Opto-optical edge defect mitigation in solution-processed WSe2 thin films for high-efficiency optical modulation Authors: Jianan Li, Zefeng Xu, Yuan Chen, Chenxuan Zeng, Jie Yang, Jin Feng Leong, Maheswari Sivan, Baoshan Tang, Zhe Wang, Yuan Gao, Hong-Lin Lin, Arjun Kesav Mugilvannan, Evgeny Zamburg, Aaron Danner, Cheng-Wei Qiu, Goki Eda, Aaron Voon-Yew Thean Published: 2026 ABSTRACT AbstractTwo-dimensional optical modulators often suffer from low efficiency due to defect-induced losses and light–matter interaction strength. Here we show that edge defects in solution-processed WSe2 nanosheets, arising from metallic 1 T domains at the edges and semiconducting 2H interiors, can be dynamically saturated through carrier trap filling. By tuning the nanosheet lateral size and applying optical or electrical excitation, defect states are progressively saturated, reducing nonradiative recombination and reshaping the carrier distribution within the film. This process enhances the intrinsic excitonic response, as reflected by the improved photoluminescence peak modulation efficiencies of 0.025 eV·V−1 (electro-optic) and 0.1 eV·mW−1 (opto-optic). More importantly, the carrier redistribution and trap filling induce changes in the complex refractive index (Δn + iΔk) through defect-state modulation and free-carrier (Drude-like) effects. Integrated with lithium-niobate-on-insulator micro-ring resonators, solution-processed WSe2 films deliver efficient C-band (1530–1565 nm) modulation with a tuning efficiency of 1.84 × 10−5 pm−1·mW·m2. Time-resolved measurements further confirm fast dynamics, with ~48.1 ns rise, ~79.4 ns fall, and sustained 200 ns operation ( ~ 5 MHz), which surpasses prior 2D-material OO modulators, establishing engineered defect regulation as a scalable route toward high-performance photonic integration.Subject terms: Nanophotonics and plasmonics, Optical materials and structures Introduction Optical modulation plays a critical role in data conversion for optical communication applications, especially in addressing the challenges posed by dramatically increasing volume and rate of data1–3, necessitating the development of energy-efficient modulation methods. Optical modulators based on two-dimensional (2D) materials have emerged as a promising solution4,5. Typical approaches for 2D material-based optical modulation can be categorized into three types: opto-optic (OO), electro-optic (EO), and thermal-optic (TO). To date, most practical OO solutions rely on 2D materials’ saturable absorption property, enabled by electron–hole pair excitations resonating with incoming photons6,7. On the other hand, EO modulators tune the refractive index through doping-induced changes in dielectric permittivity, when integrated with a waveguide platform5,8,9, while TO modulation leverages temperature-induced refractive index changes for light control10,11. 2D material-based optical modulators face important challenges, particularly the balance between modulation efficiency and optical loss associated with light–matter interaction strength. Increasing the light-material interaction length with atomically thin active materials can enhance modulation depth; however, it may also increase optical absorption and propagation loss4,5,12. Consequently, the achievable modulation efficiency depends on a careful trade-off among interaction strength, material quality, and device geometry. Ironically, defect engineering in 2D materials could be a promising approach to overcome these limitations: previous literature have shown that defects in 2D materials can induce defect-related excitons whose density can be tuned by external optical pumping13,14; such excitons can significantly change the refractive index of 2D materials, and can be accurately modeled using a multi-Lorentzian oscillator approach that incorporates exciton contributions15,16. In another application, such as mode-locked ultrafast laser, defect regulation has been proven beneficial. For example, Ni et al. demonstrated that Ar⁺ plasma treatment can regulate oxygen and selenium defect states in 2D Bi2O2Se, significantly enhancing photocarrier trapping and Auger recombination rates17. Despite these reported works that show the possibility of refractive index and photocarrier-capture modulation, it remains unclear if the use of defects is beneficial for optical modulators, especially given that defects in the bandgap are detrimental to extending carrier recombination lifetimes and modulation efficiency 18–20. To achieve defect regulation, several techniques such as grain size control have been applied to commonly used chemical vapor deposition (CVD) for 2D material deposition21,22. However, current approaches for CVD-grown 2D material thin film transfer before device fabrication still present transfer quality challenges with the risk of polymer residue, which acts as an optically lossy layer with unintended doping effects8,23, limiting the optical modulation efficiency in 2D material-based optical modulators. Alternatively, transfer-free solution-processed techniques for 2D material deposition have gained much attention, due to their ability to disperse 2D materials in solvents and subsequently deposit them directly onto various substrates24–28. However, there are abundant defects along the edges of nanosheets in solution-processed 2D material thin film29–31. Hence, there is a need for defect engineering and regulation to enable solution-processed 2D materials for achieving high-efficiency optical modulation through precise defect regulation. In this work, solution-processed WSe2 is chosen to study the impact of edge defects on optical modulation. WSe2’s strong light-matter interaction at exciton resonance make it well-suited for optic and photonic applications32–34. Through laser power-dependent photoluminescence (PL) measurements, distinct sub-bandgap emission peaks can be observed, which