Surface plasmons (SPs) have garnered significant research interests for their ability to confine light at sub-wavelength scales, thereby circumventing the diffraction limit and enabling substantial field enhancement [1–3]. Leveraging these unique properties, extensive studies have been conducted in various plasmonic systems, including plasmon–exciton [4–6], plasmon–photon [7–9], plasmon–phonon [10,11], and pure plasmon coupling [12–17]. These studies have advanced applications in laser technology [18], sensing [16], chirality [14,15,19], optical filtering [13], perfect absorption [20], surface-enhanced Raman spectroscopy [21], nonlinear frequency-conversion enhancement [22–24], etc. High quality-factor ($Q$-factor) plasmonic resonances are essential for enhancing light–matter interactions, lowering the threshold for lasing and nonlinear processes and increasing the sensitivity of sensors [25]. As a key focus in nanophotonics, high-$Q$ plasmonic resonances hold promise for the development of photonic devices, sensor technologies, and other advancements.

Localized surface plasmon resonances (LSPRs) and propagating surface plasmon polaritons (SPPs) are two fundamental modes in plasmonic systems [26–28]. The $Q$-factors of LSPRs are usually less than 10 [29]. SPPs, excited at the metal–dielectric interface, feature extended evanescent fields, but their $Q$-factors are usually under 100 [28]. The intrinsic metallic losses pose significant challenges to their practical applications.

To overcome these limitations, hybrid plasmonic modes with enhanced optical properties have been developed, including surface lattice resonances (SLRs) and Fano resonances [30,31]. SLRs, formed by the coupling of the LSPRs generated by metal nanoparticles and the Rayleigh anomalies (RAs), exhibit enhanced local fields and reduced radiation losses [32–35]. Compared to LSPRs, SLRs have higher $Q$-factors and are widely used in sensing, laser, and nonlinear optics [18,36–38]. Metal nanoparticle arrays with optimized geometry, reduced refractive index mismatches, and coherent light sources further enhance SLR excitation [39,40]. To date, $Q$-factors as high as 2300 in the communication band (under normal incidence) [39] and 790 in the visible (VIS) band (under oblique incidence) [41] have been reported for plasmonic SLRs.

SLRs can be considered as bright Fano resonances, which inherently limits their $Q$-factors. When a Fano resonance collapses, it transitions into bound states in the continuum (BICs) [42,43]. BICs are confined states that eliminate radiation loss [44,45]. Topologically, BICs are polarized vortices in momentum space with quantized topological charges [46–49]. Generally, BICs fall into two categories: symmetry-protected BICs (SP-BICs) and Friedrich-Wintgen BICs (FW-BICs) [50]. While ideal BICs are primarily theoretical, they transform into quasi-BICs (q-BICs) with finite lifetimes with broken in-plane ${\mathrm{C}}_2$ symmetry [13,16]. Though BICs have been extensively observed in dielectric systems [51–58], achieving BICs in pure plasmonic systems remains challenging, due to metallic losses and the overshadowing of BICs by other bright modes with large linewidth [59]. The reported $Q$-factors in pure plasmonic cavities have not exceeded 150 [12,13,15,16,60,61]. Hybrid plasmonic–photonic structures, consisting of metal metasurfaces and finite dielectric layers, have been proposed [7,9,62–64]. Both SLRs and BICs are advantageous for engineering radiation losses, thereby greatly expanding the available platforms for achieving high-$Q$ resonances. However, research on BICs in the communication band within plasmonic structures is still relatively scarce. Although the coexistence of SLRs and BICs has been numerically predicted before [65], direct experimental comparisons between them in identical structures and energy bands have yet to be studied. Notably, reported $Q$-factors of SLRs often surpass those of BICs, which is a counterintuitive result. The main reason is that they are typically measured in different samples under varying conditions.

