Topological exciton-polaritons with negative coupling Authors: Zixuan Yu, Feng Jin, Jiahao Ren, Subhaskar Mandal, Baile Zhang, Rui Su Published: 2025 ABSTRACT AbstractCoupling between quantum or classical degrees of freedom underpins a wide range of physical phenomena, from condensed matter systems to engineered photonic lattices. While positive coupling arising naturally from evanescent interactions has been extensively studied and employed, negative coupling unlocks unique phenomena that are challenging to realize with positive coupling alone. Exciton-polariton micropillars, celebrated for enabling topological lasers, reservoir computing, and quantum simulations, have primarily relied on positive coupling. In this work, we experimentally demonstrate negative coupling between two micropillars using an additional larger micropillar. By combining positively and negatively coupled micropillars, we construct a Su-Schrieffer-Heeger (SSH) topological lattice with a significant topological gap of approximately 15 meV, where band inversion occurs at the center of the Brillouin zone (BZ), rather than at the edges as in conventional SSH lattices. Under non-resonant excitation, we achieve polariton condensation in the in-gap topological edge states at room temperature. Our study introduces a universal method to realize negative coupling in polariton systems, paving the way for novel polaritonic devices based on lattices with arbitrarily controlled coupling signs.Subject terms: Polaritons, Topological insulators, Nanophotonics and plasmonics Introduction Exciton-polaritons, formed through the strong coupling between excitons and cavity photons, provide a unique platform for studying hybrid light-matter quasiparticles1,2. These bosonic quasiparticles inherit characteristics from both their photonic and excitonic components, leading to enhanced nonlinear interactions, reduced effective mass, ultrafast dynamics, and macroscopic coherence3,4. Such properties make them ideal for realizing polariton lasers5–9 and various all-optical information processing elements, including optical logic gates and transistors10–17. Polariton-based information processing and simulation, including neuromorphic computing and polariton simulators, often rely on coupled lattices of spatially localized sites18–21. In this context, topological concepts have emerged as a powerful tool for expanding the functionality of polaritonic devices. This has led to the realization of polariton Chern insulators22, flat band lasing23–25, topological lasers26–30, higher-order topological lasers31–33, Valley Hall topological phases34,35, and novel topological phases in optical lattices36,37, with many more intriguing phases theoretically proposed38–46. Despite these advancements, polariton lattices have predominantly been limited to positive coupling due to the natural evanescent overlap of neighboring polariton wavefunctions. To further advance the field, it is crucial to explore alternative coupling mechanisms, particularly those achievable through intelligent lattice design. Negative coupling has emerged as an important mechanism for linking two lattice sites, wherein the in-phase bonding and out-of-phase antibonding states are interchanged in energy47. This coupling plays a key role in realizing novel physical phenomena. For instance, it facilitates the emergence of quadrupole topological phases48,49, which have been proposed as a basis for topological polariton memory50. It also enables quantum simulations, such as modeling frustrated antiferromagnets in Kerr nonlinear photonic cavities51. Moreover, negative coupling is fundamental in biological systems, particularly in neural networks with inhibitory neurons52–54, making it a promising avenue for advancing polariton-based neuromorphic computing. Thus, introducing negative coupling in polariton lattices is important, however, it has not been realized yet. In this work, we experimentally demonstrate negative coupling in the perovskite exciton-polariton lattices at room temperature. By introducing a mediator micropillar between two smaller micropillars and precisely tuning its dimensions, we achieve effective negative coupling in the polaritonic platform. The validity of our approach is supported by the consistency of the theoretical predictions and experimental observations. Furthermore, we experimentally construct a staggered lattice composed of alternating positive- and negative-coupling micropillars and observe the negative-coupling-supported topological edge states through angle-resolved photoluminescence measurements at room temperature. Notably, a topological gap of approximately 15 meV emerges, accompanied by a robust edge state localized at the Γ point. Under non-resonant pulsed excitation, we further demonstrate the condensation of exciton-polaritons into these Γ point topological edge states, highlighting the potential of negative coupling for advanced active polaritonic devices. Results Observation