are attributed to localized states introduced by edge defects. Additionally, we find that the edge defect density can be modulated by the lateral size (0.5 μm and 10 μm) of the WSe2 nanosheets and the optical effects of the localized states can be further tuned by optical (0.0-0.8mW) and electrical (0.0–2.0 V) pumping at room temperature. We integrated the solution-processed WSe2 size-specific-distributed nanosheet thin film with a lithium niobate on insulator (LNOI) micro-ring resonator waveguide platform. Tunable C-band (1530–1565 nm) transmission was achieved, driven by refractive index variations originating from edge defect regulation. Notably, the WSe2–LN modulators exhibit a tuning efficiency of 1.84 × 10−5 pm−1·mW·m2. Time-resolved measurements confirm fast modulation dynamics, with a rise time of ~48.1 ns, a fall time of ~79.4 ns, and stable operation at 200 ns intervals ( ~ 5 MHz), surpassing previously reported 2D material-based all-optical modulators. These results demonstrate that defect regulation provides an effective route to scalable, effective optical modulation, paving the way toward advanced optoelectronic and photonic applications. Results We implemented multi-layer WSe2 thin film to achieve C-band optical modulation by integrating with a lithium-niobate-on-insulator (LNOI) micro-ring resonator (MRR) platform (Fig. 1a) using Langmuir-Schaefer (LS) technique (“see Methods”). Fig. 1b shows a top-view optical image of the WSe2–LN MRR, where a window is etched in the SiO2 cladding to allow the WSe2 thin film to directly contact with the LN MRR. As shown in the cross-sectional view of the platform in Fig. 1c, the optical mode is jointly confined within the WSe2 film and the LN waveguide. Lumerical MODE’s FDE eigensolver simulations estimate the optical mode overlap with the WSe2 layer to be ~0.8% (Supplementary Note 1). This hybrid configuration allows the optical mode to interact directly with the WSe2 film while still being guided by the LN core, thereby enabling efficient light–matter coupling. Owing to their resonant nature, such MRR structures are highly sensitive to changes in absorption and refractive index, making them ideal for detecting the subtle optical perturbations induced in the WSe2 film. To achieve OO modulation, a 532-nm laser is directly applied onto the WSe2 thin film, serving as the modulation signal. From the transmission measurements of waveguides with varying lengths, we extracted propagation losses of 7.2 dB/cm for the TE mode and 5.8 dB/cm for the TM mode, with the facet coupling loss estimated to be ~1.5 dB per facet (Supplementary Fig. 5). According to our optical simulations (Supplementary Fig. 6a), the coupling loss is predicted to be ~2 dB per facet. The slightly lower experimentally measured value is attributed to structural refinements introduced during the FIB facet polishing process. As shown in Supplementary Fig. 6b, the FIB-formed trench near the taper tip likely improves vertical mode matching between the lensed fiber (MFD ~ 3 µm) and the expanded waveguide facet, thereby reducing scattering and reflection at the interface. After depositing the WSe2 film, a total insertion loss of ~2.1 dB was introduced. Notably, this value shows good quantitative agreement with the WSe2 propagation loss directly obtained from Lumerical MODE simulations, confirming the consistency between experiment and simulation (Supplementary Note 2 and Supplementary Note 3). The measured transmission spectra (Fig. 1d) reveal a resonance near 1550 nm that undergoes a distinct blueshift and reduced transmission with increasing 532 nm laser power (0–0.23 mW·mm−2). Specifically, the resonance shifts from 1550.20 nm to 1550.14 nm, while the transmission depth decreases from 13.3 dB to 7.9 dB, indicating a reduced effective refractive index and enhanced absorption within the WSe2-LN MRR waveguide. Control experiments on bare LN MRR show negligible resonance shifts (Supplementary Fig. 7), confirming that the modulation originates from the WSe2 thin film rather than photorefractive effects in the LN substrate. Furthermore, we characterized the WSe2–LN MRR transmission spectra using WSe2 thin films of different thicknesses and found that both the resonance shift and transmission depth depend strongly on film thickness (Supplementary Fig. 8). Based on this analysis, a 5 nm WSe2 film was selected to further study the changes in resonance shift and transmission depth. To further understand the optical response of the WSe2 film, we performed broadband optical characterization, including reflectance and absorption measurements (Supplementary Fig. 9). The spectra show clear absorption in the visible region, confirming that the WSe2 thin film can be effectively excited under optical pumping using a 532 nm laser. This reflection and absorption provide the basis for the optically induced modification of the complex refractive index, as illustrated in Fig. 1e. In the laser-off state (top-left panel), the WSe2 thin film exhibits its intrinsic complex refractive index, determined by the equilibrium carrier population. The evanescent optical mode circulating in the MRR partially overlaps with the WSe2 layer, leading to a baseline effective refractive index of the hybrid waveguide. This corresponds to a specific resonance condition, shown as the initial transmission dip in the spectrum. In the laser-on state (bottom-left panel), optical pumping generates additional carriers in the WSe2 thin film and modifies the occupation of defect states. This induces a change in the material’s complex refractive index (Δn+iΔk) through carrier-induced effects. As a result, the effective refractive index of the hybrid waveguide is altered. Consequently, this refractive index modulation shifts the resonance condition of the MRR, producing a spectral shift in the transmission spectrum (right panel). This forms the basis of