizontal axis corresponds to the tuning parameter r, which may correspond to FIG. 8. (a) Variational free energies Fv plotted against the vari- pressure or isotope concentration. In the paraelectric phase, ational parameter reff with decreasing tuning parameter r from top decreasing the value of r corresponds to approaching the tran- to bottom panels. The red and blue solid curves correspond to the sition point. We can then interpret the results in Fig.8(b) results for the cavity setting with the insulator thicknesses d = 1 as the cavity-induced stabilization of the ferroelectric phase, and d = 3, respectively, while the green solid curves correspond to which should be compared to the bulk transition point indi- the bulk results. The minima of the variational energies determine r∗ = Ω2 cated by the black dashed line. We emphasize that this theo- the equilibrium values eff eff of the effective phonon frequency including the nonlinear feedback. (b) Phase diagram in the cavity set- retical result has direct experimental implications. Consider, ting at different nonlinear strengths U. As the tuning parameter r is for example, a pressure-tuned transition into the ferroelectric decreased, the transition from the paraelectric phase (PE) to the fer- phase in materials such as SrTiO3 or KTaO3, which are para- roelectric phase (FE) occurs when the equilibrium effective phonon ∗ electric at ambient conditions. Our results then predict that, frequency reff goes down to zero. The green dashed curve is the when the sample is placed inside the cavity, transition into the phase boundary in the cavity geometry, while the black (resp. red) ferroelectric phase should take place at lower pressure. vertical dashed line is that in the bulk geometry (resp. at U = 0). We This favoring of the ferroelectric phase in the quantum elec- set c = ~ = β = 1, ωp = 5, g = 3.5, and the momentum cutoff to 2 tromagnetic environment is robust against changes in specific be Λc = 10 . 15 the higher plasma frequency. the modulation wavevector perpendicular to the direction of The enhanced ferroelectricity found here can be understood the dipoles. The tendency to having an ordered state at large from the interplay between the phonon nonlinearity and the q should be opposed by the elastic energy that favors the uni- dispersion of hybrid light-matter modes in the cavity geome- form value of φ. This contribution can be described by the try. To this end, we recall that the phonon repulsion included gradient term of φ, which we did not include in our model. in our model renormalizes the effective phonon frequency as Only by properly including the competition between dipolar p ∗ interactions and elastic energy, one can determine the correct Ωeff = reff . In general, low-energy modes provide the dominant contribution to the interaction-induced hardening ordering wavevector. We plan to discuss the nature of the or- of phonons. In the present cavity configuration, there are dered phase in a future publication. many low-lying dispersions, corresponding to different values of longitudinal wavenumber κ for a fixed in-plane momentum q. The transition occurs when energy of the most- V. SUMMARY AND DISCUSSIONS softened mode (RTM0 mode) reaches zero, and it is sufficient that only this mode softens to have the ferroelectric transition. Our work demonstrates a great promise of a heterostructure With decreasing the transverse size of the cavity d, the other configuration as a platform towards controlling the quantum modes (i.e., RTM1,2,3,...) are pushed to the frequency close phase of matter by modifying quantum electrodynamic envi- to the frequency of the bare phonon, thus they contribute less ronments in the absence of external pump. In particular, the to phonon hardening. In contrast, when we increase d, these present consideration suggests that the strong hybridization modes go down in frequency and thus increase their contri- between the continuum of matter excitations and cavity elec- bution to phonon hardening, which is unfavorable to soften tromagnetic fields enriches the spectrum of elementary exci- the RTM0 mode, i.e., to induce the ferroelectric transition. As tations in a way qualitatively different from the bulk material. shown in Fig.8(b), the enhanced ferroelectricity in the cavity More specifically, the main results of this paper can be sum- configuration is more pronounced as the nonlinear strength U marized as follows: is increased. In particular, we note that the enhancement is absent unless the nonlinearity is included into the analysis, as (i) Analysis of hybridized collective modes in a cavity. indicated by the red vertical line. We analyzed the collective modes in the heterostruc- Theoretical predictions of our paper can be tested in several ture configuration shown in Fig.2, i.e., the dielectric slab having an infrared active phonon at frequency Ω materials, including SrTiO3 and KTaO3 [135, 155], as well surrounded by two metallic electrodes. Hybridization as SnTe and Pb1−zSnzTe [156], where a tuning parameter can be varied by pressure, chemical composition, or isotope of the phonon modes of the dielectric, the cavity elec- substitution. This flexibility allows one to naturally prepare tromagnetic modes, and the plasmons of the surround- materials on the verge of the ferroelectric phase. Strong non- ing metallic mirrors gives rise to a rich structure of the linearity is also a ubiquitous feature of these materials. To be eigenmodes; in particular, we found the emergence of a number of energy modes at frequencies slightly below concrete, if we consider SrTiO3, one can use the transverse optical phonon mode, for which the phonon nonlinear cou- the bare phonon frequency Ω. The origin of these eigen- pling is U/M 2 ' 2.2×103 THz2 A˚ −2 kg−1 [157]. In the unit modes can be understood from the strong frequency dependence of the dielectric function close to the phonon of c = ~ = β = 1 with, e.g., the temperature T = 60 K, this nonlinear coupling strength becomes U ' 1.5, which is frequency (see Sec. III B 3). close to the value used in the present numerical calculations (ii) Cavity-enhanced ferroelectric phase transition. To d = 1 [158]. With this convention, the thickness corresponds go beyond the simple quadratic model of the photon- to a length scale of several tens of microns. In this regime, the phonon-plasmon hybridization, we included phonon expected shift of the physical phonon frequency is an order of nonlinearities in the dielectric material through the vari- THz (cf. the bottom panel in Fig.8(a)), which should corre- ational method. We found that the collective modes spond to tens of degrees Kelvin shift in the transition temper- are softened as the thickness of the dielectric slab is ature [157, 159]. decreased. We emphasize that this feature is present We remark that the model discussed here is still not suffi- only when we include phonon nonlinearities and, there- cient to determine spatial character of the ordered ferroelectric fore, its mechanism goes beyond simple screening of phase. From the symmetry of the softened RTM0 mode, we ferroelectric fluctuations by electrons in the metallic expect that it will have an in-plane polarization, however, we claddings. The softening corresponds to the enhance- cannot determine its momentum. It is common to address this ment of the ferroelectric instability, and allows a cavity question by analyzing unstable collective modes in the dis- system to undergo transition into the ferroelectric phase ordered (i.e., paraelectric) state and using r that corresponds even when the bulk material remains paraelectric. We to the ordered (i.e., ferroelectric) phase. The wavevector of suggest that such cavity-enhanced transition could be the most unstable modes is then an indication of the expected experimentally observed in several quantum paraelec- ordered state. In our consideration, this procedure suggests in- tric materials at the verge of the ferroelectric phase. stability at large q (see also AppendixE). This is not surprising since dipoles have repulsive interactions when oriented side The mechanism proposed in this paper is not a specific fea- by side, thus dipolar interactions favor stripe domains with ture of a particular setup, but should be applicable to a wide 16 class of systems supporting excitations that possess electric is not the case, one may in principle apply a more inclusive dipole moments. For instance, our theory can be used to approach such as by explicitly including bath degrees of free- analyze the coupling of cavity fields to excitonic molecules dom as bosonic reservoirs into the analysis. Meanwhile, on [160, 161] and also to artificial quantum systems such as ultra- the border of the phase transition, there are also interesting cold polar molecules [162], trapped ions [163], and Rydberg open questions on how quantum critical behavior is modified atoms [164]. The present formulation should also be applica- due to cavity confinement and how it can be affected by dis- ble to analyze polariton modes in the hexagonal boron nitride sipative nature of a cavity [135, 180–186]. It would also be (hBN) [137, 165, 166], which is a dielectric insulator material interesting to explore combining cavity-enhanced phase tran- associated with a large optical phonon frequency. We point sitions with topological aspects of open systems [187–192]. out again that the predicted changes of the elementary exci- Third, it will be intriguing to consider the situation in which tations can occur without external driving in contrast to, e.g., the primary coupling between matter and light is mediated the earlier studies of exciton-polaritons in microcavities [167– by the magnetic field. A concrete example can be Damon- 173]. Eschbach modes in ferromagnetic films [193]. The intrinsi- As by-product of the analyses given in this paper, we en- cally chiral nature of these modes may qualitatively alter the vision several potential applications of the cavity-induced character of collective modes in the cavity configuration. One changes of the dispersive excitation modes. For instance, one can also consider a situation with both electric and magnetic can consider a molecular crystal sandwiched between metallic couplings being important. This can be relevant to multifer- mirrors, where polaritons originate from the hybridization of roic systems [194], in which ferroelectric and magnetic orders singlet excitons and cavity electromagnetic modes. The corre- are intertwined. sponding triplet excitons have negligible interaction with light Finally, it is also intriguing to consider further features and and are thus unaffected by the cavity confinement. The het- functionalities that can arise from fabricating additional struc- erostructure cavity configuration then allows one to change tures at the metal-insulator interfaces. For instance, periodic relative energies of singlet and triplet excitations, which can texturing or distortion of the surfaces can lead to the forma- be applied to improve the efficiency of organic light-emitting tion of gaps in the in-plane dispersion as realized in photonic devices and realize photon-exciton converters. We expect that crystals. This will provide additional flexibility for manipula- the modification of the hybridized dispersion in the proposed tion of the hybridized light-matter excitations. It also merits cavity setup can also be used to control photochemical reac- further study to consider how changes in the phonon spectrum tions in molecular systems. The similar physics may be ob- in the metal-paraelectric heterostructure can affect many-body served in the presence of a single metal-dielectric interface. states in the metallic mirrors, such as, superconducting states. Further discussions can be found in AppendixF. This should be particularly relevant to low-density supercon- Several open questions remain for future studies. First, we ductivity mediated by plasmons as discussed in the case of recall that the whole system considered in this paper is as- doped strontium titanate, where the soft transverse optical sumed to be in thermal equilibrium, i.e., the effective tem- phonon plays a crucial role [195]. We hope that our work perature is identical for both of the matter and the cavity. This stimulates further studies in these directions. assumption may not always apply; for example, the metal mirrors and the insulator could be coupled to external bath modes at different temperatures. Moreover, driving the cavity into ACKNOWLEDGMENTS a high-intensity regime or coherent pumping of atoms should induce the cavity-mediated interactions among the underlying We are grateful to Richard Averitt, Dmitri Basov, An- elementary excitations [64, 174–177]. It remains an intrigu- toine Georges, Bertrand I. Halperin, Mikhail Lukin, Dirk ing question how our formalism can be generalized to such van der Marel, Giacomo Mazza, Marios Michael, Prineha nonequilibrium situations. Narang, Angel Rubio, and Sho Sugiura for fruitful discus- Second, a quantitative amount of the predicted modification sions. Y.A. acknowledges support from the Japan Society for of the elementary excitations and also that of the transition- the Promotion of Science through Grant Nos. JP16J03613 and point shift can in practice depend on several specifics such as JP19K23424, and Harvard University for hospitality. J.F. ac- finite thickness of metallic mirrors, the phonon-scattering rate knowledges funding from the Swiss National Science Foun- in matter, and the plasma frequency and the conductivity in dation grant 200020-192330. D.J. acknowledges support by metals. It merits further study to explore the impact of those EPSRC grant No. EP/P009565/1. D.J. and A.C. acknowledge experimental complexities on the present results. In particular, funding from the European Research Council under the Eu- it is intriguing to elucidate how dissipation effects on phonons ropean Union’s Seventh Framework Programme (FP7/2007- and cavity photons can influence the phase diagram obtained 2013)/ERC Grant Agreement No. 319286 (Q-MAC). A.C. ac- in the present work. Our preliminary phenomenological anal- knowledges support from the Deutsche Forschungsgemein- ysis in Sec. III B 4 indicates that the equilibrium properties schaft (DFG) via the Cluster of Excellence ‘The Ham- should not be qualitatively affected by dissipation; however, burg Centre for Ultrafast Imaging’ (EXC 1074–Project ID it may still cause quantitative modifications and, in particu- 194651731) and from the priority program SFB925. E.D. lar, influence nonequilibrium dynamics. These issues could acknowledges support from Harvard-MIT CUA, AFOSR- be addressed by relying on recent developments in Markovian MURI Photonic Quantum Matter (award FA95501610323), open quantum many-body systems [178, 179]. Even if this and DARPA DRINQS program (award D18AC00014). 17