of the negative coupling in the polariton system To achieve negative coupling in the polariton system, we design and fabricate a negatively coupled unit cell (NUC) by etching the spacer layer of CsPbBr3 perovskite planar microcavities, as illustrated in Fig. 1a (more fabrication details refer to the methods). Figure 1b presents an atomic force microscopy (AFM) image of the fabricated NUC lattice on the perovskite layer prior to the deposition of the top distributed Bragg reflector (DBR). The AFM image clearly reveals the profile of the NUC lattice with a homogeneous thickness of 60 nm, which consists of two small micropillars (R0=350nm) flanking a larger intermediate micropillar (RA=450nm) that serves as a mediator negative coupler. The mediator micropillar is centrally positioned between the two small micropillars with a longitudinal center-to-center spacing of 0.8R0, while the horizontal spacing between the small micropillars is 4R0, allowing them to overlap with the central mediator pillar. To elucidate the underlying negative coupling mechanism, we present the eigenstates supported by the NUC lattice and compare them with those obtained from conventional positive coupling, as shown in Fig. 1c. We denote the intrinsic energy of an isolated micropillar mode as E0. Upon coupling, this mode energy splits into two distinct eigenstates, labeled E1 and E2 (with E1EA. For our analysis, we set EM=0 and EA=−1. To explore the system’s topological properties, we calculate the winding number: where ∣ψn> is the nth eigenstate of the Hamiltonian and the integral is evaluated over the BZ. The numerically calculated winding number is shown in Fig. 2b. For the first band, the winding number remains at 0, irrespective of the t2/t1 ratio. However, for the second and third band, the winding number transitions from 0 to 1 and − 1 at around t2/t1=0.62, respectively, which indicates a trivial-to-topological phase transition. Notably, unlike the typical SSH chain, where the topological phase transition occurs in the lowest band, here the lowest band remains trivial throughout, with the topological transition occurring in the second band. Theoretical explanations and numerical results of negative coupling. a One-dimensional chain with negative coupling via mediator pillars (yellow) between main pillars (red). Unit cell in blue. b Winding number for the first two bands, showing a phase transition for the second band at t2/t1=0.62. c Band structure: trivial case (t2/t1=0.4, dashed) and phase transition (t2/t1=0.62, solid) occurring at the center of the BZ (kx=0) rather than its edge. The color bar means the spatial distribution of the state. States localized on the mediator pillars are represented by light colors, whereas states mainly residing on the main lattice are shown in dark colors. d Topological band structure at t2/t1=1. e Eigenenergies of a finite chain with edge modes between the second and third bands. f Spatial profiles of edge modes. g Schematic of the photon potential used for theoretical calculation, embodying coupled exciton-polariton micropillar chain. Unit cell in blue. h, i Simulated polaritonic band structure for periodic and finite systems. j Eigenenergy spectrum with edge modes shown in red between the second and third bands. k Spatial profiles of edge modes. The band structure analysis further highlights the distinction between trivial and topological phases. As shown in Fig. 2c, the trivial case (t2/t1=0.4) exhibits a gap, while the topological phase transition (t2/t1=0.62) leads to a band closure at the center of the BZ (Γ point), contrasting with the band closure at the BZ edges typically observed in conventional SSH systems. This unconventional behavior originates from the introduction of negative coupling into the SSH model, which is equivalent to a unitary transformation in reciprocal space eiakx→eia(kx+K), where K=π/a. The further discussion is shown in Supplementary Note 4. For t2/t1=1, the system enters a topological regime, as evidenced by the band structure in Fig. 2d and the finite eigenenergy spectrum which reveals edge states localized between the second and third bands, as shown in Fig. 2e, with their spatial profiles depicted in Fig. 2f. A similar structure is implemented using a coupled exciton-polariton micropillar lattice, where the unit cells are placed with the lattice constant of 5R0 (Fig. 2g). Using the continuous model simulation, we compute the band structures for both periodic and finite systems, presented in Fig. 2h, i, respectively. The eigenenergy spectrum of the finite micropillar chain confirms the presence of edge modes between the second and third bands (Fig. 2j), with the spatial localization of these edge states illustrated in Fig. 2k. These results demonstrate the feasibility of implementing and observing topological edge states in the exciton-polariton system. Experimentally, we fabricate the alternating positively- and negatively- coupled 1D chain by patterning the