opto-optical modulation, where the input optical pump controls the transmission of a probe signal in the telecom band. C-band opto-optical modulation using solution-processed WSe2 thin film. a Schematic illustration of the lithium niobate (LN)–WSe2 micro-ring resonator (MRR) platform under 532-nm optical pumping. b Optical micrograph of the fabricated LN–WSe2 MRR (scale bar: 100 µm). c Simulated transverse electric (TE)-mode electric field distribution in the composite LN–WSe2 waveguide cross-section (scale bar: 80 nm). d Transmission spectra of the MRR with WSe2 under increasing 532-nm pump power. e Mechanism illustration of optical pumping–induced refractive index modulation in a WSe2-LN MRR. While the WSe2 thin film introduces measurable optical loss after integration with the LNOI MRR, the observed resonance modulation cannot be understood from static absorption alone. Instead, the pump-dependent shift of the MRR resonance indicates that the optical constants of WSe2 are dynamically modified under excitation. This motivates a closer investigation of the structural and electronic origin of the WSe2 optical response. We turned to high-resolution electron microscopy. Figure 2a presents a low-magnification scanning transmission electron microscopy (STEM) image of a WSe2 nanosheet, where multiple regions of interest are marked for detailed analysis. Atomic-resolution STEM imaging of region 1 (Fig. 2b) reveals a well-ordered 2H lattice, as confirmed by the corresponding fast Fourier transform (FFT) pattern and atomic scale. In contrast, imaging of region 3 (Fig. 2c) shows distortions in the 1 T lattice periodicity, characteristics of defect incorporation or local disorder. Such structural disruptions at the atomic scale provide a physical basis for the optical inhomogeneity observed in PL mapping. Interestingly, region 2 (Fig. 2d) displays a clear 2H–1 T phase boundary within the nanosheet. The coexistence of these two phases within a single nanosheet can act as a scattering center for charge carriers and photons, as well as a site for defect state accumulation. Such phase coexistence is well-reported in previous literature35, where kinetic stabilization of metastable 1 T domains occurs. The elemental composition and bonding environment across these domains were probed using atomically resolved energy-dispersive X-ray spectroscopy (EDS) line scans. As shown in Fig. 2e, the 2H region exhibits the expected periodic modulation of W and Se intensities, consistent with the ordered semiconducting lattice. By contrast, the EDS profile in the 1 T region (Fig. 2f) shows deviations in both periodicity and elemental distribution, indicative of structural distortion and local stoichiometry fluctuation of edge and non-edge area. The accurate calculation of the defect density is provided in Supplementary Note 2. Raman spectrum of the WSe2 edge shows a clearly distinct set of vibration modes of lower intensity compared to the non-edge area (Supplementary Fig. 10), which was attributed to poor interaction of the metallic phase with light36. Low-temperature PL characterization further verifies the inhomogeneity within the WSe2 flake. Supplementary Fig. 11a shows an optical micrograph of a representative 5nm-WSe2 nanosheet deposited on a SiO2/Si substrate. In contrast, the room-temperature PL spectrum measured at the WSe2 edge region exhibits additional emission peaks at energies below the Trion emission (Supplementary Fig. 11b). Upon cooling to 30 K, these sub-Trion states become significantly more pronounced within the bandgap of WSe2, consistent with the behavior of localized excitonic states. We therefore attribute these features to bound exciton (BX) states, in agreement with prior reports on WSe237. The schematic band diagram of Supplementary Fig. 11binset illustrates this concept: edge defects introduce electronic states within the bandgap, facilitating additional radiative transitions. PL intensity mapping (Supplementary Fig. 11c), extracted at an emission energy of ~1.5 eV, reveals a pronounced contrast between the edge and non-edge regions of the WSe2 nanosheet. The non-edge regions are primarily dominated by intrinsic excitonic emission centered at ~1.6 eV, whereas the edge regions exhibit an additional defect-related PL component around ~1.5 eV. As a result, the higher PL intensity observed at the edges originates from edge-localized radiative recombination channels at this distinct emission energy, rather than from an enhancement of the overall excitonic PL yield. Importantly, we observe a systematic dependence of the PL characteristics on the lateral size of the WSe2 nanosheets (Supplementary Fig. 12). In contrast, larger flakes display PL spectra dominated by a sharp A-exciton peak with substantially suppressed sub-gap emission (Supplementary Fig. 12c). Nanosheets with smaller lateral dimensions, corresponding to a higher edge-to-area ratio, exhibit enhanced and spectrally broadened defect-related emission accompanied by a reduced A-exciton intensity. This size-dependent evolution indicates that the lateral dimension governs the effective density of edge-induced localized states, thereby modulating the relative contribution of defect-mediated recombination and intrinsic excitonic emission. The above PL behavior remains consistent regardless of the specific edge location on the flake or the polarization angle of the incident excitation light (Supplementary Fig. 13 and Supplementary Fig. 14). The PL peak energy and the relative intensity of the defect-related emission vary minimally and show no systematic dependence on either edge orientation or excitation polarization. As shown in Supplementary Fig. 15a and 15b, the PL intensities of BX (A1, A2, A3), X-, and neutral exciton (X0) all increase with rising excitation power but display distinct behaviors. The BX emission saturates at higher laser powers, while X- and X0 intensities