Arxiv:2006.08519V1 [Physics.Chem-Ph] 15 Jun 2020

Enhanced optical nonlinearities under collective strong light-matter coupling

Raphael F. Ribeiro,1 Jorge A. Campos-Gonzales-Angulo,1 Noel C. Giebink,2 Wei Xiong,1, 3 and Joel Yuen-Zhou1 1Department of Chemistry and Biochemistry, University of California San Diego, La Jolla, CA 92093 2Department of Electrical Engineering, The Pennsylvania State University, University Park, PA, 16802 3Materials Science and Engineering Program, University of California San Diego, La Jolla, CA 92093 (Dated: June 16, 2020) Optical microcavities and metallic nanostructures have been shown to significantly modulate the dynamics and spectroscopic response of molecular systems. We present a study of the nonlinear optics of a model consisting of N anharmonic multilevel systems (e.g., Morse oscillators) undergoing collective strong coupling with a resonant infrared microcavity. We find that, under experimentally accessible conditions, molecular systems in microcavities may have nonlinear phenomena significantly intensified due to the high quality of polariton resonances and the enhanced microcavity electromagnetic energy density relative to free space. Particularly large enhancement of multiphoton absorption happens when multipolariton states are resonant with bare molecule multiphoton transitions. In particular, our model predicts two-photon absorption cross section enhancements by several orders of magnitude relative to free space when the Rabi splitting ΩR is approximately equal to the molecular anharmonic shift 2∆. Our results provide rough upper bounds to resonant nonlinear response enhancement factors as relaxation to dark states is treated phenomenologically. Notably, ensembles of two-level systems undergoing strong coupling with a cavity (described by the Tavis-Cummings model) show no such optical nonlinearity enhancements, highlighting the rich phenomenology afforded by multilevel anharmonic systems. Similar conclusions are expected to hold for excitonic systems that share features with our model (e.g., molecular dyes with accessible S0 S1 S2 transitions) and strongly interact with a UV-visible cavity. → →

1. INTRODUCTION ton transport [25–27], and chemical kinetics [28–31].

Light-induced nonequilibrium phenomena is a topic of great contemporary interest due to its relevance to the energy, biochemical, and material sciences. Nonlin- ear spectroscopy provides tools for probing and controlling nonequilibrium quantum dynamics [1, 2] driven by external radiation. Applications of nonlinear optics to chemistry include investigations of the dynamics of energy and charge transport in light-harvesting complexes FIG. 1. Left: Planar microcavity consisting of two highly [3, 4], organic electronics [5], and other excitonic systems reflective mirrors filled with a molecular ensemble (e.g., [6]. Nonlinear optical processes are also basic to vari- W(CO)6 in solution) with sufficiently large collective oscilla- ous developing technologies including all-optical devices tor strength that hybrid polaritonic states are formed. Right: [7, 8], quantum information processors [9, 10], and en- Mechanism for enhancement of two-photon absorption by an ensemble of Morse oscillators (represented by the various il- hanced sensors [11]. lustrative Morse potentials) under strong coupling with an Unfortunately, the nonlinearities of molecular sys- optical cavity. An external field resonant with the lower- tems are generally weak [12]. Recently, hybrid mate- polariton (LP) drives the hybrid cavity and excites two-LP rials consisting of a molecular ensemble hosted by a states which can be tuned to be near-resonant with the an- photonic (or plasmonic) device (e.g., optical microcav- harmonically shifted doubly-excited molecular states forming ities and metallic nanostructures) have been explored the totally-symmetric 2S state. This polariton-mediated ab- as potential sources of magnified nonlinear optical re- sorption channel allows enhancement of several orders of mag- sponse [13–16]. Under accessible experimental condi- nitude of the molecular two-photon absorption cross-section.

arXiv:2006.08519v1 [physics.chem-ph] 15 Jun 2020 tions (room temperature and atmospheric pressure) the light-matter interaction in photonic materials can be- Recent experiments [32–41] have surveyed the non- come strong enough that excited states corresponding linear optics of polaritonic systems to gain further in- to superposition of (collective) material polarization and sight into the relaxation kinetics and optical response cavity excitations emerge [16–20]. The corresponding hy- of strongly coupled devices. In Refs. [36–38, 42], the brid quasiparticles (modes) are commonly denoted by transient response and relaxation to equilibrium of vibra- (cavity)-polaritons[21]. They show controllable coher- tional polaritons (those arising from the strong coupling ence and relaxation dynamics that allow modulation of of molecular infrared polarization with a resonant mi- various physicochemical properties. Molecular phenom- crocavity) were investigated with pump-probe and two- ena significantly influenced by strong light-matter inter- dimensional infrared spectroscopy. These studies demon- actions include: energy transfer[22–24], charge and exci- strated how vibrational anharmonicity is manifested in 2 the pump-probe polariton response [43]. However, the ity field and the molecular polarization N 1 p 0 i=1 h i| i| ii observed time-resolved spectra were sensitive to vari- (where pi is the effective dipole operator of the ith P ous system-dependent effects arising from the small Rabi molecule and 0i and 1i denotes states where the ith splittings of the studied materials, and significant static molecule is in the ground and first excited-state, respec- and dynamical disorder which induces ultrafast polariton tively, whereas all other molecules are in the ground- decay into the weakly-coupled (dark) molecular modes. state) is significantly stronger than the coupling of ei- In this work, we focus on universal (system- ther subsystem to external (bath) degrees of freedom, independent) features of molecular polariton nonlinear but still only a tenth or less of the bare vibrational and optics. Our aim is to provide qualitative and quanti- cavity frequencies (so considerations exclusive to ultra- tative insight on the potential to achieve giant optical strong coupling can be ignored [51–53]). nonlinearities with molecular polaritons in the collective The total Hamiltonian of the composite material is regime (which is the case in most experiments) with a given by HT (t) = HL(t) + HM + HLM, where HL(t) large number of molecules in a microcavity (nonlinear and HM are the bare cavity (driven by an external time- optical effects of single-molecule polaritonic systems have dependent field) and molecular Hamiltonians and HLM been studied within a non-adiabatic model of the dynam- contains the interaction between the cavity EM field and ical Casimir effect in Ref. [44], as well as in vibrational matter. The cavity Hamiltonian is given by: polariton spectra in Ref. [45]). H (t) = ω b† b In Sec. 2, we describe our model, provide an analytical L ~ k k k expression for the nonlinear optical susceptibility of an Xk ideal molecular ensemble under strong interaction with a κ L L + i~ b (t) † bk b† b (t) , (1) microcavity (the full derivation is in the SI Sec. 2), and 2 kin − k kin r k discuss its main features. In Sec. 3, we compare the free X n o space and the polariton-mediated two-photon absorption where this effective Hamiltonian can be obtained from (TPA) rates, and show that, especially when overtone po- input-output theory [54–56] which describes the interac- lariton transitions are resonant with multiphoton molec- tion of the optical cavity with left input and right out- L R ular transitions, nonlinearity enhancements of several or- put flux operators bkin(t) and bkout(t), respectively (SI ders of magnitude may be achieved with currently avail- Sec. I), and we include only a single cavity band and able optical cavities (Fig. 1) as a result of three main EM field polarization (as the cavity band gaps are much effects: increased electromagnetic energy density in the larger than the cavity and molecular linewidths, due to optical microcavity relative to free space [13, 46], creation the smallness of the cavity’s longitudinal length Lc, and of new optical resonances, and strong coupling induced electric field polarization conversion gives a tiny pertur- suppression of lineshape broadening [47]. A discussion bation on the results presented here especially as we con- of our main results and conclusions are given in Sec. sider isotropic molecular ensembles [57]). The frequency 4. Our article is accompanied by Supporting Informa- of the mode with (in-plane) wave-vector k = (kx, ky) is 2 2 2 2 tion (SI) containing detailed derivations of the molecular ωk = c k + m π /Lc /n (m Z is the index of the ∈ nonlinear susceptibility, and rate of nonlinear absorption cavity band; n is the index of refraction of the cavity p in free space and under strong coupling with an optical interior; hereafter n = 1), and bk is its annihilation microcavity. operator. The cavity leakage (decay) rate is κ. The Heisenberg equations of motion generated by Eq. 1 are turned into the Heisenberg-Langevin equations when the 2. MOLECULAR NONLINEAR RESPONSE replacement ωk ω˜k ωk iκ/2 is performed (SI Sec. 1). → ≡ − 2.1. Effective Hamiltonian The bare vibrational dynamics is generated by the Hamiltonian HM given by The physical system of interest consists of a molecular N N HM = ~ω0a†ai ~∆ a†a†aiai, (2) ensemble containing N molecules uniformly distributed i − i i i=1 i=1 in a region enclosed by two highly-reflective planar mir- X X rors separated by a distance Lc of the order of the wave- where the vibrational creation and annihilation opera- length of a specific material infrared excitation (Lc is tors of the ith molecule are ai† and ai, respectively. The usually between 0.1 and 20 µm) [48–50]. This setup cor- fundamental frequency of each molecule is ω0, and the responds to a Fabry-Perot (FP) microcavity [13, 46] filled anharmonic coupling is ∆ > 0. We neglect intermolec- with a homogeneous molecular system. Our description ular interactions as they are too weak relative to light- of the molecular subsystem will include explicitly only matter coupling (the situation could be different in other the modes which are nearly-resonant with the optical situations, e.g., molecular crystals and liquid-solid inter- cavity. The effects of all other molecular degrees of free- faces [21, 58]). We treat the relaxation of the molecular dom will be treated phenomenologically by introduction subsystem phenomenologically by converting the Heisen- of damping to the molecular polarization (see below). berg equations of motion (EOMs) of molecular opera- We suppose that the interaction between the cav- tors into Heisenberg-Langevin EOMs via the substitution 3

ω ω˜ = ω iγ /2, where γ is the bare molecule centered at k 0, with a small width δk. The fre- 0 → 0 0 − m m 0 ≈ fundamental transition (homogeneous) linewidth. quency ωk0 is nearly resonant with the bare molecule The light-matter interaction is treated with the multi- fundamental frequency ω0. Therefore, we shall retain polar gauge [59] in the long-wavelength limit within the only a single cavity-mode corresponding to k0 0. This rotating wave approximation[55] (see next paragraph for assumes there is no variation in the polariton≈ nonlinear a discussion of these and other approximations): response with respect to changes of magnitude δk in the | | incident wave-vector k0. N Based on the above, we simplify HL(t) and HLM and H = g a†b +g ¯ a b† + H 2 , (3) LM − ik i k ik i k P employ the following effective Hamiltonian for the hybrid i=1 Xk X cavity-matter system where g = µ Ec is the coupling constant for the inter- jk j · jk N N action between the jth molecular vibration (with effec- HT (t) =~ωcb†b + ~ω0a†ai ~∆ a†a†aiai i − i i tive transition dipole moment µj) and the cavity mode k, i=1 i=1 X X with mode profile evaluated at the position rj of the jth N c ik rj c c molecule, i.e., Ejk = i ~ωk/(20Vc)e · sin(mπzj/Lz) µ E a†b + E¯ b†a − 0 i 0 i ( is the electrical permittivity of free space, V is the i=1 0 p c X cavity quantization volume and zj is the position of the κ L † L molecule along the cavity longitudinal axis), f¯ denotes i~ bin(t) b b†bin(t) , (4) − 2 − the complex conjugate of f, and H 2 is the molecular r P n o self-polarization energy [59]. Although this term ensures L L c where ωc ωk , b = bk , b = b , and µE gjk = the existence of a ground-state for the composite system ≡ 0 0 k0in in 0 ≡ 0 [60] and it becomes essential for an appropriate treatment iµ ~ωc/(20Vc). of a system with total light-matter interaction energy ap- p proaching or surpassing the bare cavity and molecular frequencies [61], HP2 can be neglected under the strong 2.2. Nonlinear molecular polarization under strong coupling conditions assumed here. Therefore, we will dis- light-matter coupling regard this term onward. The length scale over which the cavity mode profile The optical response of a hybrid microcavity can be varies substantially (of order 0.1-20 µm) is much larger investigated by measuring the transmission, reflection than typical molecular diameters (of order 0.5 5nm). or absorption spectrum of light input into the system. Thus, under strong coupling, the k 0 cavity modes− in- ≈ For instance, transmission and reflection spectra can be teract coherently with material polarization consisting of obtained by applying the input-output relations to the a macroscopic number of molecules. This notion forms iωt steady-state cavity field b(t) = ω>0 b(ω)e− . Because the basis for neglecting spatial, orientational and ener- the cavity is weakly-coupled to the external fields, the ex- getic dispersion of the molecular excitations, since fluc- pectation value b(t) admits aP perturbative expansion in tuations of these quantities are necessarily weak effects h i (2p 1) powers of the input amplitude b(t) = ∞ b(t) − , compared to the collective light-matter interactions from h i p=1 h i (2p 1) L 2p 1 which polaritons emerge. Fluctuations about the mean where b(t) − = O bin − (onlyP odd powers of the inputh fieldi appear in| the| cavity response because values of the molecular transition frequency and dipole moment can lead to dephasing-induced polariton decay the material is assumed homogeneous and symmetric [62, 63], weak-coupling of light to states which are dark with respect to spatial inversion [1, 12]). The mate- N according to Eq. 3, as well as polariton [64, 65] and rial polarization P (t) = µ i=1 ai(t) is strongly cou- dark-state localization [64, 66]. Since we are not con- pled to the optical cavity. Therefore, molecular observ- cerned with transport phenomena, we will not include ables also admit a perturbativeP expansion in powers of L them in our model, although Sec. 4 qualitatively ana- bin . Note that the empty cavity is a linear system, and lyzes their implications to our main results. Despite ne- thus, the source of the nonlinear part of b(t) is the h i glecting these effects, we highlight that our input-output molecular subsystem (specifically, the source of b(t) (3) treatment of the material and photonic components natu- (3) N (3) h i is P (t) = µ ai(t) ; see SI Sec. 2). Therefore, rally accounts for polariton dissipation via cavity leakage h i i=1 h i P (t) (3) directly determines the amplitude of the non- and molecular homogeneous dephasing [43, 54, 55, 67]. In P linearh i optical response of a strongly coupled system as Eq. 3, we also assumed validity of the so-called rotating- measured by the output transmitted and reflected light. wave-approximation: only light-matter interactions pre- serving the total number of cavity and molecular exci- Neglecting quantum fluctuations of the input field, tations are retained. This approximation is justified by bin(t) is a complex number that we express as: N 2 i=1 gik ω0, k. | | ∀ (ω) iθin(ω) iωt qWe aim to investigate the nonlinear response of the bin(t) = i P e e− , (5) P ~ω hybrid system to an input radiation field with k R2 ω>0 r ∈ X 4 where (ω) is the power of the free space mode with nonlinear molecular susceptibility [12] under strong light- P (3) frequency ω driving the cavity, and θin(ω) is its phase matter interaction conditions. The ratio between χ (SI Sec. 1). and the bare molecular system nonlinear susceptibility It follows (SI Sec. 2) that the third-order polarization (3) χ0 provides an external-field independent measure of (3) in the frequency domain P (ωs) can be written in strong light-matter coupling effects on the optical non- terms of the input electrich fieldsi as follows linearities of an arbitrary molecular system.