Appendix A: Diagonalization of the quadratic Hamiltonian then be simpliﬁed as ! ! 1 X 1 0 πˆ⊥P Hˆ ⊥ = πˆ⊥P AˆP −kλ Here we provide technical details about the diagonaliza- tot,0 2 kλ kλ 0 1 AˆP tion of the quadratic parts of the bulk Hamiltonian in the kλ −kλ ! ! main text. To this end, we decompose the matter ﬁeld as 1 Ω2 −gΩ πˆ⊥ X ⊥ ˆ −kλ φˆ(r) = φˆk(r) + φˆ⊥(r) and πˆ(r) = πˆ k(r) + πˆ ⊥(r), whose + πˆkλ Akλ 2 2 , 2 −gΩ ω + g Aˆ−kλ explicit forms are given by kλ k (A9)

r from which one can readily obtain Eqs. (20) and (21) in the X φˆ⊥(r) = ~ φˆ⊥ eik·r + H.c. , (A1) main text. 2MNΩ kλ kλ kλ r MΩ X Appendix B: Macroscopic Maxwell equations πˆ ⊥(r) = −i ~ φˆ⊥ eik·r − H.c. , (A2) 2N kλ kλ kλ r In the linear regime, elementary excitations of the electro- X k k φˆk(r) = ~ φˆ eik·r + H.c. , (A3) magnetic fields in nonabsorbing matter can in general be ob- 2MNΩ k |k| k tained by solving the macroscopic Maxwell equations with r using boundary conditions appropriate for a physical setting MΩ X k k πˆ k(r) = −i ~ φˆ eik·r − H.c. , (A4) of interest. Specifically, the dynamics of the electromagnetic 2N k |k| k fields is governed by ∂B ∇ × E + = 0, (B1) ∂t N = V/v φˆ⊥ φˆ⊥† where is the number of modes, kλ ( kλ) is the an- 1 ∂D nihilation (creation) operator of transverse phonons with mo- ∇ × B = , (B2) µ ∂t mentum k and polarization λ, and φˆk (φˆk†) is the annihilation 0 k k ∇ · D = 0, (B3) (creation) operator of longitudinal phonons. We recall that the transverse polarization vectors kλ satisfy Eqs. (6) and (7). At ∇ · B = 0, (B4) the quadratic level (U = 0), only the transverse components where E is the electric field, B is the magnetic field, µ is the couple to the dynamical electromagnetic field while the lon- 0 vacuum permeability, and D is the macroscopic electric field gitudinal components couple to the static longitudinal electric in matter. Transforming to the frequency basis, the macro- field. In the similar manner as in the electromagnetic Hamil- scopic electric field D is related to E via tonian Hˆlight, the quadratic part Hˆ0 of the matter Hamiltonian Hˆmatter can then be diagonalized as in Eq. (12) in the main D (r) = 0(ω, r)E (r) (B5) text. with 0 being the vacuum permittivity and (ω, r) being the To diagonalize the transverse part of the total quadratic frequency- and position-dependent relative permittivity re- ˆ ⊥ Hamiltonian Htot,0, it is useful to introduce the conjugate flecting material properties. We choose the Coulomb gauge pairs of variables for the field operators of electromagnetic field πˆ ⊥ and Aˆ in the Fourier space: ∇ · A (r) = 0. (B6) In terms of the vector potential, the equation of motion can be simplified as r ⊥ ~ ˆ⊥ ˆ⊥† 2 πˆkλ = −i φkλ − φ , (A5) −ω (ω, r) 2Ω −kλ A (r) + ∇ × [∇ × A (r)] = 0, (B7) r c2 Ω πˆ⊥P = ~ φˆ⊥ + φˆ⊥† , (A6) √ kλ 2 kλ −kλ where c = 1/ 0µ0 is the vacuum light speed. In the bulk case, one can simply solve this eigenvalue problem by em- r ˆ ~ † ploying the spatial invariance and expressing the eigensolu- Akλ = aˆkλ +a ˆ−kλ , (A7) 2ωk tions in terms of the plane waves (cf. Eq. (4)). r ω AˆP = −i ~ k aˆ − aˆ† , (A8) kλ 2 kλ −kλ Appendix C: Eigenmodes in the cavity configuration where the superscript P indicates that an operator is a conju- From now on, we focus on the heterostructure configuration gate variable of the corresponding operator, i.e., each pair of as illustrated in Fig.5, which is symmetric under the spatial conjugate variables satisfies the canonical commutation rela- parity transformation along the direction perpendicular to the tions. The transverse part of the quadratic Hamiltonian can plane, i.e., under the reflection (x, y, z) → (x, y, −z). 18

TABLE II. Eigenmodes in the cavity configuration. The labels of the bands are consistent with those shown in Fig.6. The polarization of each mode is specified by Λ ∈ {TM, TE, L}, where TM and TE indicate the transverse magnetic and electric polarizations, respectively, and L ⊥ k denotes the longitudinal polarization. The parity of the mode functions uη and fη are specified by P = ± (see also the main text). We also indicate whether or not the divergence of the electric field ∇ · Eη vanishes. The last column shows the explicit functional form of the mode function for each eigenmode, where ei indicates the unit vector in the direction i = x, y, z.

Bands Λ P ∇ · Eη Mode function ⊥ I q u (z) = u cos(κηz)ex − i sin(κηz)ez [ϑ(d/2 − z) + ϑ(d/2 + z) − 1] η η κη TM2, RTMn=0,2,4..., SS (=TM0) TM + zero M P −sνη z q +u e ex + is ez ϑ(sz − d/2) η s=±1 νη ⊥ I q u (z) = u sin(κηz)ex + i cos(κηz)ez [ϑ(d/2 − z) + ϑ(d/2 + z) − 1] η η κη TM1, RTMn=1,3,5..., AS1,2 TM − zero M P −sνnz q +u e sex + i ez ϑ(sz − d/2) η s=±1 νn ⊥ I uη (z) = uη cos(κηz)ey [ϑ(d/2 − z) + ϑ(d/2 + z) − 1] TE1, RTEn=0,2,4... TE + zero M P −sνη z +uη s=±1 e eyϑ(sz − d/2) ⊥ I uη (z) = uη sin(κηz)ey [ϑ(d/2 − z) + ϑ(d/2 + z) − 1] TE0, RTEn=1,3,5... TE − zero M P −sνnz +uη s=±1 se eyϑ(sz − d/2) k I q iκη z −iκη z κη iκη z −iκη z fη (z) = f √ (e ± e )ex + √ (e ∓ e )ez η κ2 +q2 κ2 +q2 L± L ± nonzero η η × [ϑ(d/2 − z) + ϑ(d/2 + z) − 1]

1. Transverse modes present case can then be given by

2 We first consider the transverse modes. In this case, the −ωη (ωη, z) u⊥ (z)+e−iq·ρ∇ × ∇ × eiq·ρu⊥ (z) = 0 vector field can be quantized and expanded as c2 η η (C7) s X Aˆ(r) = ~ aˆ U ⊥ (r) + H.c. , (C1) 2 Aω η η with the relative permittivity being defined by η 0 η X where A is the area of the system, η ∈ {q, n, λ} labels each (ω, z) = p(ω) ϑ(sz − d/2) eigenmode in the cavity setting, aˆη is its annihilation oper- s=± ator, ω is an eigenfrequency, and U is the corresponding ! η η X ⊥ +ξ(ω) ϑ(d/2 − sz) − 1 . (C8) transverse eigenmode function satisfying ∇ · U η = 0. Here, q is the two-dimensional in-plane wavevector, n ∈ N la- s=± bels a discrete eigenmode in each sector λ = (Λ,P ) with Λ ∈ {TM, TE} specifying the transverse magnetic (TM) or Here, we use the Heaviside step function the transverse electric (TE) polarization defined by the condi- ( tions, 1 (x > 0) ϑ(x) = , (C9) 0 (x < 0) h ⊥ i ∇ × U η (r) = 0 (TM modes), (C2) z h ⊥ i and assume the relative permittivity functions in the metals U η (r) = 0 (TE modes), (C3) z (|z| > d/2) and the insulator (|z| < d/2) as and P = ± specifying the parity symmetry of the eigenmode ω2 function under the reflection, (ω) = 1 − p , (C10) p ω2 [U ⊥(x, y, −z)] = P [U ⊥(x, y, z)] , (C4) g2 η x,y η x,y ξ(ω) = 1 + . (C11) ⊥ ⊥ Ω2 − ω2 [U η (x, y, −z)]z = −P [U η (x, y, z)]z. (C5)

Because of the spatial invariance in the x and y directions, We recall that ωp is the plasma frequency in the metal mir- we can decompose the eigenfunction as rors, Ω is the frequency of the nondispersive matter excitation such as an optical phonon mode in ionic crystals, and g char- ⊥ iq·ρ ⊥ U η (r) = e uη (z), (C6) acterizes the strength of the light-matter coupling between the matter dipolar mode and cavity electromagnetic fields. It is where ρ = (x, y)T. The eigenvalue problem (B7) in the also useful to express the electric and magnetic fields in terms 19 of the eigenmodes as follows: define the skin wavenumber νη, which is real and positive, as follows: r ω ˆ X ~ η r 2 E(r) = (âηEη (r) + H.c.) , (C12) ωη 20A ν = q2 − (ω ) > 0. (C25) η η c2 p η ⊥ Eη (r) = iU η (r) , (C13) M I The coefficients uη and uη in the metals (M) and in the insu- s lator (I), respectively, satisfy the following relations imposed ˆ X ~ B(r) = (âηBη (r) + H.c.) , (C14) by the boundary conditions: 2 Aω η 0 η M I νη d/2 ⊥ uη /uη = e cos (κηd/2) Bη (r) = ∇ × U η (r) . (C15) ξ (ω ) ν η η νη d/2 = − e sin (κηd/2) . (C26) The boundary conditions require the continuity of n×Eη and p (ωη) κη n · [0Eη] across the interfaces with n being the unit vector perpendicular to the surfaces. The latter condition is equiva- Aside an irrelevant phase factor, these coefficients can lent to the continuity of the magnetic field orthogonal to the be fixed by further imposing the normalization condition R ⊥ 2 surface normal n × Bη. dz uη (z) = 1. In the same way as done above, we can also obtain the corresponding eigenmodes in the case of the odd parity P = −. a. TM modes Their functional forms are given in the second line in TableII. The corresponding coefficients satisfy the boundary condi- Without loss of generality, we here consider the TM- tions polarized eigenmodes satisfying M I νη d/2 uη /uη = e sin (κηd/2) ξ (ω ) ν [Bη] 6= 0, [Bη] = 0, (C16) η η νη d/2 y x,z = e cos (κηd/2) , (C27) p (ωη) κη [Eη]y = 0, [Eη]x,z 6= 0. (C17) R ⊥ 2 and also the normalization condition dz uη (z) = 1. The Maxwell equation ∇ · Bˆ = 0 then leads to q = 0, i.e., y These modes are labeled as TM1, RTMn=1,3,5..., AS1,2 that the eigenmodes propagate along the x axis and thus we denote are shown in Fig.6(b,e). We show typical spatial profiles of q = qx. The parity conditions (C4) and (C5) in the present the electric fields at low and high in-plane momenta q for each case can be read as of these TM modes in Fig.9.