increase gradually up to ~1.5 mW·mm−2 before exhibiting a sharp rise. To further probe the role of localized edge states, we quantified the ratio of the summed BX intensity (A1 + A2 + A3) to the X0 intensity (Supplementary Fig. 15c). This ratio first increases with laser power, then saturates near ~1.4 mW·mm−2 and subsequently decreases at higher powers, consistent with the filling and saturation of defect-induced trap states. These findings further confirm that phase heterogeneity and defect-induced lattice irregularities coexist in solution-processed WSe2 nanosheets. Scanning transmission electron microscopy (STEM) characterization of WSe2 nanosheet. a Low-magnification STEM image of a WSe2 nanosheet, with selected areas marked for atomic-resolution imaging. b, c High-resolution STEM images of regions 1 and 3, respectively, showing crystalline ordering and defect sites; insets display fast Fourier transform (FFT) patterns and atomic-scale zooms. d STEM image of region 2, revealing a 2H–1 T phase boundary. e, f Atomically resolved energy-dispersive spectroscopy (EDS) line scans across W and Se atoms in 2H and 1 T phases, respectively, confirming compositional contrast and structural differences. Interestingly, Fig. 3a, b compare modulation behaviors of WSe2–LN MRR using WSe2 thin films consisting of different lateral-size nanosheets (WSe2 TF10 and WSe2 TF0.5), i.e., 10 µm and 0.5 µm lateral-size (Supplementary Fig. 16). For WSe2 TF10 thin film, both the resonance wavelength shift and transmission change scale linearly with pumping laser power. In contrast, WSe2 TF0.5 thin film exhibits a nonlinear response: gradual modulation up to ~0.15 mW/mm2, followed by an abrupt increase in both resonance shift and transmission change at higher powers. Furthermore, the WSe2 TF0.5 film exhibits a slightly higher insertion loss of ~0.55 dB/µm compared with ~0.54 dB/µm for WSe2 TF10 (Supplementary Fig. 17). This small variation is consistent with the increased density of inter-flake junctions and edge boundaries present in the smaller-flake film, which may introduce additional scattering sites and slightly perturb the guided optical mode. Notably, the difference remains minimal, indicating comparable optical quality between the two films, and the insertion-loss variation is well within an acceptable range for reliable device operation. To directly elucidate the mechanism underlying C-band optical modulation in WSe2 thin films, we systematically investigated the effects of carrier pumping on the PL and excitonic responses of the WSe2 thin films. Figure 3c, d illustrate the contrasting PL emission of films composed of small (WSe2 TF0.5) and large (WSe2 TF10) lateral-size WSe2 nanosheets. Based on the single-flake PL spectrum in Supplementary Fig. 11, we performed a three-component spectral fitting that includes the indirect bandgap emission (IX), defect-bound emission (BX), and direct bandgap emission (X0). The corresponding peak positions were estimated from Fig. 2b to be ~1.4 eV, 1.5 eV, and 1.58 eV, respectively. These extracted energies are in good agreement with values reported in the literature for WSe238. We normalize the BX intensity by the sum of the IX and neutral X0 intensities to obtain a dimensionless PL intensity ratio that captures variations across the entire film (Fig. 3c, d), and similar PL ratios have been widely adopted in TMDC literature to assess defect densities and sample quality39–42. Herein, small nanosheets exhibit PL spectra dominated by BX emission, consistent with their high edge-to-area ratio that enhances the density of localized states. Conversely, films dominated by large nanosheets show stronger X- and X0 emission, reflecting reduced edge defect density and a more intrinsic excitonic response. Spatial PL mapping (10 µm × 10 µm) further confirms this trend: BX intensity is spatially correlated with nanosheet edges and smaller fragments, while X0 emission dominates in larger domains. This nanosheet-size dependence is a direct manifestation of edge defect density, establishing the link between structural morphology and excitonic recombination pathways. To further reveal the dynamic behavior of these excitons, we performed time-resolved photoluminescence (TRPL) measurements (Supplementary Fig. 18). BX emission lifetime of ~1.0 ns exhibits a longer lifetime compared to X0 lifetime of ~0.5 ns, reflecting the trapping of carriers at defect-associated states before radiative recombination. Effect of electrical and optical pumping on effective index of WSe2 thin film. a, b Resonance wavelength shift (blue) and transmission change (red) as a function of pumping power on WSe2 TF10 and WSe2 TF0.5, inset: schematic of WSe2 TF10 and WSe2 TF0.5 on the MRR. c, d Schematics of WSe2 thin films composed of small- and large-lateral-size nanosheets, together with their corresponding photoluminescence (PL) spectra and 10 µm × 10 µm spatial PL mappings. FWHM full width at half maximum. e Extracted changes in the real (Δn) and imaginary (Δk) parts of the effective refractive index of WSe2 thin films under optical pumping. Carrier pumping through optical excitation with a 532-nm laser provides a pathway to verify the trap-state occupancy dynamics (Supplementary Fig. 19). The PL spectrum exhibited a similar blue shift in the peak position, transitioning from ~1.52 eV to ~1.57 eV as the laser power increased. Notably, WSe2 integrated on the LNOI waveguide shows PL emission characteristics comparable to those observed on Si/SiO2 (Supplementary Fig. 20), including similar excitonic peak positions, linewidths, and power-dependent blue-shift behavior, indicating that the underlying excitonic and defect-state physics is preserved on the LNOI platform. Furthermore, by systematically comparing the Raman and PL spectra of WSe2 on LNOI waveguide with those on reference Si/SiO2 substrates, we observe negligible spectral shifts or