(3) P (3) (ω ) = χ(3)( ω ; ω , ω , ω )E(+)(ω ) To obtain χ [1] for the system described by Eq. 4, we h i s − s v − w u in v × solve perturbatively the Heisenberg-Langevin equations ω ,ω ,ω u Xv w of motion (EOM) for the molecular polarization to third- ( ) (+) E − (ω )E (ω ) + h.c., (6) L in w in u order in the driving ﬁeld bin [68–70]. The EOMs for the cavity and material polarization expectation values ad- where ωs > 0 is the signal frequency, the brackets denote mit simple solutions since the initial condition (ground- expectation values, the driving frequencies ωu, ωv, ωw are state) and the time-evolution of our system ensures that (+) all positive, the input ﬁelds Ein (ωu) are directly pro- it remains in a pure state at all times. The result is (SI (3) portional to the bin(ωu) (see SI Sec. 1), and χ is the Sec. 2):

(3) ¯ χ ( ωs; ωv, ωw, ωu) =Nµ(2~∆)Gmm(ωs)Gmm(ωw)Γmm,mm(ωu + ωv)Gmm(ωu)Gmm(ωv) − − × 2 (0) ~κ 2 ~κ ¯(0) ~κ 2 (0) µ F Gpp (ωv) µ F Gpp (ωw) µ F Gpp (ωu) δωs,ωv ωw+ωu , (7) " r π 2 #" r π 2 #" 2 r π # −

c where is the cavity ﬁnesse (the electromagnetic ﬁeld in- where ΩR = 2 µE0 √N/~ is the Rabi frequency (split- tensityF in a resonant cavity is stronger than in free space ting). The imaginary| | parts of the polariton poles by the factor 2 /π, or alternatively, = Q/m, where m in G (ω) correspond to their (linear) absorption F F mm is the aforementioned band index, and Q = ωc/κ is the linewidths. Under weak-coupling conditions, we can ne- µE 2N quality factor; SI Sec. I and Ref. [46]), and Gmm(ω) glect the cavity-induced self-energy | 0| to ob- ~ω ~ωc+iκ/2 is the Fourier transform (FT) of the retarded single- (0)− tain the bare molecule propagator Gmm(ω) = 1/(~ω molecule propagator − ~ω0 + i~γm/2). Similarly, the photon-photon correlator 1 Gpp(ω) has resonances at the polariton frequencies, as is Gmm(ω) = 2 . (8) µE0 N clear from its explicit form: ~ω ~ω0 + i~γm/2 | | − − ~ω ~ωc+iκ/2 − 1 Gpp(ω) = . (11) Note the real part of the poles of Gmm(ω) are the funda- c 2 µE0 N mental polariton resonance frequencies ~ω ~ωc + i~κ/2 ω | ω +|i γ /2 − − ~ −~ 0 ~ m In the weak-coupling limit, G (ω) approaches the empty 2 2 pp ωc + ω0 (ωc ω0) + ΩR (0) ωLP = − , (9) cavity frequency-domain propagator Gpp (ω) = 1/(~ω 2 − 2 − p ~ωc + i~κ/2). 2 2 ωc + ω0 (ωc ω0) + ΩR The function Γmm,mm(ωu +ωv) is the two-particle elas- ωUP = + − . (10) 2 p 2 tic scattering matrix element given by:

2 (~ω 2~ω˜0)(~ω ~ω˜0 ~ω˜c) (~ω 2~ω˜0)(~ω 2~ω˜c) 4g N Γmm,mm(ω) = − − − − − − , (12) D (ω)

c where D(ω) is a 4th-order polynomial of ω given by: where g = µE0 is the single-molecule light-matter cou- |(0) | pling and D (ω) = (~ω ~ω˜c ~ω˜0)(~ω 2~ω˜0 +2~∆) (0) 2 − − − × D(ω) =D (ω) 2g N(~ω 2~ω˜0)(~ω 2~ω˜0 + 2~∆) (~ω 2~ω˜c)(~ω 2~ω˜0). The roots of D(ω) correspond − − − − − 2 to the bright two-particle resonances of the hybrid sys- 2g (N 1)(~ω 2~ω˜0 + 2~∆)(~ω 2~ω˜c) − − − − tem, as can be veriﬁed by comparison to the eigenval- 2 2g (~ω 2~ω˜c)(~ω 2~ω˜0), (13) − − − 5

the resonance with frequency ωLU corresponds to that containing an LP,UP pair (see Fig. 2). Physically, Γ (ω + ω ) G (ω , ω ) (SI Sec. mm u v ∝ mm,mm u v 2), where Gmm,mm(ω) is the frequency-domain single- molecule two-excitation propagator, i.e., the FT of the probability amplitude that a molecule initially in its doubly excited-state remains in the same state after time t. Importantly, when N , Eq. 12 becomes: → ∞

(0) ω 2˜ω0 Γmm,mm(ω) Γmm,mm(ω) − ,N , ≈ ≡ ω 2˜ω0 + 2∆ → ∞ − (14)

where Γ(0)(ω) is the bare single-molecule two-particle (elastic) scattering matrix (see next subsection and SI Sec. 4). This result is expected, since Γmm,mm(t) describes the time-dependent propagation of single- molecule doubly excited-states under interaction with the optical cavity, and as we show in SI Sec. 7, the totally-symmetric doubly-excited molecular state 2m = FIG. 2. Energy level diagram for a model sys- 1 N | i i=1 2i (where 2i is the state where the ith tem with zero-detuning (ωc = ω0) and Rabi splitting √N | i | i molecule is in the 2nd excited-state while the cavity ΩR > anharmonicity 2∆, including only bright excitations P of the single and two-polariton manifolds (SI Sec. 7). and all other molecules are in the ground-state) is only weakly-coupled to two-polariton states via an interaction that is proportional to the single-molecule light-matter ues of the doubly-excited block of the Hamiltonian in coupling g. Therefore, while polaritons play an essen- L Eq. 4 with bin = 0 (see SI Sec. 7 for a discussion and tial role as intermediate states for TPA by the molecular analytical results for the ωc = ω0 case). In the large subsystem, Eq. 14 indicates the dynamics of molecular N limit appropriate to almost all experimental studies doubly excited-states is almost insensitive to their cou- of infrared strong coupling, the real parts of the two- pling to the cavity electromagnetic ﬁeld in the ensemble particle resonances in the zero-detuning case are given strong coupling limit. √ by ω2UP = 2ωUP + O(g/ N), ωLU = ωc + ω0, ω2m = To gain further insight into the molecular nonlinear 2ω0 2∆ + O(∆/N), and ω2LP = 2ωLP + O(g/√N), polarization in the strong light-matter coupling regime, where− the subscripts label the dominant character of each we now compare Eq. 7 to the nonlinear susceptibility of state, e.g., the highest-frequency resonance is dominated the bare molecules in free space (under the rotating-wave by the component with a doubly-excited UP mode while approximation) given by

(3) 3 (0) ¯(0) (0) (0) (0) χ0 ( ωs; ωv, ωw, ωu) = Nµ(2~∆)µ Gmm(ωs)Gmm(ωw)Γmm,mm(ωu + ωv)Gmm(ωv)Gmm(ωu)δωs,ωv ωw+ωu . (15) − − −

(0) By contrasting Eqs. 15 and 7, we find that the [Gmm(ω) Gmm(ω)] and two-particle molecular re- nonlinear optical response of a molecular system (e.g., → (0) sponse functions [Γmm,mm(ω) Γmm,mm(ω)] which are solution [37, 71], polymer [72, 73], etc) in an opti- non-perturbatively dressed by→ the interaction with the cal microcavity is significantly distinct from that in cavity field, as well as (ii) molecular transition dipoles free space mainly because of: (i) near-resonant in- µ which are renormalized by factors that depend on tracavity field intensity enhancement [which renormal- the cavity finesse and the bare photon propagator izes the induced molecular transition dipole moments [µ µ˜(ω)]. The renormalizationF of the induced molecu- (0) → µ µ˜(ω) = µ 2 /πi~ κ/2Gpp (ω)], and (ii) the lar dipoles is a result of the well-known enhancement of → F appearance of new (polariton) resonances correspond- the intracavity electric field relative to free space [46, 74]. p p ing to hybrid superpositions of molecular polarization In the weak coupling regime, a Purcell-like result follows and cavity modes. In other words, the molecular non- where the molecular nonlinear susceptibility in a micro- linear response under strong coupling can be written cavity (Eq. 7) is simply related to that of the bare system entirely in terms of cavity-renormalized single-particle (Eq. 15), χ(3) χ(3) intracavity field enhancement → 0 × 6 factors. For simplicity, we restrict our analysis of the nonlin- As discussed in Sec. 3, by virtue of the cavity- ear absorption spectrum to the zero-detuning case where NL matter strong coupling, the nonlinear polarization con- ωc = ω0. We also simplify W (ω) by using the follow- tribution to the energy absorbed by the molecular sub- ing conditions necessarily valid at strong coupling: Ω R system is not directly proportional to the imaginary part ~ηs ~(κ+γm), and ΩR ~η ~κγm/(κ+γm). Under of χ(3)( ω; ω, ω, ω) (in contrast to the nonlinear ab- these≡ conditions, the nonlinear ≡ component of molecular sorption− of bare− molecules; see SI Sec. 5). Therefore, we absorption under strong coupling with a cavity can be ex- 4 provide additional comments and a numerical compari- 4 (+) pressed as W NL(ω) = W NLα (ω) E (ω) , where son of the real and imaginary parts of Eqs. 7 and 15 as a α=1 in function of cavity detuning and Rabi splitting in SI Sec. P 2ηκ 1 1 6, and focus below on the nonlinear absorption spectrum NL1 W (ω) 2 2 + 2 ≈ − ~ 4(ω ω0) + κ (ΩR/~) of the strongly coupled material. − 2 Re F χ(3)(ω) , (19) 3. POLARITON-ENHANCED TWO-PHOTON × "r π # ABSORPTION

4η ω ω0 2 The steady-state rate of excitation of a molecular sys- NL2 (3) W (ω) − 2 2 Im F χ (ω) , ≈ ~ 4(ω ω0) + κ π tem driven by the electromagnetic field can be written as − "r # (SI Secs. 3 and 5) (20) 2 (2) 2 2 a a (2ω) W (ω) Im [E(ω)]† P (ω) , (16) 4∆ N i i T W NL3 (ω) η h i , (21) ≡ ~ h i 2 2 (+) h i ≈ (ω ω0) + γm/4 E (ω) 4 − | in | where E(ω) is the frequency-domain representation of 2 (2) the free space or cavity Heisenberg electric field operator. aiai (2ω) (2ω ω20)2∆N W NL4 (ω) η − h i , (22) For a molecular system in free space interacting weakly 2 (+) ≈ − (ΩR/2~) E (ω) 4 with a classical monochromatic EM field with (positive- | in | (+) frequency) amplitude E (ω), it follows that the photon (3) (3) in where χ (ω) 6χ ( ω; ω, ω, ω), ω20 2ω0 2∆, absorption rate (in the rotating-wave approximation) is and ≡ − − ≡ − [1] (2) aiai (2ω) = Γmm,mm(2ω)Gmm(ω)Gmm(ω) 2 (1) (+) 2 h i − W0(ω) Im χ0 ( ω; ω) Ein (ω) 2 ≡~ − | | 2 (0) ~κ (+) 2 h i µ F Gpp (ω) E (ω) . (23) (3) (+) 4 × π 2 in + Im χ0 ( ω; ω, ω, ω) Ein (ω) + ... " r # ~ − − | | h i (17) The expression for the nonlinear absorption by This expression is clearly invalid when the molecular en- the molecular system under strong coupling semble interacts strongly with a cavity, since in this in- with a cavity is more complicated relative to NL (+) 4 stance, the cavity field and the material electrical polar- the bare system given by W0 (ω) E (ω) = | in | ization are correlated, and therefore E(ω)P (ω) cannot 2 (3) (+) 4 Im χ0 ( ω; ω, ω, ω) Ein (ω) . For example, (in general) be factorized into E (ωh) P (ω) (wherei E ~ − − | | W NLh(ω) shows dependencei on both the real and imag- refers to the cavity EM field). Nevertheless,h i h i the external inary parts of χ(3) (Eqs. 19 and 20) in addition to the input field interacts weakly with the cavity, and the rate steady-state population of molecular doubly excited- of absorption by the strongly coupled molecular system states P T (ω) = N a a (2) (2ω) 2/2 (Eqs. 21 and admits the following perturbative expansion in powers of 2m i=1 | h i ii | the input field amplitude 22). This additional complexity of nonlinear absorption under strong couplingP conditions is expected, since 2 (2) (4) while external fields acting on the bare system drives W (ω) = Im [E (ω)]† P (ω) + [E (ω)]† P (ω) + ... h c i h c i transitions between three molecular states (ground, first ~ (18) and second excited-state), at least seven energy levels (Fig. 2) may play a role in the nonlinear response c where Ec(ω) = E0b(ω). The contribution to the ab- of a material strongly coupled to an optical cavity. sorption spectrum dependent on the nonlinear response Nevertheless, the main features of W NL(ω) can be of the molecular subsystem is given by W NL(ω) = obtained from Eqs. 19 22. Specifically, − 2 (4) NL Im E†(ω)P (ω) . In SI Sec. 3, we obtain W (ω) ~ h c i 1. The nonlinear absorption intensity is largest when by employingh the Heisenberg-Langevini EOMs following the input field is nearly-resonant with either the the same approach taken to derive Eq. 7. LP or UP, since this maximizes G (ω) 4 which | mm | 7

NL NL W (ω)/W0 (ω0-Δ) Physically, the polaritons provide the resonant optical window to efficiently drive the transitions of -1 1.5 ΩR = 40 cm interest. In the studied case of zero cavity detun- -1 ΩR = 35 cm ing, the conditions described in Eq. 25 can be sum- bare marized as the Rabi splitting being equal to the anharmonic shift, that is, ΩR = 2∆. 1.0 ± When the criteria in Eq. 25 are satisfied, we expect strong enhancement of nonlinear absorption based on the following argument: if the input field con- 0.5 sists of photons with ω = ω0 ∆ and Eq. 25 is satisfied, polaritons will be efficiently− pumped, and a fraction of those will subsequently decay by pop- 0.0 ulating molecular doubly-excited states. In other 1972 1973 1974 1975 1976 1977 1978 words, when the two-polariton resonance condi- ω(cm -1) tion is satisfied, the molecular doubly excited-states provides an efficient sink for the energy stored in FIG. 3. Ratio of two-photon absorption rate of two strongly two-polariton modes. This effect was indeed re- 1 1 coupled (ω0 = ωc = 1983 cm− , γm = κ = 3 cm− , ∆ = ported in a recent experiment [42], where evidence 1 1 1 8 cm− , and ΩR = 40 cm− or ΩR = 35 cm− ) systems rela- was given that (for systems with weak system-bath tive to that of the molecular ensemble in free space normalized interactions and slow molecular polarization de- by the maximum of the latter. phasing) the second excited vibrational state was preferentially populated over the first when the pump (input) field was resonant with LP. appears in all of Eqs. 19 22. Physically, the po- − lariton resonance condition for maximal photon ab- 4. Conversely, in the limit where two-polariton states sorption is a consequence of the optical filtering are highly off-resonant with the molecular TPA performed by a microcavity (off-resonant external ( 2∆ ΩR 0, Fig. 3), the nonlinear response fields are suppressed relative to the resonant). substantially| − | weakens. In this limit, the studied 2. W NL3 (ω) is the only component of W NL(ω) which model approaches the Tavis-Cummings [75], where is positive for all values of the input frequency. a collection of two-level systems interact strongly Therefore, it necessarily gives molecular excited- with a single-mode cavity. The nonlinear response state absorption contributions to W NL(ω). Fur- given by this system is known to become negligible ther evidence is given by the fact that W NL3 (ω) in the large N limit [76] (e.g., as we show in the is proportional to the steady-state population of SI Sec. 7, in the Tavis-Cummings model, the two- molecules in the doubly-excited state, and thus, level system nonlinearity produces a large N limit anharmonic shift proportional to µEc /√N). | 0| 8η∆2 P T (2ω) NL3 2m It follows, therefore, that the condition given in Eq. W (ω) 2 2 4 . (24) ≈ (ω ω0) + γ /4 (+) 25 allows the harnessing of the enhanced electro- − m E (ω) in magnetic field of optical cavities to enhance TPA.