In this work, we design and fabricate a series of reusable plasmonic–photonic hybrid metasurfaces that support both high-$Q$ plasmonic SLRs and BICs. Notably, the hybrid modes supported by the metasurface can improve the resonance modulation and signal-to-noise ratio (SNR) of q-BICs near the $\mathrm{\Gamma }$ point, allowing higher measured $Q$-factors. The footprint of the sample is $200\mathrm{\mu m}\times 200\mathrm{\mu m}$, significantly smaller than those used in previous studies for achieving high $Q$-factors. This size is among the best pixel sizes for display applications. First, we study the effects of lattice constant and dielectric cladding thickness on the $Q$-factors of both modes. With optimized cladding thickness, the measured $Q$-factors of BICs improve by an order of magnitude for all samples across VIS and near infrared (NIR) wavelengths. This improvement is quantitatively explained using Hopfield coefficients and the coupled oscillator model. Next, we demonstrate the influence of the light collecting fiber core diameter on $Q$-factors in the Fourier plane, showing that decreasing the core diameter can improve the resolution of the Fourier plane, which is essential for enhancing the measured $Q$-factors of BICs but has little effect on SLRs. The measured $Q$-factor of BICs at 1585 nm can reach 260. Finally, we compare the measured $Q$-factors of SLRs and BICs under identical structures and energy bands. The results show that the $Q$-factor of BICs is approximately four times that of SLR, indicating that BICs have a clear advantage over SLRs in achieving high-$Q$ resonances, while SLRs are more robust to perturbations. That is to say, any structure that supports high-$Q$ SLRs can also be optimized to support high-$Q$ BICs. This work presents the first, to the best of our knowledge, experimental comparative study of $Q$-factors for SLRs and BICs using a reusable metasurface, greatly expanding the available platforms for high-$Q$ resonances, which may find applications in sensing, lasing, nonlinear enhancement, photocatalysis, and other fields.

The reusable hybrid plasmonic–photonic metasurface proposed here consists of a rectangular array of gold nanorods embedded in a dielectric cladding, as illustrated in Fig. 1(a). While the geometry of the nanorods and the lattice constant $P_x$ are fixed, the cladding thickness $t$ and the lattice constant $P_y$ are adjustable. Polyvinyl alcohol (PVA, $n\sim 1\text{.}48$) is selected as the dielectric cladding material to create a homogeneous refractive index environment with the silica substrate ($n\sim 1\text{.}45$), facilitating the efficient excitation of SLRs. It is worth noting that the structure proposed here is similar to that in a previous work [66]; however, the previous study only employs normal incidence excitation, limiting the exploration to SLRs without considering BICs. In contrast, this work investigates both SLRs and BICs supported by the reusable hybrid metasurface, leveraging the water solubility and easy reconstruction of the PVA cladding film.

Figure 1(b) presents an intuitive description of these two modes. When the wavelength of the incident electromagnetic wave meets a specific relationship with the lattice constant $P$ ($\lambda \sim nP$, where $n$ is the effective refractive index), surface waves propagating along opposite directions can be generated under normal incidence and form two standing wave modes. One mode corresponds to the SLR, where the electric field antinodes coincide with the nanorods. The other mode, where the electric field nodes align with the nanorods, corresponds to the BIC, which is characterized by radiation suppression.

To rigorously verify the generation of coupling modes supported by the reusable hybrid metasurface, finite-difference time-domain (FDTD) simulation and eigenmode analysis are first conducted on a metasurface with $P_y=1100\mathrm{nm}$ and $t=940\mathrm{nm}$, as shown in Fig. 1(c). The dashed white box marks the SLR and BIC generated by mode coupling, and the details are zoomed in insets (left panel). The high-energy branch exhibits a breakpoint at the $\mathrm{\Gamma }$ point, corresponding to the BIC with a topological charge of $+1$ (right panel), while the lower branch represents a bright mode, i.e., the SLR (in-plane SLR under normal incidence, out-of-plane SLR under oblique incidence).