linewidth variations, suggesting the absence of significant charge transfer or substrate-induced doping (Supplementary Fig.21). These results collectively confirm that the observed modulation originates from carrier-induced optical effects within WSe2, rather than doping or electronic interaction from the LNOI substrate. We further ruled out laser heating by performing temperature-dependent PL measurements. Supplementary Fig.22 demonstrates that excitonic PL redshifts steadily as the temperature is raised from 30 K to 150 K, consistent with the Varshni‑type band‑gap shrinkage of WSe2. In contrast, our modulation experiments show a blue shift whose magnitude far exceeds the few‑meV change expected from a modest ( ~ tens of kelvin) temperature rise, and previous reports under similar laser powers find only tens‑kelvin heating43–48. Expanding the analysis, PL peak position mapping (Supplementary Fig. 23a) demonstrated that the same blue shift occurred consistently across the entire solution-processed WSe2 TF0.5 as the laser power increased from 0.1 mW to 0.8 mW. This consistent PL peak shift can be attributed to the saturation of photo-excited carrier traps by localized potential stemming from the high density of edge defects distributed throughout the thin film system, which is not observed in WSe2 TF10 (Supplementary Fig. 23b). The effect of optical pumping on the effective refractive index of WSe2 TF0.5 is shown in Fig. 3e. The variations in the real (Δn) and imaginary (Δke) components are derived from a unified Drude-model-based analysis, in which the excitation-dependent carrier density is used as the input parameter. The carrier mobility employed in the model is independently extracted from the transfer characteristics (ID–VG) of the WSe2 FET (Supplementary Note 4). Both Δn and Δke therefore originate from the same excitation conditions and represent the real and imaginary parts of the complex refractive index, respectively. As the optical pumping increases, both components exhibit a nonlinear dependence, which is consistent with trap-state filling and the subsequent redistribution of carriers. The extracted magnitude of Δn ( ~ 10−3) is sufficient to induce measurable resonance shifts in the high-Q MRR cavity, in good agreement with the modulation behavior observed in Fig. 1. To further elucidate the carrier pumping mechanism in WSe2 thin films, electrical excitation is employed as a complementary approach. A vertical ITO/WSe2/ITO structure (Supplementary Fig. 24inset) is fabricated to enable controlled carrier injection and direct probing of carrier-density-dependent excitonic behavior. Under applied bias (0–2 V), the PL spectra exhibit a consistent blue shift (Supplementary Fig. 24). Notably, measurements across 30 independently fabricated devices show highly uniform spectral profiles and peak positions (Supplementary Fig. 25), confirming good device reproducibility. To further analyze the modulation mechanism, each PL spectrum is deconvolved into IX, BX, and X0 (Supplementary Fig. 26). Importantly, no significant shift in the peak energies of these excitonic transitions is observed with increasing bias. Instead, electrical pumping primarily induces a redistribution of spectral intensity among IX, BX, and X⁰ (Supplementary Fig. 27), indicating that the modulation is governed by carrier-density-dependent population dynamics rather than a pure Stark effect. With increasing carrier injection, IX and BX initially dominate but gradually saturates beyond ~1 V, while X0 increases sharply. This crossover signifies a suppression of defect-mediated recombination, consistent with an edge-defect saturation effect. The effect of electrical pumping on the effective refractive index of WSe2 is also presented in Supplementary Fig. 28a, exhibiting a trend consistent with that observed under optical pumping (Supplementary Fig. 28b). Together, these results establish that carrier pumping modulates the effective refractive index of WSe2 through carrier-induced trap-state dynamics, enabling a unified mechanism for efficient all-optical control in the LNOI micro-ring platform. To investigate the modulation dynamics of the WSe2–LN MRR device, we employed time-resolved spectroscopy characterization. The experimental setup is illustrated in Fig. 4a. A pulsed 405 nm laser was focused onto the WSe2 thin film and served as the high-speed optical modulation source. Although different excitation wavelengths were employed in the modulation and PL measurements, both photon energies are well above the WSe2 bandgap and thus efficiently generate electron–hole pairs. Since the observed modulation behavior arises from carrier population regulation and trap-state filling, rather than wavelength-specific resonant excitation, the use of different excitation wavelengths does not alter the underlying physical mechanism or the conclusions drawn in this work. Simultaneously, probe light near 1550 nm was coupled into the MRR waveguide, and the transmitted output was recorded using a high-speed photodetector connected to a digital oscilloscope. This configuration enabled direct monitoring of the temporal evolution of the C-band resonance in response to periodic optical excitation. The modulation laser was operated with a pulse interval of 200 ns, corresponding to a repetition frequency of ~5 MHz. Figure 4b displays the time trace of the modulating laser, confirming the stable and periodic excitation delivered to the WSe2–LN MRR. Figure 4c presents the time-resolved transmission of the C-band light signal under pulsed excitation. The modulated resonance manifested as a periodic oscillation synchronized with the 200 ns laser interval. This behavior was further validated using only either the 405 nm pulsed laser or C-band light (Supplementary Fig. 29), confirming that the integrated WSe2 film is capable of tracking modulation at 200ns-pulse timescale. Importantly, the