All other contributions to the nonlinear absorp- Points 3 and 4 are the main conclusions of our work. tion can be either positive (when excited-state ab- We will now quantitatively illustrate that under exper- sorption processes dominate) or negative (when imentally accessible conditions, it is possible to employ ground-state bleach and stimulated emission pro- cavity-strong coupling to substantially enhance the TPA cesses dominate [1]) depending on ω. cross section of a resonant molecular system. 3. Based on items 1 and 2, we expect the TPA rate In Fig. 3, we present the infrared TPA spectrum will be largely enhanced relative to free space when for a molecular system in free space (we take repre- the doubly-excited molecular states are approxi- sentative parameters for W(CO)6 in solution [37, 42], 1 1 1 mately resonant with either one of the available ω0 = 1983 cm− , ∆ = 8 cm− , γ = 3 cm− ), and under two-polariton transitions (see Fig. 1), i.e., strong coupling with a microcavity (ωc = ω0, κ = γ) for 1 1 ΩR = 40 cm− and ΩR = 35 cm− . The curves are nor- 2ω 2∆ = 2ω , ∆ > 0, or malized by the maximum of the bare system TPA. Figure 0 − LP 2ω0 2∆ = 2ωUP, ∆ < 0, (25) 3 shows the strong dependence of the TPA cross-section 1 − on the light-matter interaction: when ΩR = 40 cm− 1 since in this case, all response functions showing (Ω 2∆ = 24 cm− ), the nonlinear absorption is sup- R − up in Eqs. 7 and 19 23 (namely, Gmm(ω), and pressed relative to that given by the bare system. How- − 1 the scattering amplitude Γmm,mm(2ω)) become res- ever, a slight decrease of ΩR to 35 cm− leads to enhanced onant at ω = ωLP (if ∆ > 0) or ω = ωUP (if ∆ < 0). TPA due to a stronger spectral overlap between the LP2 8

NL NL W (ω)/W0 (ω0-Δ) forΩ R =2Δ two reasons: (a) the intracavity enhancement factor (represented by the cavity finesse) varies spatially accord- -1 6000 κ= 3 cm (Lc) ing to the cavity mode profile (sin(πz/L) in the sim- -1 κ= 6 cm (Lc/2) plest case), whereas we assumed that all molecules are 5000 -1 κ= 12 cm (Lc/4) within a small region around an antinode of the cavity 4000 field (so that the cavity field enhancement factor is maximal), and (b) inhomogeneous broadening of the molecu- 3000 lar subsystem which allows for potentially fast polariton relaxation into reservoir (dark) modes, as well as reduc- 2000 tion in efficiently of polariton pumping due to photonic 1000 intensity borrowing. Although we recognize the impor- tance of these approximations, we disregard them in our 0 explorations, since the inhomogeneity of the cavity mode 1970 1972 1974 1976 1978 1980 profile is expected to change the nonlinear response prop- ω(cm -1) erties by factors of order 1 (alternatively, spacers may be introduced between the molecular system and the op- FIG. 4. Enhancement of TPA rate of strongly coupled sys- tical cavity so that the molecules occupy only a small 1 tem when ΩR = 2∆, with ω0 = ωc, γ = 3 cm− , ∆ = region around the antinode of the cavity mode profile), 1 8 cm− for optical microcavities with different cavity lengths while (lower) polariton decay can be slowed down by in- (Lc,Lc/2,Lc/4) and corresponding decay rates. creasing the Rabi splitting and (or) lowering the temperature. Moreover, polariton transitions are well-known to be homogeneously broadened within their lifetimes [47], mode and the molecular doubly excited-state transition and therefore, for molecular systems with significant in- from the ground-state. homogeneously broadened transitions, we expect polari- In Fig. 4, we explore the great potential for obtain- ton lineshapes to be significantly narrower than that of ing polariton-enhanced TPA with optical microcavities the bare system (given the polariton “hole-burning” ef- of varying longitudinal lengths Lc,Lc/2 and Lc/4 (with fect [78] yielding subnatural linewidths). In this instance, Lc = 10 µm, and cavity-mode indices m = 4, m = 2 and the mechanism for polariton-mediated TPA presented in m = 1, respectively, which would require cavity mirrors our article would become even more efficient than in the with transmissivity t 2 0.01%). We assume the cav- model considered here. ities are resonant with| | the≈ molecular fundamental tran- 1 While our study focused on infrared polaritonics, we sition, and the TPA condition ΩR = 2∆ = 16 cm− is note that the phenomenology observed in molecular vi- valid (the remaining bare molecule parameters are the brational and electronic strong coupling can be very same as in Fig. 3). The two main conclusions from Fig. similar depending on the system. For instance, ultra- 4 are that: (a) the polariton-mediated TPA cross section fast pump-probe transmission recorded for a microcavity predicted by our model can be larger than the bare one strongly coupled to an organic semiconductor in Ref. [33] by close to 4 orders of magnitude for accessible parame- showed qualitative features identical to the first reported ters, and (b) a decrease in cavity length leads to stronger vibrational polariton pump-probe data [36]. The exciton- nonlinear signals, so that the cavity-mediated TPA will biexciton ladder described in in Ref. [33] is also notable in be maximally efficient when the strongly coupled cavity the studied context because the energy of the correspond- mode has the lowest possible longitudinal quantum num- ing biexciton is more than twice of the excitonic, which ber and mirrors with highest available reflectivity. These would enable verification of the ∆ < 0 case of Eq. 25. In conditions, in fact, maximize the intracavity electromag- fact, whenever electronic transitions are only weakly cou- netic field enhancement relative to free space (see SI Sec. pled to high-frequency vibrational modes, and the elec- 1). tronic S1 S2 (first to second excited-state) transition A similar increase in nonlinear response signal strength is dipole-allowed→ and slightly red- or blue-shifted from with decreasing molecular concentration (Fig. 3) or cav- the S0 S1 (ground to first excited-state), we expect ity longitudinal length (Fig. 4), was observed and quali- electronic→ TPA rates to have similarly appealing poten- tatively analyzed in a different context in Ref. [77]. tial for enhancement in optical cavities under the strong coupling regime as discussed in Sec. 3. It is also notable that, although our model and ex- 4. DISCUSSION AND CONCLUSIONS pressions allow us to derive quantitative nonlinear properties of molecular systems described as three-level sys- The computed enhanced polariton-mediated TPA pro- tem ensembles, the enhancement of molecular nonlinear vides an upper bound estimate to future measurements polarization by intracavity field effects and subnatural of TPA under strong coupling conditions. Experiments polariton linewidths thoroughly discussed in Sec. 2.2 performed on analogous systems could give reduced en- are universal features of the nonlinear susceptibilities of hancements relative to those presented here for at least molecular ensembles under strong coupling. In fact, our 9 work qualitatively corroborates recently reported nonlin- linewidths. Our results suggest an increase of several ear response enhancement induced by strong coupling orders of magnitude can potentially be achieved for the of microcavities with organic semiconductor materials polaritonic nonlinear optical response, especially, when a [79, 80] whose effective Hamiltonian is not given by Eq. multipolariton transition is resonant with a multiphonon 4. Specifically, Barachati and coworkers [79] ascribed (or multi-electronic state) absorption of the molecular third-harmonic generation efficiency gains under cavity system. Our work suggests new application of molecular strong coupling to intracavity field energy density en- polaritonics in two-photon imaging [81], and efficient hancement, whereas the recent Z-scan measurements re- generation of hot molecular excited-state distributions ported by Wang et al. [80] showed that polariton res- via (polariton) ladder climbing [82–84], possibly by- onance effects were also essential to obtain increases in passing deleterious intramolecular vibrational relaxation. the magnitude of the nonlinear index of refraction and absorption. Acknowledgments R.F.R., J.C.G., N.C.G, and In summary, we have derived and analyzed the nonlin- J.Y.Z. acknowledge funding− support from the Defense ear optical susceptibility and TPA rates for a molecular Advanced Research Projects Agency under Award No. system under strong coupling with an infrared micro- D19AC00011 for the calculations of nonlinear suscepti- cavity. By contrasting the polaritonic response with bilities under strong coupling and the Air Force Office that of bare molecules in free space, we found that en- of Scientific Research award FA9550-18-1-0289 for the hanced nonlinearities in the strong coupling regime may analysis of optimal conditions for multiphoton absorp- emerge due to intracavity field enhancement, creation of tion under strong light-matter coupling. W.X. thanks suitable optical resonances, and subnatural polaritonic the support of NSF CAREER Award DMR1848215.

[1] Shaul Mukamel, Principles of Nonlinear Optical Spec- [11] Dhabih V. Chulhai, Zhongwei Hu, Justin E. Moore, troscopy (Oxford University Press on Demand, 1999). Xing Chen, and Lasse Jensen, “Theory of Linear [2] J. Yuen-Zhou, J.J. Krich, A. Aspuru-Guzik, I. Kassal, and Nonlinear Surface-Enhanced Vibrational Spectro- and A.S. Johnson, Ultrafast Spectroscopy: Quantum In- scopies,” Annual Review of Physical Chemistry 67, 541– formation and Wavepackets, IOP Expanding Physics (In- 564 (2016), eprint: https://doi.org/10.1146/annurev- stitute of Physics Publishing, 2014). physchem-040215-112347. [3] Tobias Brixner, Jens Stenger, Harsha M. Vaswani, Min- [12] R.W. Boyd and D. Prato, Nonlinear Optics (Elsevier Sci- haeng Cho, Robert E. Blankenship, and Graham R. ence, 2008). Fleming, “Two-dimensional spectroscopy of electronic [13] Alexey V Kavokin, Jeremy J Baumberg, Guillaume couplings in photosynthesis,” Nature 434, 625 (2005). Malpuech, and Fabrice P Laussy, Microcavities, Vol. 21 [4] E. Thyrhaug, R. Tempelaar, M. J. P. Alcocer, K. Z´ıdek,ˇ (Oxford University Press, 2017). D. B´ına,J. Knoester, T. L. C. Jansen, and D. Zigmantas, [14] Thomas W. Ebbesen, “Hybrid Light–Matter States in a “Identification and characterization of diverse coherences Molecular and Material Science Perspective,” Accounts in the Fenna-Matthews-Olson complex.” Nature chem- of Chemical Research 49, 2403–2412 (2016). istry 10, 780–786 (2018). [15] Maxim Sukharev and Abraham Nitzan, “Optics of [5] Bo Xiang, Yingmin Li, C. Huy Pham, Francesco Pae- exciton-plasmon nanomaterials,” Journal of Physics: sani, and Wei Xiong, “Ultrafast direct electron transfer Condensed Matter 29, 443003 (2017). at organic semiconductor and metal interfaces,” Science [16] Raphael F. Ribeiro, Luis A. Mart´ınez-Mart´ınez, Advances 3, e1701508 (2017). Matthew Du, Jorge Campos-Gonzalez-Angulo, and Joel [6] Ilana Breen, Roel Tempelaar, Laurie A. Bizimana, Yuen-Zhou, “Polariton chemistry: Controlling molecu- Benedikt Kloss, David R. Reichman, and Daniel B. lar dynamics with optical cavities,” Chemical Science 9, Turner, “Triplet Separation Drives Singlet Fission af- 6325–6339 (2018). ter Femtosecond Correlated Triplet Pair Production in [17] Maxim Sukharev, Tamar Seideman, Robert J. Gordon, Rubrene,” Journal of the American Chemical Society Adi Salomon, and Yehiam Prior, “Ultrafast Energy 139, 11745–11751 (2017). Transfer between Molecular Assemblies and Surface Plas- [7] Daniele Sanvitto and StéphaneKéna-Cohen,“The road mons in the Strong Coupling Regime,” ACS Nano 8, 807– towards polaritonic devices,” Nature Materials 15, 1061– 817 (2014). 1073 (2016). [18] Parinda Vasa and Christoph Lienau, “Strong [8] Anton V. Zasedatelev, Anton V. Baranikov, Darius Ur- Light–Matter Interaction in Quantum Emitter/Metal bonas, Fabio Scafirimuto, Ullrich Scherf, Thilo Stöferle, Hybrid Nanostructures,” ACS Photonics 5, 2–23 (2018). Rainer F. Mahrt, and Pavlos G. Lagoudakis, “A room- [19] D. S. Dovzhenko, S. V. Ryabchuk, Yu P. Rakovich, and temperature organic polariton transistor,” Nature Pho- I. R. Nabiev, “Light–matter interaction in the strong cou- tonics 13, 378 (2019). pling regime: Configurations, conditions, and applica- [9] C. Monroe, “Quantum information processing with tions,” 10, 3589–3605 (2018). atoms and photons,” Nature 416, 238 (2002). [20] Felipe Herrera and Jeffrey Owrutsky, “Molecular polari- [10] Samuel L. Braunstein and Peter van Loock, “Quantum tons for controlling chemistry with quantum optics,” The information with continuous variables,” Reviews of Mod- Journal of Chemical Physics 152, 100902 (2020). ern Physics 77, 513–577 (2005). [21] Vladimir M Agranovich, Excitations in Organic Solids, 10

