betaFeedback

WorksPhysics and Astronomy

Synchronization-Induced Limit Cycles and Hysteresis in Spinor Exciton-Polaritons

Abstract

We study the nonlinear dynamics of spinor exciton-polaritons in resonantly pumped microcavities. Using steady-state analysis and computing eigenfrequencies via Bogoliubov-like linearization, we show that multiple Hopf bifurcations beyond conventional bistability occur within specific power regimes and are not associated with stable solutions of the system. Owing to nonlinear excitonic interactions, these transitions occur at different pump powers during forward and backward pump sweeps, leading to pronounced hysteresis between fixed-point and limit-cycle states. In the spinor regime, multiple limit-cycle branches and associated dynamical hysteresis loops emerge due to the interplay of polarization coupling and spin-dependent nonlinearities. We show how the locations of the bifurcation points can be tuned by the pump frequency and other system parameters. Our results identify synchronization as the organizing principle underlying limit-cycle formation, dynamical hysteresis, and bifurcation cascades in driven spinor exciton-polariton microcavities.

Introduction

The resonant interaction between Wannier-Mott excitons and cavity photons in optical microcavities produces quantum quasiparticles known as exciton-polaritons [1]. Nonlinearities arising in this interaction under coherent optical pumping are important for a variety of applications ranging from superfluidity [1, 2] and condensation [3, 4, 5], to optical bistability and limiting [6, 7]. Polaritonic optical bistability has been theoretically predicted [8] and experimentally verified [9], enabling simulation of classical Ising models for optimization [10], logical gates [11], optical switching [12], and many other applications [13]. While early studies did not consider phase shifts due to the evolution of different transverse electric (TE) and magnetic (TM) fields in microcavity (known as the TE-TM splitting [14, 15]) and the spin degree of freedom in the exciton-exciton interaction, these effects are now known to play an important role in the formation of topological states and spin textures [13].

Optically pumped microcavities, in which cavity photons and excitons are linearly coupled while excitons interact nonlinearly, constitute a fundamental class of interacting exciton–polariton systems that can be quantitatively described in the lower and upper polariton basis [1]. When upper polaritons are taken into account, higher cavity and excitonic states become accessible and, depending on the pump power and frequency detuning, the system may exhibit chaotic dynamics and complex pattern formation [16]. The polariton dynamics become further enriched when cavity polarization and exciton spin degrees of freedom are included [14, 15]. In both cases of neglecting and considering spin and polarization degrees of freedom of cavity photons and exciton, the physical mechanism underlying the chaotic origin of the excited-state formation remains an open question. Despite the chaotic dynamics of this nonlinear interacting system, exciton and/or cavity-photon bistability—namely, the coexistence of two stable population states for fixed pump power and frequency—commensurate with Hopf bifurcation formation have already been predicted theoretically [17] and confirmed experimentally [18, 19].

In this paper, we show that resonantly driven spinor exciton-polaritons exhibit a hierarchy of synchronization-induced dynamical transitions beyond conventional optical bistability. Multiple Hopf bifurcations give rise to stable limit-cycle oscillations and hysteresis between fixed-point and limit-cycle attractors. Through forward and backward pump power sweeps, we uncover dynamical hysteresis loops associated with the creation and annihilation of these oscillatory states. Although the critical powers depend on pump frequency and other control parameters, synchronization between excitonic and photonic modes emerges as the universal mechanism governing limit-cycle formation and hysteretic transitions.

Model

A sketch of the microcavity exciton-polariton system is shown in Figure 1(a). It consists of a distributed Bragg mirror with a central defect layer designed to maximize electromagnetic-field localization at the defect center (z=0z = 0), for which an excitonic material in a quantum well structure is placed. We work in a rotated frame with respect to the external field’s probe frequency ω\omega. Assuming γc±\gamma_{\mathrm{c}\pm} and γx±\gamma_{\mathrm{x}\pm} as the cavity and exciton decay rates, we denote the left (right) circularly polarized photon wavefunctions as ψcj(r,t)exp⁡[(iω−γc/2)t]\psi_{\mathrm{c}j}(\boldsymbol{r},t)\exp[(i\omega-\gamma_{\mathrm{c}}/2)t], and spin-up (down) exciton wavefunction as ψxj(r,t)exp⁡[(iω−γx/2)t]\psi_{\mathrm{x}j}(\boldsymbol{r},t)\exp[(i\omega-\gamma_{\mathrm{x}}/2)t] where j∈{−,+}j\in\{-,+\}. Furthermore, we consider the coherent near-resonant and spatially homogeneous pump in circular basis as P±exp⁡(iωt)P_\pm\exp(i\omega t) [1]. Considering longitudinal (transverse) cavity photon mass as mlm_l (mtm_t), respectively, the relative mass mrm_r and effective mass mm can be expressed as [15]

Schematic of microcavity exciton–polariton system

Figure 1. (a) Schematic of microcavity exciton–polariton system driven by circularly polarized light. Distributed Bragg mirrors with a central defect layer exhibiting linear polarization splitting Ω\Omega, while the excitonic layer at z=0z = 0 hosts delocalized electron–hole pairs. Inset: cavity polarizations ψc±\psi_{\mathrm{c}\pm}, spinor excitons ψx±\psi_{\mathrm{x}\pm}, Rabi coupling ΩR\Omega_{\mathrm{R}}, and inter (intra) exciton–exciton interactions α1\alpha_1 (α2\alpha_2).

mr=mlmtml−mt,(1)m_r = \frac{m_lm_t}{m_l-m_t}, \tag*{(1)}
m=mlmtml+mt.(2)m = \frac{m_lm_t}{m_l+m_t}. \tag*{(2)}

We then represent the TE-TM splitting term as β=ℏ22mr\beta= \frac{\hbar^2}{2m_r} [15]. Furthermore, we assume exciton nonlinearities as intra-spin (inter-spin) interactions characterized by α1\alpha_1 (α2\alpha_2) [1]. We include exciton and cavity photon decays through γi±\gamma_{i\pm} coefficient for i∈{c,x}i\in\{\mathrm{c},\mathrm{x}\} describing cavity (exciton) decay rates and assume γi+=γi−=:γi\gamma_{i+}=\gamma_{i-}=:\gamma_i. Then we employ a single-particle Hamiltonian approximation for exciton–exciton interaction and obtain the equations governing the dynamics of the cavity and exciton wavefunctions as [15]

iℏ∂ψc±∂t=Δcψc±−ℏ22m∇2ψc±+ΩRψx±+[β(∂y∓i∂x)2−Ω2]ψc∓+P±,(3)i\hbar\frac{\partial\psi_{\mathrm{c}\pm}}{\partial t}=\Delta_{\mathrm{c}}\psi_{\mathrm{c}\pm}-\frac{\hbar^2}{2m}\nabla^2\psi_{\mathrm{c}\pm}+\Omega_{\mathrm{R}}\psi_{\mathrm{x}\pm}+\left[\beta(\partial_y\mp i\partial_x)^2-\frac{\Omega}{2}\right]\psi_{\mathrm{c}\mp}+P_\pm, \tag*{(3)}
iℏ∂ψx±∂t=Δxψx±+ΩRψc±+G(ψx±)ψx±,(4)i\hbar\frac{\partial\psi_{\mathrm{x}\pm}}{\partial t}=\Delta_{\mathrm{x}}\psi_{\mathrm{x}\pm}+\Omega_{\mathrm{R}}\psi_{\mathrm{c}\pm}+G(\psi_{\mathrm{x}\pm})\psi_{\mathrm{x}\pm}, \tag*{(4)}

where the effective cavity (exciton) detuning is

Δi=ℏ(ωi−ω−iγi2),(5)\Delta_i=\hbar\left(\omega_i-\omega-i\frac{\gamma_i}{2}\right), \tag*{(5)}

ωi\omega_i is the bare cavity (exciton) frequency, and

G(ψx±)=α1∣ψx±∣2+α2∣ψx∓∣2,(6)G(\psi_{\mathrm{x}\pm})=\alpha_1\lvert\psi_{\mathrm{x}\pm}\rvert^2+\alpha_2\lvert\psi_{\mathrm{x}\mp}\rvert^2, \tag*{(6)}

is the nonlinear term due to exciton–exciton interactions. To explore exciton and cavity photon dynamics, we consider typical GaAs microcavity parameters; however, without loss of generality, we assume that the polarization and spin detunings and the decay rates have the same values and drop the ±\pm signs. Throughout the text, we use meV and meV/μ\mum to identify the physical units of the excitation frequency [together with detunings] and pump power, respectively. Hence, cavity and exciton detunings are set as ωc=2\omega_{\mathrm{c}}=2 and ωx=0\omega_{\mathrm{x}}=0 while corresponding decay rates are γc≈66×10−3\gamma_{\mathrm{c}}\approx66\times10^{-3} and γx=6.6×10−3\gamma_{\mathrm{x}}=6.6\times10^{-3}, respectively, longitudinal (transverse) photon mass as ml=5×10−5mem_l=5\times10^{-5}m_{\mathrm{e}} (mt=10mlm_t=10m_l), linear polarization splitting as Ω=0.2\Omega=0.2 [15, 20], exciton–polariton coupling as ΩR=4\Omega_{\mathrm{R}}=4 and intra-spin interaction coefficient as α1=6×10−3\alpha_1=6\times10^{-3}.