modulated output exhibited minimal delay relative to the input excitation. To quantify the device dynamics, the measured time trace (blue curve) was fitted with a damped sinusoidal model (red curve). From this analysis, we determined the switching times using the 10% – 90% rule to extract a rise time of ~48.1 ns and a fall time of ~79.4 ns. Furthermore, Fig. 4d suggests the WSe2-LN MRR device shows good stability with the endurance characterization of ~107 operation cycle without noticeable degradation. Under pulsed excitation, the maximum C-band transmission change was ~3 dB, which is lower than the ~5.5 dB observed with continuous-wave excitation. This reduction is attributed to the short laser pulse width (200 ns), which likely limits exciton generation compared to the CW case. Modulation speed characterization of WSe2-LN MRR. a Schematic of the experimental setup for modulation speed measurement of the WSe2-LN MRR under 405-nm pulse laser pumping. TLS: tuneable laser source, HS-PD: high-speed photodetector, OSCP: oscilloscope. b Time-resolved intensity of the reflected modulation laser. The interval time of the pulsed modulating laser is 200 ns. Inset: enlarged view of a single pulse showing a pulse width of 24.58 ns. c Measured time-resolved modulated C-band signal under the pulsed modulating laser. d Endurance test on operation cycle WSe2-LN MRR. The horizontal dashed lines indicate the average transmission levels of the pristine state and the state under pulsed laser excitation. Discussion In this study, we successfully demonstrated defect-engineered optical modulation in solution-processed WSe2 films integrated with an LN MRR. Remarkably, with only a ~ 5 nm thin film, we achieved electro-optic and opto-optic modulation efficiencies of 0.025 eV V⁻¹ and 0.1 eV mW⁻¹, respectively. Time-resolved measurements further revealed fast modulation dynamics, with a rise time of ~48.1 ns, a fall time of ~79.4 ns, and stable operation at 200 ns intervals ( ~ 5 MHz), and modulation efficiency was carefully evaluated by normalizing the resonance shift to the effective optical power density incident on the device, ensuring an accurate representation of the light–matter interaction strength. Based on this calibrated analysis, the achieved modulation efficiency surpasses that of previously reported 2D-material-based modulators10,49–57 (Fig. 5) and more details are included in Supplementary Table 4. Performance comparison of 2D-based optical modulator. Benchmark table of current 2D material based optical modulators using for all-optical modulation, in terms of switching time and modulation efficiency. While the modulation principle may appear similar to conventional refractive-index tuning by optical pumping, the efficiency here is significantly enhanced by a defect-saturation mechanism. Specifically, metallic 1 T domains located at nanosheet edges introduce localized trap states within the bandgap. Under electrical or optical excitation, the trap states become dynamically filled and no longer serve as recombination centers. As a result, carriers redistribute into delocalized excitonic states, thereby enabling high-efficiency refractive-index modulation. Compared with previously reported PL blue shifts in TMDCs (Supplementary Table 5), most of which rely on creating high densities of chalcogen vacancies through Ar-plasma treatment or electron-beam irradiation, but at the cost of lattice damage, crystallinity degradation, and reduced material quality58–64. Our work achieves comparable or even larger shifts by leveraging intrinsically engineered 2H/1 T phase boundaries at flake edges. Crucially, we demonstrate that this controlled phase–defect architecture can be directly harnessed for active optical and electro-optical modulation in integrated LNOI photonic devices, avoiding the material degradation typically associated with vacancy-generation methods. Film thickness plays a key role in balancing modulation efficiency and insertion loss. Both resonance shift and transmission depth scale with film thickness, with the ~5 nm regime offering the most favorable trade-off between effective index modulation and excess absorption. This underscores the importance of thickness and lateral size engineering of nanosheets to optimize device performance. While compared with the state-of-the-art EO Si and LN modulators (Supplementary Table 6), TFLN modulators routinely exceed 100 GHz and Si modulators achieve 25–40 GHz, our device is limited to ~5 MHz because the modulation relies on defect-mediated carrier trapping and release lifetimes in WSe2, a process inherently slower than the ultrafast Pockels effect in LN or the plasma-dispersion effect in Si. Several routes could mitigate these limitations. First, the current free-space optical pumping is inefficient because the illumination spot is much larger than the waveguide width, resulting in weak absorption. Focusing the pump with a lensed fiber or directly coupling it into the waveguide would confine the optical energy within the guided mode, thereby enhancing both the modulation depth and the response speed. A higher pump fluence can further accelerate the exciton dynamics by inducing exciton–exciton annihilation, bimolecular recombination, and trap saturation, which collectively shorten the exciton rise and decay times65–67 (from ~ns at low power to tens of ps at high power) and thus could increase the modulation bandwidth. Second, resonator engineering, such as increasing the Q-factor or the active WSe2 thickness, which could amplify the effective refractive-index change and absorption, paving the way toward higher-speed, more efficient modulators. Our results demonstrate that edge defects, typically regarded as detrimental, can be engineered into functional elements for active photonic modulation through a defect-saturation mechanism in which optical pumping saturates edge-related