Vol. 142 (OUP Oxford, 2009). mons in metal nanostructures with J-aggregates,” Nature [22] David M. Coles, Niccolo Somaschi, Paolo Michetti, Cas- Photonics 7, 128–132 (2013). par Clark, Pavlos G. Lagoudakis, Pavlos G. Savvidis, [35] Thibault Chervy, Jialiang Xu, Yulong Duan, Chun- and David G. Lidzey, “Polariton-mediated energy trans- liang Wang, Lo¨ıcMager, Maurice Frerejean, Joris A. W. fer between organic dyes in a strongly coupled optical Münninghoff, Paul Tinnemans, James A. Hutchison, microcavity,” Nature Materials 13, 712 (2014). Cyriaque Genet, Alan E. Rowan, Theo Rasing, and [23] Xiaolan Zhong, Thibault Chervy, Lei Zhang, Anoop Thomas W. Ebbesen, “High-Efficiency Second-Harmonic Thomas, Jino George, Cyriaque Genet, James A. Hutchi- Generation from Hybrid Light-Matter States,” Nano Let- son, and Thomas W. Ebbesen, “Energy Transfer be- ters 16, 7352–7356 (2016). tween Spatially Separated Entangled Molecules,” Ange- [36] A. D. Dunkelberger, B. T. Spann, K. P. Fears, B. S. Simp- wandte Chemie International Edition 56, 9034–9038 kins, and J. C. Owrutsky, “Modified relaxation dynam- (2017). ics and coherent energy exchange in coupled vibration- [24] Matthew Du, Luis A. Mart´ınez-Mart´ınez, Raphael cavity polaritons,” Nature Communications 7 (2016), F. Ribeiro, Zixuan Hu, Vinod M. Menon, and Joel 10.1038/ncomms13504. Yuen-Zhou, “Theory for polariton-assisted remote energy [37] Bo Xiang, Raphael F. Ribeiro, Adam D. Dunkelberger, transfer,” Chemical Science 9, 6659–6669 (2018). Jiaxi Wang, Yingmin Li, Blake S. Simpkins, Jeffrey C. [25] E. Orgiu, J. George, J. A. Hutchison, E. Devaux, J. F. Owrutsky, Joel Yuen-Zhou, and Wei Xiong, “Two- Dayen, B. Doudin, F. Stellacci, C. Genet, J. Schachen- dimensional infrared spectroscopy of vibrational polari- mayer, C. Genes, G. Pupillo, P. Samor`ı, and T. W. Ebbe- tons,” Proceedings of the National Academy of Sciences sen, “Conductivity in organic semiconductors hybridized , 201722063 (2018). with the vacuum field,” Nature Materials 14, 1123–1129 [38] Adam D. Dunkelberger, Roderick B. Davidson, Wonmi (2015). Ahn, Blake S. Simpkins, and Jeffrey C. Owrutsky, “Ul- [26] Johannes Feist and Francisco J. Garcia-Vidal, “Extraor- trafast Transmission Modulation and Recovery via Vi- dinary Exciton Conductance Induced by Strong Cou- brational Strong Coupling,” The Journal of Physical pling,” Physical Review Letters 114, 196402 (2015). Chemistry A 122, 965–971 (2018). [27] David Hagenmüller, Johannes Schachenmayer, Stefan [39] Aleksandr G. Avramenko and Aaron S. Rury, “Quantum Schütz, Claudiu Genes, and Guido Pupillo, “Cavity- Control of Ultrafast Internal Conversion Using Nanocon- Enhanced Transport of Charge,” Physical Review Letters fined Virtual Photons,” The Journal of Physical Chem- 119, 223601 (2017). istry Letters 11, 1013–1021 (2020). [28] Anoop Thomas, Jino George, Atef Shalabney, Mar- [40] Courtney A. DelPo, Bryan Kudisch, Kyu Hyung Park, ian Dryzhakov, Sreejith J. Varma, Joseph Moran, Saeed-Uz-Zaman Khan, Francesca Fassioli, Daniele Thibault Chervy, Xiaolan Zhong, Elo¨ıseDevaux, Cyri- Fausti, Barry P. Rand, and Gregory D. Scholes, “Po- aque Genet, James A. Hutchison, and Thomas W. Ebbe- lariton Transitions in Femtosecond Transient Absorption sen, “Ground-State Chemical Reactivity under Vibra- Studies of Ultrastrong Light–Molecule Coupling,” The tional Coupling to the Vacuum Electromagnetic Field,” Journal of Physical Chemistry Letters 11, 2667–2674 Angewandte Chemie International Edition 55, 11462– (2020). 11466 (2016). [41] Daniel Finkelstein-Shapiro, Pierre-Adrien Mante, Sema [29] A. Thomas, L. Lethuillier-Karl, K. Nagarajan, R. M. A. Sarisozen, Lukas Wittenbecher, Iulia Minda, Sinan Vergauwe, J. George, T. Chervy, A. Shalabney, E. De- Balci, Tonu Pullerits, and Donatas Zigmantas, “Radia- vaux, C. Genet, J. Moran, and T. W. Ebbesen, “Tilting tive Transitions and Relaxation Pathways in Plasmon- a ground-state reactivity landscape by vibrational strong Based Cavity Quantum Electrodynamics Systems,” coupling,” Science 363, 615–619 (2019). arXiv:2002.05642 [physics, physics:quant-ph] (2020), [30] Jorge A. Campos-Gonzalez-Angulo, Raphael F. Ribeiro, arXiv:2002.05642 [physics, physics:quant-ph]. and Joel Yuen-Zhou, “Resonant catalysis of thermally [42] Bo Xiang, Raphael F. Ribeiro, Liying Chen, Jiaxi Wang, activated chemical reactions with vibrational polaritons,” Matthew Du, Joel Yuen-Zhou, and Wei Xiong, “State- Nature Communications 10, 1–8 (2019). Selective Polariton to Dark State Relaxation Dynamics,” [31] Tao E. Li, Abraham Nitzan, and Joseph E. Subotnik, The Journal of Physical Chemistry A 123, 5918–5927 “On the Origin of Ground-State Vacuum-Field Catalysis: (2019). Equilibrium Consideration,” arXiv:2002.09977 [physics] [43] Raphael F. Ribeiro, Adam D. Dunkelberger, Bo Xiang, (2020), arXiv:2002.09977 [physics]. Wei Xiong, Blake S. Simpkins, Jeffrey C. Owrutsky, and [32] Nche T. Fofang, Nathaniel K. Grady, Zhiyuan Fan, Joel Yuen-Zhou, “Theory for Nonlinear Spectroscopy of Alexander O. Govorov, and Naomi J. Halas, “Plexciton Vibrational Polaritons,” The Journal of Physical Chem- Dynamics: Exciton-Plasmon Coupling in a J-Aggregate- istry Letters 9, 3766–3771 (2018). Au Nanoshell Complex Provides a Mechanism for Non- [44] Juan B. Pérez-Sánchez and Joel Yuen-Zhou, “Polariton linearity,” Nano Letters 11, 1556–1560 (2011). Assisted Down-Conversion of Photons via Nonadiabatic [33] T. Virgili, D. Coles, A. M. Adawi, C. Clark, P. Michetti, Molecular Dynamics: A Molecular Dynamical Casimir S. K. Rajendran, D. Brida, D. Polli, G. Cerullo, and Effect,” The Journal of Physical Chemistry Letters 11, D. G. Lidzey, “Ultrafast polariton relaxation dynamics in 152–159 (2020). an organic semiconductor microcavity,” Physical Review [45] Federico J. Hernández and Felipe Herrera, “Multi-level B 83, 245309 (2011). quantum Rabi model for anharmonic vibrational polari- [34] Parinda Vasa, Wei Wang, Robert Pomraenke, Melanie tons,” The Journal of Chemical Physics 151, 144116 Lammers, Margherita Maiuri, Cristian Manzoni, Giulio (2019). Cerullo, and Christoph Lienau, “Real-time observation [46] Daniel A Steck, Classical and Modern Optics, revision of ultrafast Rabi oscillations between excitons and plas- 1.7.4 ed. (available online at http://steck.us/teaching, 11

2017). of molecular vibrations with a microcavity mode,” New [47] R. Houdré,R. P. Stanley, and M. Ilegems, “Vacuum-field Journal of Physics 17, 053040 (2015). Rabi splitting in the presence of inhomogeneous broad- [63] Zhedong Zhang, Kai Wang, Zhenhuan Yi, M. Suhail ening: Resolution of a homogeneous linewidth in an in- Zubairy, Marlan O. Scully, and Shaul Mukamel, homogeneously broadened system,” Physical Review A “Polariton-Assisted Cooperativity of Molecules in Micro- 53, 2711–2715 (1996). cavities Monitored by Two-Dimensional Infrared Spec- [48] Jino George, Atef Shalabney, James A. Hutchison, Cyr- troscopy,” The Journal of Physical Chemistry Letters iaque Genet, and Thomas W. Ebbesen, “Liquid-Phase (2019), 10.1021/acs.jpclett.9b00979. Vibrational Strong Coupling,” The Journal of Physical [64] V. M. Agranovich, M. Litinskaia, and D. G. Lidzey, Chemistry Letters 6, 1027–1031 (2015). “Cavity polaritons in microcavities containing disordered [49] J. P. Long and B. S. Simpkins, “Coherent Coupling be- organic semiconductors,” Physical Review B 67, 085311 tween a Molecular Vibration and Fabry–Perot Optical (2003). Cavity to Give Hybridized States in the Strong Coupling [65] V. M. Agranovich and Yu. N. Gartstein, “Nature and dy- Limit,” ACS Photonics 2, 130–136 (2015). namics of low-energy exciton polaritons in semiconductor [50] Shaelyn R. Casey and Justin R. Sparks, “Vibrational microcavities,” Physical Review B 75, 075302 (2007). Strong Coupling of Organometallic Complexes,” The [66] Thomas Botzung, David Hagenmüller, Stefan Schütz, Journal of Physical Chemistry C 120, 28138–28143 JérômeDubail, Guido Pupillo, and Johannes Schachen- (2016). mayer, “Dark state localization of quantum emitters in [51] Cristiano Ciuti, GéraldBastard, and Iacopo Carusotto, a cavity,” arXiv:2003.07179 [cond-mat, physics:quant-ph] “Quantum vacuum properties of the intersubband cavity (2020), arXiv:2003.07179 [cond-mat, physics:quant-ph]. polariton field,” Physical Review B 72, 115303 (2005). [67] Crispin W Gardiner and Hermann Haken, Quantum [52] Y. Todorov, A. M. Andrews, R. Colombelli, S. De Liber- Noise, Vol. 26 (Springer Berlin, 1991). ato, C. Ciuti, P. Klang, G. Strasser, and C. Sirtori, “Ul- [68] Jan A. Leegwater and Shaul Mukamel, “Exciton- trastrong Light-Matter Coupling Regime with Polariton scattering mechanism for enhanced nonlinear response of Dots,” Physical Review Letters 105, 196402 (2010). molecular nanostructures,” Physical Review A 46, 452– [53] Artem Strashko and Jonathan Keeling, “Raman scatter- 464 (1992). ing with strongly coupled vibron-polaritons,” Physical [69] Ningjun Wang, Vladimir Chernyak, and Shaul Mukamel, Review A 94, 023843 (2016). “Cooperative ultrafast nonlinear optical response of [54] C. W. Gardiner and M. J. Collett, “Input and output in molecular nanostructures,” The Journal of Chemical damped quantum systems: Quantum stochastic differen- Physics 100, 2465–2480 (1994). tial equations and the master equation,” Physical Review [70] Vladimir Chernyak, Ningjun Wang, and Shaul Mukamel, A 31, 3761–3774 (1985). “Four-wave mixing and luminescence of confined excitons [55] Daniel A Steck, Quantum and Atom Optics, revision in molecular aggregates and nanostructures. many-body 0.12.0 ed. (available online at http://atomoptics- green function approach,” Physics Reports 263, 213–309 nas.uoregon.edu/˜dsteck/teaching/quantum- (1995). optics/quantum-optics-notes.pdf, 2017). [71] Merav Muallem, Alexander Palatnik, Gilbert D. Nessim, [56] Hao Li, Andrei Piryatinski, Jonathan Jerke, Ajay and Yaakov R. Tischler, “Strong Light-Matter Coupling Ram Srimath Kandada, Carlos Silva, and Eric R. and Hybridization of Molecular Vibrations in a Low-Loss Bittner, “Probing dynamical symmetry breaking us- Infrared Microcavity,” The Journal of Physical Chem- ing quantum-entangled photons,” Quantum Science and istry Letters 7, 2002–2008 (2016). Technology 3, 015003 (2018). [72] Kishan S. Menghrajani, Geoffrey R. Nash, and [57] Marina Litinskaya and Peter Reineker, “Loss of coher- William L. Barnes, “Vibrational Strong Coupling with ence of exciton polaritons in inhomogeneous organic mi- Surface Plasmons and the Presence of Surface Plasmon crocavities,” Physical Review B 74, 165320 (2006). Stop Bands,” ACS Photonics 6, 2110–2116 (2019). [58] Jan Philip Kraack, Angelo Frei, Roger Alberto, and [73] A. Shalabney, J. George, J. Hutchison, G. Pupillo, Peter Hamm, “Ultrafast Vibrational Energy Transfer in C. Genet, and T. W. Ebbesen, “Coherent coupling of Catalytic Monolayers at Solid–Liquid Interfaces,” The molecular resonators with a microcavity mode,” Nature Journal of Physical Chemistry Letters 8, 2489–2495 Communications 6 (2015), 10.1038/ncomms6981. (2017). [74] Max Born and Emil Wolf, Principles of Optics: Elec- [59] D.P. Craig and T. Thirunamachandran, Molecular Quan- tromagnetic Theory of Propagation, Interference and tum Electrodynamics: An Introduction to Radiation- Diffraction of Light (2013). Molecule Interactions, Dover Books on Chemistry Series [75] Michael Tavis and Frederick W. Cummings, “Exact So- (Dover Publications, 1998). lution for an N-Molecule-Radiation-Field Hamiltonian,” [60] Vasil Rokaj, Davis M. Welakuh, Michael Ruggenthaler, Physical Review 170, 379–384 (1968). and Angel Rubio, “Light–matter interaction in the long- [76] G. V. Varada, M. Sanjay Kumar, and G. S. Agar- wavelength limit: No ground-state without dipole self- wal, “Quantum effects of the atom-cavity interaction on energy,” Journal of Physics B: Atomic, Molecular and four-wave mixing,” Optics Communications 62, 328–332 Optical Physics 51, 034005 (2018). (1987). [61] Simone De Liberato, “Light-Matter Decoupling in the [77] Bo Xiang, Raphael F. Ribeiro, Yingmin Li, Adam D. Deep Strong Coupling Regime: The Breakdown of the Dunkelberger, Blake B. Simpkins, Joel Yuen-Zhou, Purcell Effect,” Physical Review Letters 112, 016401 and Wei Xiong, “Manipulating optical nonlinearities (2014). of molecular polaritons by delocalization,” Science Ad- [62] Javier del Pino, Johannes Feist, and Francisco J. Garcia- vances 5, eaax5196 (2019). Vidal, “Quantum theory of collective strong coupling [78] Chiao-Yu Cheng, Rijul Dhanker, Christopher L. Gray, 12

Sukrit Mukhopadhyay, Eric R. Kennehan, John B. As- tation Fluorescence Microscopy,” Annual Review of bury, Anatoliy Sokolov, and Noel C. Giebink, “Charged Biomedical Engineering 2, 399–429 (2000), eprint: Polaron Polaritons in an Organic Semiconductor Micro- https://doi.org/10.1146/annurev.bioeng.2.1.399. cavity,” Physical Review Letters 120, 017402 (2018). [82] Jan Philip Kraack and Peter Hamm, “Vibrational ladder- [79] Fábio Barachati, Janos Simon, Yulia A. Getmanenko, climbing in surface-enhanced, ultrafast infrared spec- Stephen Barlow, Seth R. Marder, and StéphaneKéna- troscopy,” Physical Chemistry Chemical Physics 18, Cohen, “Tunable Third-Harmonic Generation from Po- 16088–16093 (2016). laritons in the Ultrastrong Coupling Regime,” ACS Pho- [83] Ikki Morichika, Kei Murata, Atsunori Sakurai, Kazuyuki tonics 5, 119–125 (2018). Ishii, and Satoshi Ashihara, “Molecular ground-state dis- [80] Kuidong Wang, Marcus Seidel, Kalaivanan Nagarajan, sociation in the condensed phase employing plasmonic Thibault Chervy, Cyriaque Genet, and Thomas W. field enhancement of chirped mid-infrared pulses,” Na- Ebbesen, “Large optical nonlinearity enhancement under ture Communications 10, 3893 (2019). electronic strong coupling,” arXiv:2005.13325 [physics] [84] Dominik M. Juraschek, Tomáˇs Neuman, Johannes (2020), arXiv:2005.13325 [physics]. Flick, and Prineha Narang, “Cavity control of [81] Peter T. C. So, Chen Y. Dong, Barry R. Mas- nonlinear phononics,” arXiv:1912.00122 [cond-mat, ters, and Keith M. Berland, “Two-Photon Exci- physics:physics] (2019), arXiv:1912.00122 [cond-mat, physics:physics]. Supporting Information for Enhanced optical nonlinearities under strong light-matter coupling