To investigate the hybridization between exciton and cavity modes, we then calculate the photoluminescence dispersion by setting pump power and nonlinear terms to zero, namely, α1=α2=0\alpha_1=\alpha_2=0 and P±=0P_\pm=0 in Eqs. (3a) and (3b), and calculating the eigenvalues of the linearized 4×44\times4 Hamiltonian. In general, the photoluminescence spectrum implies four separate branches; however, for the parameters considered here, the weak polarization splitting produces nearly degenerate lower and upper polariton branches, shown in Figure 2.

Photoluminescence dispersion showing lower and upper polariton branches

Figure 2. Photoluminescence dispersion showing lower (LP) and upper (UP) polariton branches with Rabi splitting ΩR=4\Omega_{\mathrm{R}}=4 meV for ω\omega the eigenfrequency and k:=(kx,kz)\boldsymbol{k}:=(k_x,k_z) the probe wavevector in momentum space. We assume a one-dimensional cavity and set ky=0k_y=0. We are interested in analyzing cavity and exciton population for frequencies close to the lower branch ω1=−3\omega_1=-3 meV and in the gap ω2=2\omega_2=2 meV. See the text for other parameters.

The dispersion implies anti-crossing of the polaritonic branches due to the strong exciton–polariton hybridization (ΩR≫γt\Omega_{\mathrm{R}}\gg\gamma_t). In the following, we investigate the time-steady-state and time-averaged dynamics, obtainable from Eqs. (3a) and (3b), for both two-component (neglecting the spin degree of freedom) and four-component (taking spin splitting into account) exciton–polariton evolutions. We focus on two frequencies: ω1=−3\omega_1=-3, below the lower edge of the lower polariton branch, where the wavefunction is predominantly photonic, and ω2=2\omega_2=2, in the gap between the lower and upper polariton branches, where both photonic and excitonic components are significant [1].

Stationary solutions

We are interested in the stationary solutions of the homogeneous (i.e., r\mathbf{r} independent) condensate under resonant pumping described by Eqs. (3a) and (3b). Setting ∂x=∂y=∂t=0\partial_x=\partial_y=\partial_t=0 and introducing the four-component vector of the stationary solution ψ(s)=(ψc+,ψc−,ψx+,ψx−)T\boldsymbol{\psi}^{(s)}=(\psi_c^+,\psi_c^-,\psi_x^+,\psi_x^-)^T we require F(ψ(s))=0\mathbf{F}(\boldsymbol{\psi}^{(s)})=0, where

F=((Δcσ0−Ω2σ1ei3ϕΩRσ0ΩRσ0Δxσ0+G(ψx±)))ψ(s)+(Pe+0).(7)\mathbf{F}=\left(\begin{pmatrix}\Delta_c\sigma_0-\frac{\Omega}{2}\sigma_1e^{i3\phi}&\Omega_R\sigma_0\\ \Omega_R\sigma_0&\Delta_x\sigma_0+\mathbf{G}(\psi_{x^\pm})\end{pmatrix}\right)\boldsymbol{\psi}^{(s)}+\begin{pmatrix}P\mathbf{e}_+\\ 0\end{pmatrix}. \tag*{(7)}

and we have defined

G(ψx±)=diag⁡(G+(ψx±),G−(ψx±)),\mathbf{G}(\psi_{x^\pm})=\operatorname{diag}\left(G_+(\psi_{x^\pm}),G_-(\psi_{x^\pm})\right),

with e+=(1,0)T\mathbf{e}_+=(1,0)^T and 0=(0,0)T\mathbf{0}=(0,0)^T. The linear stability of a stationary state is then defined using the ansatz

ψ(t)=ψs+δψ(t)(8)\boldsymbol{\psi}(t)=\boldsymbol{\psi}^{s}+\delta\boldsymbol{\psi}(t) \tag*{(8)}

with (∣δψ(t)∣≪∣ψ(s)∣)(|\delta\boldsymbol{\psi}(t)|\ll|\boldsymbol{\psi}^{(s)}|) which after the linearization of the system (3a), (3b) yields the following standard eigenvalue problem [21]

λϕ=Jϕ,(9)\lambda\boldsymbol{\phi}=\mathbf{J}\boldsymbol{\phi}, \tag*{(9)}

where the Jacobian matrix has entries Jij:=∂Fij/∂ψjJ_{ij}:=\partial F_{ij}/\partial\psi_j. Stability and instability of the fixed point ψ(s)\boldsymbol{\psi}^{(s)} is identified by Re⁡λP>0\operatorname{Re}\lambda_P>0 and Re⁡λP<0\operatorname{Re}\lambda_P<0, respectively. We then compute the eigenvalues of the J\mathbf{J} for both two-component and four-component exciton-polariton dynamics.

Two-component limit

For the outlined system parameters but with α2=0\alpha_2=0, and Ω=0\Omega=0, Eqs. (3a) and (3b) reduce to a two-component exciton-cavity system, where cavity polarization and exciton spin degrees of freedom are negligible. In this case, Eq. (7) has an exact solution [17]. Considering only ψx+\psi_{x+} component and dropping the subindex “+”, the cavity wavefunction acquires the form ψcs=−(P+ΩRψxs)/Δc\psi_c^s=-(P+\Omega_R\psi_x^s)/\Delta_c which upon substituting the exciton characteristic equation (Δc+α1∣ψxs∣2)ψxs+ΩRψcs=0(\Delta_c+\alpha_1|\psi_x^s|^2)\psi_x^s+\Omega_R\psi_c^s=0 and defining nx=∣ψxs∣2n_x=|\psi_x^s|^2 and ϖTC:=(ΔcΔx−ΩR2)\varpi_{\mathrm{TC}}:=(\Delta_c\Delta_x-\Omega_R^2), gives the following characteristic equation

α12∣Δc∣2nx3+2α1Re⁡[ΔcϖTC∗]nx2+∣ϖTC∣2nx−(ΩRP)2=0.(10)\alpha_1^2|\Delta_c|^2n_x^3+2\alpha_1\operatorname{Re}[\Delta_c\varpi_{\mathrm{TC}}^*]n_x^2+|\varpi_{\mathrm{TC}}|^2n_x-(\Omega_RP)^2=0. \tag*{(10)}

Now the cavity population nc:=∣ψcs∣2n_c:=|\psi_c^s|^2 is obtained as

nc=1ΩR2∣Δx+α1nx∣2nx.(11)n_c=\frac{1}{\Omega_R^2}|\Delta_x+\alpha_1n_x|^2n_x. \tag*{(11)}

Eq. (10) can be rewritten in a simpler form

(α1Δcnx+ϖTC)2nx−(ΩRP)2=0.(12)(\alpha_1\Delta_cn_x+\varpi_{\mathrm{TC}})^2n_x-(\Omega_RP)^2=0. \tag*{(12)}

We define η=(α1Δc/ϖTC)nx\eta=(\alpha_1\Delta_c/\varpi_{\mathrm{TC}})n_x and A=α1Δc(ΩRP)2/ϖTC3A=\alpha_1\Delta_c(\Omega_RP)^2/\varpi_{\mathrm{TC}}^3 to obtain (η+1)2η−A=0(\eta+1)^2\eta-A=0 whose discriminant is

Δ=−A(27A+4).(13)\Delta=-A(27A+4). \tag*{(13)}

If Δc>0\Delta_c>0, then A>0A>0 and the equation has only one real root. Bistability emerges at A=−4/27A=-4/27. We note that AA is power-dependent, whose range depends on the frequency of the pump field (see Eq. (20)) for ω=ω2\omega=\omega_2 and the power range satisfying (13).