trap states, thereby suppressing nonradiative recombination and enhancing excitonic transitions. This mechanism enables refractive index modulation in solution-processed WSe2 thin films integrated with LN MRR. While OO modulation with an external pump is not a practical scheme for communication systems, we applied it as a characterization method to probe the intrinsic defect-assisted response independent of the device’s structural influences or operations. Moreover, recent demonstrations of waveguide-delivered optical excitation in graphene–Si hybrid modulators68, TMDC–Si waveguides69, and WSe2–SiN emitter–waveguide systems70, which highlight credible routes to replace free-space pumping with scalable on-chip light sources such as heterogeneous III–V/Si nano-lasers or µLEDs. Importantly, we also demonstrate that the defect-saturation-induced modulation can be readily extended to electrical carrier injection. Electrical injection provides a practical pathway for device implementation, and our results confirm that the same defect-saturation–driven modulation mechanism is applicable under purely electrical or hybrid electro-optical pumping. While OO modulation serves here as a proof-of-concept validation, the established materials platform and mechanism provide a realistic foundation for future high-speed electro-optic modulators based on solution-processed 2D semiconductors. Methods Preparation of WSe2 nanosheets dispersions We adopted liquid-phase exfoliation (LPE) to obtain WSe2 nanosheets. The WSe2 single crystal was intercalated using a tetrahexylammonium tetrafluoroborate (Sigma-Aldrich) solution in propylene carbonate (5 mg mL⁻¹). A voltage of −6 V was applied for 2 h to introduce THA⁺ cations into the WSe2 crystal (Supplementary Fig. 30a). After electrochemical intercalation, the crystal was rinsed three times with DMF, followed by ultrasonication for 2 h. The resulting WSe2 dispersion was centrifuged at 67 g for 5 min to remove unexfoliated particles. To obtain nanosheets of large lateral size ( ~ 10 µm), the supernatant was collected and subjected to a second round of ultrasonication for 1 h, then centrifuged at 151 g for 5 min. The sediment from this step was collected and re-dispersed in DMF. For smaller nanosheets ( ~ 0.5 µm), the initial supernatant was processed separately, undergoing a second ultrasonication for 2 h followed by centrifugation at 1677 g for 5 min. The sediment from this step was also collected and re-dispersed in DMF (Supplementary Fig. 30b). The lateral size distribution is shown in Supplementary Fig. 30c. Raman spectra are characterized to evaluate the WSe2 material quality under different sonication duration (Supplementary Fig. 31). The Raman signal intensity decreases when the sonication time exceeds 4 h, indicating possible material degradation. To further verify the material quality, FTIR spectra were also characterized to confirm the removal of undesirable chemical residues (Supplementary Fig. 32). Solution-processed WSe2 thin film formation by Langmuir−Schaefer method The Langmuir-Schaefer setup utilizes a glass gooch filter crucible with a fritted disc, positioned over a Kitasato flask. A Si/SiO2 or LN MRR on Si/SiO2 substrate (1.5 cm × 1.5 cm) is placed on the fritted disc (Supplementary Fig. 33). To create a water/hexane interface, ~20 mL of distilled hexane is drop-casted onto the surface of deionized water, with the Si/SiO2 substrate submerged beneath this interface. WSe2 nanosheets dispersions are drop-cast onto the water/hexane surface until a continuous, gap-free interface is formed. Following the self-assembly of WSe2 nanosheets, hexane is carefully removed using a pipette, and the water is extracted via vacuum through the fritted disc. This allows the solution-processed WSe2 thin film to deposit onto the Si/SiO2 substrate. The deposited thin film is then annealed at 120 °C for 1 h on a hot plate inside an N2 glovebox to enhance the adhesion of the solution-processed WSe2 thin film to the Si/SiO2 substrate. Finally, the film is immersed in dichloromethane to remove any remaining residues, followed by a rinse with isopropanol (IPA) and drying with a nitrogen (N2) stream. The large-area atomic force microscopy (AFM) image reveals a uniform thickness (Supplementary Fig. 34). Device fabrication The ITO/WSe2/ITO device was fabricated by depositing a WSe2 nanosheet dispersion onto a Si/SiO2 substrate with a pre-patterned bottom ITO contact layer (30-nm thick). The deposition was performed using the Langmuir-Schaefer method at room temperature. Afterward, an additional ITO layer (30-nm thick) was deposited using a sputtering system, which was deposited under low-power RF sputtering (20 W, 1 × 10⁻⁶ Torr Ar ambient, deposition rate <0.2 nm/s). These conditions were chosen to minimize plasma/ion bombardment. The preserved PL and electrical response of the WSe2 films after deposition confirm negligible sputtering-induced damage. The solution-processed WSe2 FET was fabricated began with the patterning of 30-nm tungsten bottom gate (BG) electrodes on a Si/SiO2 (90 nm) substrate using e-beam lithography. Subsequently, a 10-nm HfO2 dielectric layer was grown by atomic layer deposition (ALD). WSe2 nanosheet was deposited onto the bottom gate by Langmuir−Schaefer method. A channel width of 5.0 μm was defined through e-beam lithography, followed by reactive ion etching. Ag/Au (10 nm/25 nm) source/drain contacts were patterned using e-beam lithography and e-beam evaporation to define channel lengths of 1.0 μm. Finally, the device was annealed at 200°C in Ar/H2 gas for 30 min. For the WSe2-LNOI MRR waveguide platform, the WSe2 nanosheet dispersion was similarly deposited using the Langmuir-Schaefer method. The deposition was applied onto an LN MRR waveguide structure. This waveguide was fabricated using electron-beam lithography followed by inductively coupled plasma (ICP) etching to define the