Raphael F. Ribeiro,1 Jorge A. Campos-Gonzalez-Angulo,1 Noel C. Giebink,2 Wei Xiong,1, 3 and Joel Yuen-Zhou1 1Department of Chemistry and Biochemistry, University of California San Diego, La Jolla, CA 92093 2Department of Electrical Engineering, The Pennsylvania State University, University Park, PA, 16802 3Materials Science and Engineering Program, University of California San Diego, La Jolla, CA 92093 (Dated: June 16, 2020)

CONTENTS

1. Basic deﬁnitions for empty microcavity[1–3]1

2. Nonlinear susceptibility of strongly coupled molecular system3

3. Nonlinear absorption spectrum under strong coupling7 3.1. Zero detuning 10

4. Nonlinear response of bare molecular system 10

5. Nonlinear absorption spectrum of bare molecular system 11

6. Quantitative comparison of molecular nonlinear susceptibility of bare and strongly coupled systems 13

7. Energy eigenvalues and eigenstates of non-dissipative system 14

References 17

1. BASIC DEFINITIONS FOR EMPTY MICROCAVITY[1–3]

We employ input-output theory[1,2] to describe the open quantum system dynamics of a planar optical microcavity consisting of two highly-reflective symmetric mirrors [4] separated by a distance Lc. The input radiation is taken to have zero momentum along the transverse direction to the cavity longitudinal axis. Since we work in the limiting case where a single cavity mode interacts with the material system, we only consider the free space electromagnetic modes to the right and left of living in a 1D space with length L sufficiently large for the corresponding field operators to satisfy periodic boundary conditions. We suppose the system is probed in transmission geometry, where the incident light irradiates the “left” mirror and the optical signal is generated by the photon flux traversing the “right” mirror. The output photon flux is given by

bR (t) † b (t) , (1) h out out i where the output annihilation operator bout(t) is written in terms of the “right” free space modes at a future time t1 > t [1]:

arXiv:2006.08519v1 [physics.chem-ph] 15 Jun 2020 ∞ R i R iω0(t t1) b (t) = dω0b (ω0)e− − , t1 > t (2) out √2π 1 Z−∞ R where b1 (ω0) is the Heisenberg annihilation operator for a photon with frequency ω0 in the free space to the right of the optical cavity at t1. In the absence of any input on the system from the right mirror, the input-output relations allow us to directly relate the (right) output EM power at time t with the state of the cavity at the same moment. In particular[1,2],

κ bR (t) = b(t), (3) out 2 r 2 where b is the cavity mode annihilation operator and κ is the total cavity leakage rate (including field decay through both mirrors). The latter is proportional to the mirrors transmission probability t 2, as well as inversely related to | | the cavity round-trip time τc = 2Lc/c,(Lc is the cavity longitudinal length) [3], i.e., t 2 c κ = | | = t 2 . (4) τc | | 2Lc In this work, we suppose the microcavity is driven by a superposition of coherent state fields which are nearly- resonant and weakly interact with the cavity (κ is much smaller than the cavity photon frequency). The rotating-wave approximation is employed throughout, as is customary in an input-output treatment [2]. We suppose the electric field of the external source which drives the system is expressed as:

L (+) iωt ( ) iωt Ein(t) = Ein (ω)e− + Ein− (ω)e ω>0 X h i ~ω iωt iωt = i α (ω)e− α† (ω)e , (5) 2 V in − in ω>0 r 0 X h i where 0 is the free space permittivity, V = SL is the quantization volume of the l.h.s (or r.h.s.) free space, and αin C is a coherent state amplitude characterizing the phase and intensity of the input external ﬁeld mode with ∈ 2 1 frequency ω. The photon ﬂux corresponding to each frequency in5 is given by αin(ω) c/L . The photon input operator is thus 2 | |

L (ω) iθin(ω) bin(ω) = i P e , (8) r ~ω 2 iθin(ω) where (ω) = ~ω αin(ω) c/L and θin(ω) is determined by the relationship αin(ω) = αin(ω) e We concludeP this| section| by reviewing some relationships between the cavity electromagnetic| | field intensity in the presence of steady driving, and the corresponding free space intensity. This identification will be essential for the comparison of the nonlinear optical response of a hybrid cavity with that of the bare molecular material. First, we recall that the mirrors of a good cavity have nearly vanishing photon transmission probability t 2 0. The cavity is usually characterized by (a) the total photon leakage rate κ (Eq.4) dependent on both geometric| | → parameters (e.g., the cavity length) and the quality of the mirrors (via its dependence on t 2), and (b) its finesse | | coefficient = π r /(1 r ) (where r is the field reflection probability amplitude) [3], which depends only on the quality of theF mirrors.| | As− we | demonstrate| below, the finesse provides a simple measure of the steady-state intracavity p (resonant) electromagnetic field intensity Ic enhancement compared to free space. In particular, at a cavity antinode, it follows that[3], 2 I F I , (9) c ≈ π 0 where I0 is the free space electromagnetic field intensity [3]. In terms of the finesse, the cavity leakage rate κ can be written as πc κ = . (10) L cF Alternatively, is a given as a simple function of the cavity quality factor Q = ω /κ [3]: F c πc ω = c F Lcωc κ Q = , (11) m

1 Here, we used the following expression for the mean photon flux T/2 S0c L † L φin = Ein(t) Ein(t) ~ωT T/2 h i Z− h i = α (ω) 2c/L (6) | in | ω>X0 which is thus given in units of photon number per unit time. 2 Note that our input fields are obtained from superpositions of coherent states of the electromagnetic field in the l.h.s. free space defined by i L ∞ L iω0(t t0) L iωt b (t) = dω0b0 (ω0)e− − = b (ω)e− , (7) in √2π in Z−∞ ω L X where b0 (ω0) is the annihilation operator of the left mode with frequency ω0 evaluated at a time t0 < t.[1] 3 where m Z is the longitudinal quantum number of the cavity mode, and we used that the symmetric planar cavity ∈ mode frequency corresponding to m is given by ωc = cmπ/Lc. We conclude this section by deriving the cavity electric field enhancement factor from input-output theory. Consider 2 an empty cavity driven by an external field with power (ω) = ~ω αin(ω) c/L, where αin(ω) C. Using the previously L iθin(ω) P | | ∈ defined parametrization bin(ω) = i (ω)/~ωe for the input field, it follows from the input-output treatment of an empty driven cavity that the steady-stateP positive-frequency component of the empty cavity electric field (in the (+)p 3 rotating-wave approximation) Ec (ω) is given by :

κ ibL (ω) E(+)(ω) = Ec − in c 0 2 ω ω + iκ/2 r − c ~ωc 2c κ/2 = i α (ω) eiθin(ω) 2 SL | in | κL ω ω + iκ/2 r 0 c r − c 2 κ/2 F E(+)(ω). (14) ≈ π ω ω + iκ/2 in r − c c where we used E = i ~ωc/20SLc. In the last line we employed ω = ωc + δ and the limit where δω/ωc 0, i.e., 0 → ω ωc. This approximation is consistent with the weak-coupling and near-resonant assumptions of input-output theory[≈ 1,2], and is usuallyp satisﬁed when ω is resonant with polaritons in the strong coupling limit (with Rabi splitting signiﬁcantly weaker than the relevant bare molecule and cavity frequencies).

Equation 14 demonstrates the well-known results that under resonant driving (ω = ωc) (a) the cavity electromagnetic field intensity is enhanced by a factor of 2 /π (at cavity antinodes) compared to free space, and (b) the cavity field is phase-shifted by π/2 relative to the phaseF of the external field. −

2. NONLINEAR SUSCEPTIBILITY OF STRONGLY COUPLED MOLECULAR SYSTEM

In this section, we derive the steady-state third-order polarization induced by continuous-wave input fields acting on a molecular system strongly-coupled to an optical cavity as described in the main text. This polarization is the source of the nonlinear optical signal discussed above. The results obtained here are essential for the computation of the nonlinear molecular absorption under strong light-matter coupling which is described in the next section. To obtain the material nonlinear polarization we solve perturbatively the equations of motion (EOM) for the expectation value of the molecular polarization operator in terms of the driving input fields. We perform this procedure in this subsection assuming the system remains in a pure state at all times. From the effective Hamiltonian introduced in the main text (Eq. 4), we obtain the Heisenberg-Langevin EOM for the expectation value of the cavity-photon annihilation operator using i~∂tb(t) = [b(t),H] i~κb(t)/2: − N d κ L ¯c i~ ~ω˜c b(t) = i~ b (t) µE ai(t) , (15) dt − h i − 2 in − 0 h i r i=1 X where the cavity leakage rate κ (derived within input-output theory) was introduced by the replacement ω ω˜ = c → c ωc iκ/2. Eq. 15 describes the time-dependent response of the cavity to a collective molecular polarization and to the− driving by the input field. Similarly, the molecular response to the cavity electromagnetic field is expressed by the analogous Heisenberg-Langevin EOMs satisfied by the time-dependent single-molecule and collective material

3 This equation can be simply derived by using the Heisenberg equations of motion for the driven cavity mode operator

κ L (i~∂t ~ω˜c)b(t) = i~ b (t) − − 2 in r L κ bin(ω) = b(ω) = i~ , (12) ⇒ − 2 ~ω ~ωc + iκ/2 r − L (+) ~ωc ~ωc κ bin(ω) = Ec (ω) = i b(ω) = , (13) ⇒ s 20SLc s 20SLc 2 ω ωc + iκ/2 r − iωt where we usedω ˜c = ωc iκ/2, and b(t) = b(ω)e . − ω − P 4 polarizations:

d 2 c i~ ~ω˜0 µai(t) = µ E b(t) 2~∆µ a†(t)ai(t)ai(t) dt − h i − 0 h i − h i i 2 c = µ E b(t) 2~∆µ a†(t) ai(t)ai(t) , (16) − 0 h i − h i i h i N N d 2 c i~ ~ω˜0 µ ai(t) = Nµ E b(t) 2~∆µ a†(t)ai(t)ai(t) dt − h i − 0 h i − h i i i=1 i=1 X X N 2 c = Nµ E b(t) 2~∆µ a†(t) ai(t)ai(t) , (17) − 0 h i − h i i h i i=1 X whereω ˜0 = ω0 iγm/2, and we obtained the final equations in each case using the factorization property of normal-ordered (all− annihilation operators are to the right of the creation) pure-state correlation functions which, 3 3 4 for a†(t)a(t)a(t) is valid to O E (see e.g., Ref. [5]). Hence, Eqs. 16 and 17 are valid to O E . h i | in| | in| Each of the time-dependent expectation values appearing in the Eqs. 15 and 17 admits an expansion in powers of the L input field bin (since the cavity is only weakly-coupled to the external modes). For instance, we can write ai(t) = (ph) i p L p h i p ai(t) , where ai(t) = O bin . Hereafter, we will employ the following frequency-domain expansion of the expectationh i value ofh time-dependenti operators, e.g., P iωt (+) iωt ( ) iωt O (t) = O (ω)e− = O (ω)e− + O − (ω)e . (18) h i h i h i h i ω ω>0 X X Performing an expansion of both sides of Eqs. 15 and 17 in powers of the input electric field amplitude we find the third-order contribution to the cavity and molecular annihilation operator expectation values satisfy the following coupled equations in the frequency domain: N (3) ¯c (3) (~ω ~ω˜c) b (ω) = µE ai (ω), (19) − h i − 0 h i i=1 X N N (3) 2 c (3) (1) (2) (~ω ~ω˜0) µai (ω) = Nµ E0 b (ω) 2~∆µ ai† ( ωa) aiai (ωb)δω, ωa+ωb . (20) − h i − h i − h i − h i − i=1 ω ω i=1 X Xa Xb X The positive frequency material third-order polarization component with frequency ω is given by P (3) (ω) = N (3) (3) h i µ i=1 ai (ω). As shown above, it can be expressed in terms of the photonic variable b (ω) and lower-order molecularh correlators.i Inserting the formal solution of Eq. 20 into Eq.19, we find: h i P N (1) (2) (3) ¯c ai† ( ωa) aiai (ωb) b (ω) = 2~∆µE0 h i − h i δω, ωa+ωb , (21) h i ( ω ω˜ )( ω ω˜ ) N µEc 2 − ω ω i=1 ~ ~ c ~ ~ 0 0 Xa b X − − − | | 4 N (3) ~ω ~ω˜c (1) (2) µ ai (ω) = 2~∆µ − ai† ( ωa) aiai (ωb)δω, ωa+ωb . (22) h i − ( ω ω˜ )( ω ω˜ ) N µEc 2 h i − h i − i=1 ω ω i=1 ~ ~ c ~ ~ 0 0 X Xa b X − − − | |

4 More explicitly, we used,

a†(t)a (t)a (t) = ψ a†(t)a (t)a (t) ψ h i i i i h | i i i | i = ψ(t) a†a a ψ(t) h | i i i| i = ψ(t) a† n n a a ψ(t) h | i | iih i| i i| i n Xi = c (t) (n + 1)(n + 1)(n + 2)c (t) n∗ i+1 i i i n+2 n Xi p 5 = c∗(t)√2c2(t) +O E 1 | in| =O( E 3) | in| 5 = |ψ(t){za† 0 0} a a ψ(t) + O E h | i | ih | i i| i | in| ψ(t) a† ψ(t) ψ(t) a a ψ(t) ≈ h | i | i h | i i| i =O( E ) =O( E 2) | in| | in| |a†(t){za (t)a} |(t) , {z } ≈ h i ih i i i where we used the resolution of the identity for the ith oscillator Ii = ni ni in terms of Fock states ni , and kept contributions 3 | ih | {| i} to the wavefunction ψ(t) to O( Ein ). | i | | P 5

(1) The first-order molecular expectation values µai(ωa) describe the linear polarization induced on each molecule. By solving the coupled cavity-matter equationsh (Eqs. i15 and 17) to first-order in the input field, we can obtain the linear molecular polarization in the strongly coupled device. In the frequency domain the equations to be solved are:

N (1) κ L ¯c (1) (~ω ~ω˜c) b (ω) = i~ b (ω) µE ai (ω), (23) − h i − 2 in − 0 h i r i=1 X N (1) c (1) (~ω ~ω˜0) ai (ω) = NµE b (ω). (24) − h i − 0 h i i=1 X The explicit solution for the linear polarization P (1) (ω) N µa (1) (ω) induced by the input ﬁeld is given by: h i ≡ i=1 h ii κ PNµ2EcbL (ω) P (1) (ω) = i 0 in ~ c 2 h i 2 (~ω ~ω˜c)(~ω ~ω˜0) µE N r − − − | 0| c (0) κ L = NµGmm(ω)µ E G (ω)i~ b (ω) , (25) 0 pp 2 in r (0) where Gpp (ω) = 1/(~ω ~ω˜c) is the bare cavity photon frequency-domain propagator, and Gmm(ω) is the single- molecule response function− renormalized due to the material strong interaction with the optical cavity

1 G (ω) = . (26) mm c 2 µE0 N ~ω ~ω˜0 | ω | ω˜ − − ~ −~ c (0) Note the light-matter weak-coupling limit for the molecular response function Gmm = 1/(hbarω ~ω˜0) can be − c straightforwardly obtained from the above expression by performing a power series expansion in terms of µE0 . The linear response induced by the external ﬁeld on the cavity photon is similarly given by: | |