ZEP520A spacer layer of the CsPbBr3 perovskite microcavity, with the lattice x-axis precisely aligned along the perovskite crystal axes (Fig. 3a). Figure 3b presents the AFM image of this lattice on the perovskite layer, showing a uniform height of 60 nm prior to top DBR deposition and clearly featuring an intentional alternation of positive and negative couplings. In order to demonstrate the emergence of the negative-coupling-supported topological edge state, we initially probe the band structure of the alternating 1D chain through mapping the angle-resolved photoluminescence spectrum (Methods). The entire 1D chain sample with a detuning of ~ − 170 meV is non-resonantly excited using a continuous-wave laser at 2.713 eV at room temperature. By collecting the emission from the bulk and the edge areas, framed by the white and red dashed lines in Fig. 3b, respectively, we observe distinct spectral features. Figure 3c illustrates the folded dispersion in the first BZ from the bulk area along K→Γ→K path. Inside the dispersion, we observe a pronounced bandgap of ~ 15 meV at around 2.24 eV (framed by the black dashed lines). Notably, the negative-coupling supported topological gap in our system (~ 15 meV) is larger than those in earlier SSH-type polariton systems, and the quantitative comparison is summarized in Supplementary Note 11. Subsequently, we move the collection area to the lattice edge and characterize its band structure behavior along the same K→Γ→K path. As shown in Fig. 3d, the edge area dispersion closely resembles that of the bulk area, but notably, an additional discrete state emerges within the topological gap, corresponding to the negative-coupling-supported topological edge state. Moreover, we observe that this edge state is localized at the Γ point of the BZ, aligning well with the theoretical calculation (Fig. 2i). This is in sharp contrast to the conventional SSH model, where edge states typically localize at the BZ boundaries (K points). The non-trivial topology nature can be further experimentally confirmed by the real-space photoluminescence characterizations. Specifically, we collect the energy-resolved spatial image along the positively coupled micropillars sites. As shown in Fig. 3e, the topological edge state is clearly observed inside the topological gap at 2.240 eV, highlighted by black dashed lines. In addition, Fig. 3f illustrates the real-space emission profile of the topological state, which is sharply confined to the lattice edge with an absence of bulk emission. Our results collectively validate the emergence of a negative-coupling-supported topological edge state localized at the Γ point of the BZ in the alternating positively- and negatively-coupled 1D chain lattice. Characterization of topological edge states supported by negative coupling. a Schematic diagram of the staggered coupling chain. b AFM image of the chain on the perovskite layer before the deposition of the top DBR. c, d Energy-resolved momentum-space polariton dispersions of the bulk and edge of the chain lattice folded in the first BZ. The black dashed lines in (c) highlights an opening of a topological gap of 15 meV. The topological edge states locate in the energy gap at the Γ point, in good agreement of the calculated energy band. e Energy-resolved spatial images collected along the x-axis of the lattice. The energy gap is also presented by a dashed cube, and the topological state localized at the edge is exhibited. f Real-space image at the edge state, with localized states at both ends shown. Exciton-polariton condensation into negative-coupling-supported topological edge state A notable feature of CsPbBr3 perovskite exciton-polaritons is their ability to realize non-equilibrium condensation at room temperature, accompanied by coherent, laser-like emission. In particular, thanks to their intrinsic driven-dissipative nature and finite lifetime, polaritons can condense into various non-ground states by tuning the microcavity detuning. To facilitate efficient polariton condensation into the negative-coupling-supported topological edge state, we employ a staggered positive-negative coupling chain sample with a detuning of ~ − 95 meV and non-resonantly excite it with a femtosecond pulsed laser at 3.1 eV (Methods). As shown in Fig. 4a–c, we present the evolution of the topological edge state lasing behavior in momentum space along K→Γ→K path under varying excitation powers. Below the condensation threshold (P=0.6Pth, Pth=14μJ/cm2), the measured dispersion exhibits similar features as the linear-regime dispersion (Fig. 4a). As the excitation power are increased to the threshold (P=Pth), we observe a higher occupation of the topological edge state within the dispersion, marking the onset of the polariton condensation (Fig. 4b). When the excitation power is further increased to 1.6Pth (beyond the condensation threshold), the topological edge state becomes dominant in the spectrum, accompanied by a narrower linewidth and