core structure. At the chip facet, a taper structure was incorporated to function as a mode converter. The taper width was gradually expanded from 1 µm to 3 µm to improve mode matching and enable efficient excitation of both fundamental TE and TM modes. The facet was subsequently polished using focused ion beam (FIB) milling to form a smooth and well-defined coupling interface. For optical characterization, a lensed fiber was employed at the input. The lensed fiber has a mode field diameter (MFD) of ~3 µm, closely matching the expanded taper tip, thereby enhancing coupling efficiency and reducing insertion loss. Optical characterization Low-temperature PL measurements were conducted using a confocal system (Andor Solis spectrometer with an iDus Si CCD detector) with a 473 nm continuous-wave (CW) laser source, housed within a Janis ST-500 cryostat, with an integration time of 0.1 s. Low-temperature time-resolved PL measurements were characterized using a 405 nm picosecond pulsed laser (PicoQuant, PDL 800-D). The emitted signal was detected using a time-correlated single-photon counting module (PicoHarp 300) with an avalanche photodiode. Room-temperature and bias-dependent PL measurements were performed using a confocal Raman/PL system (Alpha 300 R, Witec) equipped with a 532 nm laser source, with an integration time of 0.1 s. For room-temperature bias-dependent PL measurements, ITO/WSe2/ITO devices were mounted on a substrate carrier (KD-78516-C, Evergreen Semiconductor Materials) and electrically connected to a DC source meter (Keithley) via wire bonding. C-band transmission spectra were measured using a tunable laser source (Keysight N7776C) with a wavelength sweep from 1549.8 nm to 1550.7 nm, an optical power of 5 dBm, a scan speed of 20 nm min−1, and a step size of 1 pm. The light is coupled into a single-mode optical fiber and edge-coupled into the device via a 6-axis fiber alignment system (EOULU). The optical output of the device is then collected by another optical fiber followed by a power meter (Keysight N7748C) for analysis. The transmission spectrum is generated by Keysight PAS LS engine software. The external laser source (iLASER 532 nm) is fixed in a customized holder to maintain ~30 cm between the laser output and the sample, with an exposure time of 10 s for each measurement. To analyze the response time, time-resolved spectroscopy was employed using an oscilloscope (Tektronix, OS054) with InGaAs switchable gain amplified Detector (THORLABS, PDA20CS). In this time-resolved experiment, WSe2 thin film was excited by a pulsed laser with a wavelength of 405 nm (PicoQuant, PDL 800-D). UV–Vis extinction spectra of WSe2 films deposited on quartz substrates were measured in the 200–1100 nm range using a Shimadzu UV-1800 spectrophotometer. Reflectance spectra were acquired using a confocal micro-reflectance configuration integrated with a Raman/PL system (Alpha 300 R, Witec). A broadband white-light source was focused onto the sample through a microscope objective, and the reflected signal was collected by the same objective in a backscattering geometry and directed to a spectrometer equipped with a CCD detector. STEM characterization Multilayer WSe2 sample was prepared by drop-casting solution-processed multilayer WSe2 onto holey SiN TEM substrate with holes size of 1 m (TED PAELLA). The sample was then cleaned using acetone vapor and kept in a vacuum. An aberration-corrected STEM (JEOL-ARM200F) equipped with a cold field emission gun and a DCOR probe corrector operating at 80 kV was used for HAADF-STEM imaging. HAADF images were captured at semi angles ranging between 68-280 mrad with a probe convergence of ≈ 30 mrad. Supplementary information Supplementary information Transparent Peer Review file Author contributions J.L., Z.X., and A.V.-Y.T. conceived the idea. J.L. fabricated the ITO/WSe2/ITO device, J.L. and Z.X. fabricated the LN-WSe2 MRR modulator, C.Y. and G.E. performed low temperature power-dependent PL, TRPL and STEM characterization, B.T. performed UV-Vis absorption characterization, C.Z. performed Lumerical mode simulation, A.K.M. performed FITR characterization, J.L. performed bias-dependent PL characterization and reflectance measurement assisted by Z.W. and C.Q, Z.X., J.Y. and J.L. performed C-band light transmission characterization assisted by Y.G., H.-L. L., and A.D., J.F.L, M.S. and E.Z assisted in manuscript preparations. J.L wrote the manuscript with suggestions from all co-authors. All the authors discussed the results and commented on the manuscript. Peer review Peer review information Nature Communications thanks Haitao Chen, Alberto Montanaro, and the other, anonymous, reviewer for their contribution to the peer review of this work. A peer review file is available. Funding This work is supported in part by the Ministry of Education, Singapore, under its Academic Research Fund (AcRF) Tier 2 (MOE-T2EP50123-0033), Semiconductor Research Corporation (contract 3145.001), as well as NRF’s Medium Sized Center: Singapore Hybrid-Integrated Next-Generation μ-Electronics (SHINE) Center funding program. Data availability Relevant data supporting the key findings of this study are available within the article and the Supplementary Information file. All raw data generated during the current study are available from the corresponding authors 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: Jianan Li, Zefeng Xu. Supplementary information The online version contains supplementary material available at (10.1038/s41467-026-73576-1). References Associated Data Supplementary Materials Supplementary information Transparent Peer Review file Data Availability Statement Relevant data supporting the key findings of this study are available within the article and the Supplementary Information file. All raw data generated during the current study are available from the corresponding authors upon request.