Equations (10) and (11) are consistent with ψc(x)s\psi_{c(x)}^s satisfying F(ψc(x)s)=0\mathbf{F}(\psi_{c(x)}^s)=0. We then obtain J\mathbf{J} and corresponding eigenvalues by substituting (8) into ∂tψ(t)=F(ψ(t))\partial_t\boldsymbol{\psi}(t)=\mathbf{F}(\boldsymbol{\psi}(t)), performing a Taylor expansion up to first order of δψ\delta\boldsymbol{\psi}, and making use of the stationary solution. We then find the total number of solutions NTN_T and the number of stable solutions NSN_S, shown in Figure 3.

Total number of solutions and number of stable solutions

Figure 3. (a) Total number of solutions (NTN_T) and (b) number of stable solutions NsN_s [right panel] as a function of pump frequency ω\omega and pump power PP for two-component exciton polariton calculated numerically by the two-component version of the Eq. (9). Here we have used α2=0\alpha_2=0 and Ω=0\Omega=0.

Near the lower polariton branch ω1\omega_1, we observe bistability, for which (10) represents lower and upper stable states and middle unstable states [22], as clearly shown via the orange and green regions of Figure 3 for ω<2\omega<2. In the mid-gap where both lower and upper polaritons contribute to the interaction, the system loses its stability, and all the solutions become unstable (the dark blue region in Figure 3). This corresponds to a region of parametric instability in which nonlinear wave mixing leads to simultaneous excitation of the upper and lower polariton branches [17]. We finally recover the bistability for lower pump and high probe frequencies (see green region of Figure 3 with ω>3\omega>3 and P<100P<100).

Four-component model

Inspired by the steady-state solution of the two-component limit, we now consider the four-component case described by Eq. (7). Defining ψcs=(ψc+s,ψc−s)T\psi_c^s=(\psi_{c+}^s,\psi_{c-}^s)^{\mathrm{T}}, we obtain

ψcs=−Δcσ0+(Ω/2)eiσ3σ1Δc2+(Ω/2)2(ΩRψxs+Pe+).(14)\psi_c^s=-\frac{\Delta_c\sigma_0+(\Omega/2)e^{i\sigma_3}\sigma_1}{\Delta_c^2+(\Omega/2)^2}\left(\Omega_R\psi_x^s+Pe_+\right). \tag*{(14)}

Upon substituting (14) into Δxσ0+G(ψxs)\Delta_x\sigma_0+\boldsymbol{G}(\psi_x^s), and defining n±:=∣ψx±∣2n_\pm:=|\psi_{x\pm}|^2, D:=Δc2−(Ω/2)2D:=\Delta_c^2-(\Omega/2)^2, ϖFC:=ΔxD−ΔcΩR2\varpi_{\mathrm{FC}}:=\Delta_xD-\Delta_c\Omega_R^2, α+:=α1n++α2n−\alpha_+:=\alpha_1n_++\alpha_2n_- and α−:=α1n−+α2n+\alpha_-:=\alpha_1n_-+\alpha_2n_+ we obtain amplitude of the exciton wavefunction in component form as

(ϖFC+Dα+)ψx+−ΩR2Ω2e−iϕψx−=ΔcΩRP,(15)\left(\varpi_{\mathrm{FC}}+D\alpha_+\right)\psi_{x+}-\frac{\Omega_R^2\Omega}{2}e^{-i\phi}\psi_{x-}=\Delta_c\Omega_RP, \tag*{(15)}
(ϖFC+Dα−)ψx−−ΩR2Ω2eiϕψx+=ΩR2Ω2eiϕP.(16)\left(\varpi_{\mathrm{FC}}+D\alpha_-\right)\psi_{x-}-\frac{\Omega_R^2\Omega}{2}e^{i\phi}\psi_{x+}=\frac{\Omega_R^2\Omega}{2}e^{i\phi}P. \tag*{(16)}

We observe that Eqs. (14a) and (14b) are nonlinear functions of the exciton populations n±n_\pm and, dissimilar to (10), may contain higher powers of the exciton populations n±>3n_\pm^{>3}. Thereby, by computing the eigenvalues of JJ and identifying the total number of stationary solutions NTN_{\mathrm{T}} and the number of stable stationary solutions NSN_{\mathrm{S}}, we observe that for the frequencies close to the lower polariton branch ω1\omega_1, or at higher probe frequencies but lower powers, the number of stable solutions can exceed two (see the orange dark red colored regions in Figure 4). Thus, multistability can occur.

Total number of solutions and number of stable solutions for the four-component exciton-polariton

Figure 4. (a) Total number of solutions (NTN_{\mathrm{T}}) and (b) number of stable solutions NSN_{\mathrm{S}} calculated numerically using Eq. (9) as a function of pump frequency ω\omega and pump power PP for four-component exciton polariton (see analytical formulae Eqs. (14a) and (16)). Here we have used α2=−0.1α1\alpha_2=-0.1\alpha_1 and Ω=0.2\Omega=0.2, see the text for the details.

The total number and number of stable solutions in Figure 3 and Figure 4 depend on the pump power PP and frequency ω\omega. Tuning these parameters and computing the eigenvalues of JJ we then identify saddle-node bifurcation (the relevant eigenvalues remain negative) and Hopf bifurcations (Re⁡[λmax⁡,P]\operatorname{Re}[\lambda_{\max,P}] changes sign), denoted by SN and HB, respectively. Similar to the two component case, a region of parametric instability emerges in the gap between the lower and upper polariton branches.

Saddle node and Hopf bifurcations

Next we address the long-time dynamics of the system, computing time evolution of the cavity and exciton wavefunctions. In particular, considering γt:=2∑λγλ\gamma_t:=2\sum_\lambda\gamma_\lambda as the total decay rate of the system, we solve Eqs. (3a) and (3b) within t∈[0,T]t\in[0,T]; for T:=103γt−1T:=10^3\gamma_t^{-1}, and perform time-averaging over the long-time dynamics of the cavity and exciton wavefunctions as ⟨f⟩T:=(2/T)∫T/2Tdtf(t)\langle f\rangle_T:=(2/T)\int_{T/2}^{T}dt f(t) [23]. Considering P+:=PP_+:=P and P−=0P_-=0, we define the cavity (exciton) population as

Ix(P):=⟨∣ψx+∣2+∣ψx−∣2⟩t.(17)I_x(P):=\langle|\psi_{x+}|^2+|\psi_{x-}|^2\rangle_t. \tag*{(17)}

We use this equation to identify whether the system relaxes to a stable fixed point, enters a limit cycle, or follows a distinct history-dependent branch for backward pump variation. Numerically, this can be done by employing the pump continuation method, using the final state obtained at the previous pump power step as the initial condition for the next step. This continuation procedure allows tracking of history-dependent attractors and hysteretic branches. In what follows, we compute (17) for both two-component and four-component exciton-polariton systems.

Two component limit

We start our analysis by assuming a two-component problem (namely, with α2=0\alpha_{2}=0 and Ω=0\Omega=0). In this simpler case, critical points in the power spectrum (exciton and cavity population for various pump powers) for specific probe frequency ω\omega satisfy

3α12∣Δc∣2nx2+4α1Re⁡[ΔcϖTC∗]nx+∣ϖTC∣2=0.(18)3\alpha_{1}^{2}|\Delta_{c}|^{2}n_{x}^{2}+4\alpha_{1}\operatorname{Re}[\Delta_{c}\varpi_{\mathrm{TC}}^{*}]n_{x}+|\varpi_{\mathrm{TC}}|^{2}=0. \tag*{(18)}