a significant intensity enhancement exceeding three orders of magnitude (Fig. 4c). The polariton topological edge state condensation can be further confirmed by the real-space characterization (Fig. 4d), where we observe the localized emission at the end sites of our negative coupling 1D lattice chain with gradual decay into neighboring pillars. To directly characterize the evolution of the topological polariton condensation process, we demonstrate the integrated emission intensity spectrum under varying excitation power (0.6Pth, 0.8Pth, 1.0Pth, 1.5Pth and 2.0Pth), as shown in Fig. 4e. The intensities of the spectra under 0.6Pth, 0.8Pth are multiplied by a factor of 10, and the spectra intensity under 1.0Pth is multiplied by a factor of 5 to make the comparison observable. The green arrows indicate the negative-coupling-supported topological edge state, which highlights the significant increase in intensity and the obvious blueshift. In addition, to characterize this transition quantitatively, we further examine the integrated emission energy, linewidth, and peak energy position of the topological edge state as a function of pumping fluence. As shown in Fig. 4f, we observe that the integrated emission intensity exhibits a pronounced nonlinear increase as the excitation power surpasses the threshold. Concurrently, the linewidth of the topological edge state displays a rapid decrease from 5.68 meV to 1.91 meV across the threshold. In addition, the peak energy of the topological state exhibits a continuous blueshift trend with increasing excitation, resulting from the strong nonlinear repulsive interactions inherited from the excitonic component of exciton-polaritons (Fig. 4g). Collectively, all of these experimental demonstrations conclusively confirm the emergence of the polariton condensation into the negative-coupling-supported topological edge state. Exciton-polariton condensation of the edge state. a–c Energy-resolved momentum-space polariton dispersion with pumping power of P=0.6Pth, P=1.0Pth and P=1.6Pth. When pumping energy increases, we observe the edge state becoming dominant. d Real-space emission of the topological edge state at P=1.6Pth. e Waterfall figure of polariton emission spectrum under different pumping power P=0.6Pth, P=0.8Pth, P=1.0Pth, P=1.5Pth and P=2.0Pth. f, g Evolution of integrated output intensity, linewidth and peak position of the edge state as a function of the pump fluence. The abruptly increased intensity and decreased linewidth mark the exciton-polariton condensation. The trend of the peak position shows a continuous blueshift until saturation. Discussion Our study demonstrates the engineering of negative coupling in a perovskite polariton lattice, leading to the realization of large-gap topological edge states and their condensation at room temperature. By incorporating a mediator site, we successfully achieve universal sign control, enabling negative coupling between the smaller micropillars within the unit cell. We observe the inversion of bonding and anti-bonding states, as well as the altered edge state position in the lattice energy band, providing strong evidence of the effectiveness of our negative coupling design. In addition, a staggered positive-negative coupling chain was shown to support robust topological edge states with a large bandgap at the Γ point. Our experiments confirmed that exciton-polariton condensation and the resulting lasing can occur in the topological edge state at room temperature, with low energy consumption as low as P=14μJ/cm2. One can further realize CW-pumped topological polariton condensation in this platform by integrating structural engineering approaches, such as introducing nano-ridge potential landscapes in the perovskite to enhance polariton relaxation55 or employing optimized cavity designs to improve thermal dissipation56. The Γ point topological states have also been realized in grating-type SSH metasurface37,57, but the mechanism of negative coupling can further provide a broadly applicable framework for engineering complex polaritonic lattices. Extending this architecture to higher-dimensional configurations such as Kagome, Lieb, or honeycomb lattices will enable the realization of more exotic topological phases that are inaccessible with conventional positive coupling. Moreover, the negative coupling can also introduce a universal and highly tunable degree of freedom for controlling interaction signs in driven-dissipative quantum systems. Such a coupling design not only enriches the band structure engineering of exciton-polaritons but also lays the foundation for advanced quantum simulation and neuromorphic computing, where inhibitory or sign-reversed links are essential. Methods Continuous model simulation of exciton-polariton All details regarding the tight-binding calculations are provided in the main text. To study polaritons using the continuous model, we employ the following Schrödinger Hamiltonian in the linear regime, where