L (1) κ (~ω ~ω˜0)b (ω) b (ω) = i − in ~ c 2 h i − 2 (~ω ~ω˜c)(~ω ~ω˜0) µE N r − − − | 0| κ L = Gpp(ω) i~ b (ω) , (27) − 2 in r where Gpp(ω) is the frequency-domain representation of the cavity photon retarded propagator under strong coupling conditions 1 G (ω) = . (28) pp c 2 µE0 N ~ω ~ω˜c | ω | ω˜ − − ~ −~ 0 Note that the hybrid cavity linear response ﬁeld amplitude given by Eq. 27 has the same form as that for an empty cavity (12). The bare cavity result is obtained trivially by simply taking µ 0 in Eq. 27. The following relationship between the cavity and molecular polarization retarded Green functions will→ be useful later:

(0) Gmm(ω) Gpp(ω) = Gpp (ω) (0) (29) Gmm(ω)

The last expectation value which we need to compute in order to obtain the hybrid cavity third-order response is a (t)a (t) (2) (see Eq. 21). The time-dependence of this function is coupled to the other totally-symmetric (with h i i i respect to permutation of the molecular indices) 2-particle variables of the system, namely, b(t)b(t) (2) which describes (2) h i the evolution of the two-cavity photon state, ai(t)b(t) which probes the correlated propagation of a photon and (2) h i the ith molecule phonon, and ai(t)aj(t) , i = j, that describes propagation of vibrational excited-states in distinct molecules. h i 6 The system of Heisenberg-Langevin equations for the bright two-particle variables mentioned above can be derived using the operator equations of motion generated Hamiltonian in Eq. (4) of the main text, together with the same replacements eﬀected above ω ω iγ /2, and ω ω iκ/2. It follows from the input-output treatment [2] 0 → 0 − m c → c − 6 that under the assumptions of Markovian molecular bath, and in the absence of an input molecular polarization, the resulting two-particle EOMs are given by:

d (2) c (2) i~ 2 (~ω˜0 ~∆) ai(t)ai(t) = 2µE0 ai(t)b(t) , (30) dt − − h i − h i N d (2) c (2) ¯c (2) κ L (1) i~ (~ω˜c + ~ω˜0) ai(t)b(t) = µE0 b(t)b(t) µE ai(t)aj (t) i~ bin(t) ai(t) , (31) dt − h i − h i − 0 h i − 2 h i j=1 r X d (2) ¯c (2) κ (1) i~ 2~ω˜c b(t)b(t) = µE ai(t)b(t) 2i~ bin(t) b(t) , (32) dt − h i − 0 h i − 2 h i i r X d (2) c (2) (2) i~ 2~ω˜0 ai(t)aj (t) = µE0 ai(t)b(t) + aj (t)b(t) , j = i. (33) dt − h i − h i h i 6 h i These equations show, as expected, that two-particle states are driven by the input field only in the presence of non- (1) (1) vanishing first-order photonic or molecular polarization (represented by b(t) and ai(t) ). To solve this system in the frequency domain, we note that the electromagnetic field interactsh equallyi withh eachi molecule, and therefore, ai(t)b(t) = aj(t)b(t) , for all i, j 1, ..., N . From the same argument, it also follows that the correlators ha (t)a (ti) ,h and a (it)a (t) are independent∈ { of} the molecular indices. These considerations imply that, while the h i j ii=j h i i i system of two-particle6 eqs. given above has (N + 1)2 unknowns, only four of those are independent. In order to proceed, we need a a (2) (ω) which can be written as: h i ii (2) 2(~ω 2~ω˜0) 2 2(~ω 2~ω˜c)(~ω 2~ω˜0) a a (ω) = − (µEc) f bb (ω) − − µEcf mb(ω), (34) h i ii D(ω) 0 ext − D(ω) 0 ext

(2) bb κ L (1) (2) mb κ L (1) where fext(ω) = 2i~ 2 binb (ω) and fext (ω) = i~ 2 binai (ω), and D(ω) is a 4th-order polynomial, with its roots corresponding− h to thei bright resonances of− the doubly-excitedh i manifold of the system. Denoting by D(0)(ω) the bare noninteractingp 2-particle resonances p

(0) D (ω) = (~ω ~ω˜c ~ω˜0)(~ω 2~ω˜0 + 2~∆)(~ω 2~ω˜c)(~ω 2~ω˜0), (35) − − − − − it follows that the interacting complex two-particle energy eigenvalues are given by the roots of (0) 2 2 D(ω) =D (ω) 2g N(~ω 2~ω˜0)(~ω 2~ω˜0 + 2~∆) 2g (N 1)(~ω 2~ω˜0 + 2~∆)(~ω 2~ω˜c) − − − − − − − 2 2g (~ω 2~ω˜c)(~ω 2~ω˜0), (36) − − − 2 c 2 where g = µE0 as in the main text. We can also write Eq. 34 in terms of retarded single and two-particle Green functions in| the frequency| domain:

(2) κ c 2 L (1) κ c L (1) aiai (ω) = 2i~ Gmm,pp(ω)(µE ) b b (ω) + i~ Gmm,mp(ω)µE b a (ω), h i − 2 0 h in i 2 0 h in i i r r 2 κ (0) c 2 L L = ~ 2Gmm,pp(ωu + ωv)Gpp(ωu) + Gmm,mp(ωu + ωv)Gmm(ωu)G (ωu) (µE ) b (ωv)b (ωu)δω,ωu+ωv − 2 pp 0 in in uv h i X (37) where Gmm,pp(ω) corresponds to the Fourier transform of the probability amplitude for a two-cavity photon state to undergo a transition into a state where a given molecule is doubly excited, and Gmm,mp(ω) is the transition amplitude into the doubly-excited state of a given molecule from an initial state containing a photon and a single vibrational excitation of the same molecule. These propagators can be written explicitly as:

2(~ω 2~ω0 + i~γm) G (ω) = − , (38) mm,pp D(ω) 2(~ω 2~ωc + i~κ)(~ω 2~ω0 + i~γm) G (ω) = − − . (39) mm,mp D(ω) Using the relation introduced in Eq. 29, we rewrite the two-particle molecular response as: 1 1 1 1 a a (2) (ω) = 2G (ω + ω ) G(0) (ω ) − G(0)(ω ) − + G (ω + ω ) G(0)(ω ) − h i ii 2 mm,pp u v mm u pp v mm,mp u v pp v uv X h i h i h i c 2 κ (0) L κ (0) L Gmm(ωu)(µE ) i~ G (ωv)b (ωv) i~ G (ωu)b (ωu) δω,ωu+ωv . (40) × 0 − 2 pp in − 2 pp in r r 7

By symmetrizing the summand of the previous equation, we obtain:

(2) c κ (0) L c κ (0) L aiai (ω) = Γmm,mm(ωu + ωv)Gmm(ωu)Gmm(ωv) µE i~ G (ωv)b (ωv) µE i~ G (ωu)b (ωu) h i − 0 2 pp in − 0 2 pp in uv r r X δ , (41) × ω,ωu+ωv where Γ is the two-particle scattering matrix, and Γmm,mm is the amplitude for the elastic scattering of two excitations in the same molecule. It may be written as:

2 (~ω 2~ω˜0)(~ω ~ω˜0 ~ω˜c) (~ω 2~ω˜0)(~ω 2~ω˜c) 4g N Γmm,mm(ω) = − − − − − − . (42) D(ωu + ωv) (3) N (3) The nonlinear component of the molecular polarization P (ωs) = µ i=1 ai(ωs) can now be given the explicit form: h i h i P (3) 4 ¯ P (ωs) = 2N~∆µ Gmm(ωs)Gmm(ωw)Γmm,mm(ωu + ωv)Gmm(ωv)Gmm(ωu) h i × ω ω ω uXv w 3 (0) (0) (0) ~κ 2 (+) ( ) (+) ¯ − Gpp (ωv)Gpp (ωw)Gpp (ωu) F Ein (ωv)Ein (ωw)Ein (ωu)δωs,ωv ωw+ωu , (43) 2 r π ! −

c κ L κ 2 (+) where we used iE b (ω) F E (ω) (from Eq. 14). From the above expression and the deﬁnition of the − 0 2 in ≈ 2 π in molecular nonlinear susceptibilityp [6q], we ﬁnd (3) ¯ χ ( ωs; ωv, ωw, ωu) = 2~∆NµGmm(ωs)Gmm(ωw)Γmm,mm(ωu + ωv)Gmm(ωu)Gmm(ωv) − − × 2 ~κ (0) 2 ~κ ¯(0) 2 ~κ (0) µ F Gpp (ωv) µ F Gpp (ωw) µ F Gpp (ωu) δωs,ωv ωw+ωu . " r π 2 #" r π 2 #" r π 2 # − (44)

3. NONLINEAR ABSORPTION SPECTRUM UNDER STRONG COUPLING

In this section, we compute the nonlinear part of the absorption spectrum of an optical microcavity strongly coupled to the molecular polarization. In particular, we will calculate the nonlinear part (in the input electric field amplitude) of the external field power dissipated by the molecular system under steady-state conditions. Mathematically, the steady-state regime is characterized by a time-independent molecular excited-state population N N i.e., ∂t i=1 ai†(t)ai(t) = 0. Using the Heisenberg-Langevin equation for i=1 ai†(t)ai(t), we find that steady-state implies: P h i P N N N ¯c c 0 = i~γm a†(t)ai(t) + µ E b†(t) ai(t) E a†(t)b(t) , − i 0 − 0 i i=1 " i=1 i=1 # X X X N N ¯c µE0 = γ a†a = 2Im b†a (45) ⇒ m i i i i=1 "i=1 ~ # X X The last equality expresses the balance between the steady-state rate of molecular excited-state decay (l.h.s.) and driving by the external field mediated by the cavity (r.h.s). Hence, the photon absorption rate by the molecular system can be written as: 2 W = Im Ec†P , (46) ~ h iss ¯c where Ec† = E0b† and P are the (complex conjugate) cavity electric field amplitude and collective molecular polarization in steady-state, respectively. Both Ec and P admit power series expansions in the external fields (see Sec.2). The L first non-vanishing nonlinear terms scales cubically with the input field bin. Therefore, it follows that the nonlinear response contribution to the photon absorption rate W scales as E 4. To obtain this quantity, we will solve the | in| 8 coupled Heisenberg-Langevin EOMs for population and coherence variables in the presence of driving by the external input fields. From now on, we will denote steady-state quantities by the usual expectation value notation without the subscript “ss”, as we will always work under steady-state conditions. Moreover, we will disregard the frequency dependence of all quantities until we obtain the final expression for the nonlinear absorption. In this section, we take the input field to be a monochromatic beam, i.e., b (ω0) = 0 for all ω0 = ω. in 6 In steady-state, the cavity-molecular polarization coherence E†P satisfies h c i N N 2 c 2 κ ¯c L (~ω˜∗ ~ω˜0) E†P = µ E N b†b a†aj 2~∆µ E†a†aiai i~ E b † P . (47) c − h c i − | 0|  h i − h i i − h c i i − 2 h 0 in i ij=1 i=1 r X X   L 4 Because we only care about the O bin absorption component, and we have assumed our system is in a pure-state, it follows by the same argument employed in Sec.2 that b†a†aiai = b†a† aiai . Thus, h i i h i i h i N N (4) 2 c 2 (4) (4) (2) (2) κ ¯c L (3) (~ω˜∗ ~ω˜0) E†P = µ E N b†b a†aj 2~∆µ E†a† aiai i~ E b † P , c − h c i − | 0|  h i − h i i  − h c i i h i − 2 0 in h i ij=1 i=1 r X X   (48) where we also used that the input fields are classical states uncorrelated with the cavity. Our task is now to express N the steady-state cavity photon number N = b†b , total molecular excited-state population N = a†a and p h i m i=1 h i ii intermolecular coherences a†a in terms of the input field operators to the desired orders. The steady-state cavity h i jii=j P photon number satisfies 6

2 (4) 2 (4) (4) L † Np = Im Ec†P Re bin b , (49) −~κ h i − rκ h i whereas the total molecular excited-stated population and intermolecular coherences are given by:

(4) 2 (4) Nm = Im Ec†P , (50) ~γm h i N N (4) (4) 2(N 1) (4) 4∆ (2) (2) a†a + a†a = − Im E†P + Im a†a† a a , (51) h i ji h j ii γ h c i γ h j ji h j ii i>j j=1 ~ m m ij=1 X X X h i 2 (2) (2) (2) where to obtain the last line, we used Im a†a† a a = Im a†a† = 0. Using the last three results, we h j ji h j ji h j ji h i ﬁnd the intermediate result

n N (4) (4) (4) 2N (4) 2 L 4∆ (2) (2) NN a†a = Im E†P N Re b † b Im a†a† a a , (52) p − h i ji − η h c i − κ h in i − γ h j ji h j ii ij=1 ~ r m ij=1 X X h i 1 1 1 where η− κ− + η− . We now have all of the quantities required to obtain the rate of nonlinear absorption W . In particular,≡ it follows from inserting our last result in Eq. 48 that

2 2 2 N (4) ΩR (4) ΩR 2 L (3) ΩR∆ (~ω˜∗ ~ω˜0) E†P = Im E†P + Re b † b + Im a†a† ajai c − h c i 2 η h c i 4 κ in h i Nγ h j ji h i ~ r m ij=1 h i h i X h i N (2) (2) κ ¯c L (3) 2~∆µ aiai a†E† i~ E b † P . (53) − h i h i c i − 2 0 h ini h i i=1 r X where we used Ω = 2 µEc √N. Note the l.h.s of the previous equation can be written as: R | 0| (4) (4) (4) ~ηs (4) (~ω˜∗ ~ω˜0) Re E†P + iIm E†P =(~ωc ~ω0)Re E†P Im E†P c − h c i h c i − h c i − 2 h c i h i h i h i ~ηs (4) (4) + i Re E†P + (~ωc ~ω0)Im E†P . (54) 2 h c i − h c i 9 where we introduced the notation ηs = κ + γm. Using the above to equate the imaginary part of the left and right-hand-side of Eq. 53, we ﬁnd that:

(4) (ωc ω0) 4∆µN (2) √2κ ~ (4) (2) ¯c L † (3) Re Ec†P + − W = Im aiai ai†Ec† Re E0 bin P , (55) h i ηs − ηs h i h i − ηs h i h i h i

2 2 N (4) ~ηs NL ΩR NL ΩR 2 L (3) 4∆ (2) (2) (ω ω )Re E†P W = W + Re b † b + Im a†a† a a c − 0 h c i − 4 4 η 4  κ in h i Nγ h j ji h j ii  ~ ~ r m ij=1  h i X h i N  (2) (2) κ c L (3)  2∆µ Re a a a†E† + Im E¯ b † P , (56) − h i ii h i c i 2 0 h ini h i i=1 r X h i h i

NL (4) 2 (4) (4) NL where we used W W = Im E†P . We can now eliminate Re E†P and solve for W in terms ≡ ~ h c i h c i of the input ﬁeld variables. This procedure gives

2 NL 2η~ηs ΩR 2 L (3) 4∆(N 1) (2) (2) † † † W = 2 2 2 Re bin b + − Im ajaj ajai j=i (2ω) − 2η [(~ωc ~ω0) + (~ηs) /4] + Ω ηs/2 4~ κ h i γm h i h i 6 − R (r ) h i h i 2η~ηs (2) (2) κ c L (3) † ¯ † + 2 2 2 2∆µNRe aiai ai Ec† + Im E0 bin P 2η [(~ωc ~ω0) + (~ηs) /4] + ΩRηs/2 h i h i 2 h i h i − h i r h i 2η~ηs(ωc ω0) 4∆µN (2) √2κ c L (3) † ¯ † + 2 − 2 2 Im aiai ai Ec† + Re E0 bin P , (57) 2η [(~ωc ~ω0) + (~ηs) /4] + Ω ηs/2 ηs h i h i ηs h i − R ( ) h i h i where we used that Im a†a† a a = 0 when i = j. Thus, our ﬁnal expression for the total nonlinear absorption h j ji h j ii is given by: h i