[Bη(x, y, −z)]y = −P [Bη(x, y, z)]y , (C18)

[Eη(x, y, −z)]x = P [Eη(x, y, z)]x , (C19) b. TE modes

[Eη(x, y, −z)]z = −P [Eη(x, y, z)]z . (C20) In the case of the TE polarization, without loss of generality ⊥ In terms of the mode function uη , the TM modes considered we can assume the following conditions: here satisfy [Eη]y 6= 0, [Eη]x,z = 0, (C28) ⊥ uη (z) = 0, (C21) y [Bη]y = 0, [Bη]x,z 6= 0. (C29) u⊥(−z) = P u⊥(z) , (C22) η x η x From the transversality condition, ⊥ ⊥ uη (−z) = −P uη (z) . (C23) iq·ρ ⊥ z z ∇ · (e uη ) = 0, (C30) We first consider the case of the even parity P = +. Solv- and the Maxwell equation ∇ · Dˆ = 0, we obtain the condition ing the eigenvalue equation (C7) for the mode function u⊥ η q = 0, i.e., the TE modes considered here also propagate iq·ρ ⊥ y with the transversality condition, i.e., ∇ · (e uη ) = 0, along the x axis. The electromagnetic fields of the TE modes ⊥ we obtain the explicit functional form of uη for this class of transform under the parity transformation as eigenmodes as shown in the first line of TableII. This set in- [E (x, y, −z)] = P [E (x, y, z)] , (C31) cludes the bands labeled as TM2, RTMn=0,2,4..., SS (=TM0) η y η y that are shown in Fig.6(a,d). In these modes, the longitudinal [Bη(x, y, −z)]x = −P [Bη(x, y, z)]x , (C32) wavenumber κη in the insulator region is defined by [Bη(x, y, −z)]z = P [Bη(x, y, z)]z . (C33) 2 2 ωη 2 These conditions can be read in terms of the mode function κη = 2 ξ (ωη) − q . (C24) ⊥ c uη as

We emphasize that κη takes either a real value or a purely ⊥ uη (z) = 0, (C34) imaginary value. In the latter case, the eigenmodes are lo- x,z u⊥(−z) = P u⊥(z) . (C35) calized at the interfaces and represent the surface modes. We η y η y 20

FIG. 9. Spatial profiles of the TM modes in the cavity configuration at low and high in-plane momenta q. The black dashed vertical lines ⊥ indicate the metal-insulator interfaces. The blue solid curve shows the spatial profile of the out-of-plane component [iuη (z)]z of the mode function (corresponding to the z component of the electric field). The red solid curve shows the spatial profile of the in-plane component ⊥ [uη (z)]x of the mode function (corresponding to the x component of the electric field). The parameters are the same as in Fig.6 while the in-plane momentum q for low- and high-momentum results is chosen to be q = 0.1 Ω/c and q = 7 Ω/c, respectively.

The eigenvalue equation then leads to the functional form in Because of our choice of the Coulomb gauge Aˆk = 0, the lon- the third (fourth) line in TableII in the case of the even par- gitudinal field is directly linked with the longitudinal mode of ity P = + (the odd parity P = −). The boundary conditions the matter field φˆk as discussed in the main text (cf. Eq. (28)). ⊥ require that the value of the mode function uη and its first spa- From the Maxwell equation ∇ · Dˆ = 0, we can see that tial derivative must be continuous across the interfaces. More this mode oscillates in time at a constant frequency Ωk cor- specifically, in the case of the even parity they lead to the con- responding to the vanishing permittivity in the insulator ditions on the coefficients uM,I, η ξ Ωk = 0, (C39) M I νη d/2 uη /uη = e cos (κηd/2) k p 2 2 νη Ω = Ω + g . (C40) = − eνη d/2 sin (κ d/2) , (C36) κ η η We can expand the longitudinal matter field in terms of the while in the odd-parity case we have to impose the conditions mode function as r M I νη d/2 X k uη /uη = e sin (κηd/2) ˆk ~ ˆ iq·ρ k φ (r) = φqnλe fη (z) + H.c. , ν 2MNdΩ η νη d/2 η = e cos (κηd/2) . (C37) κη (C41)

R ⊥ 2 The normalization condition dz uη (z) = 1 can be also where the longitudinal mode function satisfies imposed on the coefficients. In Fig.6(c,f), the TE modes with iq·ρ k the even parity are labeled as TE1, RTEn=0,2,4... while the ∇ × (e fη (z)) = 0. (C42) ones with the odd parity are labeled as TE0, RTEn=1,3,5.... ˆ ∂Bˆ In contrast to some of the TM modes (such as the AS1,2 and Meanwhile, the Maxwell equation ∇×E = − ∂t ensures the the RTM0 modes), the spatial profiles of the electric fields are vanishing magnetic field Bˆ = 0. Thus, it suffices to only con- qualitatively insensitive to the in-plane momenta for all of the sider the longitudinal electric field and, without loss of gener- TE modes; their typical results are shown in Fig. 10. ality, we here consider the modes satisfying

[Eη]y = 0, [Eη]x,z 6= 0, (C43) 2. Longitudinal mode Bη = 0. (C44)

The other type of solutions is the longitudinal mode that is The corresponding in-plane momenta point to the x direction, characterized by the longitudinal electric ﬁeld i.e., qy = 0 and thus q = qx. The boundary conditions at the interfaces lead to the vanishing electromagnetic ﬁelds in the ∇ × Eη = 0. (C38) metal mirrors. The resulting explicit functional form of the 21

(a) (b) in-plane in-plane RTE RTM0 low q RTM0 high q 3 ] ] d d [a.u.] [a.u.] [ [ TE 1

TE 0 RTE2 z z ﬁeld ﬁeld RTE5 RTE1

RTE4 RTE0 x [1/q] x [1/q] (c) (d) out-of-plane z [d] z [d] in-plane out-of-plane field field in-plane FIG. 10. Spatial profiles of the TE modes in the cavity configuration. The black dashed vertical lines indicate the metal-insulator z [d] z [d] k interfaces. The mode functions fη of all the TE modes have only the in-plane components, whose spatial profiles are shown by the red FIG. 11. (a,b) Electrical flux lines and (c,d) spatial profiles of the solid curve (corresponding to the y component of the electric field). mode function (i.e., the electric field) for the most-softened phonon The parameters are the same as in Fig.6. We use q = 0.1 Ω/c for mode denoted as the RTM0 mode. Without loss of generality, we the sake of concreteness while the spatial profiles of the TE modes consider the mode propagating in the x direction with the in-plane are qualitatively insensitive to a specific choice of the in-plane mo- momentum q = qx and the vanishing electric field in the y direction mentum q. Ey = 0. (a,b) The black solid curved arrows indicate the electrical flux lines on the xz plane at (a) low and (b) high in-plane momenta q. The color indicates the magnitude of the electric field Ez perpendic- mode function is given in the last line in TableII. Depending ular to the interfaces. (c,d) The spatial profiles of the in-plane (red solid curve) and the out-of-plane (blue solid curve) components of on the parity P = ± defined via the relations the electric field at x = 0 in (a) and (b) are plotted in (c) and (d) in the h k i arbitrary unit, respectively. The black dashed vertical lines indicate fη (z) = 0, (C45) the interfaces at |z| = d/2. The parameters are d = c/Ω, g = 3.5 Ω, y and ωp = 5 Ω. We set q = 0.1 Ω/c in (a,c) and q = 7 Ω/c in (b,d). h k i h k i fη (−z) = P fη (z) , (C46) x x h k i h k i fη (−z) = −P fη (z) , (C47) down to zero if and only if the bare phonon frequency also z z vanishes Ω → 0, indicating that the divergence of the Drude there are two classes of solutions, which are labeled as the L± permittivity is necessarily compensated by that of ξ(ω) (cf. modes in Fig.6(a). The corresponding longitudinal wavenum- Eq. (D1)). Secondly, it is worthwhile to note that the Kramers- ber κη can be simply obtained by the boundary conditions Kronig relations can in principle be recovered even in the loss- h k i less limit by extending the permittivity as follows [196]: f η(±d/2) = 0. x 1 (ω) = 1 − ω2 + iπδ(1)(ω) , (D2) p p ω2 Appendix D: Loss effect on the softened polariton mode where δ(1) is the first derivative of the Dirac delta function. We here provide further details about the analysis of the While this additional contribution can be of physical impor- cavity loss. To be specific, we employ the Drude model and tance when one is interested in real-time dynamics, it is irrel- use a phenomenological complex permittivity. The resulting evant to the main results discussed in the present paper. complex eigenvalue equation at q = 0 is

q 2 2 κ ωp/(ω + iωγ) − 1 Appendix E: Details about physical properties of the hybridized ω = , (D1) modes ξ(ω) tan(κd/2) where we recall that γ > 0 represents the cavity-loss rate. We here discuss full details about physical properties of The results in Fig.7 in the main text are obtained by solving the hybridized elementary modes. Firstly, to gain further in- Eq. (D1) for the solution that reduces to the RTM0 mode in sights into the most-softened mode, in Fig. 11 we plot the the loss-less limit γ → 0. spatial profiles of the electric field E of the RTM0 mode. Fig- Two remarks are in order. Firstly, while the imaginary part ures 11(a) and (b) show the electrical flux lines of the RTM0 of the complex Drude permittivity diverges in ω → 0, this mode at low and high in-plane momenta q, respectively, with does not mean that an eigenmode should have a vanishingly the color indicating the magnitude of the out-of-plane com- short lifetime. The reason is that an eigenfrequency can go ponent Ez of the electric field. We note that our convention 22

of denoting the z-component as the out-of-plane one is in ac- 0.94 g/Ω=0.5 cordance with the planar cavity setting whose metal-insulator interfaces are parallel to the xy plane. Without loss of generality, here we focus on the mode propagating in the x direc- ] Ω [ 0.935 tion, i.e. q = qx, leading to the condition Ey = 0 (see Ap- 0.73 pendixC). Figures 11(c) and (d) show the low- q and high-q g/Ω=1.5 spatial proﬁles of the in-plane (red solid curve) and the out- 0.72