In this scenario, the system operating near the dashed white box can be approximated by a $3\times 3$ Hamiltonian [13]: $$H=\left[\begin{array}{ccc}{E}_{\mathrm{LSPR}}-i{\gamma }_1& g_{\mathrm{str}}& g_{\mathrm{str}}\\ g_{\mathrm{str}}& E_{{\mathrm{RA}}_1}-i{\gamma }_2& g_{\mathrm{bg}}-i\sqrt{{\gamma }_2{\gamma }_3}\\ g_{\mathrm{str}}& g_{\mathrm{bg}}-i\sqrt{{\gamma }_2{\gamma }_3}& E_{{\mathrm{RA}}_{-1}}-i{\gamma }_3\end{array}\right],$$(1)where $g_{\mathrm{str}}$ represents the coupling strength between the LSPR and the RAs and $g_{\mathrm{bg}}$ is the coupling strength between the RAs, forming a bandgap. $E_{\mathrm{LSPR}}$ denotes the resonant frequency of the LSPR, while $E_{{\mathrm{RA}}_1}$ and $E_{{\mathrm{RA}}_{-1}}$ denote the resonant frequencies of the RAs (the subscripts 1 and $-1$ indicate the diffraction orders). ${\gamma }_1$, ${\gamma }_2$, and ${\gamma }_3$ are the radiation loss, and $\sqrt{{\gamma }_2{\gamma }_3}$ represents the radiation coupling between the RAs. At the $\mathrm{\Gamma }$ point (let $E_{\mathrm{LSPR}}=E_1$, $E_{{\mathrm{RA}}_1}=E_{{\mathrm{RA}}_{-1}}=E_2$, and ${\gamma }_2={\gamma }_3$), the corresponding eigenvalues are $$E_a=E_2-g_{\mathrm{bg}},$$(2)$$E_b=\frac{E_1+E_2+g_{\mathrm{bg}}-i(2{\gamma }_2+{\gamma }_1)-\sqrt{{[E_2-E_1+g_{\mathrm{bg}}-i(2{\gamma }_2-{\gamma }_1)]}^2+8g_{\mathrm{str}}^2}}{2},$$(3)$$E_c=\frac{E_1+E_2+g_{\mathrm{bg}}-i(2{\gamma }_2+{\gamma }_1)+\sqrt{{[E_2-E_1+g_{\mathrm{bg}}-i(2{\gamma }_2-{\gamma }_1)]}^2+8g_{\mathrm{str}}^2}}{2}.$$(4)Notably, the eigenvalue $E_a$ has no imaginary part, indicating no radiation loss, which is consistent with the property of BICs. The hybrid modes $E_b$ and $E_c$ are associated with SLRs in this metasurface, which proves that BICs and SLRs have the same origin. By the way, recall that the resonance $E_1$ depends on the geometry of the nanorods and the effective refractive index of the surrounding environment, while $E_2$ is related to the lattice constant of the structure and also the effective refractive index of the surrounding environment. These factors can be used to tune the resonances of the two high-$Q$ modes, providing a foundation for sensor applications based on the two modes.

The transmission spectra from Fig. 1(c) at incident angles $\theta =0°$ and 2° are plotted in Fig. 1(d). As the incident angle increases, the SLR (highlighted in red) undergoes a red shift and an increase in the $Q$-factor, while the BIC transitions into a q-BIC (highlighted in blue) with a finite $Q$-factor. The insets display the simulated magnitude of electric field $|\mathit{E}|$ for the unit cell of both SLR ($\theta =0°$) and q-BIC ($\theta =2°$). Oblique incidence will break the symmetry of electric field distribution, resulting in slight differences from the ideal symmetry as depicted in Fig. 1(b). To further understand the nature of SLRs and BICs, a multipole expansion of the scattering cross section of the resonances in Fig. 1(d) is performed [67]. Figures 2(a) and 2(d) illustrate the changes in the results of the multipole expansion as the incident angle $\theta$ increases from 0° to 2°. For the SLR, the electric dipole remains consistently dominant. In contrast, for the BIC, the dominant composition transitions from the electric dipole to a combination of both the electric and magnetic dipoles. While the scattering cross section of the electric dipole shows minimal change, that of the magnetic dipole increases significantly. This indicates that the excitation of the BIC at non-zero angles is closely related to the magnetic dipole. Further analysis of the electric and magnetic dipole components reveals that the SLR is predominant of an electric dipole mode along the $x$ direction, as shown in Figs. 2(b) and 2(e), while the BIC corresponds to a magnetic dipole mode along the $z$ direction, as depicted in Figs. 2(c) and 2(f).