2 ηΩ ηs 1 NL R L † (3) W (ω) = 2 2 2 Re bin(ω) b (ω) − 2η [(~ωc ~ω0) + (~ηs) /4] + Ω ηs/2 2κ h i − R "r # 2 ηΩRηs 2∆(N 1) (2) (2) † † 2 2 2 − Im ajaj ( 2ω) ajai j=i (2ω) − 2η [(~ωc ~ω0) + (~ηs) /4] + ΩRηs/2 γm h i − h i 6 − h i 2η(2~∆N)ηs (2) (2) † † + 2 2 2 Re ai ai ( 2ω) µaiEc (2ω) 2η [(~ωc ~ω0) + (~ηs) /4] + Ω ηs/2 h i − h i − R ηηs h i √ ¯c L † (3) + 2 2 2 Im ~ 2κE0 bin(ω) P (ω) 2η [(~ωc ~ω0) + (~ηs) /4] + Ω ηs/2 h i − R h i 2η(~ωc ~ω0)4∆N (2) (2) † + 2 − 2 2 Im aiai (2ω) µai Ec† ( 2ω) 2η [(~ωc ~ω0) + (~ηs) /4] + ΩRηs/2 h i h i − − h i 2η( ωc ω0)√2κ ~ ~ ¯c L † (3) + 2 − 2 2 Re E0 bin(ω) P (ω) . (58) 2η [(~ωc ~ω0) + (~ηs) /4] + Ω ηs/2 h i − R h i Each of the above terms can be further simplified by using results obtained in Sec.2. For instance, the identities (3) c (0) (3) c L κ 2 ( ) b (ω) = E¯ Gpp (ω) P (ω) and iE¯ b (ω) † = F E − (ω) (see Eq. 14) can be employed to simplify h i − 0 h i 0 in 2 π in the first line of the last equation, while the 2nd, 3rd, and 5th lines can be simplified using the following results from p q Eqs. 30 and 33

c (2) µE0 (2) (2) aiaj i=j (ω) = ajb (ω) + aib (ω) , and h i 6 −~ω 2~ω˜0 h i h i − h i (2) ~ω 2~ω˜0 + 2~∆ (2) aib (ω) = − c aiai (ω), which imply that h i − µE0 h i (2) (~ω 2~ω0 + 2~∆ + i~γm)(~ω 2~ω0 i~γm) (2) = aiaj i=j (ω) = − 2 2−2 − 2 aiai (ω). (59) ⇒ h i 6 (~ω 2~ω0) + ~ γ × h i − m 10

3.1. Zero detuning

The physical content of the terms in Eq. 58 becomes clearer in the zero-detuning case where ωc ω0, in which case the last two lines of Eq. 58 vanish. Taking advantage also that when the strong coupling condition≈ is satisﬁed ΩR ~η and ΩR ~ηs, the nonlinear absorption can be written as a sum of four simple contributions 4 4 NL NLα (+) W (ω) = W (ω) Ein (ω) , (60) α=1 X where

NL1 2ηκ 1 1 2 (3) W (ω) 2 2 + 2 Re F χ (ω) , (61) ≈ − ~ 4(ω ω0) + κ (ΩR/~) π − "r #

NL2 4η ω ω0 2 (3) W (ω) − 2 2 Im F χ (ω) , (62) ≈ ~ 4(ω ω0) + κ π − "r # 2 (2) 4∆2N aiai (2ω) W NL3 (ω) η h i , (63) 2 2 (+) ≈ (ω ω0) + γm/4 E (ω) 4 − | in | 2 (2) aiai (2ω) (2ω ω20)2∆N W NL4 (ω) η − h i . (64) 2 (+) ≈ − (ΩR/2~) E (ω) 4 | in |

4. NONLINEAR RESPONSE OF BARE MOLECULAR SYSTEM

In order to describe the free space nonlinear polarization induced on a bare molecule ensemble driven by external continuous-wave fields, we employ an effective Hamiltonian that is similar to that used to model the molecular system in an optical cavity. The main difference is that each molecule now interacts with several EM modes (with vanishing momentum along the x, y directions) quantized with periodic boundary conditions. The total Hamiltonian in the rotating-wave approximation is:

N N iωzi/c ¯ iωzi/c H = ~ωb† bω + ~ω0a†ai ~∆a†a†aiai µ E0ωa†bωe + E0ωaib† e− , (65) ω i − i i − i ω ω i=1 i=1 ω>0 X X X X where ω = ck, k = 2πm/L, m Z, and zi is the projection of the position of molecule i on the field direction of ∈ propagation, and E = i ~ω . The input field which drives the material polarization is introduced as a boundary 0ω 20V condition to the electromagneticq mode operators in the Heisenberg picture, i.e., the input field satisfies the homogeneous part of the EM field equations. We assume the E0ω are classical variables, as in the computation performed with the optical cavity in the previous sections.

The equation of motion for the expectation value of the molecular polarization is given by:

2 iωzi/c (i~∂t ~ω˜0) µai(t) = 2~∆µ a†(t)ai(t)ai(t) µ E0ω bω(t) e . (66) − h i − h i i − h i ω X Assuming the system is always in a pure state, the equation of motion for the third-order component of ai(t) is given by: h i

(3) (1) (2) (i~∂t ~ω˜0) ai(t) = 2~∆ a†(t) ai(t)ai(t) . (67) − h i − h i i h i The time evolution of the relevant ﬁrst and second-order molecular expectation values are given by the solutions of the equations

(1) (i~∂t ~ω˜0) ai(t) = µ Eω in(t), (68) − h i − 0 Xω0 (2) (1) iω0zi/c (i~∂t ~ω˜20) ai(t)ai(t) = 2µ ai(t) Eω in(t)e− , (69) − h i − h i 0 Xω0 11 where we made the replacement Eωin(t) = E0ω bω(t) . Using the long wavelength limit, and thus disregarding the spatial dispersion of the electromagnetic ﬁeld (ash in thei computations performed for a molecular system in a cavity), the frequency-domain solutions of the prior equations are: E a (1) (ω) = µ ωuin δ , (70) h ii − ω ω˜ ωu,ω ω ~ ~ 0 Xu − 2µ2E E a a (2) (ω) = ωuin ωv in δ , (71) h i ii ( ω + ω ω˜ )( ω ω˜ ) ω,ωu+ωv ω ω ~ u ~ v ~ 20 ~ u ~ 0 Xu Xv − − where ~ω˜20 = ~ω20 i~γm, and ~ω20 = 2~ω0 2~∆ is the energy diﬀerence between the doubly-excited vibrational state and the ground-state.− These results can also− be written in terms of bare molecule single-particle and two-particle retarded response functions in the frequency domain:

a (1) (ω) = µG(0) (ω)E , h ii − mm ωin a a (2) (ω) = µ2 G(0) (ω + ω )G(0) (ω )E E δ , (72) h i ii mm,mm u v mm u ωuin ωv in ω,ωu+ωv ω ,ω Xu v (0) (0) where Gmm(ω) = 1/(~ω ~ω˜0) and Gmm,mm(ω) = 2/(~ω ~ω˜20) are the Fourier transform of the single-particle and two-particle retarded− molecular Green functions, respectively.− In the time domain, they measure the probability amplitude that a single and a two-phonon state exist for a time t after their creation. Note that the last equation may (0) also be written in terms of a vibration-vibrational scattering matrix element Γmm,mm(ω) = (~ω 2~ω˜0)/(~ω 2~ω˜20) as follows − −

a a (2) (ω) = µ2 Γ(0) (ω + ω )G(0) (ω )G(0) (ω )E E δ . (73) h i ii mm,mm u v mm u mm v ωuin ωv in ω,ωu+ωv ω ,ω Xu v Direct insertion of Eqs. 70 and 71 into the frequency-domain representation of Eq. 67 gives the following solution: ¯ (3) 3 EωuinEωwinEωv in ai (ωs) = 4µ ~∆ δωs,ωu+ωv ωw . (74) h i ( ω ω˜ )( ω ω˜ )( ω + ω ω˜ )( ω ω˜ ) − ω ω ω ~ s ~ 0 ~ w ~ 0∗ ~ u ~ v ~ 20 ~ u ~ 0 uXv w − − − − In terms of the bare molecule Green functions and phonon-phonon scattering amplitudes, the bare third-order molecular nonlinear polarization P (3)(ω ) = µ N a (ω ) (3) can be written as 0 s i=1 h i s i (3) 4 (0) ¯P(0) (0) ¯ P (ωs) = 2~∆Nµ Gmm(ωs)Gmm(ωw)Γmm,mm(ωu + ωv)Gmm(ωv)Gmm(ωu)EωuinEωwinEωv inδωs,ωu+ωv ωw , h i0 − ωuωv ωw X (75) which implies the bare nonlinear susceptibility

(3) 4 (0) ¯(0) (0) (0) (0) χ0 ( ωs; ωv, ωw, ωu) = 2~∆Nµ Gmm(ωs)Gmm(ωw)Γmm,mm(ωu + ωv)Gmm(ωv)Gmm(ωu)δωs,ωv ωw+ωu . (76) − − −

5. NONLINEAR ABSORPTION SPECTRUM OF BARE MOLECULAR SYSTEM

The steady-state rate of photon absorption by the molecular system in free space can be computed from the Hamiltonian in Eq. 65. In particular, the steady-state condition stipulates that in the presence of an external radiation N field, the rate of excitation of the molecular system is equal to its rate of decay, and therefore ∂t i=1 ai†(t)ai(t) = 0, where t is an arbitrary time during which the system satisfies the condition given above. h i Using Heisenberg-Langevin equations of motion for the description of the response of the molecularP system to the external electromagnetic field we find that:

N N 2µE¯0ω γ a†(t)a (t) = Im b† (t)a (t) . (77) m h i i i h ω i i i=1 i=1 ω ~ X X X The l.h.s of the above equality corresponds to energy extracted from (or transferred to) the molecular system by the bath, whereas the r.h.s describes the pumping of the molecular system by the electromagnetic field. Assuming the 12 usual weak coupling condition to be valid in free space, and taking the external field to be given by a macroscopic coherent state with negligible quantum fluctuations, it follows that the bare rate of photon absorption is given by: 2 W = Im E¯ (t) P (t) , (78) 0 ωin h i0 ~ ω X where P (t) refers to the free space (weakly coupled to the EM field) molecular polarization, i.e., P (t) = h i0 h i0 N µa (t) . Thus, the nonlinear contribution to the molecular absorption spectrum is given by: h i=1 i i0 P NL 2 ¯ (3) W0 = Im Ein(tss) P (tss) , (79) ~ h i0 h i where tss is sufficiently long that the system is in steady-state. Equivalently, we can write 2 W NL(ω) = Im E¯ P (ω ) (3) 0 ωsin h s i0 ~ ω Xs 2 (3) ¯ ¯ = Im χ0 ( ωs; ωv, ωw, ωu)EωsinEωuinEωwinEωv in δωs,ωv ωw+ωu (80) − − − ~ ω ω ω ω Xs uXv w h i The nonlinear absorption spectrum for photons with frequency ω is given by: 4 NL 2 Eωin (3) W0 (ω) = | | Im χ0 ( ω; ω, ω, ω) . (81) ~ − − h i Using Eq. 76, we find the nonlinear rate of absorption of photons by the molecular system is given by: 4 NL 2 4~∆µ ~γm(~ω0 ~ω) 4 W0 (ω) = N 2 −2 2 2 Eωin ~ [( ω ω )2 + 2γ2 /4] (2~ω ~ω20) + ~ γ | | ~ ~ 0 ~ m − m − 4 2 4~∆µ ~γm(~ω20/2 ~ω) 4 + N − Eωin . (82) 2 2 2 2 2 2 2 ~ [(~ω ~ω0) + ~ γ /4] (2~ω ~ω20) + ~ γm | | − m − Each of the two terms in the above rate of nonlinear absorption correspond to a distinct nonlinear absorption resonance. This can be seen by noting that the first term vanishes when ω = ω0, whereas the second vanishes when 2ω is resonant with the two-photon transition with frequency ω20 = 2ω0 2∆. When ∆/γm 1, the lineshapes corresponding to − NL the two possible nonlinear absorption resonances are well separated, and we can isolate the contribution to W0 (ω) corresponding to two-photon absorption:

4 TPA 2 4~∆µ ~γm(~ω0 ~ω) 4 W (ω) N − Eωin . (83) 0 2 2 2 2 2 2 2 ≡ ~ [(~ω ~ω0) + ~ γ /4] (2~ω ~ω20) + ~ γm | | − m − The textbook expression for the two-photon absorption rate [6] follows from the last result by taking the limit where ∆ γ, and by assuming only probe frequencies ω around the TPA resonance at ω0 ∆ (so that no other quantum transitions interference with the absorption) In this case, it follows that5: − 4 TPA 2πN 2µ 4 W0 (ω) 2 ρ2(2ω) Eωin , (85) ≈ ~ (~ω ~ω0) | | − 1 (2) where ρ (2ω) = Im Gmm,mm(2ω) : 2 − π h i 1 2 1 2~γ ρ2(2~ω) = Im = 2 2 2 . (86) −π 2~ω ~ω20 + i~γ π (2~ω ~ω20) + ~ γ − − m

5 Speciﬁcally, letting ω = ω0 ∆ with 0 and γm/∆ 0, we have − − → → 2 2 ~∆(~ω0 ~ω) ~ ∆ (1 + /∆) − 2 2 2 2 2 2 [(~ω ~ω0) + ~ γ /4] ≈ [(~ω ~ω0) ] − m − ~2∆2 (1 + /∆) 2 2 2 2 ≈ (~ω ~ω0) ~ ∆ (1 + /∆) − 1 = [1 + O(/∆)] . (84) 2 (~ω ~ω0) − 13

Im[(3)(-;,-,)] Re[(3)(-;,-,)]

0.6 1.0 0.4 0.2 0.5 0.0 -0.2 0.0 -0.4 -0.6 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 (cm-1) (cm-1)

(3) (3) FIG. 1. Left (Right): Imaginary (Real) parts of χ ( ω; ω, ω, ω) and χ0 ( ω; ω, ω, ω) for a system with ω0 = 1 1 1 1 − − 1 − − 1983 cm− , γm = 3 cm− , κ = 6 cm− , ΩR = 40 cm− , and ∆ = 8 cm− . The grey curve corresponds to results obtained 1 1 for the bare molecular system. Purple (orange) corresponds to ωc ω0 = 7 cm− (ωc ω0 = 7 cm− ), and blue describes the − − − results obtained for ωc = ω0.