of-plane (blue solid curve) components of the electric field ω at x = 0. At low q before crossing the lower bulk dispersion, the field predominantly points to the in-plane direction 0.71 and extends over the insulator region |z| ≤ d/2 (Fig. 11(c)). g/Ω=4 In contrast, after crossing the bulk dispersion, the longitudinal 0.4 wavenumber κη turns out to take a purely imaginary value and the field profiles qualitatively change. Specifically, at high q 0.39 the fields are localized around the interfaces and the z com- 0 1 2 3 4 5 ponent of the electric field exhibits significant discontinuities q [Ω/c] due to the surface charges concentrated on the metal interfaces (Fig. 11(d)); these features are characteristic of surface FIG. 12. The RTM0 mode at different light-matter coupling polariton modes. As shown in the main text, it is the soft- strengths g with a low plasma frequency. The strong hybridization ening of the RTM0 mode discussed here that will trigger the with plasmons leads to the emergence of the minimum of the energy structural instability of materials, thus ultimately inducing the dispersion at finite in-plane momentum q∗ > 0, indicating the for- paraelectric-to-ferroelectric phase transition. mation of roton-type excitations. The parameters are d = c/Ω and ω = 2 Ω. We next consider the symmetric and antisymmetric surface p modes denoted as SS(=TM0) and AS2 in Figs.6(a) and (b), respectively. The assignment of the label TM0 to the SS mode is in accordance with the convention of studies in plasmonics low the plasma frequency and has a physical origin similar [147]. At a high in-plane wavevector q, the SS and AS2 modes to the TM2 mode, but there still exists a difference that the become bound modes that are localized around the interfaces; TM1 mode sustains the nonvanishing hybridization with the ⊥ the field amplitudes take maxima at the interfaces and expo- electromagnetic fields at low q. The mode functions uη of nentially decay away from them. In the limit of q → ∞, the TM1,2 modes and the corresponding dipole moments pre- the energy dispersions for both of the SS and AS2 modes be- dominantly point to the directions parallel to the interfaces at come flat and saturate to the same, finite frequency below ωp. low q. The longitudinal wavenumber κη remains real over the Meanwhile, at q = 0 the AS2 mode starts from a frequency entire in-plane momentum space, and thus the amplitudes of lower than that of the SS mode, and thus the former exhibits the TM1,2 modes extend over the insulator region (see also a larger dispersion. The SS mode can have a real longitudinal Fig.9 in AppendixC). wavenumber κη at low q and thus it can be radiative, i.e., it When the plasma frequency is close to the phonon fre- can be coupled to light modes outside the cavity when metal quency and the hybridization with plasmons becomes partic- mirrors have finite thicknesses. As a result, the SS mode can ularly prominent, the above crossover behavior of the RTM0 contribute to optical properties of the system. With increasing mode also manifests itself as the emergence of roton-type ex- q, the longitudinal wavenumber κη changes from a real value citations. To see this explicitly, in Fig. 12 we show the energy to a purely imaginary value and the SS mode changes into the dispersion of the RTM0 mode at different light-matter cou- nonradiative mode. In contrast, the AS2 mode is nonradia- pling strengths g with a low plasma frequency. It is evident tive over the entire in-plane momentum and can potentially that the energy dispersion takes the minimum value at finite be applied to nanoscale plasmonic waveguides owing to its in-plane momentum q∗ > 0, indicating the formation of roton localized and dispersive nature. The difference between two excitations. Physically, this minimum roughly corresponds to surface modes at low q also results in their qualitatively dif- the vanishing point of the longitudinal wavenumber κη = 0, ferent low-energy polarization behavior; the dipole moment of at which the crossover from low to high in-plane momentum the SS mode points to the in-plane direction while that of the behavior occurs (compare Figs. 11(c) with (d)). While a spe- ∗ AS2 mode is purely out-of-plane. It is worthwhile to mention cific value of the finite in-plane momentum q depends on that in the context of plasmonics the nanoscale localization of system parameters such as the thickness of the insulator, it is such surface modes has previously been utilized for squeezing a general feature that roton excitations eventually disappear light waves below the diffraction limit [197, 198]. (i.e., q∗ → 0) as the plasmon component in the eigenmode

The highest band (denoted as TM2 in Fig.6(a)) starting becomes less significant by, for instance, increasing either the from the plasma frequency ωp at q = 0 is essentially the upper light-matter coupling strength g or the plasma frequency ωp. branch of the standard plasmon polaritons. This hybridized The lowest mode in the TM polarization, which is denoted mode reduces to the bare plasmon (photon) excitation in the as AS1, lies slightly below the lower bulk dispersion (black limit of q → 0 (q → ∞). The next highest band (denoted dotted curve) and converges to the linear photon dispersion as TM1 in Fig.6(b)) initiates from a frequency slightly be- at q → 0 in contrast to the other modes featuring quadratic 23 asymptotes. This mode essentially shares the similar proper- (a) (b) ties with the AS2 mode; both of them are localized modes over the entire in-plane momentum q and point out of the plane at S’ low q due to the surface charges. The AS1 mode asymptoti- Metal cally becomes a purely photon mode at q → 0. ωS’ =2ω T The longitudinal modes L± in Fig.6(a) have the parity ω Molecular solid P = ± and lie at the constant frequency Ωk = pΩ2 + g2. These nondispersive modes arise from the vanishing insulator S permittivity and is characterized by the nonzero divergence of T iq·ρ k Metal S’’ the mode function ∇ · (e f η) 6= 0. Both the in-plane and out-of-plane amplitudes of the longitudinal electric field ex- q tend over the insulator region while the amplitudes vanish in the metal regions due to the boundary conditions. At low q, FIG. 13. Possible applications of the cavity setting to molecular the out-of-plane component becomes dominant in the insula- and excitonic systems. (a) Schematic figure illustrating the proposed tor region due to the surface charges. setup, in which molecular solids or assembled molecules interact Finally, we discuss the TE-polarized bands of both parities with light modes confined between metal mirrors. (b) Schematic fig- P = ± plotted in Fig.6(c). The corresponding magnified ure illustrating applications of the elementary excitations in the cav- plot around the phonon frequency is also shown in Fig.6(f). ity setting to an organic light-emitting device and a photon-exciton These TE modes have simpler structures than the TM ones, conversion. The black horizontal solid (dotted) line denoted as T (S) i.e., each of all the modes lies slightly above the higher or indicates a bare nondispersive triplet (singlet) energy of a molecular lower branches of the bulk dispersions (black dotted curves) state or an electron-hole pair. Only the singlet state S couples to con- due to energy enhancements by the cavity confinement. We fined light modes, resulting in the split into the multiple branches of label the higher set of the TE modes as TE while the lower the in-plane dispersions. A lower dispersion (red solid curve denoted 0,1 as S00) can potentially be applied to enhance the efficiency of organic set of a large number of resonant TE modes as RTEn=0,1,2.... light-emitting devices. Tuning the in-plane momentum q or the thick- Because the mode functions (and thus the electric fields) for ness of the layer, a higher dispersion (blue solid curve denoted as S0) all the modes are continuous even at the interfaces, the ori- can be used to realize the resonance condition ωS0 = 2ωT, leading gin of the enhanced energies in these TE modes can simply to a conversion of a photon into a pair of triplet excitons as denoted be understood as the acquisition of nonzero longitudinal mo- by two dashed black arrows. mentum associated with the cavity confinement. Indeed, the corresponding longitudinal wavenumber κη remains real for all the TE modes over the entire region of the in-plane mo- be utilized to improve the efficiency of organic light-emitting mentum q. This real-valuedness of κη also indicates that the devices (LEDs) and to realize exciton-photon converters. TE modes are essentially radiative photonic modes that can We consider a crystal of polar molecules surrounded by directly couple to light modes outside the system when the metallic electrodes as shown in Fig. 13. The relevant infrared thicknesses of the metal mirrors are taken to be finite. active mode can be either an electronic excitation, such as the E1 electronic transition, or a vibrational excitation of the molecular crystal. While these two types of excitations have Appendix F: Applications of cavity-based control of polariton very different frequencies (i.e., optical vs mid-infrared), the dispersions general formalism presented in Sec. III should be applicable to both cases. Either one of these modes can be introduced as Substantial modifications of the polariton dispersions in the a nondispersive mode having the frequency Ω in Eq. (12). To cavity geometry (see e.g., Fig.6 in Sec. III) are generic fea- be concrete, we focus on the case of electronic excitations, and tures for collective dipolar modes coupled to confined light argue that the proposed cavity configuration can be applied to modes. In this respect, the cavity-based control of energy- improve efficiencies in molecular LEDs. level structures can potentially be applied to a wide range A major challenge in organic LEDs is that the lowest exci- of systems such as ionic crystals, molecular solids, assem- tonic state is a spin triplet, which is optically dark (i.e., it re- bled molecules, and excitonic systems. Organic molecules are combines through nonradiative channels). The reason for this particularly promising for such applications since their exci- is that Hund’s rule for electrons in a molecule favors the spin tations have large dipole moments and crystals can achieve alignment and thus the spin triplet excitonic states are lower high-molecular densities [199]. Recent experiments have in energy than the singlet ones. Having a practical method of demonstrated the strong coupling between a single molecule reversing the order of excitonic states by making the singlet and a cavity mode even for the modest cavity quality fac- state of an exciton lower in energy will have important impli- tors [131, 200]. This means that the strength of the cou- cations for improving the efficiency of organic LEDs [122]. In pling between the molecular optical excitation and the cav- the cavity configuration, since triplet excitonic states are dark, ity mode exceeds the decoherence rate of the excitations they do not couple to light and are not strongly modified. In [69, 90, 106, 201]. Besides a case study of the cavity- contrast, the singlet excitonic state hybridizes with confined enhanced ferroelectricity in Sec.IV, we discuss below that light modes and forms the softened collective modes such as the cavity-induced changes in the hybridized dispersion can the RTM0 mode shown in Fig.6(a). This should make it pos- 24 sible to have the lowest excitonic state as a spin singlet, thus The second term δEeff,1 contributes from an expectation value realizing efficient organic LEDs (see the red solid curve in of the quadratic terms in the energy difference Hˆtot(r) − Fig. 13(b)). While this type of hybridizations have been dis- Hˆtot,0(reff ) with respect to the variational state, resulting in cussed in the literature before [122], the novelty of the present consideration is the inclusion of the continuum of both collec- δEeff,1 (r, reff ) = tive matter excitations and cavity light modes in a slab of a fi- ~ (r − reff ) X β~ωqnλ(reff ) nite thickness as well as taking into account the dispersive na- √ cqnλ coth . (G3) 4 reff 2 ture of the hybridized excitations there. In general, the disper- qnλ sive nature could be utilized to further tune the selective con- Here, positive coefficients cqnλ are defined by trol of chemical reactivity landscapes in order to enhance or Z d/2 suppress certain types of photochemical reactions. Remark- 2 cqnλ ≡ ζ(ωqnλ) dz |fqnλ(z)| , (G4) ably, an inversion of the singlet-triplet polaritons has recently −d/2 been demonstrated in organic microcavities [125]; it is inter- 4 4 2 2 where ζ(ω) = g /[g + (reff − ω ) ]. The mode func- esting to consider whether such a technique can be transferred ⊥ to continuum collective excitations associated with nontrivial tion fqnλ(z) represents fqnλ for transverse modes with Λ = k dispersions as discussed here. TM, TE, while it represents fqnλ for longitudinal modes with We next discuss a complementary idea of using a cavity to Λ = L; their explicit functional forms are given in Ap- facilitate the conversion of photons into pairs of triplet exci- pendixC. The final term δEeff,2 results from the nonlinear tons. This process could be useful for generating entangled term in the microscopic Hamiltonian; after using the Cauchy- pairs of excitations. At ambient conditions, such conversion Schwartz inequality, we can bound it as process is challenging to realize because the resonant absorp- δE (r ) ≤ tion of light occurs at the energy of singlet excitons and this eff,2 eff energy is only slightly above the energy of the triplet excitons;  2 2U β ω (r ) typical energy spacing is characterized by the Hund’s splitting ~ X ~ qnλ eff 2  c˜qnλ coth ,(G5) 4M reff N 2 and is an order of a few meV. qnλ This limitation can be overcome by utilizing the proposed cavity setup. As discussed before, when a molecular solid where we introduce positive coefficients s is confined in the cavity, the singlet exciton hybridizes with Z d/2 4 light modes while the triplet one does not. This can lead to c˜qnλ ≡ ζ(ωqnλ) d dz |fqnλ(z)| . (G6) the formation of a hybridized singlet mode above the origi- −d/2 nal exciton energy (see e.g., the blue solid curve S’ above the Summing up the right-hand sides of Eqs. (G2), (G3), and black dashed horizontal line S in Fig. 13(b)). One may, for (G5), we obtain the variational free energy. While strictly example, use the SS(=TM0) and AS2 modes for this purpose. Owing to the nontrivial dispersion of such a hybridized mode, q d one can tune the in-plane momentum (and the thickness or 2 the plasma frequency ωp if necessary) such that a hybridized singlet energy ω 0 becomes equal to twice of the triplet energy S 1.5 d decrease 2ωT and thus the conversion process of a single photon into two triplet excitons becomes resonant (cf. the dashed black arrows in Fig. 13(b)). One can create singlet exciton-polaritons ω 1 at finite in-plane momenta by, e.g., having incident light at nonzero angle or constructing a grating on the surface of the 0.5 metals. 0 0 1 2 3 4 5 6 7 Appendix G: Details of the variational analysis q