To experimentally confirm the hybrid metasurface’s ability to generate high-$Q$ resonances, a series of plasmonic metasurfaces are fabricated using electron-beam lithography (EBL) and spin-coating. Each array has the same size ($200\mathrm{\mu m}\times 200\mathrm{\mu m}$) but different lattice constants, ranging from $P_y=500\mathrm{nm}$ to 1100 nm (samples A–I). The cladding layer can be repeatedly cleaned and reapplied by spin-coating, allowing changes in layer thickness without damaging the metasurfaces. This enables the creation of a series of metasurfaces with varying cladding thicknesses from the same sample at a relatively low cost. This capability allows us to systematically study the effect of cladding thickness on the $Q$-factors of BICs and SLRs under consistent conditions, minimizing ambiguity in the analysis. Figure 1(e) shows an image of a typical fabricated metasurface (sample E) with $P_y=800\mathrm{nm}$ before cladding is applied.

Subsequently, the fabricated samples are tested using a home-built experimental setup with an incoherent light source. The $k$-space information is obtained by point-by-point scanning in the Fourier plane using a fiber [12]. Figures 3(a)–3(c) show the measured angle-resolved transmission spectra of the three representative samples (A, E, I) with a cladding thickness $t=340\mathrm{nm}$. As the lattice constant $P_y$ increases, the excited BICs (indicated by dashed white ellipses) and SLRs (indicated by solid and dashed black arrows) redshift. The $Q$-factors of the BICs in the three metasurfaces range from 15 to 35, comparable to or even lower than those of the SLRs at the $\mathrm{\Gamma }$ point in the same metasurfaces (10 to 55).

When the cladding thickness increases to 940 nm, as shown in Figs. 3(d)–3(f), richer mode coupling occurs, resulting in multiple new resonances with noticeable reduced loss in the transmission spectra. Notably, the resonances near the BICs become sharper with a pronounced increase in the $Q$-factor. Figures 3(g)–3(i) show the $Q$-factors of the band where the BICs are located (named the BIC band) extracted from Figs. 3(a)–3(f). As the cladding layer thickness increases from 340 to 940 nm, the $Q$-factors improve significantly by a factor of 6 to 11. The maximum $Q$-factors observed in these three samples are 129 at 1.68 eV ($\sim 737\mathrm{nm}$), 198 at 1.07 eV ($\sim 1163\mathrm{nm}$), and 218 at 0.78 eV ($\sim 1585\mathrm{nm}$), respectively, covering the spectrum range from the VIS to NIR. The $Q$-factors for all samples are summarized in Appendix A. For SLRs, the resonances of the lower branch and their $Q$-factors (10–66) at the $\mathrm{\Gamma }$ point do not change substantially with cladding thickness. However, the upper branches show significant improvement at the $\mathrm{\Gamma }$ point (77–123), although these values remain generally lower than those of the BICs. By closely examining the results in Fig. 3, we can qualitatively explain why the $Q$-factors of the BICs are greatly enhanced with a thicker cladding layer. To achieve a high $Q$-factor experimentally, the q-BIC should be as close as possible to the ideal BIC. For a thin cladding layer, the q-BIC near the ideal BIC is weak and lossy. In contrast, a thicker cladding layer involves more waveguide modes (WGMs), making the modulation resonance of the q-BIC stronger and closer to the ideal BIC, leading to a higher $Q$-factor.

To further investigate the origin of these new resonance modes and provide a quantitative explanation for the improvement of $Q$-factors of BICs, detailed simulations and corresponding band fitting are performed, with parameters chosen to match sample E ($P_y=800\mathrm{nm}$, $t=340$, 940, $\infty \mathrm{nm}$). Figure 4(a) plots the measured and simulated transmission spectra under normal incidence. Additional resonances are observed with increasing cladding thickness. Next, cross-sectional profiles of magnitude of the electric field $|\mathit{E}|$ for selected resonances indicated by red marks with $t=940\mathrm{nm}$ are simulated, as shown in Fig. 4(b). Obviously, they are all bright SLRs. The electric field at 0.96 eV ($\sim 1292\mathrm{nm}$) is primarily confined to the surface of the nanorods, explaining why its resonance energy is almost unaffected by the cladding thickness. For resonances above 0.96 eV, as the cladding thickness increases, the electric field gradually shifts toward the dielectric layer, showing a different number of nodes in the $z$ direction. This shift indicates the introduction of additional photonic WGMs with reduced absorption loss.