We now investigate the critical points for probe frequencies ω1\omega_{1} and ω2\omega_{2}. For ω=ω1=−3\omega=\omega_{1}=-3 and the outlined system parameters, we obtain Δc=1−0.033i\Delta_{c}=1-0.033i, Δx=3−0.003i\Delta_{x}=3-0.003i and ϖTC=−13.0−0.103i\varpi_{\mathrm{TC}}=-13.0-0.103i [in the units of (meV)2(\mathrm{meV})^{2}]. Plugging these numerical values into (18), we obtain the critical points

nx,cr(1)=476.75,nx,cr(2)=1425.44,(19)n_{x,\mathrm{cr}}^{(1)}=476.75,\qquad n_{x,\mathrm{cr}}^{(2)}=1425.44, \tag*{(19)}

which upon substitution into Eq. (10) and using P>0P>0, it follows that

Pcr,1=5.03,Pcr,2=47.30,(20)P_{\mathrm{cr},1}=5.03,\qquad P_{\mathrm{cr},2}=47.30, \tag*{(20)}

for Pcr,1P_{\mathrm{cr},1} (Pcr,2P_{\mathrm{cr},2}) being the decreasing (increasing) power saddle node bifurcation, respectively. Furthermore, by calculating Jacobian matrix, we obtain max⁡Re⁡[λPcr,2]=−2.914×10−14\max\operatorname{Re}[\lambda_{P_{\mathrm{cr},2}}]=-2.914\times10^{-14} and max⁡Re⁡[λPcr,1]=1.506×10−12\max\operatorname{Re}[\lambda_{P_{\mathrm{cr},1}]}=1.506\times10^{-12}, indicating that these critical points correspond to a saddle-node bifurcation whose Jacobian eigenvalues satisfy max⁡Re⁡[λSN]→0\max\operatorname{Re}[\lambda_{\mathrm{SN}}]\to0. For the mid-gap frequency ω=ω2=2\omega=\omega_{2}=2, we obtain Δc=−3−0.033i\Delta_{c}=-3-0.033i, Δx=−1−0.003i\Delta_{x}=-1-0.003i and ϖTC=−8.0−0.079i\varpi_{\mathrm{TC}}=-8.0-0.079i. Plugging these values into (18), we obtain two negative solutions for nxn_{x}, indicating the destabilization of the saddle node.

We then exploit Eq. (17) to calculate time-averaged exciton and polariton populations using increasing- and decreasing-pump-power sweeps. To this aim, we solve (3) and (4) by assuming noise to exciton and cavity wavefunctions as the initial condition. For ω=ω1\omega=\omega_{1} late-time dynamics follow different stable branches, see dark-blue and dark-violet lines in Figure 5(a) and 5(b), respectively. Specifically, we observe optical bistability for P≈47P\approx47 (green dot SN1 in Figure 5(a) and 5(b)). For decreasing pump power, the cavity and exciton wavefunctions possess different memory due to the exciton-exciton interaction, and we obtain optical bistability at P≈7P\approx7 (red dot SN1 in Figure 5(a) and 5(b)). These findings are in good agreement with the steady-state critical-power calculation [see Eq. (20)].

Time-averaged cavity and exciton populations during increasing and decreasing pump power sweeps

Figure 5. (a) Time-averaged cavity population IcI_c for both increasing (dark-blue solid line) and decreasing (dashed-dotted dark-blue line) pump power sweeps. (b) Time-averaged exciton population IxI_x for increasing (dark-blue solid line) and decreasing (dashed-dotted dark-blue line) pump power sweeps. For these panels, we have used ω1=−3\omega_1=-3, while the other parameters are the same as in Figure 2. We characterize saddle-node(points) via SN1(2)_{1(2)} of the increasing (decreasing) pump with green (red) dots, respectively. (c,d) Same as (a,b), but with ω2=2\omega_2=2. Critical points HB1_1 and HB2_2 characterized by green (red) stars denote Hopf bifurcations.

Applying Eq. (18) for ω=ω2\omega=\omega_{2} yields negative values for nxn_{x}, indicating that SN points lose stability. To investigate the system’s evolution, we use Eq. (17) to calculate exciton and cavity populations for both increasing and decreasing pump power sweeps. For increasing pump power, we see a non-chaotic increment of Il(P)I_{l}(P) up to P≈63P\approx63, after which cavity and exciton populations experience a jump to a higher state, evolving with near-plateau population amplitude (up to P≈182P\approx182) [see light-blue and magenta, solid lines in Figure 5(c) and 5(d)]. Both populations evolve similarly to Figure 5(a) and 5(b) for P>182P>182. For decreasing pump power, populations start to evolve as a semi-plateau for P≈167P\approx167, and this higher-state evolution lasts up to P≈16P\approx16, for which both cavity and exciton populations undergo a jump and relax to the ground state.

To characterize the criticality behind this evolution, we calculate the eigenvalues of the J\boldsymbol{J} matrix: for increasing pump power and PF1=62.475P_{F1}=62.475 (green HB1_{1} star in Figure 5(c) and 5(d)) the eigenvalue changes from λPF1=−2.6251×10−5+4.2252i\lambda_{PF1}=-2.6251\times10^{-5}+4.2252i to λPF1=4.5873×10−5+4.2290i\lambda_{PF1}=4.5873\times10^{-5}+4.2290i, while for PF2=182.134P_{F2}=182.134 (green HB2_{2} star), we observe a change from λPF2=1.5875×10−5+7.985i\lambda_{PF2}=1.5875\times10^{-5}+7.985i to λPF2=−2.8099×10−5+7.8980i\lambda_{PF2}=-2.8099\times10^{-5}+7.8980i. Thereby, compared to fixed SN points, in this case, max⁡Re⁡[λ]\max\operatorname{Re}[\lambda] changes sign around critical points: the dynamics turns from oscillatory decay (max⁡Re⁡[λ]<0\max\operatorname{Re}[\lambda]<0) to a growth oscillation (max⁡Re⁡[λ]>0\max\operatorname{Re}[\lambda]>0) and vice versa, whose population growth is bounded by the nonlinearity of the system indicating that these points, indeed, are related to the Hopf bifurcations. Despite the applicability of the eigenvalue approach to characterize the Hopf bifurcation for increasing pump power, this approach cannot be applied for decreasing pump power.

To get a deeper insight into the phenomena and to describe the power evolution of the exciton and polariton during the power sweeps, we also investigate the cavity and exciton populations in the (∣ψl(t;P)∣2,∂t∣ψl(t;P)∣2)(|\psi_{l}(t;P)|^{2},\partial_{t}|\psi_{l}(t;P)|^{2}) phase space. Introducing fluctuations about the time-averaged state as δψl2(t;P)=∣ψl(t;P)∣2−⟨∣ψl(t;P)∣2⟩t\delta\psi_{l}^{2}(t;P)=|\psi_{l}(t;P)|^{2}-\langle|\psi_{l}(t;P)|^{2}\rangle_{t} and δ∂t∣ψl(t;P)∣2=∂t∣ψl(t;P)∣2−∂t⟨∣ψl(t;P)∣2⟩t\delta\partial_{t}|\psi_{l}(t;P)|^{2}=\partial_{t}|\psi_{l}(t;P)|^{2}-\partial_{t}\langle|\psi_{l}(t;P)|^{2}\rangle_{t}, we define the phase-space radius using as the root mean square distance from the time-averaged state

Rl(P)=⟨δψl2(t;P)+δ∂t∣ψl(t;P)∣2⟩t.(21)R_{l}(P)=\sqrt{\left\langle\delta\psi_{l}^{2}(t;P)+\delta\partial_{t}|\psi_{l}(t;P)|^{2}\right\rangle_{t}}. \tag*{(21)}

For the increasing sweep, we see a negligible RlR_{l} up to the Hopf bifurcation at P≈62P\approx62, where the distant function starts to grow smoothly, possesses a jump at P≈67P\approx67, reaches to its maximum at P≈105P\approx105 and gradually decreases up to P≈182P\approx182, at which the system relaxes to fixed point through HB2_{2} Hopf bifurcation; see Figure 6(a,Figure 6c). The non-zero radius of the distant function RlR_{l} within P∈[67,182]P\in[67,182] establishes the excitation of the cavity and exciton population to a stable limit-cycle state bounded by the nonlinearity. The downward evolution shows a transition from a fixed point to a limit cycle at P≈167P\approx167 (see red HB1_{1} stars in Figure 6(b,Figure 6d)), the limit cycle approaches the maximum at P≈105P\approx105 and relaxes back to a stable fixed point at P≈16P\approx16 (red HB2_{2} stars in Figure 6). Interestingly, we notice that the trajectories of the cavity photons and excitons are different for increasing and decreasing power sweeps; in the P∈[16,67]P\in[16,67], the system has two different attractors: a fixed point (limit cycle) for increasing (decreasing) power, which establishes the extension of the concept of hysteresis loop for polaritons excited to a higher limit-cycle states.