We here provide technical details about the variational anal- FIG. 14. Softening of the phonon mode relevant to the paraelectric- ysis. As explained in the main text, we minimize the following to-ferroelectric phase transition, which is labeled as the RTM0 mode variational free energy with respect to a variational parameter in Fig.6. The solid colored curves represent the in-plane dispersions reff (cf. Eq. (47)): of the softened phonon mode with decreasing the insulator thickness d from top to bottom curves. Each dispersion is obtained by set- 1 2 F = − ln Z + δE + δE . ting the phonon-frequency parameter Ω equal to the effective value v 0 eff,1 eff,2 (G1) ∗ β reff determined from the variational analysis for each thickness (cf. Here, the first term corresponds to the free energy for the non- Fig.8(a)). The black dashed line represents the effective phonon frequency in the bulk case. We set c = ~ = β = 1, ωp = 5, interacting theory with 2 U = 1, g = 3.5, and Λc = 10 . The tuning parameter is fixed to h i Y −βωqnλ(reff ) be r = −45. From top to the bottom solid curves, the thickness is Z0 (reff ) = 1 − e . (G2) varied from d = 4 to d = 1 with step δd = 1. qnλ 25 speaking this gives the upper bound of Eq. (47), in this pa- parameter r (see Fig. 14). It is this softening that destabilizes per we term this summation as the variational free energy Fv the paraelectric phase when it goes down to zero energy, trig- for the sake of notational simplicity. gering the phase transition. Thus, the unstable dipolar phonon One possible way to infer the quality of the present varia- mode at the verge of the transition is precisely the transverse tional theory is to check the convexity of the variational free one and points to the in-plane direction (cf. Fig. 11(a,c)). We energy and its variance around the global minimum reff = emphasize that the results in Fig. 14 are obtained by using ∗ ∗ p ∗ reff . In general, when the free-energy landscape turns out to the renormalized effective phonon frequency Ωeff = reff be nonconvex and has several local optima, it should raise the for each thickness d, which is determined from identifying question about the validity of a choice of variational states. the minimum of the variational free energy including the non- However, in the present analysis, the free-energy landscape is linear interaction in a self-consistent manner (cf. Fig.8(a)). very well behaved and has only a single global minimum. In We note that all the contributions from the ‘spectator’ modes, particular, the landscape around the global minimum becomes i.e., the modes other than the RTM0, are also included in the ∗ sharper as we approach to the transition point at reff = 0. variational analysis to determine the renormalized phonon fre- These features can be inferred from Fig.8(a) in the main text. quency. As shown in Fig. 14, to realize the ultimate softening In addition to plotting the variational free energies and the of the RTM0 mode, a sparser band structure as realized with a resulting phase diagrams as shown in Fig.8, it is also useful thinner layer is preferable since the phonon hardening due to to plot the most-softened phonon mode (i.e., the RTM0 mode) the nonlinear repulsive interaction between the energetically with varying the insulator thickness d, but at the fixed tuning dense elementary modes can be mitigated.

∗ [email protected] trafast route to lattice control, Nat. Phys. 7, 854 (2011). [1] C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg, Pho- [15] R. Mankowsky, A. Subedi, M. Forst,¨ S. O. Mariager, M. Chol- tons and Atoms (Wiley, New York, 1989). let, H. Lemke, J. S. Robinson, J. M. Glownia, M. P. Minitti, [2] T. Kampfrath, K. Tanaka, and K. A. Nelson, Resonant and A. Frano, M. Fechner, N. A. Spaldin, T. Loew, B. Keimer, nonresonant control over matter and light by intense terahertz A. Georges, and A. Cavalleri, Nonlinear lattice dynamics as transients, Nat. Photon. 7, 680 (2013). a basis for enhanced superconductivity in YBa2Cu3O6.5, Na- [3] P. W. Anderson and A. H. Dayem, Radio-frequency effects in ture 516, 71 (2014). superconducting thin film bridges, Phys. Rev. Lett. 13, 195 [16] R. Matsunaga, N. Tsuji, H. Fujita, A. Sugioka, K. Makise, (1964). Y. Uzawa, H. Terai, Z. Wang, H. Aoki, and R. Shimano, [4] A. F. G. Wyatt, V. M. Dmitriev, W. S. Moore, and F. W. Light-induced collective pseudospin precession resonating Sheard, Microwave-enhanced critical supercurrents in con- with Higgs mode in a superconductor, Science 345, 1145 stricted tin films, Phys. Rev. Lett. 16, 1166 (1966). (2014). [5] G. M. Eliashberg, Film superconductivity stimulated by a [17] A. F. Kemper, M. A. Sentef, B. Moritz, J. K. Freericks, and high- frequency field, JETP LETT. 11, 114 (1970). T. P. Devereaux, Direct observation of Higgs mode oscillations [6] B. I. Ivlev, S. G. Lisitsyn, and G. M. Eliashberg, Nonequilib- in the pump-probe photoemission spectra of electron-phonon rium excitations in superconductors in high-frequency fields, mediated superconductors, Phys. Rev. B 92, 224517 (2015). J. Low Temp. Phys. 10, 449 (1973). [18] M. Mitrano, A. Cantaluppi, D. Nicoletti, S. Kaiser, A. Peruc- [7] A. Schmid, Stability of radiation-stimulated superconductiv- chi, S. Lupi, P. Di Pietro, D. Pontiroli, M. Ricco,` S. R. Clark, ity, Phys. Rev. Lett. 38, 922 (1977). D. Jaksch, and A. Cavalleri, Possible light-induced super- [8] T. Oka and H. Aoki, Photovoltaic hall effect in graphene, conductivity in K3C60 at high temperature, Nature 530, 461 Phys. Rev. B 79, 081406 (2009). (2016). [9] T. Kitagawa, T. Oka, A. Brataas, L. Fu, and E. Demler, Trans- [19] E. Pomarico, M. Mitrano, H. Bromberger, M. A. Sentef, A. Al- port properties of nonequilibrium systems under the applica- Temimy, C. Coletti, A. Stohr,¨ S. Link, U. Starke, C. Cacho, tion of light: Photoinduced quantum hall insulators without R. Chapman, E. Springate, A. Cavalleri, and I. Gierz, En- landau levels, Phys. Rev. B 84, 235108 (2011). hanced electron-phonon coupling in graphene with periodi- [10] N. H. Lindner, G. Refael, and V. Galitski, Floquet topological cally distorted lattice, Phys. Rev. B 95, 024304 (2017). insulator in semiconductor quantum wells, Nat. Phys. 7, 490 [20] A. A. Patel and A. Eberlein, Light-induced enhancement (2011). of superconductivity via melting of competing bond-density [11] Y. H. Wang, H. Steinberg, P. Jarillo-Herrero, and N. Gedik, wave order in underdoped cuprates, Phys. Rev. B 93, 195139 Observation of floquet-bloch states on the surface of a topo- (2016). logical insulator, Science 342, 453 (2013). [21] M. Knap, M. Babadi, G. Refael, I. Martin, and E. Demler, Dy- [12] J. W. McIver, B. Schulte, F.-U. Stein, T. Matsuyama, G. Jotzu, namical Cooper pairing in nonequilibrium electron-phonon G. Meier, and A. Cavalleri, Light-induced anomalous hall ef- systems, Phys. Rev. B 94, 214504 (2016). fect in graphene, Nat. Phys. 16, 38 (2020). [22] A. Komnik and M. Thorwart, BCS theory of driven supercon- [13] M. Rini, N. Dean, J. Itatani, Y. Tomioka, Y. Tokura, R. W. ductivity, Eur. Phys. J. B 89, 244 (2016). Schoenlein, and A. Cavalleri, Control of the electronic phase [23] M. A. Sentef, A. Tokuno, A. Georges, and C. Kollath, The- of a manganite by mode-selective vibrational excitation, Na- ory of laser-controlled competing superconducting and charge ture 449, 72 (2007). orders, Phys. Rev. Lett. 118, 087002 (2017). [14]M.F orst,¨ C. Manzoni, S. Kaiser, Y. Tomioka, Y. Tokura, [24] M. A. Sentef, Light-enhanced electron-phonon coupling from R. Merlin, and A. Cavalleri, Nonlinear phononics as an ul- nonlinear electron-phonon coupling, Phys. Rev. B 95, 205111 26