Band fitting for the simulated angle-resolved transmission spectra is performed using a coupled oscillator model. The RAs and WGMs follow the dispersion relations derived from Brillouin zone folding, as shown below [68]: $${\omega }_{\mathrm{RA}}=\pm \frac{c}{n_{\mathrm{RA}}}(k_{\parallel }\pm \frac{2\pi }{P_y}),$$(5)$${\omega }_{\mathrm{WGM}}=\frac{c}{n_{\mathrm{WGM}}}\sqrt{{\left(\frac{m\pi }{L_{\mathrm{cav}}}\right)}^2+{\left(k_{\parallel }\pm \frac{2\pi }{P_y}\right)}^2},$$(6)where ${\omega }_{\mathrm{RA}}$ and ${\omega }_{\mathrm{WGM}}$ are the resonance frequencies of the two modes, $c$ is the speed of light in vacuum, and $k_{\parallel }$ and $P_y$ represent the in-plane wave vector and lattice constant in the $y$ direction, respectively. $n_{\mathrm{RA}}$ and $n_{\mathrm{WGM}}$ are the effective refractive indices, $m$ is the order of WGMs, and $L_{\mathrm{cav}}$ is the geometric thickness of the cladding layer.

Initially, a three-mode coupled oscillator model consisting of the LSPR, ${\mathrm{RA}}_1$, and ${\mathrm{RA}}_{-1}$ is applied in Fig. 4(c) to describe an infinite cladding layer, and the hybrid branches (solid pink curve and solid white curves) are in good agreement with the simulated results. It is notable that the upper branch in this region is very broad and weak. The Hopfield coefficients of the BIC band corresponding to pink curve are further given in Fig. 4(f), demonstrating that the BIC results from the coupling of ${\mathrm{RA}}_1$ and ${\mathrm{RA}}_{-1}$ with opposite phases and equal contributions at the $\mathrm{\Gamma }$ point (solid red and blue curves). For this case, even though the ideal BIC consists of only two RAs, the LSPR plays a crucial role in the q-BIC nearby [12,13], which is a key reason why the reported $Q$-factors of the plasmonic q-BIC are not very high.

As the cladding thickness changes from infinity to $t=940\mathrm{nm}$, the coupling of the periodic metasurface and upper cladding interface can produce a series of WGMs, as shown in Fig. 4(d), and Fano resonances occur in the spectra where multiple WGMs overlap with the hybrid branches in Fig. 4(c). The hybrid branches marked by the solid black ellipses in Fig. 4(d) are not clearly reflected in the simulated and measured angle-resolved transmission spectra due to high loss and weak strength. Figure 4(g) displays the corresponding Hopfield coefficients of the BIC band in Fig. 4(d). Away from the $\mathrm{\Gamma }$ point, the contributions of ${\mathrm{WGM}}_1$ and ${\mathrm{WGM}}_{-1}$ become dominant very quickly, while the contribution of the LSPR is suppressed, resulting in an increased $Q$-factor of the q-BIC. With a further reduction of cladding thickness to 340 nm, only ${\mathrm{WGM}}_1$ and ${\mathrm{WGM}}_{-1}$ are retained, but they are far from RAs. The modal hybridization is shown in Fig. 4(e). The Hopfield coefficients of BIC band in Fig. 4(h) show that the contribution of LSPR remains almost the same as in Fig. 4(f) in the range of $k_{\parallel }$ from 0 to 1, indicating that the introduction of WGMs does not significantly improve the $Q$-factors.

As we know, the $Q$-factor of an ideal BIC should approach infinity. However, in practice, every experimental setup has finite resolution, which prevents it from reaching the ideal BIC. What is actually obtained is a q-BIC convolved with the system’s resolution. For a given sample, the higher the resolution, the higher the measured $Q$-factor will be. In angle-resolved measurements, the Fourier plane contains information from each angle after the incident beam passes through the sample, as shown in Fig. 5(a). Therefore, the resolution of the Fourier plane, determined by the core diameter $d$ of the collecting fiber in this work, can significantly affect the measurement of $Q$-factors of BICs.