Root-mean-square phase-space radius of the cavity and exciton populations

Figure 6. Root-mean-square phase-space radius Rα(P)R_{\alpha}(P) of the cavity and exciton populations as a function of pump power PP for ω=ω2\omega=\omega_{2}. Panels (a) and (c) show the increasing pump power sweep for the cavity and exciton components, respectively, whereas panels (b) and (d) show the decreasing power sweep. The green stars mark the forward critical powers PHB1≈62P_{\mathrm{HB}_{1}}\approx62 and PHB2≈182P_{\mathrm{HB}_{2}}\approx182, where the fixed point loses and regains stability through Hopf bifurcations. The red stars indicate the corresponding backward dynamical transition powers, P≈167P\approx167 and P≈16P\approx16, delimiting the persistence of the limit-cycle branch during the downward sweep. A vanishing radius identifies a fixed point, whereas a finite radius signals stable limit-cycle oscillations. The mismatch between the forward and downward transition powers demonstrates hysteretic coexistence of fixed-point and limit-cycle attractors.

Four component model

We now investigate the four-component dynamics corresponding to α2≠0\alpha_{2}\ne0 and Ω≠0\Omega\ne0. We start our analysis by calculating the exciton and cavity photon populations for both increasing and decreasing power sweeps. Similar to the two-component model, we plot in Figure 7 Ii(P)I_{\mathrm{i}}(P) for the probe frequencies close to the lower polariton and in the mid-gap of the PL spectrum. For ω=ω1\omega=\omega_{1}, excitons and cavity photons are close to the condensation; the interaction does not change the attractor, and for forward and downward pump continuation, we observe bistability and hysteresis loop formation. Similar to the two-component case, we find the critical powers as Pcr,1=47.5P_{\mathrm{cr},1}=47.5, Pcr,2=5P_{\mathrm{cr},2}=5, which is very close to Eq. (20). The physical origin is that the four-component PL spectrum for our system parameters and at kx=0k_{x}=0 is very similar to the two-component spectrum. Then we consider the attractor corresponding

Time-averaged cavity and exciton populations

Figure 7. Time-averaged cavity and exciton populations as functions of pump power PP for the four-component model. Panels (a) and (b) show Ic(P)I_{\mathrm{c}}(P) and Ix(P)I_{\mathrm{x}}(P), respectively, for ω1=−3\omega_{1}=-3, exhibiting a bistable hysteresis loop bounded by the saddle-node points SN1_{1} and SN2_{2}. Panels (c) and (d) show the corresponding cavity and exciton populations for ω2=2\omega_{2}=2. Green and red stars mark the dynamical transition points obtained from the increasing and decreasing power sweeps, respectively, while the black star denotes the high-power transition common to both sweeps. Solid and dash-dotted curves indicate increasing and decreasing power sweeps, respectively.

to these critical pumps as the SN.

For forward pump tuning, at P≈58P\approx58, the fixed point loses stability due to dominant inter-exciton-exciton interaction; here, cavity and excitation populations exhibit the first Hopf bifurcation (HB1_{1}, green stars). Similar to a two-component model, the system evolves into a limit cycle. Dissimilarly, as the power increases further, a second Hopf bifurcation occurs at P≈120P\approx120 (HB2_{2}, green stars), associated with enhanced intra-exciton-exciton nonlinear phase shifts, driving the system into a distinct limit cycle. At higher pump powers, P≈150P\approx150, the cavity and pump terms overcome nonlinear phase shifts, which destabilize the limit cycle.

We now confirm these observations by analyzing the eigenvalues of J\mathcal{J} for the increasing power sweep; for ω=ω2\omega=\omega_{2} the two Hopf bifurcations occur at PF1≈58P_{\mathrm{F1}}\approx58 and PF2≈152P_{\mathrm{F2}}\approx152. To characterize the dynamical features of these jumps, we then calculate the maximum eigenvalues of the Jacobian matrix: for PF1P_{\mathrm{F1}} maximum eigenvalue change from λPF1=−6.988×10−5+4.230i\lambda_{P_{\mathrm{F1}}}=-6.988\times10^{-5}+4.230i to λPF1=2.3209×10−4+4.230i\lambda_{P_{\mathrm{F1}}}=2.3209\times10^{-4}+4.230i, and for PF2P_{\mathrm{F2}} we observe a change from λPF1=1.5527×10−4+7.2475i\lambda_{P_{\mathrm{F1}}}=1.5527\times10^{-4}+7.2475i to λPF1=−3.7767×10−5+7.2532i\lambda_{P_{\mathrm{F1}}}=-3.7767\times10^{-5}+7.2532i. Hence, for PF1P_{\mathrm{F1}} the attractor changes from a fixed point to a limit cycle by excitation to the higher polarization and excitonic states, while for $P_{\mathrm{F2}}, pump power inclusion becomes dominant, the limit-cycle attractor loses its stability and relaxes back to the fixed point. This establishes that the steady-state analysis of the forward pump continuation can be used as a universal framework for characterizing HBs and SNs of the forward pump continuation.

For the decreasing power sweep, the dynamics broadly reverse those of the increasing power sweep; however, nonlinear phase shifts induce a distinct history dependence. In particular, at P≈150P \approx150, the fixed point loses stability and, due to the increasing role of the cavity field, undergoes a continuous transition to a limit cycle via spiral trajectories. As the pump power decreases, a stable limit cycle persists down to P≈77P \approx77, where inter-exciton-exciton interactions induce a transition to a distinct limit-cycle regime (HB1_1, red stars in Figure 7). This state remains stable down to P≈14P \approx14, where intra-exciton-exciton interactions dominate the nonlinear phase shifts, leading to a second transition (HB2_2, red stars) from a stable limit cycle to an attracting spiral trajectory, which subsequently relaxes to a stable fixed point. Both cavity and excitonic wavefunctions remain stable for very-low or very high power, as constant pump power is the only dominant mechanism that governs system evolution.

To deepen insight into the attractors of the downward pump continuation, we use the phase space analysis (Iℓ(P),I˙ℓ(P))(I_{\ell}(P),\dot{I}_{\ell}(P)), where Iℓ(P)=Iℓ+(P)+Iℓ−(P)I_{\ell}(P)=I_{\ell+}(P)+I_{\ell-}(P): for each pump PP, we use Eq. (17) to calculate the populations and their time derivatives for both increasing and decreasing power sweeps, and we exploit Eq. (21) to study the trajectory radius of the cavity photons and exciton evolution in Figure 8. Similar to the two-component case, we check the validity of this analysis by considering the increasing pump: we observe a small trajectory radius for both photons and exciton populations up to P≈47P \approx47 for which Rℓ(P)R_{\ell}(P) starts to increase. The trajectory radius then possesses the same cycle as the two-component up to P≈120P \approx120 where the second HB happens, and it relaxes to zero value for P≈150P \approx150. This analysis matches well with the eigenvalue description of the increasing power sweep.

Phase-space trajectory radii of cavity and exciton populations

Figure 8. Phase-space trajectory radii of cavity and exciton populations as functions of pump power PP for the four-component model. (a) and (c) show Rc(P)R_{\mathrm{c}}(P) and Rx(P)R_{\mathrm{x}}(P) for increasing pump power, respectively, whereas (b) and (d) show for decreasing the pump power. Green and red stars mark the forward and backward dynamical transition points, respectively, while black stars indicate the high-power transition common to both sweeps. A negligible radius identifies a fixed-point attractor, whereas a finite radius signals limit-cycle dynamics. Different transition powers and intermediate limit-cycle branches for the two sweep directions reveal multiple hysteresis between distinct dynamical attractors.

For the decreasing pump power, we find a bifurcation (transition of the attractor from fixed-point to a limit cycle) at P≈150P \approx150 for both cavity photon and exciton population, then Rℓ(P)R_{\ell}(P) evolves similarly to the downward pump of the two-component case up to P≈120P \approx120, where the attractor changes from limit-cycle to another limit-cycle, and the radius decreases up to P≈15P \approx15 where the limit-cycle attractor loses its stability, radius becomes negligible and we observe a fixed-point formation. Due to the different classical memory and the formation of different limit cycles, the Hopf bifurcations become different for increasing and decreasing power sweeps; hence, we observe a multiple hysteresis loop of the limit cycles for each limit-cycle attractor.