(2017). (2009). [25] D. M. Kennes, E. Y. Wilner, D. R. Reichman, and A. J. Mil- [48] J. Keeling, Coulomb interactions, gauge invariance, and lis, Transient superconductivity from electronic squeezing of phase transitions of the dicke model, J. Phys. Cond. Matt. 19, optically pumped phonons, Nat. Phys. 13, 479 (2017). 295213 (2007). [26] N. Tancogne-Dejean, M. A. Sentef, and A. Rubio, Ultrafast [49] T. Grießer, A. Vukics, and P. Domokos, Depolarization shift modification of Hubbard U in a strongly correlated material: of the superradiant phase transition, Phys. Rev. A 94, 033815 Ab initio high-harmonic generation in NiO, Phys. Rev. Lett. (2016). 121, 097402 (2018). [50] T. Jaako, Z.-L. Xiang, J. J. Garcia-Ripoll, and P. Rabl, [27] J. Zhang and R. Averitt, Dynamics and control in complex Ultrastrong-coupling phenomena beyond the dicke model, transition metal oxides, Annu. Rev. Mat. Res. 44, 19 (2014). Phys. Rev. A 94, 033850 (2016). [28] A. Cavalleri, Photo-induced superconductivity, Contemp. [51] D. De Bernardis, T. Jaako, and P. Rabl, Cavity quantum elec- Phys. 59, 31 (2018). trodynamics in the nonperturbative regime, Phys. Rev. A 97, [29] H. Patel, L. Huang, C.-J. Kim, J. Park, and M. W. Graham, 043820 (2018). Stacking angle-tunable photoluminescence from interlayer ex- [52] G. Mazza and A. Georges, Superradiant quantum materials, citon states in twisted bilayer graphene, Nat. Commun. 10, 1 Phys. Rev. Lett. 122, 017401 (2019). (2019). [53] K. Lenk and M. Eckstein, Collective excitations of the U(1)- [30] A. Cartella, T. F. Nova, M. Fechner, R. Merlin, and A. Caval- symmetric exciton insulator in a cavity, arXiv:2002.12241 leri, Parametric amplification of optical phonons, Proc. Natl. (2020). Acad. Sci. U.S.A. 115, 12148 (2018). [54] L. Chirolli, M. Polini, V. Giovannetti, and A. H. MacDonald, [31] Y. K. Srivastava, M. Manjappa, H. N. S. Krishnamoorthy, and Drude weight, cyclotron resonance, and the Dicke model of R. Singh, Accessing the high-Q dark plasmonic Fano reso- graphene cavity QED, Phys. Rev. Lett. 109, 267404 (2012). nances in superconductor metasurfaces, Adv. Opt. Mater. 4, [55] D. De Bernardis, P. Pilar, T. Jaako, S. De Liberato, and 1875 (2016). P. Rabl, Breakdown of gauge invariance in ultrastrong- [32] M. Fleischhauer, A. Imamoglu, and J. P. Marangos, Electro- coupling cavity qed, Phys. Rev. A 98, 053819 (2018). magnetically induced transparency: Optics in coherent media, [56] G. M. Andolina, F. M. D. Pellegrino, V. Giovannetti, A. H. Rev. Mod. Phys. 77, 633 (2005). MacDonald, and M. Polini, Cavity quantum electrodynamics [33] B. Y. Zeldovich, N. F. Pilipetsky, and V. V. Shkunov, Princi- of strongly correlated electron systems: A no-go theorem for ples of Phase Conjugation (Springer, Berlin, 1985). photon condensation, Phys. Rev. B 100, 121109 (2019). [34] S. Sugiura, E. A. Demler, M. Lukin, and D. Podolsky, Reso- [57] G. M. Andolina, F. M. D. Pellegrino, V. Giovannetti, nantly enhanced polariton wave mixing and Floquet paramet- A. H. MacDonald, and M. Polini, Theory of Photon ric instability, arXiv:1910.03582 (2019). Condensation in a Spatially-Varying Electromagnetic Field, [35] R. H. Dicke, Coherence in spontaneous radiation processes, arXiv:2005.09088 (2020). Phys. Rev. 93, 99 (1954). [58] A. Stokes and A. Nazir, Uniqueness of the phase transition in [36] M. Tavis and F. W. Cummings, Exact solution for an N- many-dipole cavity QED systems, arXiv:1905.10697 (2019). molecule—radiation-field hamiltonian, Phys. Rev. 170, 379 [59] F. Dimer, B. Estienne, A. S. Parkins, and H. J. Carmichael, (1968). Proposed realization of the Dicke-model quantum phase tran- [37] K. Hepp and E. H. Lieb, On the superradiant phase transition sition in an optical cavity QED system, Phys. Rev. A 75, for molecules in a quantized radiation field: the Dicke maser 013804 (2007). model, Ann. Phys. 76, 360 (1973). [60] E. G. D. Torre, S. Diehl, M. D. Lukin, S. Sachdev, and [38] F. T. Hioe, Phase transitions in some generalized Dicke mod- P. Strack, Keldysh approach for nonequilibrium phase tran- els of superradiance, Phys. Rev. A 8, 1440 (1973). sitions in quantum optics: Beyond the Dicke model in optical [39] C. Emary and T. Brandes, Quantum chaos triggered by precur- cavities, Phys. Rev. A 87, 023831 (2013). sors of a quantum phase transition: The Dicke model, Phys. [61] Z. Zhiqiang, C. H. Lee, R. Kumar, K. J. Arnold, S. J. Mas- Rev. Lett. 90, 044101 (2003). son, A. S. Parkins, and M. D. Barrett, Nonequilibrium phase [40] C. Emary and T. Brandes, Chaos and the quantum phase tran- transition in a spin-1 Dicke model, Optica 4, 424 (2017). sition in the Dicke model, Phys. Rev. E 67, 066203 (2003). [62] Z. Zhang, C. H. Lee, R. Kumar, K. J. Arnold, S. J. Masson, [41] M. Gaudin, Diagonalisation d’une classe d’hamiltoniens de A. L. Grimsmo, A. S. Parkins, and M. D. Barrett, Dicke-model spin, J. Phys. (France) 37, 1087 (1976). simulation via cavity-assisted Raman transitions, Phys. Rev. A [42] R. Richardson, A restricted class of exact eigenstates of the 97, 043858 (2018). pairing-force Hamiltonian, Phys. Lett. 3, 277 (1963). [63] C. Maschler, I. B. Mekhov, and H. Ritsch, Ultracold atoms in [43] O. Tsyplyatyev, J. von Delft, and D. Loss, Simplified deriva- optical lattices generated by quantized light fields, Eur. Phys. tion of the Bethe-ansatz equations for the Dicke model, Phys. J. D 46, 545 (2008). Rev. B 82, 092203 (2010). [64] K. Baumann, C. Guerlin, F. Brennecke, and T. Esslinger, [44] K. Rzazewski,˙ K. Wodkiewicz,´ and W. Zakowicz,˙ Phase tran- Dicke quantum phase transition with a superfluid gas in an sitions, two-level atoms, and the A2 term, Phys. Rev. Lett. 35, optical cavity, Nature 464, 1301 (2010). 432 (1975). [65] C. Hamner, C. Qu, Y. Zhang, J. Chang, M. Gong, C. Zhang, [45] P. Nataf and C. Ciuti, No-go theorem for superradiant quan- and P. Engels, Dicke-type phase transition in a spin-orbit- tum phase transitions in cavity QED and counter-example in coupled Bose–Einstein condensate, Nat. Commun. 5, 1 circuit QED, Nat. Commun. 1, 72 (2010). (2014). [46] G. Chen, Z. Chen, and J. Liang, Simulation of the superra- [66]L.P alov´ a,´ P. Chandra, and P. Coleman, Quantum critical diant quantum phase transition in the superconducting charge paraelectrics and the casimir effect in time, Phys. Rev. B 79, qubits inside a cavity, Phys. Rev. A 76, 055803 (2007). 075101 (2009). [47] J. Koch and K. Le Hur, Superfluid–Mott-insulator transition of [67] E. M. Purcell, H. C. Torrey, and R. V. Pound, Resonance ab- light in the Jaynes-Cummings lattice, Phys. Rev. A 80, 023811 sorption by nuclear magnetic moments in a solid, Phys. Rev. 27

69, 37 (1946). F. Mucklich,¨ M. Markham, A. M. Edmonds, and [68] E. T. Jaynes and F. W. Cummings, Comparison of quantum C. Becher, Nanoimplantation and purcell enhancement of sin- and semiclassical radiation theories with application to the gle nitrogen-vacancy centers in photonic crystal cavities in di- beam maser, Proc. IEEE 51, 89 (1963). amond, Appl. Phys. Lett. 106, 221103 (2015). [69] H. J. Kimble, Strong interactions of single atoms and photons [87] A. Sipahigil, R. E. Evans, D. D. Sukachev, M. J. Burek, J. Bor- in cavity QED, Phys. Scr. 76, 127 (1998). regaard, M. K. Bhaskar, C. T. Nguyen, J. L. Pacheco, H. A. [70] J. M. Raimond, M. Brune, and S. Haroche, Manipulating Atikian, C. Meuwly, R. M. Camacho, F. Jelezko, E. Biele- quantum entanglement with atoms and photons in a cavity, jec, H. Park, M. Loncar,ˇ and M. D. Lukin, An integrated dia- Rev. Mod. Phys. 73, 565 (2001). mond nanophotonics platform for quantum-optical networks, [71] R. Miller, T. E. Northup, K. M. Birnbaum, A. Boca, A. D. Science 354, 847 (2016). Boozer, and H. J. Kimble, Trapped atoms in cavity QED: cou- [88] D. Riedel, I. Sollner,¨ B. J. Shields, S. Starosielec, P. Appel, pling quantized light and matter, J. Phys. B 38, S551 (2005). E. Neu, P. Maletinsky, and R. J. Warburton, Deterministic [72] H. Walther, B. T. H. Varcoe, B.-G. Englert, and T. Becker, enhancement of coherent photon generation from a nitrogen- Cavity quantum electrodynamics, Rep. Prog. Phys. 69, 1325 vacancy center in ultrapure diamond, Phys. Rev. X 7, 031040 (2006). (2017). [73] A. Boca, R. Miller, K. M. Birnbaum, A. D. Boozer, J. Mc- [89] P. Lodahl, S. Mahmoodian, and S. Stobbe, Interfacing sin- Keever, and H. J. Kimble, Observation of the vacuum Rabi gle photons and single quantum dots with photonic nanostruc- spectrum for one trapped atom, Phys. Rev. Lett. 93, 233603 tures, Rev. Mod. Phys. 87, 347 (2015). (2004). [90] B. Hacker, S. Welte, G. Rempe, and S. Ritter, A photon– [74] K. M. Birnbaum, A. Boca, R. Miller, A. D. Boozer, T. E. photon quantum gate based on a single atom in an optical Northup, and H. J. Kimble, Photon blockade in an optical resonator, Nature 536, 193 (2016). cavity with one trapped atom, Nature 436, 87 (2005). [91] S. Sun, H. Kim, Z. Luo, G. S. Solomon, and E. Waks, A [75] C. Hamsen, K. N. Tolazzi, T. Wilk, and G. Rempe, Two- single-photon switch and transistor enabled by a solid-state photon blockade in an atom-driven cavity QED system, Phys. quantum memory, Science 361, 57 (2018). Rev. Lett. 118, 133604 (2017). [92] I. Aharonovich, D. Englund, and M. Toth, Solid-state single- [76] G. Khitrova, H. Gibbs, M. Kira, S. W. Koch, and A. Scherer, photon emitters, Nat. Photon. 10, 631 (2016). Vacuum Rabi splitting in semiconductors, Nat. Phys. 2, 81 [93] X. Ding, Y. He, Z.-C. Duan, N. Gregersen, M.-C. Chen, S. Un- (2006). sleber, S. Maier, C. Schneider, M. Kamp, S. Hofling,¨ C.-Y. Lu, [77] J. P. Reithmaier, G. Sek, A. Loffler,¨ C. Hofmann, S. Kuhn, and J.-W. Pan, On-demand single photons with high extrac- S. Reitzenstein, L. Keldysh, V. Kulakovskii, T. Reinecke, tion efficiency and near-unity indistinguishability from a res- and A. Forchel, Strong coupling in a single quantum dot– onantly driven quantum dot in a micropillar, Phys. Rev. Lett. semiconductor microcavity system, Nature 432, 197 (2004). 116, 020401 (2016). [78] T. Yoshie, A. Scherer, J. Hendrickson, G. Khitrova, H. Gibbs, [94] H. Snijders, J. A. Frey, J. Norman, V. P. Post, A. C. Gos- G. Rupper, C. Ell, O. Shchekin, and D. Deppe, Vacuum sard, J. E. Bowers, M. P. van Exter, W. Loffler,¨ and Rabi splitting with a single quantum dot in a photonic crys- D. Bouwmeester, Fiber-coupled cavity-QED source of iden- tal nanocavity, Nature 432, 200 (2004). tical single photons, Phys. Rev. Applied 9, 031002 (2018). [79] K. Hennessy, A. Badolato, M. Winger, D. Gerace, M. Atature,¨ [95] S. Daiss, S. Welte, B. Hacker, L. Li, and G. Rempe, Single- S. Gulde, S. Falt,¨ E. L. Hu, and A. Imamoglu,˘ Quantum na- photon distillation via a photonic parity measurement using ture of a strongly coupled single quantum dot–cavity system, cavity QED, Phys. Rev. Lett. 122, 133603 (2019). Nature 445, 896 (2007). [96] S.-B. Zheng and G.-C. Guo, Efficient scheme for two-atom [80] M. T. Rakher, N. G. Stoltz, L. A. Coldren, P. M. Petroff, entanglement and quantum information processing in cavity and D. Bouwmeester, Externally mode-matched cavity quan- QED, Phys. Rev. Lett. 85, 2392 (2000). tum electrodynamics with charge-tunable quantum dots, Phys. [97] L.-M. Duan and H. J. Kimble, Scalable photonic quantum Rev. Lett. 102, 097403 (2009). computation through cavity-assisted interactions, Phys. Rev. [81] L. Greuter, S. Starosielec, A. V. Kuhlmann, and R. J. Warbur- Lett. 92, 127902 (2004). ton, Towards high-cooperativity strong coupling of a quantum [98] M. Aspelmeyer, T. J. Kippenberg, and F. Marquardt, Cavity dot in a tunable microcavity, Phys. Rev. B 92, 045302 (2015). optomechanics, Rev. Mod. Phys. 86, 1391 (2014). [82] C.-H. Su, A. D. Greentree, and L. C. L. Hollenberg, Towards [99] W. L. Barnes, A. Dereux, and T. W. Ebbesen, Surface plasmon a picosecond transform-limited nitrogen-vacancy based single subwavelength optics, Nature 424, 824 (2003). photon source, Opt. Express 16, 6240 (2008). [100] J. M. Pitarke, V. M. Silkin, E. V. Chulkov, and P. M. [83] A. Faraon, C. Santori, Z. Huang, V. M. Acosta, and R. G. Echenique, Theory of surface plasmons and surface-plasmon Beausoleil, Coupling of nitrogen-vacancy centers to photonic polaritons, Rep. Prog. Phys. 70, 1 (2006). crystal cavities in monocrystalline diamond, Phys. Rev. Lett. [101] M. S. Tame, K. McEnery, S¸. Ozdemir,¨ J. Lee, S. Maier, and 109, 033604 (2012). M. Kim, Quantum plasmonics, Nat. Phys. 9, 329 (2013). [84] R. Albrecht, A. Bommer, C. Deutsch, J. Reichel, and [102] J. A. Hutchison, T. Schwartz, C. Genet, E. Devaux, and T. W. C. Becher, Coupling of a single nitrogen-vacancy center in Ebbesen, Modifying chemical landscapes by coupling to vac- diamond to a fiber-based microcavity, Phys. Rev. Lett. 110, uum fields, Angew. Chem. Int. Ed. 51, 1592 (2012). 243602 (2013). [103] J. Galego, F. J. Garcia-Vidal, and J. Feist, Cavity-induced [85] L. Li, T. Schroder,¨ E. H. Chen, M. Walsh, I. Bayn, J. Gold- modifications of molecular structure in the strong-coupling stein, O. Gaathon, M. E. Trusheim, M. Lu, J. Mower, M. Cot- regime, Phys. Rev. X 5, 041022 (2015). let, M. L. Markham, D. J. Twitchen, and D. Englund, Coher- [104] T. W. Ebbesen, Hybrid light–matter states in a molecular ent spin control of a nanocavity-enhanced qubit in diamond, and material science perspective, Acc. Chem. Res. 49, 2403 Nat. Commun. 6, 6173 (2015). (2016). [86] J. Riedrich-Moller,¨ S. Pezzagna, J. Meijer, C. Pauly, [105] J. Flick, M. Ruggenthaler, H. Appel, and A. Rubio, 28