Figure 5(b) summarizes the measured $Q$-factors of the BICs for all samples using three fibers with different core diameters (Appendix B). It is evident that the smaller the core diameter, the higher the measured $Q$-factor. However, if the core diameter is too small, the collected signal intensity becomes too weak, leading to a poor SNR and unreliable measurements. As a result, the $Q$-factors for samples A, B, H, and I are not provided for $d=105\mathrm{\mu m}$. For subsequent measurements, we use the fiber with core diameter $d=400\mathrm{\mu m}$ instead of $d=105\mathrm{\mu m}$, with sample I at the Γ point serving as a reliable demonstration of improved $Q$-factor. The results are presented in Figs. 5(c)–5(f), where the $Q$-factor of SLR remains nearly constant ($Q$-factor $\sim 66$), while the $Q$-factor of the BIC increases by 40 (up to $\sim 260$) when the core diameter $d$ is reduced to 400 μm. The $Q$-factor of 260 ($\sim 1585\mathrm{nm}$) is among the highest measured values of BICs achieved near the communication band with a sample size of $200\mathrm{\mu m}\times 200\mathrm{\mu m}$ and measured with an incoherent light source. We estimate that this value could be further improved with a smaller core diameter, a spectrometer with higher SNR, a thicker cladding layer, and a larger Fourier plane.

As mentioned earlier, in a plasmonic system, the reported $Q$-factors of SLRs are generally higher than those of BICs. However, these measurements are often based on different structures and wavelength regions. In Fig. 3, we have already shown the evolution of the $Q$-factors of q-BICs with varying $t$. Here we further compare it with that of SLRs and discuss their respective advantages. Only sample I is used for this comparison, because the frequencies of q-BIC and SLR at the $\mathrm{\Gamma }$ point are very close to each other (difference: 0.021 eV), and since they are obtained from the same sample under identical excitation condition, the comparison is free of ambiguity.

Figures 6(a) and 6(b) display part of the measured and simulated transmission spectra of the SLR band and BIC band in detail, where the highlighted red (blue) regions indicate the evolution trend of SLR (BIC) resonances. The corresponding $Q$-factors are extracted and shown in Fig. 6(c). At the $\mathrm{\Gamma }$ point, the measured $Q$-factors of SLR and BIC are 65 and 260, respectively, and the simulated $Q$-factors can be increased to 174 and 603 at an incident angle $\theta \sim 0.9°$. Compared to SLR, the $Q$-factor of BIC is increased by 4 times (measured) and 3.5 times (simulated). These results represent significant progress for structures with similar configuration and footprint [39]. However, plasmonic BICs are more sensitive to structure parameter, defects, operating wavelengths, etc., which explains the relatively low $Q$-factors reported in the literature. Basically, all the structures that support high-$Q$ SLRs can be optimized to exhibit high-$Q$ BICs. With advancements in nanofabrication technology, optimization of structure parameters, and the improvements of measurement accuracy, plasmonic BICs with $Q$-factors exceeding those of SLRs can be achieved. Meanwhile, SLRs also have significant advantages, such as their ability to enhance the interaction with the far field and maintain high-$Q$ characteristics under large excitation angles.

In this work, we have demonstrated the capability of a reusable hybrid plasmonic–photonic metasurface to support high-$Q$ resonances through both SLRs and BICs. By systematically varying the cladding thickness, the introduced WGMs enhance the resonance modulation of q-BICs near the $\mathrm{\Gamma }$ point through mode hybridization. This approach allowed us to experimentally achieve significantly improved $Q$-factors for BICs, enhanced by a factor of 6 to 11, within a compact device footprint of $200\mathrm{\mu m}\times 200\mathrm{\mu m}$. Our results also highlight the critical role of Fourier plane resolution in accurately measuring the $Q$-factors of BICs. Using fibers with smaller core diameters, we achieved a $Q$-factor of 260, among the highest for BICs near the communication band (Appendix C). Notably, all experiments were performed using an incoherent light source, suggesting that these results could be further optimized [39]. We also conducted a comparative study of SLRs and BICs, showing that BICs can achieve up to four times the $Q$-factor of SLRs under identical conditions. However, BICs are more sensitive to structural parameters and defects, whereas SLRs exhibit greater robustness and stronger interaction with the far field. This work provides a comprehensive framework for the development of high-$Q$ metasurfaces, with potential applications in sensing, display, lasing, nonlinear and quantum optics, and plasmon modulated chemical reaction [69].

Acknowledgment. J.S. acknowledges financial support from NSFC. The authors thank Yang Li for help with nanofabrication and Prof. Shumei Sun for help with optical measurement.