Synchronization

As discussed in Sec. IV, power tuning of the exciton and cavity photons yields various attractors: close to lower polariton band edge (ω=ω1\omega=\omega_1) we achieve saddle node bifurcations, while inside the band gap (ω=ω2\omega=\omega_2) Hopf bifurcations appear, which can be explained by eigenvalue and trajectory radius analysis. Despite agreement with numerical solutions of Eqs. (3a) and (3b), there are jump-like points in the power spectrum, such as jumps near P≈75P \approx75 in Figure 5, which the change of attractors cannot explain. For these points, the attractor is a limit cycle and does not change due to the jump in the power spectrum. Qualitatively, these jumps should be related to nonlinear excitonic phase shifts that act as a saturation of a limit-cycle orbit; quantitatively, these jumps change the limit cycle of the cavity and exciton populations to other limit-cycle states. Through underpinning the theory based on the Floquet multiplier and numerical confirmation by the fast-Fourier transform analysis, we prove that synchronization is the origin of these anomalous jumps.

To underpin a perturbation theory, we set kx=0k_x = 0, neglect spatial variation and assume ψij(t)=Re⁡[ψij(t)]+iIm⁡[ψij(t)]\psi_{ij}(t) = \operatorname{Re}[\psi_{ij}(t)] + i\operatorname{Im}[\psi_{ij}(t)] to separate the real and imaginary parts of Eqs. (3a) and (3b). Introducing Re⁡[ψij(t)]:=ψijr\operatorname{Re}[\psi_{ij}(t)] := \psi_{ij}^{r} and Im⁡[ψij(t)]:=ψiji\operatorname{Im}[\psi_{ij}(t)] := \psi_{ij}^{i} as short-hand notations, and defining the state vector as

u(t)=(ψc+r,ψc+i,ψc−r,ψc−i,ψx+r,ψx+i,ψx−r,ψx−i),(22)\boldsymbol{u}(t) = \left(\psi_{c+}^{r}, \psi_{c+}^{i}, \psi_{c-}^{r}, \psi_{c-}^{i}, \psi_{x+}^{r}, \psi_{x+}^{i}, \psi_{x-}^{r}, \psi_{x-}^{i}\right), \tag*{(22)}

equations (3) and (4) can be written as a vectorial form

u˙(t)=F(u).(23)\dot{\boldsymbol{u}}(t) = \mathcal{F}(\boldsymbol{u}). \tag*{(23)}

Considering a limit-cycle solution uLC(t)\boldsymbol{u}_{\mathrm{LC}}(t) with initial state u0=uLC(0)\boldsymbol{u}_0 = \boldsymbol{u}_{\mathrm{LC}}(0) satisfying uLC(t+T)=uLC(t)\boldsymbol{u}_{\mathrm{LC}}(t + T) = \boldsymbol{u}_{\mathrm{LC}}(t); for TT the period of the orbit, the nearby trajectory in the phase space can be written as

u(t)=uLC(t)+δu(t).(24)\boldsymbol{u}(t) = \boldsymbol{u}_{\mathrm{LC}}(t) + \delta\boldsymbol{u}(t). \tag*{(24)}

Substituting Eq. (24) into Eq. (23) and using u˙LC(t)=F(uLC)\dot{\boldsymbol{u}}_{\mathrm{LC}}(t) = \mathcal{F}(\boldsymbol{u}_{\mathrm{LC}}), we obtain

δu˙(t)=J(t)δu(t),(25)\delta\dot{\boldsymbol{u}}(t) = \mathcal{J}(t)\delta\boldsymbol{u}(t), \tag*{(25)}

with matrix entries Jij=∂Fi(u)/∂uj\mathcal{J}_{ij} = \partial\mathcal{F}_i(\boldsymbol{u})/\partial u_j evaluated at u=uLC(t)\boldsymbol{u} = \boldsymbol{u}_{\mathrm{LC}}(t). We also express δu(t)\delta\boldsymbol{u}(t) in terms of the ordinary uLC(0)\boldsymbol{u}_{\mathrm{LC}}(0) and perturbed uLC(0)+δu(0)\boldsymbol{u}_{\mathrm{LC}}(0) + \delta\boldsymbol{u}(0) trajectories as

δu(t)=u(t,uLC(0)+δu(0))−u(t,uLC(0)),(26)\delta\boldsymbol{u}(t) = \boldsymbol{u}(t,\boldsymbol{u}_{\mathrm{LC}}(0) + \delta\boldsymbol{u}(0)) - \boldsymbol{u}(t,\boldsymbol{u}_{\mathrm{LC}}(0)), \tag*{(26)}

which upon linearizing the system, and defining Mij(t):=∂uj(0)ui(t,u(0))\mathcal{M}_{ij}(t) := \partial_{u_j(0)}u_i(t,\boldsymbol{u}(0)) [evaluated at $\boldsymbol{u}(0) = \boldsymbol{u}_{\mathrm{LC}}(0)] can be written as

δu(t)=M(t)δu(0),(27)\delta\boldsymbol{u}(t) = \mathcal{M}(t)\delta\boldsymbol{u}(0), \tag*{(27)}

with Mij(0)=δij\mathcal{M}_{ij}(0) = \delta_{ij} as the initial condition. Finally, substituting Eq. (27) into Eq. (25) we obtain

M˙(t)=J(t)M(t).(28)\dot{\mathcal{M}}(t) = \mathcal{J}(t)\mathcal{M}(t). \tag*{(28)}

Computationally, we first use Eq. (28) and calculate the M(t)\mathcal{M}(t) matrix and plug it into Eq. (27) to find the perturbed wavefunction. To solve Eq. (28), we make J(t+T)=J(t)\mathcal{J}(t + T) = \mathcal{J}(t) as a sole assumption of our methodology. After a period TT, one can write monodromy matrix M(T)\mathcal{M}(T) as a eigenfunction vj\boldsymbol{v}_j and eigenenergies μj\mu_j as

M(T)vj=μjvj.(29)\mathcal{M}(T)\boldsymbol{v}_j = \mu_j\boldsymbol{v}_j. \tag*{(29)}

Therefore vj\boldsymbol{v}_j and μj\mu_j characterize the direction of the perturbation and its Floquet multipliers, respectively. We then expand the arbitrary perturbation δu(0)\delta\boldsymbol{u}(0) in the vj\boldsymbol{v}_j basis as

δu(0)=∑jajvj.(30)\delta\boldsymbol{u}(0) = \sum_j a_j\boldsymbol{v}_j. \tag*{(30)}

Then we obtain the final state of the perturbation as δu(mT)=∑jμjmajvj\delta\boldsymbol{u}(mT) = \sum_j \mu_j^m a_j\boldsymbol{v}_j. The trivial Floquet multiplier μ0=1\mu_0 = 1 reflects the neutral time-translation mode of the collective limit cycle. We use this, along with frequency locking ωc=ωx\omega_c = \omega_x and phase locking δθ∼arg⁡[ψc(t)ψx∗(t)]\delta\theta\sim\arg[\psi_c(t)\psi_x^*(t)], to demonstrate synchronized cavity–exciton oscillations.

Logarithmic power spectra of the cavity and exciton populations

Figure 9. Logarithmic power spectra of the cavity and exciton populations across the population jumps for the (a) increasing- and (b) decreasing power sweeps. The upper panels show the spectra before the corresponding jump, where multiple sidebands and appreciable relative-phase fluctuations indicate weak frequency-comb locking. The lower panels show the spectra after the jump, where the cavity and exciton peaks strongly overlap and only a few dominant sidebands remain, demonstrating enhanced frequency and phase locking. In panel (a), the spectra are evaluated at Pbe=75.4P_{\mathrm{be}} = 75.4 and Paf=75.9P_{\mathrm{af}} = 75.9. The dominant peaks occur near ∣ωp∣=3|\omega_p| = 3, 6, and 9, consistent with nonlinear wave mixing and the interaction-induced frequency shift. In panel (b), the pre-jump downward-pump state is evaluated at P=167P = 167; after the jump, the coincident spectra are accompanied by vanishing relative-phase fluctuations, indicating complete synchronization of the cavity and exciton populations.

Two component limit {#sec:Two component limit}

We now test the feasibility of our approach for investigating the jump that appears around P=75P = 75 in Figure 5(c,Figure 5d). For Pbe=75.4P_{\mathrm{be}} = 75.4 and Paf=75.9P_{\mathrm{af}} = 75.9, we find μ0=0.995\mu_0 = 0.995 and μ0=0.997\mu_0 = 0.997, respectively, indicating that both cavity and exciton populations oscillate collectively. Also, the orbit frequencies of cavity and exciton are locked at ωorc=ωorx=4\omega_{\mathrm{or}}^{c} = \omega_{\mathrm{or}}^{x} = 4, suggesting that these states might be synchronized. However, by looking at the logarithmic power spectrum before (see the upper panel of Figure 10(a)) the jump, we observe that the exciton and cavity populations contain multiple sidebands. After the jump (lower panel Figure 10(a)), there are only a couple of sidebands excited with a strong overlap. Thereby, despite the synchronized oscillation of the exciton and cavity photons before and after the jump, jumps cause strongly locked frequency combs.