6. QUANTITATIVE COMPARISON OF MOLECULAR NONLINEAR SUSCEPTIBILITY OF BARE AND STRONGLY COUPLED SYSTEMS

In this section, we provide a quantitative discussion of the molecular nonlinear susceptibility under strong coupling 1 with a cavity. In Fig.1, we employ a prototypical system of W(CO) 6 molecules in hexane [7] with ω0 = 1983 cm− , 1 1 γm = 3 cm− , ∆ = 8 cm− to illustrate and compare the real and imaginary parts of the bare nonlinear susceptibility 1 (Eq. 76) to that obtained for the same system under strong coupling with an optical cavity (Eq. 44) with κ = 6 cm− , 1 1 1 1 ΩR = 40 cm− , and the following cavity frequencies: ωc = 1977 cm− , 1983 cm− , 1990 cm− [8–10]. For the sake of simplicity, we show results for a monochromatic input field with frequency ω (thus, ωu = ωw = ωw = ω). Fig.1 shows that the bare and strongly coupled molecular system display strikingly contrasting nonlinear polarization. The imaginary part of the bare nonlinear susceptibility shows absorptive lineshapes, whereas dispersive behavior can be observed for the polaritonic. The opposite is true for the corresponding real parts. The absorptive lineshapes (3) 1 for Im χ ( ω; ω, ω, ω) centered at ω and ω ∆ (see small bump of grey curve around ω = 1975 cm− ) are 0 − − 0 0 − expectedh since this functioni is directly proportional to the nonlinear absorption rate (Eq. 81) by the bare molecules. The weak resonance at ω ∆ corresponds to two-photon absorption by the molecular subsystem which absorb 0 − two input photons with ω = ω0 ∆ to generate a population of molecules with energy ~ω = 2~ω0 2~∆ in the − − doubly-excited state, whereas the resonance at ω0 results from stimulated emission by excited-state population and ground-state bleach which contribute to the reduced nonlinear photon absorption probability at the fundamental frequency ω0 (thus giving rise to the observed negative amplitude). It is harder to interpret Im χ(3)( ω; ω, ω, ω) . As discussed in Sec.3, by virtue of the cavity-matter strong coupling, the nonlinear polarization− contribution− to the energy absorbed by the molecular subsystem is not directly proportional to the imaginary part of χ(3)( ω; ω, ω, ω). Nevertheless, the most obvious features of the molecular nonlinear susceptibility under strong coupling− are visible− from Fig.1. For instance, the absorptive lineshapes displayed by Re χ(3)( ω; ω, ω, ω) are all centered at the LP and the UP frequencies for each of the studied systems. Stronger − − nonlinear polarization always happens at ω = ωLP in comparison to ω = ωUP. This happens because, while for N 1, the nonlinear response mediated by LP and UP arises mainly from their interaction with molecular doubly excited- states (see Sec.7), larger spectral overlap exists between the molecular two-photon transition and the LP 2 resonance (for the parameters here chosen). As a result, energy or amplitude transfer between polaritons and molecular doubly excited-states is more efficient when the LP is resonantly driven by the external field (see detailed discussion and connection to experiments [9] in Sections 3 and 4 of the main text). Note also that, for the parameters chosen to obtain Fig.1, the maxima of the nonlinear susceptibility obtained for the molecular system inside and outside of an optical cavity are of the same order of magnitude. However, we expect that if ΩR is modified so that two-polariton states (LP2 in this example) become nearly-resonant with molecular doubly excited-states, the molecular nonlinear susceptibility under strong coupling will likely undergo significant 14

(3) FIG. 2. Left (Right): Imaginary (Real) parts of χ ( ω; ω, ω, ω) and χ0( ω; ω, ω, ω) for a system with equal cavity and molecular fundamental frequencies and decay rates and− varying− Rabi splitting.− The− barely visible grey curve corresponds to 1 results obtained for the bare molecular system, whereas the purple, blue, and orange correspond to ΩR = 20, 16, and 12 cm− .

enhancement, since in this case, spectral overlap between LP2 and molecular doubly excited-states will be large, and the latter will provide an efficient sink for energy disposal by the former (this is not the case for any of the scenarios shown in Fig.1). We conclude this section by presenting in Fig.2 the behavior of the strongly coupled molecular nonlinear suscep- 1 1 tibility for ΩR = 20, 16, and 12 cm− for a system with zero real and imaginary detuning (ωc = ω0 = 1983 cm− and 1 1 κ = γ = 3 cm− , respectively) and ∆ = 8 cm− . Our expectation of an enhanced molecular nonlinear susceptibility under strong coupling with a moderate quality cavity is now verified. Figure2 shows that as Ω 2∆ 0, the R − → nonlinear polarization of the molecular subsystem becomes larger, especially when the two-LP frequency 2ω0 ΩR approaches the TPA resonance at 2ω = 2ω 2∆. We can observe enhancement of both real and imaginary parts− of 0 − χ(3) relative to χ(3) by two orders of magnitude at ω = ω 2∆ when the condition Ω = 2∆ is satisfied. Note that 0 0 − R Im χ(3)( ω; ω, ω, ω) has absorptive lineshapes at the TPA transition. This feature suggests that the enhanced − − signal at ω = ωLP is due to two-LP decay into molecular doubly excited-states. This channel is discussed in detail in Secs. 3 and 4 of the main manuscript.

7. ENERGY EIGENVALUES AND EIGENSTATES OF NON-DISSIPATIVE SYSTEM

In this section, we obtain the optical spectrum of the hybrid system discussed in the main manuscript. For this purpose, we neglect the eﬀects of dissipation, so that the obtained transition frequencies are real. In fact, the Hamiltonian of the hybrid system can be written in this case as:

N N N H =~ωcb†b + ~ω0a†ai ~∆ a†a†aiai ~g a†b + b†ai , (87) i − i i − i i=1 i=1 i=1 X X X where g is the single-molecule light-matter coupling constant. two conservation laws follow from the effective Hamil- tonian given in the main text. First, the Hamiltonian is invariant under permutation of the molecules. Thus, the eigenstates of H can be classified according to the irreducible representation fo the permutation group of N symbols (SN ), and time-dependent evolution only allows transitions between states which belong to the same irrep. Second, it follows from the RWA approximation to the light-matter interaction that the Hamiltonian evolution of the com- N posite system preserves the total number of excitations of the photonic and matter subsystems M = i=1 ai†ai + b†b. Therefore, the eigenstates of H may also be classified according to the total number of excitations in the molecular and photonic subsystems. For instance, the ground-state of the system (M = 0) has all molecules in theP ground-state, while the cavity field is in its vacuum state. The states with M = 1 contain either a single excited vibration ( 1i where 1 i N), or a single-photon ( 1 ), etc. | i ≤ ≤ | 0i 15

Of the many irreps of SN , only the totally-symmetric is relevant in our case. In the manifold of states with M = 1, this feature is well-known: only the totally-symmetric superposition of states with a single excited molecule exchanges energy with the cavity ﬁeld. The non-totally-symmetric states are dark and thus provide no contribution to the optical response of the hybrid system (in the studied ideal model). The lower and upper polariton states are denoted by LP and UP . They can be written in terms of the local-mode basis states as follows: | i | i LP = sin(θ/2) 1 + cos(θ/2) 1 , (88) | i − | 0i | Si UP = cos(θ/2) 1 + sin(θ/2) 1 , (89) | i | 0i | Si 1 1/2 N where 2θ = tan− 2g√N/(ω ω ) and 1 = N − 1 is the molecular singly-excited bright state, and we c − 0 | Si i=1 | ii denote by g the single-molecule light-matter coupling. h i P The bright subspace of the doubly-excited state (M = 2) manifold contains the four two-particle (hybrid) states which are totally-symmetric under permutation of the molecular labels. These states are the only which can be accessed via two-photon transitions in our model (dark modes are never accessed since they require molecular permutational symmetry-breaking operators which are disregarded in our treatment). They are given by:

1 N 2 1 N 20 , 101m = 101a , 1m1m0 m=m = 1a1b , 2m = 2a . (90) | i | i √ | i | i 6 0 N(N 1) | i | i √ | i N a=1 s N a=1 X − Xa>b X Figure S1 illustrates how Hamiltonian evolution induces transitions between these states. From this ﬁgure, we can also see that these four states are the only which can be accessed from a two-photon initial state. In the subspace spanned by the priorly deﬁned states, the total Hamiltonian is given by:

2~ωc ~g√2N 0 0 B ~g√2N ~ω0 + ~ωc ~g 2(N 1) ~g√2 H2 (N) =  −  , (91) 0 ~g 2(N 1) 2~ω0 0 − p  0 g√2 0 2 ω 2 ∆  p~ ~ 0 ~   −  where the matrix was ordered in the same way as the basis states in Eq. 90. From now on, we will focus on the case where ωc ω0 since this gives the simplest analytical results, and is also the most relevant. If the molecular≈ oscillators were two-level systems, we would obtain the restriction of the Tavis-Cummings Hamil- tonian to the M = 2 Hamiltonian, which is given by:

2~ωc ~g√2N 0 B H2TC(N) = ~g√2N ~ω0 + ~ωc ~g 2(N 1) . (92)  −  0 g 2(N 1) 2 ω ~ p ~ 0  −  When ω0 = ωc, the TC eigenstates can be readily obtainedp since the secular equation can be written in the simple form: (2ω λ) (2ω λ)2 2(N 1)g2 2g2N(2ω λ) = 0, (93) 0 − 0 − − − − 0 − h i which has solutions: TC g λ = 2ω0 + 2g N 1/2 2ωUP , (94) UP2 − ≈ − 2√N TC p g λ = 2ω0 2g N 1/2 2ωLP + , (95) LP2 − − ≈ 2√N TC p λLU = 2ω0, (96) where the approximate expressions resulted from taking the limit where N . In terms of the bare states 20 , 1 1 and 1 1 , the eigenstates corresponding to the above energies are→ given ∞ by: | i | 0 mi | m m0 i N 1 N 1 UPTC = 2 + 1 1 + − 1 1 (97) | 2 i 4N 2 | 0i 2 | 0 mi 4N 2 | m m0 i r − r r − N 1 N 1 LPTC = 2 1 1 + − 1 1 (98) | 2 i 4N 2 | 0i − 2 | 0 mi 4N 2 | m m0 i r − r r − N 1 N LUTC = − 2 1 1 . (99) | i 2N 1 | 0i − 2N 1 | m m0 i r − r − 16

g√2N g 2(N 1) 20 101m − 1m1m | i | i p | 0 i

g√2

2m | i

FIG. 3. Scheme representing the bright (totally-symmetric matter and cavity states) two-particles states which play a role in the nonlinear spectroscopy of vibrational polaritons discussed here. Above each arrow connecting a pair of states we provide the corresponding Hamiltonian matrix elements (coupling constants).

Using these states along with the 2i as the new basis vectors for the totally-symmetric doubly-excited manifold of the system allowing two excitations,| i the Hamiltonian matrix (with the row and column indices in the order LUTC , UPTC , LPTC , 2 ), acquires the simple form: | i | 2 i | 2 i | ii

2~ω0 0 0 0 0 λTC 0 g B ~ UP2 ~ H (N) =  TC  , ~ωc = ~ω0. (100) 2 0 0 ~λ ~g LP2 −  0 ~g ~g 2~ω0 2~∆  − −    From this, we can see that the state LUTC is also an eigenstate of the complete Hamiltonian, and that despite its | i delocalization, the totally-symmetric doubly-excited molecular state 2S is only weakly-coupled to polaritons (the corresponding coupling constant is given by the single-molecule light-matter| i interaction energy g). If we take the single- TC molecule light-matter coupling to be very weak compared to the energy diﬀerences λ (2ω0 2∆) = g√4N 2+2∆ UP2 − − − and λTC (2ω 2∆) = g√4N 2 + 2∆, we can obtain reasonable approximate eigenstates and eigenvalues of LP2 0 B − − − − H2 (N). This will almost always be a valid assumption, even if 2∆ is nearly equal to g√4N 2, since the single- molecule-light coupling constant g is generally too small compared to the energy scale of vibrational− motion, and there exists a large number of (non-totally symmetric) molecular doubly excited-states with energy 2ω 2∆ that 0 − provides an eﬃcient decay channel for LP2 states. In other words, the TC eigenstates will almost always be very good B approximations to the eigenstates of H2 (N). The leading-order perturbatively-corrected eigenvalues are given by:

1 2g2 g 1 2g2 ωUP 2ω0 + g√4N 2 + 2ωUP + , (101) 2 ≈ − 2 g√4N 2 + 2∆ ≈ − 2√N 2 g√4N 2 + 2∆ − − 1 2g2 g 1 2g2 ωLP 2ω0 g√4N 2 2ωLP + , (102) 2 ≈ − − − 2 g√4N 2 2∆ ≈ 2√N − 2 g√4N 2 2∆ − − − − g2 ω 2ω 2∆ + ∆ , (103) 2S ≈ 0 − g2(N 1/2) ∆2 − − ωLU = 2ω0, (104) where we included the exact eigenvalue of the LU state for completeness. The corresponding approximate eigenstates can be written as: | i

N 1 N 1 g UP2 20 + 101m + − 1m1m + 2m , (105) | i ≈ 4N 2 | i 2 | i 4N 2 | 0 i g√4N 2 + 2∆ | i r − r r − − N 1 N 1 g LP2 20 101m + − 1m1m + 2m , (106) | i ≈ 4N 2 | i − 2 | i 4N 2 | 0 i g√4N 2 2∆ | i r − r r − − − g2 √2g∆ 2 2 √N 2 + √N 1 1 1 + 1 1 , (107) | Si ≈ | mi − g2(2N 1) 2∆2 | 0i − | m m0 i g2(2N 1) 2∆2 | 0 mi − − − − N 1 N LU = − 2 1 1 . (108) | i 2N 1 | 0i − 2N 1 | m m0 i r − r − 17

[1] C. W. Gardiner and M. J. Collett, “Input and output in damped quantum systems: Quantum stochastic differential equations and the master equation,” Physical Review A 31, 3761–3774 (1985). [2] Daniel A Steck, Quantum and Atom Optics, revision 0.12.0 ed. (available online at http://atomoptics- nas.uoregon.edu/˜dsteck/teaching/quantum-optics/quantum-optics-notes.pdf, 2017). [3] Daniel A Steck, Classical and Modern Optics, revision 1.7.4 ed. (available online at http://steck.us/teaching, 2017). [4] Alexey V Kavokin, Jeremy J Baumberg, Guillaume Malpuech, and Fabrice P Laussy, Microcavities, Vol. 21 (Oxford University Press, 2017). [5] Shaul Mukamel, Principles of Nonlinear Optical Spectroscopy (Oxford University Press on Demand, 1999). [6] R.W. Boyd and D. Prato, Nonlinear Optics (Elsevier Science, 2008). [7] Bo Xiang, Raphael F. Ribeiro, Adam D. Dunkelberger, Jiaxi Wang, Yingmin Li, Blake S. Simpkins, Jeffrey C. Owrutsky, Joel Yuen-Zhou, and Wei Xiong, “Two-dimensional infrared spectroscopy of vibrational polaritons,” Proceedings of the National Academy of Sciences , 201722063 (2018). [8] Raphael F. Ribeiro, Adam D. Dunkelberger, Bo Xiang, Wei Xiong, Blake S. Simpkins, Jeffrey C. Owrutsky, and Joel Yuen-Zhou, “Theory for Nonlinear Spectroscopy of Vibrational Polaritons,” The Journal of Physical Chemistry Letters 9, 3766–3771 (2018). [9] Bo Xiang, Raphael F. Ribeiro, Liying Chen, Jiaxi Wang, Matthew Du, Joel Yuen-Zhou, and Wei Xiong, “State-Selective Polariton to Dark State Relaxation Dynamics,” The Journal of Physical Chemistry A 123, 5918–5927 (2019). [10] Bo Xiang, Raphael F. Ribeiro, Yingmin Li, Adam D. Dunkelberger, Blake B. Simpkins, Joel Yuen-Zhou, and Wei Xiong, “Manipulating optical nonlinearities of molecular polaritons by delocalization,” Science Advances 5, eaax5196 (2019).