the photonic and excitonic components are coupled: Here, the diagonal elements represent the photonic and excitonic parts, while the off-diagonal term, g0, corresponds to the Rabi splitting. The parameter mph denotes the bare cavity photon mass, and Vph(x,y) represents the potential profile for photons. The exciton energy, ϵex, is assumed to be dispersionless. By numerically diagonalizing the Hamiltonian and performing a Fourier transform of the wave functions, we obtain the band structure and mode profiles. We set mph=1.1×10−5me, where me is the free electron mass. The potential depth for photons is taken as 150 meV, the Rabi splitting g0 is set to 120 meV, and the exciton energy is chosen as 2.4077 eV. Perovskite unit cell and lattice fabrication Our microcavity is formed by two DBR layers sandwiching an all-inorganic perovskite (CsPbBr3) layer with a spacer layer on top. The DBR layers are composed of alternating pairs of TiO2 (62.7 nm) and SiO2 (92.1 nm), with 15.5 pairs on the bottom and 8.5 pairs on the top, fabricated using an electron beam evaporator (Cello 50D). The single-crystal CsPbBr3 perovskite was grown on mica using a chemical vapor deposition method (CVD chamber pressure is maintained at 42.5 Torr with a nitrogen flow of 25 sccm, and the CVD tube is heated to 590 °C within 5 min, held for 10 min, and then naturally cool down to room temperature58,59), and then transferred onto the bottom DBR using Scotch tape. The perovskite layer thickness was controlled to be 65 nm. A 60 nm-thick ZEP520A layer was then spin-coated on top, serving as the spacer and providing resistance to electron beam exposure. The lattice design was patterned on the spacer layer using electron beam lithography. Finally, the lattice fabrication was completed by evaporating the top DBR layer. Optical spectroscopy characterizations We obtained the energy-resolved momentum-space spectra and real-space photoluminescence profiles using a home-built angle-resolved spectroscopy setup with Fourier optics60,61. The desired optical characteristics were detected through the photon component of the cavity polariton, which spontaneously leaks out of the perovskite cavity. The optical emission was collected using a 50 × objective lens with a numerical aperture of NA=0.75. The emission signal was then directed to a 550 mm focal length spectrometer (Horiba iHR550) equipped with a grating (600 lines/mm) and a liquid nitrogen-cooled CCD (256 × 1024 pixels). A 457 nm continuous wave laser was used to extract data from the sample in the linear region, with a pumping spot approximately 25 μm in diameter. In the nonlinear region, we switched to a 400 nm pulsed laser, setting the homogeneous pumping spot to about 15 μm in diameter. Supplementary information Supplementary Information Source data Source Data Acknowledgements R.S. and B.L.Z. gratefully acknowledge funding support from the Singapore National Research Foundation via a Competitive Research Program (grant no. NRF-CRP23-2019-0007). R.S. gratefully acknowledges funding support from the Singapore Ministry of Education via the AcRF Tier 2 grant (MOE-T2EP50222-0008), AcRF Tier 3 grant (MOE-MOET32023-0003) “Quantum Geometric Advantage” and Tier 1 grant (RG90/25). R.S. also gratefully acknowledges funding support from the Nanyang Technological University via a Nanyang Assistant Professorship start-up grant. S.M. acknowledges funding support from the Industrial Research and Consultancy Center (IRCC), Indian Institute of Technology Bombay, Seed Grant No. RD/0525-IRCCSH0-006. Author contributions R.S., B.L.Z., and S.M. supervised the project. Z.X.Y. and F.J. fabricated the devices with the help of J.H.R., F.J., and Z.X.Y. conducted all the optical characterization experiments. S.M. conceived the lattice model and performed the theoretical calculations with input from B.L.Z. Z.X.Y., F.J., S.M., R.S., and B.Z. collectively wrote the manuscript with input from all authors. Peer review Peer review information Nature Communications thanks Helgi Sigurðsson, and the other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Data availability The source data generated in this study are provided in the Source Data file with this paper. Source data are provided in this paper. Code availability The codes are available from the corresponding author upon request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Zixuan Yu, Feng Jin. Contributor Information Subhaskar Mandal, Email: subhaskar@iitb.ac.in. Baile Zhang, Email: blzhang@ntu.edu.sg. Rui Su, Email: surui@ntu.edu.sg. Supplementary information The online version contains supplementary material available at 10.1038/s41467-025-68025-4. References Associated Data Supplementary Materials Supplementary Information Source Data Data Availability Statement The source data generated in this study are provided in the Source Data file with this paper. Source data are provided in this paper. The codes are available from the corresponding author upon request.