Logarithmic power spectra of the four-component exciton–polariton system

Figure 10. Logarithmic power spectra of the four-component exciton–polariton system across the discontinuities obtained under forward and downward pump continuation. Panels (a)–(d) show the cavity spectra, while panels (e)–(h) show the corresponding exciton spectra. In each panel, the upper and lower subpanels represent the states before and after the associated jump, respectively. The total cavity and exciton spectra are shown by the black and blue solid lines, while the positive-polarization components are indicated by the orange and red dash-dotted lines and the negative-polarization components by the magenta and green dashed lines, respectively. Panels (a) and (e) correspond to the first forward discontinuity near P≃62P \simeq62, where the prejump spectra display a clear mismatch between the corresponding cavity and exciton components and a finite phase fluctuation, whereas after the jump the spectral peaks overlap and the system becomes perfectly synchronized. Panels (b) and (f) correspond to the second forward discontinuity near P≃120P \simeq120; here, the negative-polarization components become dominant, producing a nonlinear phase shift and finite phase fluctuation before the jump, while the post-jump state forms a stable, frequency-locked comb. Panels (c) and (g) show the downward-pump discontinuity near P≃75P \simeq75, for which the dominant negative components similarly evolve from a phase-lagged state into a perfectly synchronized frequency comb. Panels (d) and (h) correspond to the lower downward-pump transition near P≃15P \simeq15: the post-jump oscillatory state exhibits well-defined and mutually locked spectral peaks, whereas the lower-power state relaxes toward a fixed-point attractor and the comb structure is replaced by weak broadband fluctuations.

We also denote an abnormal phase change through this strong synchronization: before the jump and at P=74P = 74, the relative phase shows fluctuation δθ=0.2π\delta\theta= 0.2\pi, indicating ing that the phases are not locked, while after the jump δθ=0\delta\theta= 0 and both populations become synchronized. The logarithmic spectrum also shows the multiple peaks excited due to the wave-mixing of the cavity and exciton populations due to the exciton-exciton interaction. Here, we have a strong pump at ∣ωc∣=3|\omega_{\mathrm{c}}| = 3, and the nonlinear red-shift detuning causes an excitation of the sidebands at ∣ωp∣=6|\omega_{\mathrm{p}}| = 6 and ∣ωp∣=9|\omega_{\mathrm{p}}| = 9.

We now test these considerations for the decreasing power sweep. We consider P=167P = 167 for this specific case, as is clearly shown in the upper panel of Figure 10(b), the relative phases exhibit strong fluctuations δθ=0.4π\delta\theta= 0.4\pi, leading to de-synchronization of the exciton and polariton populations. After the jump, however, we recover δθ=0\delta\theta= 0 and observe a perfect synchronization of the system.

Four component model

We now investigate the synchronization for the four-component case and for both forward and downward pump continuation. For increasing power, we observe discontinuity of power spectrum at P≈62P \approx62 (see Figure 10(a,Figure 10e) for cavity and exciton spectrum, respectively) and P≈120P \approx120 (indicated at Figure 10(b,Figure 10f) for cavity and exciton spectrum, respectively), while for decreasing power a discontinuity occurs for P≈75P \approx75 (see Figure 10(c,Figure 10g)) and P≈15P \approx15 (see Figure 10(d,Figure 10h) for power spectrum). In what follows, we discuss the forward and downward cases separately.

For the forward pump continuation, we employ the Floquet formalism commensurate with the frequency-locking and phase fluctuation to study the synchronization among exciton and cavity photons. For P≈63P \approx63, we achieve ∣μ0∣=1|\mu_0| = 1, while the phase of the cavity photons and excitons before P=62P = 62 and after P=64P = 64 jumps by π\pi. Furthermore, we obtain phase fluctuations for P=62P = 62 as δθ=0.15π\delta\theta= 0.15\pi while for P=64P = 64, δθ=0\delta\theta= 0 indicating a perfect synchronization. On the other hand, around P≈120P \approx120 we achieve ∣μ0∣=1|\mu_0| = 1, and observe the same phase jump π\pi before and after the jump. Then, for P≈119P \approx119 we find δθ=0.04π\delta\theta= 0.04\pi, while for P≈120P \approx120 we observe a perfect synchronization. These considerations justify that the state of the excitons and cavity photons before and after jumps are synchronized; however, the jumps cause a minimum in phase fluctuations, yielding a perfect synchronization in the system.

To gain a deeper insight into the origin of the jumps, we next discuss the logarithmic power spectrum of the components as well as total cavity photon-exciton evolution in Figure 10. Since only the positive cavity component is pumped by external fields, positive components of the cavity and exciton spectra (shown by orange and red dashed-dotted lines in Figure 10) share similar features to the total power spectrum of the cavity and exciton (represented by black and blue solid lines in Figure 10). Before the first forward discontinuity P=62P = 62 [Figure 10(a) and Figure 10(e)], positive and negative components of the cavity photons evolve dissimilar to the corresponding exciton components, yielding a noticeable phase fluctuation [see upper panels of Figure 10(a) and Figure 10(e)]. After jump P=64P = 64, we observe perfect matching between the spectral components of the cavity photons and excitons (as shown in the lower panels of Figure 10(a) and Figure 10(e)). The same analysis is valid for the jump that occurred at P=120P = 120; however, for this case, negative cavity and exciton components become dominant (see strong variations of the dashed magenta and green lines around ωp=9\omega_{\mathrm{p}} = 9 meV that appear in Figure 10(b) and Figure 10(f), respectively); before jump, the nonlinear phase shift due to this spectral evolution yield a phase fluctuation in the total spectrum which becomes a stable frequency combs after perfect synchronization.

Despite the inaccuracy of the Floquet approach for the downward pump, the outlined spectral analysis represents a powerful approach for investigating the origin of the jumps for this driven system. For the jump that occurred at P=78P = 78, negative components become dominant (as clearly shown in Figure 10(c) and Figure 10(g), respectively); the system possesses a phase lag due to this nonlinear shift, which causes a phase fluctuation in the cavity and exciton component; through the jump, the system become perfectly synchronized and the nonlinear phase shift evolve into the stable frequency comb. The interacting system evolves as perfectly synchronized states up to P=15P = 15 (see bottom panels of Figure 10(d) and Figure 10(f), respectively), for which the system changes the attractor and relaxes to a fixed-point attractor; the frequency combs lose their stability and evolve to the weak broadband noise (see upper panels of Figure 10(d) and Figure 10(f) for P=10P = 10, respectively).

For forward pumping with positive cavity polarization, we observe a clear difference between ψc+\psi_{\mathrm{c}+} (black dot-dashed line in Figure 10(a)) and ψx+\psi_{\mathrm{x}+} (green dot-dashed line) prior to the HB1_1 transition. After this Hopf bifurcation, the system evolves into a stable limit cycle, where the positive components become frequency-locked with a well-defined phase difference (blue and red dot-dashed lines for ψc+\psi_{\mathrm{c}+} and ψx+\psi_{\mathrm{x}+}, respectively). Following HB1_1, the negative components also become frequency-locked, leading to full synchronization between all cavity and exciton components, consistent with the emergence of a limit cycle observed in the power spectrum [Figure 7(a),(Figure 7b)]. As the pump power approaches HB2_2, the positive components remain synchronized; however, ψc−\psi_{\mathrm{c}-} and ψx−\psi_{\mathrm{x}-} exhibit clear asynchronism (black and green dot-dashed lines in Figure 10(b)). This behavior precedes a weaker transition, consistent with α2≪α1\alpha_2 \ll\alpha_1, where the system undergoes a second bifurcation into another limit cycle that restores synchronization among all components (see also Figure 7(a),(Figure 7b)).

Conclusion

We have shown that resonantly driven exciton-polariton microcavities support a rich hierarchy of dynamical states extending beyond conventional optical bistability. While saddle-node bifurcations govern the familiar bistable response near the lower polariton branch, tuning the pump frequency into the regime where excitonic and photonic contributions are comparable gives rise to multiple Hopf bifurcations and stable limit-cycle oscillations. These oscillatory states generate pronounced dynamical hysteresis, with distinct transition powers observed during forward and backward pump sweeps.