Kohn–sham approach to quantum electrodynamical density- matter interactions, Nat. Commun. 9, 2273 (2018). functional theory: Exact time-dependent effective potentials in [123] D. M. Coles, Y. Yang, Y. Wang, R. T. Grant, R. A. Taylor, real space, Proc. Natl. Acad. Sci. U.S.A. 112, 15285 (2015). S. K. Saikin, A. Aspuru-Guzik, D. G. Lidzey, J. K.-H. Tang, [106] J. Flick, M. Ruggenthaler, H. Appel, and A. Rubio, Atoms and J. M. Smith, Strong coupling between chlorosomes of pho- and molecules in cavities, from weak to strong coupling in tosynthetic bacteria and a confined optical cavity mode, Nat. quantum-electrodynamics (QED) chemistry, Proc. Natl. Acad. Commun. 5, 5561 (2014). Sci. U.S.A. 114, 3026 (2017). [124] L. A. Martinez-Martinez, E. Eizner, S. Kena-Cohen, and [107] J. Feist, J. Galego, and F. J. Garcia-Vidal, Polaritonic chem- J. Yuen-Zhou, Triplet harvesting in the polaritonic regime: istry with organic molecules, ACS Photonics 5, 205 (2017). A variational polaron approach, J. Chem. Phys. 151, 054106 [108] M. A. Sentef, M. Ruggenthaler, and A. Rubio, Cavity (2019). quantum-electrodynamical polaritonically enhanced electron- [125] E. Eizner, L. A. Mart´ınez-Mart´ınez, J. Yuen-Zhou, and phonon coupling and its influence on superconductivity, Sci. S. Kena-Cohen,´ Inverting singlet and triplet excited states Adv. 4 (2018). using strong light-matter coupling, Sci. Adv. 5, eaax4482 [109] F. Schlawin, A. Cavalleri, and D. Jaksch, Cavity-mediated (2019). electron-photon superconductivity, Phys. Rev. Lett. 122, [126] D. Polak, R. Jayaprakash, T. P. Lyons, L. A. Martnez-Martnez, 133602 (2019). A. Leventis, K. J. Fallon, H. Coulthard, D. G. Bossanyi, [110] J. B. Curtis, Z. M. Raines, A. A. Allocca, M. Hafezi, and K. Georgiou, A. J. Petty, II, J. Anthony, H. Bronstein, J. Yuen- V. M. Galitski, Cavity quantum eliashberg enhancement of su- Zhou, A. I. Tartakovskii, J. Clark, and A. J. Musser, Manip- perconductivity, Phys. Rev. Lett. 122, 167002 (2019). ulating molecules with strong coupling: harvesting triplet ex- [111] E. Orgiu, J. George, J. Hutchison, E. Devaux, J. Dayen, citons in organic exciton microcavities, Chem. Sci. 11, 343 B. Doudin, F. Stellacci, C. Genet, J. Schachenmayer, (2020). C. Genes, G. Pupillo, P. Samori, and T. W. Ebbesen, Conduc- [127] A. Thomas, J. George, A. Shalabney, M. Dryzhakov, S. J. tivity in organic semiconductors hybridized with the vacuum Varma, J. Moran, T. Chervy, X. Zhong, E. Devaux, C. Genet, field, Nat. Mater. 14, 1123 (2015). J. A. Hutchison, and T. W. Ebbesen, Ground-state chemical [112] J. Feist and F. J. Garcia-Vidal, Extraordinary exciton con- reactivity under vibrational coupling to the vacuum electro- ductance induced by strong coupling, Phys. Rev. Lett. 114, magnetic field, Angew. Chem. Int. Ed. 55, 11462 (2016). 196402 (2015). [128] A. Thomas, L. Lethuillier-Karl, K. Nagarajan, R. M. A. [113] J. Schachenmayer, C. Genes, E. Tignone, and G. Pupillo, Vergauwe, J. George, T. Chervy, A. Shalabney, E. Devaux, Cavity-enhanced transport of excitons, Phys. Rev. Lett. 114, C. Genet, J. Moran, and T. W. Ebbesen, Tilting a ground-state 196403 (2015). reactivity landscape by vibrational strong coupling, Science [114] D. Hagenmuller,¨ J. Schachenmayer, S. Schutz,¨ C. Genes, and 363, 615 (2019). G. Pupillo, Cavity-enhanced transport of charge, Phys. Rev. [129] L. A. Mart´ınez-Mart´ınez, M. Du, R. F. Ribeiro, S. Kena-´ Lett. 119, 223601 (2017). Cohen, and J. Yuen-Zhou, Polariton-assisted singlet fission [115] X. Zhong, T. Chervy, L. Zhang, A. Thomas, J. George, in acene aggregates, J. Phys. Chem. Lett 9, 1951 (2018). C. Genet, J. A. Hutchison, and T. W. Ebbesen, Energy trans- [130] D. M. Juraschek, T. Neuman, J. Flick, and P. Narang, Cavity fer between spatially separated entangled molecules, Angew. control of nonlinear phononics, arXiv:1912.00122 (2019). Chem. Int. Ed. 56, 9034 (2017). [131] R. Chikkaraddy, B. De Nijs, F. Benz, S. J. Barrow, O. A. [116] D. Hagenmuller,¨ S. Schutz,¨ J. Schachenmayer, C. Genes, and Scherman, E. Rosta, A. Demetriadou, P. Fox, O. Hess, and G. Pupillo, Cavity-assisted mesoscopic transport of fermions: J. J. Baumberg, Single-molecule strong coupling at room tem- Coherent and dissipative dynamics, Phys. Rev. B 97, 205303 perature in plasmonic nanocavities, Nature 535, 127 (2016). (2018). [132] A. Thomas, E. Devaux, K. Nagarajan, T. Chervy, M. Sei- [117] G. L. Paravicini-Bagliani, F. Appugliese, E. Richter, F. Val- del, D. Hagenmuller,¨ S. Schutz,¨ J. Schachenmayer, C. Genet, morra, J. Keller, M. Beck, N. Bartolo, C. Rossler,¨ T. Ihn, G. Pupillo, and T. W. Ebbesen, Exploring Superconductiv- K. Ensslin, C. Ciuti, G. Scalari, and J. Faist, Magneto- ity under Strong Coupling with the Vacuum Electromagnetic transport controlled by Landau polariton states, Nat. Phys. 15, Field, arXiv:1911.01459 (2019). 186 (2019). [133] H. Yu, Y. Peng, Y. Yang, and Z.-Y. Li, Plasmon-enhanced [118] J. Li, D. Golez, G. Mazza, A. Millis, A. Georges, and light–matter interactions and applications, npj Comput. M. Eckstein, Electromagnetic coupling in tight-binding mod- Mater. 5, 1 (2019). els for strongly correlated light and matter, arXiv:2001.09726 [134] J. J. Baumberg, J. Aizpurua, M. H. Mikkelsen, and D. R. (2020). Smith, Extreme nanophotonics from ultrathin metallic gaps, [119] R. J. Holmes and S. R. Forrest, Strong exciton-photon coupling Nat. Mater. 18, 668 (2019). and exciton hybridization in a thermally evaporated polycrys- [135] S. Rowley, L. Spalek, R. Smith, M. Dean, M. Itoh, J. Scott, talline film of an organic small molecule, Phys. Rev. Lett. 93, G. Lonzarich, and S. Saxena, Ferroelectric quantum critical- 186404 (2004). ity, Nat. Phys. 10, 367 (2014). [120]S.K ena-Cohen´ and S. Forrest, Room-temperature polariton [136] D. E. Chang, A. S. Sørensen, E. A. Demler, and M. D. Lukin, lasing in an organic single-crystal microcavity, Nat. Photon. A single-photon transistor using nanoscale surface plasmons, 4, 371 (2010). Nat. Phys. 3, 807 (2007). [121] T. Chervy, J. Xu, Y. Duan, C. Wang, L. Mager, M. Frerejean, [137] A. J. Giles, S. Dai, I. Vurgaftman, T. Hoffman, S. Liu, L. Lind- J. A. Mnninghoff, P. Tinnemans, J. A. Hutchison, C. Genet, say, C. T. Ellis, N. Assefa, I. Chatzakis, T. L. Reinecke, J. G. A. E. Rowan, T. Tasing, and T. W. Ebbesen, High-efficiency Tischler, M. M. Fogler, J. H. Edgar, D. N. Basov, and J. D. second-harmonic generation from hybrid light-matter states, Caldwell, Ultralow-loss polaritons in isotopically pure boron Nano Lett. 16, 7352 (2016). nitride, Nat. Mater. 17, 134 (2018). [122] K. Stranius, M. Hertzog, and K. Borjesson,¨ Selective manip- [138] R. Ruppin and R. Englman, Optical phonons of small crystals, ulation of electronically excited states through strong light- Rep. Prog. Phys. 33, 149 (1970). 29

[139] G. D. Mahan, Polarons in Ionic Crystals and Polar Semicon- Phys. Rev. Lett. 106, 196405 (2011). ductors, edited by J. T. Devreese (North-Holland, Amsterdam, [162] D. I. Schuster, L. S. Bishop, I. L. Chuang, D. DeMille, and 1972). R. J. Schoelkopf, Cavity qed in a molecular ion trap, Phys. [140] X. Li, M. Bamba, Q. Zhang, S. Fallahi, G. C. Gardner, W. Gao, Rev. A 83, 012311 (2011). M. Lou, K. Yoshioka, M. J. Manfra, and J. Kono, Vacuum [163] V. Buzek,ˇ G. Drobny,´ M. S. Kim, G. Adam, and P. L. Knight, Bloch–Siegert shift in Landau polaritons with ultra-high co- Cavity qed with cold trapped ions, Phys. Rev. A 56, 2352 operativity, Nat. Photon. 12, 324 (2018). (1997). [141] T. G. Philbin, Canonical quantization of macroscopic electro- [164] G. S. Agarwal, Vacuum-field rabi splittings in microwave ab- magnetism, New J. Phys. 12, 123008 (2010). sorption by rydberg atoms in a cavity, Phys. Rev. Lett. 53, [142] P. Y. Yu and M. Cardona, Fundamentals of Semiconductors 1732 (1984). (Springer, New York, 2010). [165] C. R. Dean, A. F. Young, I. Meric, C. Lee, L. Wang, S. Sorgen- [143] J. Vuckovic, M. Loncar, and A. Scherer, Surface plasmon frei, K. Watanabe, T. Taniguchi, P. Kim, K. L. Shepard, and enhanced light-emitting diode, IEEE J. Quantum Electron. 36, J. Hone, Boron nitride substrates for high-quality graphene 1131 (2000). electronics, Nat. Nanotech. 5, 722 (2010). [144] D. Koller, A. Hohenau, H. Ditlbacher, N. Galler, F. Reil, [166] G. Constantinescu, A. Kuc, and T. Heine, Stacking in bulk and F. Aussenegg, A. Leitner, E. List, and J. Krenn, Organic bilayer hexagonal boron nitride, Phys. Rev. Lett. 111, 036104 plasmon-emitting diode, Nat. Photon. 2, 684 (2008). (2013). [145] J. Lambe and S. L. McCarthy, Light emission from inelastic [167] Y. Yamamoto, F. Tassone, and H. Cao, Semiconductor Cavity electron tunneling, Phys. Rev. Lett. 37, 923 (1976). Quantum Electrodynamics (Springer, New York, 2002). [146] S. Astilean, P. Lalanne, and M. Palamaru, Light transmission [168] A. V. Kavokin, J. J. Baumberg, G. Malpuech, and F. P. Laussy, through metallic channels much smaller than the wavelength, Microcavities (Oxford Science Publications, Oxford, 2007). Opt. Commun. 175, 265 (2000). [169] J. Kasprzak, M. Richard, S. Kundermann, A. Baas, P. Jeam- [147] S. Hayashi and T. Okamoto, Plasmonics: visit the past to know brun, J. Keeling, F. Marchetti, M. Szymanska,´ R. Andre,´ the future, J. Phys. D 45, 433001 (2012). J. Staehli, V. Savona, P. Littlewood, B. Deveaud, and L. Dang, [148] E. N. Economou, Surface plasmons in thin films, Phys. Rev. Bose–Einstein condensation of exciton polaritons, Nature 443, 182, 539 (1969). 409 (2006). [149] P. Pretre,ˆ L.-M. Wu, R. A. Hill, and A. Knoesen, Characteri- [170] R. Frank, Quantum criticality and population trapping of zation of electro-optic polymer films by use of decal-deposited fermions by non-equilibrium lattice modulations, New J. Phys. reflection fabry–perot microcavities, J. Opt. Soc. Am. B 15, 15, 123030 (2013). 379 (1998). [171] T. Byrnes, N. Y. Kim, and Y. Yamamoto, Exciton–polariton [150] P. Pilar, D. De Bernardis, and P. Rabl, Thermodynamics of condensates, Nat. Phys. 10, 803 (2014). ultrastrongly coupled light-matter systems, arXiv:2003.11556 [172] D. Sanvitto and S. Kena-Cohen,´ The road towards polaritonic (2020). devices, Nat. Mater. 15, 1061 (2016). [151] M. Schuler, D. De Bernardis, A. M. Lauchli,¨ and P. Rabl, [173] A. Lubatsch and R. Frank, Quantum many-body theory for The Vacua of Dipolar Cavity Quantum Electrodynamics, exciton-polaritons in semiconductor mie resonators in the arXiv:2004.13738 (2020). non-equilibrium, Appl. Sci. 10, 1836 (2020). [152] T. Sagawa, Second law-like inequalities with quantum rela- [174] I. D. Leroux, M. H. Schleier-Smith, and V. Vuletic,´ tive entropy: An introduction, edited by M. Nakahara and Orientation-dependent entanglement lifetime in a squeezed S. Tanaka (World Scientific, Singapore, 2012). atomic clock, Phys. Rev. Lett. 104, 250801 (2010). [153] R. P. Feynman, Statistical Mechanics (CRC Press, Florida, [175] J. Borregaard, E. J. Davis, G. S. Bentsen, M. H. Schleier- 1998). Smith, and A. S. Sørensen, One- and two-axis squeezing of [154] N. S. Gillis, Phase transitions in a simple model ferroelec- atomic ensembles in optical cavities, New J. Phys. 19, 093021 tric. ii. comments on the self-consistent phonon approxima- (2017). tion, Phys. Rev. B 11, 309 (1975). [176] J. Hu, W. Chen, Z. Vendeiro, A. Urvoy, B. Braverman, and [155] J. M. Edge, Y. Kedem, U. Aschauer, N. A. Spaldin, and V. Vuletic,´ Vacuum spin squeezing, Phys. Rev. A 96, 050301 A. V. Balatsky, Quantum critical origin of the superconduct- (2017). ing dome in SrTiO3, Phys. Rev. Lett. 115, 247002 (2015). [177] M. A. Norcia, R. J. Lewis-Swan, J. R. K. Cline, B. Zhu, A. M. [156] W. Jantsch, A. Bussmann-Holder, H. Bilz, and P. Vogl, Dy- Rey, and J. K. Thompson, Cavity-mediated collective spin- namical Properties of IV-VI Compounds, edited by G. Holer exchange interactions in a strontium superradiant laser, Sci- (Springer, Berlin, 2001). ence 361, 259 (2018). [157] M. Kozina, M. Fechner, P. Marsik, T. van Driel, J. M. Glow- [178] A. J. Daley, Quantum trajectories and open many-body quan- nia, C. Bernhard, M. Radovic, D. Zhu, S. Bonetti, U. Staub, tum systems, Adv. Phys. 63, 77 (2014). and M. C. Hoffmann, Terahertz-driven phonon upconversion [179] Y. Ashida, Z. Gong, and M. Ueda, Non-Hermitian Physics, in SrTiO3, Nat. Phys. 15, 387 (2019). arXiv preprint arXiv:2006.01837 (2020). [158] We note that the temperature is specified to be a concrete value [180] J. A. Hertz, Quantum critical phenomena, Phys. Rev. B 14, in order to fix a length scale in the setup of interest. 1165 (1976); A. J. Millis, Effect of a nonzero temperature [159]D.B auerle,¨ D. Wagner, M. Wohlecke,¨ B. Dorner, and on quantum critical points in itinerant fermion systems, Phys. H. Kraxenberger, Soft modes in semiconducting SrTiO3: II. Rev. B 48, 7183 (1993); T. Moriya, Spin Fluctuations in Itin- The ferroelectric mode, Z. Phys. B 38, 335 (1980). erant Electron Magnetism (Springer-Verlag, New York, 1985). [160] J. R. Tischler, M. S. Bradley, V. Bulovic, J. H. Song, and [181] A. Mitra, S. Takei, Y. B. Kim, and A. J. Millis, Nonequi- A. Nurmikko, Strong coupling in a microcavity LED, Phys. librium quantum criticality in open electronic systems, Phys. Rev. Lett. 95, 036401 (2005). Rev. Lett. 97, 236808 (2006). [161] T. Schwartz, J. A. Hutchison, C. Genet, and T. W. Ebbesen, [182] T. Prosen and I. Pizorn,ˇ Quantum phase transition in a far- Reversible switching of ultrastrong light-molecule coupling, from-equilibrium steady state of an XY spin chain, Phys. Rev. 30

Lett. 101, 105701 (2008). X 8, 011035 (2018). [183] S. Diehl, M. Baranov, A. J. Daley, and P. Zoller, Observability [193] R. W. Damon and J. R. Eshbach, Magnetostatic modes of a of quantum criticality and a continuous supersolid in atomic ferromagnet slab, J. Phys. Chem. Solids 19, 308 (1961). gases, Phys. Rev. Lett. 104, 165301 (2010). [194] W. Eerenstein, N. Mathur, and J. F. Scott, Multiferroic and [184] Y. Ashida, S. Furukawa, and M. Ueda, Parity-time-symmetric magnetoelectric materials, Nature 442, 759 (2006). quantum critical phenomena, Nat. Commun. 8, 15791 (2017); [195] J. Ruhman and P. A. Lee, Superconductivity at very low den- Quantum critical behavior influenced by measurement back- sity: The case of strontium titanate, Phys. Rev. B 94, 224515 action in ultracold gases, Phys. Rev. A 94, 053615 (2016). (2016). [185] L. M. Sieberer, S. D. Huber, E. Altman, and S. Diehl, Dynam- [196] A. H. Zemanian, Distribution Theory and Transform Analysis ical critical phenomena in driven-dissipative systems, Phys. (Dover Publications, New York, 2010). Rev. Lett. 110, 195301 (2013). [197] L. Novotny and B. Hecht, Principles of Nano-Optics (Cam- [186] J. Marino and S. Diehl, Driven markovian quantum criticality, bridge University Press, Cambridge, 2006). Phys. Rev. Lett. 116, 070407 (2016). [198] C. Maissen, G. Scalari, F. Valmorra, M. Beck, J. Faist, [187] C.-E. Bardyn, M. A. Baranov, C. V. Kraus, E. Rico, S. Cibella, R. Leoni, C. Reichl, C. Charpentier, and A. Imamo˙ glu,˘ P. Zoller, and S. Diehl, Topology by dissipa- W. Wegscheider, Ultrastrong coupling in the near field of com- tion, New J. Phys. 15, 085001 (2013). plementary split-ring resonators, Phys. Rev. B 90, 205309 [188] H. Dehghani, T. Oka, and A. Mitra, Dissipative floquet topo- (2014). logical systems, Phys. Rev. B 90, 195429 (2014). [199] D. G. Lidzey, D. Bradley, M. Skolnick, T. Virgili, S. Walker, [189] L. Lu, J. D. Joannopoulos, and M. Soljaciˇ c,´ Topological pho- and D. Whittaker, Strong exciton–photon coupling in an or- tonics, Nat. Photon. 8, 821 (2014). ganic semiconductor microcavity, Nature 395, 53 (1998). [190] T. Karzig, C.-E. Bardyn, N. H. Lindner, and G. Refael, Topo- [200] H. Bahsoun, T. Chervy, A. Thomas, K. Brjesson, M. Hertzog, logical polaritons, Phys. Rev. X 5, 031001 (2015). J. George, E. Devaux, C. Genet, J. A. Hutchison, and T. W. [191] Z. Gong, Y. Ashida, K. Kawabata, K. Takasan, S. Hi- Ebbesen, Electronic light–matter strong coupling in nanoflu- gashikawa, and M. Ueda, Topological phases of non- idic fabry–perot´ cavities, ACS Photonics 5, 225 (2017). hermitian systems, Phys. Rev. X 8, 031079 (2018). [201] Y. Kaluzny, P. Goy, M. Gross, J. M. Raimond, and S. Haroche, [192] C.-E. Bardyn, L. Wawer, A. Altland, M. Fleischhauer, and Observation of self-induced Rabi oscillations in two-level S. Diehl, Probing the topology of density matrices, Phys. Rev. atoms excited inside a resonant cavity: The ringing regime of superradiance, Phys. Rev. Lett. 51, 1175 (1983).