In the spinor system, weak polarization splitting and spin-dependent nonlinear interactions further enrich the dynamics, producing multiple limit-cycle branches and hysteresis loops between different dynamical attractors. Linear stability analysis, long-time time-domain simulations, and phase-space diagnostics consistently reveal transitions between stable fixed points and self-sustained oscillatory states. The locations of these transitions can be controlled through the pump frequency and system parameters.

Most importantly, our analysis identifies synchronization of the cavity-photon and exciton degrees of freedom as the organizing principle underlying the observed dynamics. The onset of limit cycles, discontinuous transitions between oscillatory branches, stabilization of frequency combs, and the resulting dynamical hysteresis all coincide with changes in frequency and phase locking between the constituent fields. These results provide a unified framework for understanding nonlinear dynamical phenomena in driven spinor polariton systems and suggest new opportunities for controlling oscillatory states and frequency-comb generation using exciton-polaritons [24, 25, 26].

While our model captures the essential nonlinear spinor dynamics responsible for the observed Hopf bifurcations and dynamical hysteresis, it neglects spatial effects that may become important in realistic devices. Future work could investigate how synchronization-driven transitions manifest in spatially extended polariton systems including lattices, where propagation effects, disorder, and topology provide additional degrees of freedom [27].

Acknowledgements

D.L. acknowledges support from the Ministry of Education, Singapore, under its SUTD Kickstarter Initiative (Grant No. SKI 20210501). V.V.K was supported by the Fundação para a Ciência e Tecnologia under the project UID/00618/2025 (DOI https://doi.org/10.54499/UID/00618/2025).

References

References

  1. [1]I. Carusotto and C. Ciuti, Quantum fluids of light, Rev. Mod. Phys. 85, 299 (2013).
  2. [2]S. Ghosh, R. Su, J. Zhao, A. Fieramosca, J. Wu, T. Li, Q. Zhang, F. Li, Z. Chen, T. Liew, D. Sanvitto, and Q. Xiong, Microcavity exciton polaritons at room temperature, Photonics Insights 1, R04 (2022).
  3. [3]H. Deng, G. Weihs, C. Santori, J. Bloch, and Y. Yamamoto, Condensation of semiconductor microcavity exciton polaritons, Science 298, 199 (2002).pubmed.ncbi.nlm.nih.gov/12364801
  4. [4]Z. Jalali-mola, T. Grass, V. Kasper, M. Lewenstein, and U. Bhattacharya, Topological bogoliubov quasiparticles from bose-einstein condensate in a flat band system, Phys. Rev. Lett. 131, 226601 (2023).
  5. [5]Z. Jalali-Mola, N. Käming, L. Asteria, U. Bhattacharya, R. W. Chhajlany, K. Sengstock, M. Lewenstein, T. Grass, and C. Weitenberg, Anomalous fluctuations of bose-einstein condensates in optical lattices, Phys. Rev. Lett. 136, 083401 (2026).
  6. [6]A. Kavokin, J. J. Baumberg, G. Malpuech, and F. P. Laussy, Microcavities, 2nd ed. (Oxford University Press, Oxford, 2017).
  7. [7]A. Delteil, T. Fink, A. Schade, S. Höfling, C. Schneider, and A. Imamoglu, Towards polariton blockade of confined exciton–polaritons, Nature Materials 18, 219 (2019).
  8. [8]A. Tredicucci, Y. Chen, V. Pellegrini, M. Börger, and F. Bassani, Optical bistability of semiconductor microcavities in the strong-coupling regime, Phys. Rev. A 54, 3493 (1996).pubmed.ncbi.nlm.nih.gov/9913876
  9. [9]F. Claude, M. J. Jacquet, R. Usciați, I. Carusotto, E. Giacobino, A. Bramati, and Q. Glorieux, High-resolution coherent probe spectroscopy of a polariton quantum fluid, Phys. Rev. Lett. 129, 103601 (2022).DOI
  10. [10]O. Kyriienko, H. Sigurdsson, and T. C. H. Liew, Probabilistic solving of NP-hard problems with bistable nonlinear optical networks, Phys. Rev. B 99, 195301 (2019).DOI
  11. [11]T. C. H. Liew, A. V. Kavokin, T. Ostatsnický, M. Kaliteevski, I. A. Shelykh, and R. A. Abram, Exciton-polariton integrated circuits, Phys. Rev. B 82, 033302 (2010).DOI
  12. [12]L. Pickup, K. Kalinin, A. Askitopoulos, Z. Hatzopoulos, P. G. Savvidis, N. G. Berloff, and P. G. Lagoudakis, Optical bistability under nonresonant excitation in spinor polariton condensates, Phys. Rev. Lett. 120, 225301 (2018).DOI
  13. [13]H. Yang and N. Y. Kim, Microcavity exciton-polariton quantum spin fluids, Advanced Quantum Technologies 5, 2100137 (2022).DOI
  14. [14]S. Dufferwiel, F. Li, E. Cancellieri, L. Giriunas, A. A. P. Trichet, D. M. Whittaker, P. M. Walker, F. Fras, E. Clarke, J. M. Smith, M. S. Skolnick, and D. N. Krizhanovskii, Spin textures of exciton-polaritons in a tunable microcavity with large TE-TM splitting, Phys. Rev. Lett. 115, 246401 (2015).DOI
  15. [15]H. Terras, H. Flayac, D. D. Solnyshkov, and G. Malpuech, Non-Abelian gauge fields in photonic cavities and photonic superfluids, Phys. Rev. Lett. 112, 066402 (2014).
  16. [16]A. Werner, O. A. Egorov, and F. Lederer, Exciton-polariton patterns in coherently pumped semiconductor microcavities, Phys. Rev. B 89, 245307 (2014).DOI
  17. [17]D. Sarchi, I. Carusotto, M. Wouters, and V. Savona, Coherent dynamics and parametric instabilities of microcavity polaritons in double-well systems, Phys. Rev. B 77, 125324 (2008).DOI
  18. [18]M. Furman, A. Opala, M. Król, K. Tyszka, R. Mirek, M. Muszyński, B. Seredyński, W. Pacuski, J. Szczytko, M. Matuszewski, and B. Pietka, Inverted optical bistability and optical limiting in coherently driven exciton–polaritons, APL Photonics 8, 046105 (2023).DOI
  19. [19]S. Luo, H. Zhou, L. Zhang, and Z. Chen, Nanophotonics of microcavity exciton–polaritons, Appl. Phys. Rev. 10, 011316 (2023).DOI
  20. [20]L. Kłopotowski, M. Martín, A. Amo, L. Viña, I. Shelykh, M. Glazov, G. Malpuech, A. Kavokin, and R. André, Optical anisotropy and pinning of the linear polarization of light in semiconductor microcavities, Solid State Communications 139, 511 (2006).DOI
  21. [21]J. D. Crawford, Introduction to bifurcation theory, Rev. Mod. Phys. 63, 991 (1991).
  22. [22]A. Baas, J. P. Karr, H. Eleuch, and E. Giacobino, Optical bistability in semiconductor microcavities, Phys. Rev. A 69, 023809 (2004).DOI
  23. [23]T should be set long enough that the both T/2 and T represents the long-time dynamics of the cavity and exciton population.
  24. [24]K. Rayanov, B. L. Altshuler, Y. G. Rubo, and S. Flach, Frequency combs with weakly lasing exciton-polariton condensates, Phys. Rev. Lett. 114, 193901 (2015).DOI
  25. [25]O. A. Egorov and D. V. Skryabin, Frequency comb generation in a resonantly pumped exciton-polariton microring resonator, Opt. Express 26, 24003 (2018).DOI
  26. [26]S. Kim, Y. G. Rubo, T. C. H. Liew, S. Brodbeck, C. Schneider, S. Höfling, and H. Deng, Emergence of microfrequency comb via limit cycles in dissipatively coupled condensates, Phys. Rev. B 101, 085302 (2020).DOI
  27. [27]C. Schneider, K. Winkler, M. D. Fraser, M. Kamp, Y. Yamamoto, E. A. Ostrovskaya, and S. Höfling, Exciton-polariton trapping and potential landscape engineering, Reports on Progress in Physics 80, 016503 (2016).DOI

Paper details

vlnc:W_2endvt8p
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 8

downloads · 1

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents