Evanescent-Field-Guiding 2D Ultrathin Waveguides for Seamlessly Integrated Devices

Abstract

Seamless integration of active devices into photonic integrated circuits remains a challenge due to the limited accessibility of the optical field in conventional waveguides, which tightly confine light within their cores. In this study, we propose a two-dimensional (2D) ultrathin waveguide as a photonic platform that enables efficient interaction between guided light and surface-mounted devices by supporting optical modes dominated by evanescent fields. We show that the guided light in a monolayer MoS₂ film propagates over millimeter-scale distances with more than 99.9% of its energy residing in the evanescent field. As a representative active device, we develop a WSe₂ photodetector and directly integrate it onto the MoS₂ waveguide. The integrated photodetector demonstrates nanowatt-level sensitivity, microsecond-scale response times, and a polarization selectivity of 0.94. These results establish 2D ultrathin waveguides as a compact, robust, and versatile photonic platform that leverages evanescent-field guiding for the seamless integration of passive and active photonic components.

Photonic integrated circuits (PICs) constitute the foundational infrastructure for a wide range of optical technologies, including high-speed data transmission, − optical signal processing, , and on-chip computing. ,− In these systems, passive elementsprimarily waveguidesfunction as channels for guiding and routing light, while active components such as lasers, , modulators, − and photodetectors control or convert optical signals. − A key enabler of compact and efficient PICs is the use of high-index contrast waveguides. Typically fabricated from materials such as silicon-on-insulator, silicon nitride, or III–V semiconductors, the waveguide’s characteristic dimensions (d) are tailored to specific operating wavelengths (λ) and the refractive index (n) of the constituent materials, ensuring that the ratio nd/λ remains on the order of unity. , This design provides tight optical confinement and enables low-loss propagation and thus supports dense integration and high-performance photonic functionalities.

However, this tight confinement presents challenges for active device integration. Since the guided mode remains largely trapped within the waveguide core, achieving efficient coupling between passive and active elements requires sophisticated alignment and bonding procedures on spatial scales finer than the core dimensions. , To reduce fabrication complexity and expand design flexibility, seamlessly integrated devices (SIDs) have been proposed as an alternative approach. SIDs, commonly employing two-dimensional (2D) materials, ,,− ,− phase-change materials, ,,,, or organic electro-optic materials − can be placed directly on the waveguide surface without modifying the core structure. Nevertheless, in core-confined waveguides, only a small fraction of the optical field extends into the surrounding cladding. This limited evanescent-field fundamentally restricts the performance of SIDs, which rely on evanescent-field interactions to couple with guided light.

A key strategy for enhancing evanescent-field interactions is to reduce the waveguide thickness so that nd/λ ≪ 1. As the waveguide core is deliberately thinned, the optical mode redistributes more broadly into the cladding, thereby enhancing the evanescent-fielda direct manifestation of the uncertainty principle in which tighter confinement in wavevector space leads to broader spatial delocalization. Figure a illustrates this physical picture, plotting the evanescent-field factor (EFF), defined as the fraction of the optical mode extending beyond the core, as a function of normalized thickness nd/λ. EFF remains low (∼0.1) when the core thickness is comparable to the wavelength (nd/λ ∼ 1; core-confined regime) but gradually increases as the waveguide becomes thinner, approaching EFF ∼ 1 when the core thickness is vanishingly small compared to the wavelength (nd/λ ∼ 0; evanescent-field dominant regime) (Figure S1). This indicates that reducing the waveguide thickness enhances the optical field overlap with surrounding materials and thus enables efficient SID interaction.

1.

Evanescent-field-guiding 2D ultrathin waveguides. a) Numerical analysis of evanescent field intensity as a function of waveguide thickness, illustrating the transition from a core-confined regime to an evanescent-field-dominant regime. b) Schematic illustration of a seamlessly integrated device (SID) on the 2D ultrathin waveguide. The inset shows an image of the fabricated device. c) Top-view image of guided light in a 400 nm-thick silicon nitride waveguide (nd/λ ∼ 1.1 at λ = 750 nm). The top panel shows the scattered light distribution along the propagation direction (scale bar: 1 mm). The bottom panel displays the corresponding signal decay extracted from the averaged intensity along the waveguide. The yellow line represents a linear fit to the measured decay. d) Top-view image of guided light propagation in a monolayer MoS₂ waveguide (nd/λ ∼ 0.003). The top panel shows the propagation of the scattered light (scale bar: 1 mm). The bottom panel shows the averaged intensity profile with the yellow line indicating the fitted linear decay. e) Schematic illustration of end-face imaging of the optical mode in the waveguide. f) Experimentally measured end-face image of a 400 nm-thick silicon nitride waveguide with the focal plane on the waveguide facet. The top inset shows the calculated mode profile (scale bar, 1 μm). The bottom inset shows a magnified image of the guided mode, where the optical mode intensity profile is shown. The arrows mark the positions used to determine the full width at half-maximum (FWHM; scale bar, 5 μm). g) End-face image of the 2D ultrathin waveguide with the focal plane on the end of the monolayer waveguide. The top inset shows the corresponding mode profile (scale bar: 1 μm). The bottom inset shows a magnified view of the cross-section-guided mode and the optical mode intensity profile (scale bar, 5 μm).

Recently, δ waveguides, realized using a monolayer MoS₂ film (d ∼ 0.6 nm, n ∼ 4) encapsulated in a silicone elastomer cladding, have been experimentally shown to operate in an evanescent-field dominant regime for visible and near-infrared light (nd/λ ∼ 10^–3). This type of waveguide traps light with its electric field polarized parallel to the film plane (y-axis), and the trapping mechanism along the confining z-direction (E _y (z)) is attributed to an optical analog of quantum mechanical δ-potential well. The propagation loss of this guided mode along the x-axis is then primarily determined by the material quality (e.g., density of scattering centers). Benefiting from the atomic scale thinness of the MoS₂ monolayer, 2D ultrathin waveguides have demonstrated millimeter-scale propagation length with losses as low as ∼0.5 dB/mm. Moreover, the atomically flat, dangling bond-free surface of van der Waals (vdW) materials allows for the direct transfer integration of SIDs, potentially avoiding the complexities associated with epitaxial growth and monolithic integration of active components (Figure b).

This study introduces 2D ultrathin waveguides as a photonic platform that leverages evanescent-field interactions with active materials for enhancing the performance of SIDs. For this, we developed a fabrication process compatible with electrically functional SIDs and established a design strategy to optimize optical coupling between the guided mode and the active device. We then applied these procedures to the proposed platform by employing a WSe₂ photodetector as a model system for the seamless integration of SIDs onto 2D ultrathin waveguides. Once integrated, the device exhibited highly sensitive, polarization-selective detection of guided modes, achieving detection limits down to the nanowatt range and polarization selectivity approaching unity. The fabrication and optimization principles established in this work are broadly applicable to other types of SIDs, such as modulators, emitters, and nonlinear elements, suggesting 2D ultrathin waveguides as a robust and versatile photonic platform capable of unifying passive and active components for efficient on-chip control of light waves.

We measured the propagation loss and spatial distribution of modes in two distinct waveguidesa 400 nm-thick silicon nitride waveguide (nd/λ = 1.1) and a monolayer MoS₂ waveguide (nd/λ = 0.003)to evaluate the effect of the waveguide core on guided mode properties. Both structures were designed with symmetric claddings to ensure that the guiding behavior could be attributed primarily to differences in the core. The silicon nitride waveguide was fabricated by depositing a 400 nm silicon nitride layer onto a fused silica substrate using plasma-enhanced chemical vapor deposition (PECVD), followed by deposition of a 500 nm silicon dioxide top cladding. The monolayer MoS₂ waveguide was prepared by transferring a continuous CVD-grown MoS₂ film via an epoxy-based delamination method, resulting in a monolayer encapsulated between continuous epoxy claddings (see Table S1 for fabrication procedures).

Propagation loss was measured by using a top-view optical microscopy configuration. A 750 nm laser beam was edge-coupled into the input facet of each waveguide, and the guided light was imaged on the xy-plane with a wide-field objective lens (NA = 0.04). Figures c and d show the recorded intensity distributions for the 400 nm-thick silicon nitride waveguide and the monolayer MoS₂ waveguide, respectively. In both cases, the guided mode appears as a bright signal at the input facet that gradually attenuates with distance, exhibiting the exponential decay characteristic of propagation loss in a waveguide. The propagation loss was extracted from the recorded intensity distribution by evaluating the average slope of exponential decay. Figure c shows the propagation of the guided mode in the silicon nitride slab waveguide with a measured propagation loss of 0.29 dB/mm. Applying the same procedure to the monolayer MoS₂ waveguide yielded a propagation loss of 0.72 dB/mm (Figure d), which is on the same order of magnitude as the value obtained from the silicon nitride slab. These results demonstrate that millimeter-scale guiding with a preserved fundamental mode is maintained even in the ultrathin limit (Figure S2), with the measured propagation loss primarily governed by the intrinsic wavelength-dependent absorption of MoS₂ (Figure S3).

To analyze the transverse mode profiles (along the z-direction), we employed end-face imaging microscopy in which an optical microscope was aligned parallel to the propagating axis (Figure S4). This allows direct visualization of the cross-sectional distribution of guided-field intensity in the yz-plane (I(y, z)), as illustrated in Figure e. The same 750 nm laser beam was edge-coupled into the input facet of each waveguide, and the output mode was captured through an objective lens (NA = 0.8). This setup enables a direct comparison of guided mode distribution in both the core-confined waveguides and evanescent-field dominant waveguides.

Figure f presents the experimentally measured mode profile of the 400 nm thick silicon nitride waveguide. The top inset shows the mode distribution obtained from finite-difference eigenmode (FDE) analysis (Materials and Methods). The calculation predicts a full width at half-maximum (FWHM) of approximately 270 nm, corresponding to an EFF of 0.05 (Figure S4). The main panel displays the end-plane image of the guided mode through the output facet. The image shows a distinct, bright red horizontal signal against a dark background. The location of the bright signal corresponds to the position of the silicon nitride core and appears only under waveguide-coupled illumination (i.e., transverse electric (TE) polarization). The lower inset provides a magnified view of the dashed rectangle and reveals the physical extent of the optical field. The experimentally measured FWHM is approximately 0.64 μm, corresponding to an EFF of 0.19. The simulation results and the experimental measurements both verify the tight optical confinement of the guided mode in the core-confined waveguide.

Figure g presents the mode profile of the monolayer MoS₂ waveguide. The top inset shows the mode distribution obtained from FDE calculations, predicting an FWHM exceeding 2.6 μm. The main panel shows the experimentally measured mode profile with a measured FWHM of 2.8 μm. In contrast to the tightly confined mode of the 400 nm thick silicon nitride waveguide, the monolayer MoS₂ waveguide supports a substantially delocalized mode, with the majority of the field extending outside the core (EFF = 0.99). These results confirm that the 2D ultrathin waveguide operates in an evanescent-field dominant regime in which the substantial extension of the optical field offers an ideal platform for SID integration.

The realization of electrically functional devices on a 2D ultrathin waveguide requires strict refractive index matching and defect-free, atomically flat interfaces along the optical path. The refractive index mismatch between the top and bottom cladding needs to be controlled with a precision of 0.25% (Figures S5 and S6). Existing transfer techniques involving chemical etching or sacrificial polymers (e.g., poly­(methyl methacrylate) (PMMA)) often leave chemical and/or polymeric residues, which introduce scattering centers. , Furthermore, integrating electrical contacts requires a low-temperature, precise assembly process to preserve the waveguide’s optical properties while forming micrometer-scale electrodes.

Figure illustrates our strategy for the integration of SIDs on a 2D ultrathin waveguide. We introduce three experimental advances: (1) implementing a microscopy-based precision platform for active device integration, (2) using ultraviolet (UV)-curable epoxy as both the cladding and exfoliation medium, and (3) employing vdW electrode transfer. Figure a schematically depicts the instrumental setup for integrating the SID onto a 2D ultrathin waveguide. The process includes MoS₂ delamination using epoxy, active material transfer, vdW electrode transfer, and wire bonding.

2.

SID fabrication process compatible with 2D ultrathin waveguides. a) Schematic of the customized setup for SID fabrication on the 2D ultrathin waveguide. b) Schematic depicting the epoxy-based transfer method. The top panel shows the alignment procedure between the epoxy droplet and the MoS₂ prior to contact; the yellow arrow indicates the direction of contact. The bottom panel depicts transferred MoS₂ on epoxy after the curing process. c) Schematic of the electrical device integration on 2D ultrathin waveguides. d) Photograph of the exfoliated MoS₂ on epoxy, corresponding to the bottom panel of Figure b. The inset shows the donor substrate after the MoS₂ exfoliation process. e, f) Optical microscopy images of the donor substrate and epoxy substrate following the exfoliation process (scale bar, 200 μm). g) Optical microscopy image of a SID on the 2D ultrathin waveguide encapsulated in epoxy cladding (scale bar, 20 μm).

Figure b shows the MoS₂ transfer procedure that utilizes UV-curable epoxy as the cladding material and exfoliation medium. Liquid-phase epoxy was applied to a slide glass mounted on a precision manipulator. The distance between the slide glass and the MoS₂ substrate was controlled, and once the desired alignment was achieved, the epoxy was cured under UV light. The MoS₂ film was then mechanically exfoliated from the substrate, forming the bottom half of the waveguide structure. Figure c illustrates the integration of electrically functional SIDs onto the prefabricated waveguide. First, 2D materials, serving as the active component of SID, were dry-transferred onto the waveguide, followed by electrode integration. For this step, we employed a vdW electrode transfer method, in which prepatterned metal films are transferred using a micromanipulator-assisted probe tip. This technique enables precise alignment of a metal film without the need for high-temperature processing or lithographic patterning. After electrode transfer, wire bonding was performed, followed by epoxy encapsulation.

Figures d and f present photographic and microscopic images of the MoS₂ exfoliation and transfer process. Figure d shows the MoS₂-on-epoxy sample with the inset displaying the donor substrate after exfoliation. Figure e provides an optical microscopy image of the donor substrate, while Figure f shows the corresponding image of the recipient epoxy substrate. These images reveal a clean surface of the donor substrate and a continuous MoS₂ film on the epoxy. Atomic force microscopy (AFM) measurements further confirm the successful transfer, indicating a uniform surface morphology across the MoS₂ region (Figure S7). This process is enabled by engineered interfacial forces: strong adhesion between epoxy and MoS₂, combined with the relatively weak vdW interaction between MoS₂ and the donor substrate, allows the monolayer to detach cleanly from the substrate. Figure g presents an optical microscopy image of an integrated SID demonstrating the successful assembly of the active device.

Efficient coupling between the evanescent field and the integrated device requires a careful design that accounts for both optical and electrical considerations. Figure a illustrates two competing optical factors that need to be simultaneously optimized: the insertion efficiency (η_insertion) at the device interface and the mode overlap factor (η_overlap) with the evanescent field. The thickness of the active material, t, is a key design parameter for balancing these two factors. Reducing t decreases the optical mode overlap with the active materials, leading to a reduction in η_overlap. Conversely, increasing t introduces scattering of the guided mode, thereby degrading the η_insertion. Therefore, identifying the optimal device thickness is critical for maximizing the overall device efficiency (η_device). From an electrical standpoint, implementing advanced fabrication strategies such as vdW electrode transfer and epoxy encapsulation requires experimental validation in terms of electrical performance.

3.

Optical and electrical design of WSe ₂ -based photodetector. a) Illustration of evanescent coupling between the optical mode and an integrated device. The E _y (z) field profile (yellow dashed curve) and the corresponding optical mode overlap, η_overlap (vertical yellow arrow), and insertion efficiency, η_insertion, are shown. b) FDTD simulation of the guided mode profile in the 2D ultrathin waveguide before device integration (top), with an integrated 10 nm thick device (middle), and with an integrated 80 nm thick device (bottom). Propagation direction from left to right of the image (scale bar, 5 μm). c) Calculated optical mode overlap factor (black) and insertion loss (red) as functions of device thickness. d) Calculated η_device as a function of device thickness. The shaded region (pink) indicates the optimized thickness range. e) Current–voltage (I–V) curve of the device under dark conditions (black) and 750 nm free space light illumination (yellow). The inset shows an optical microscopy image of the fabricated SID-on-epoxy (scale bar: 20 μm). g) Time-dependent current signal of the device under 750 nm light source, toggled on and off every 10 s.

We determine the optical device thickness by performing finite-difference time-domain (FDTD) simulations on a model in which a device is seamlessly placed on a 2D ultrathin waveguide. Figure b shows the electric field intensity distributions of light propagating (from left to right) along the MoS₂ monolayer waveguide under three different conditions: without an integrated device (top), with a 10 nm thick device (middle), and with an 80 nm thick device (bottom). In the absence of a device, the guided mode remains continuous over the entire propagation length. When a 10 nm thick device is introduced, scattering is observed at the device edges; however, the majority of the guided energy remains confined and propagates beyond the device. In contrast, the 80 nm thick device induces significant insertion loss, resulting in a substantial reduction of the transmitted guided mode.

Figure c presents a quantitative analysis of the trade-off between optical mode overlap and insertion loss, showing the simulated optical mode overlap factor (η_overlap; black line) and insertion loss (η_insertion ; red line) as functions of device thickness. The mode overlap factor was obtained from FDE simulations by calculating the fraction of the guided mode confined within the device relative to the total mode. Insertion loss was quantified by comparing the electric field intensities before and after the guided mode propagated through the device. As shown in Figure c, insertion loss increases with a greater overlap factor, indicating an inverse relationship. We define the overall device efficiency as η_device = η_overlap × η_insertion to determine the optimal device thickness. Figure d shows the calculated η_device as a function of device thickness, which reaches a maximum between 10 and 20 nm. We selected a device thickness of 20 nm for subsequent experiments, which balances sufficient optical interaction with minimal insertion loss.

To implement the optimized design, we fabricated a photodetector as a model SID, using WSe₂ as the active material due to its narrow bandgap and well-established optoelectronic performance. , A 20 nm thick WSe₂ flake was dry-transferred onto a flat epoxy substrate, and two Au electrodes were transferred at each end. This device serves as a testbed for validating the electrical viability of the SID prior to optical coupling. We measured the current–voltage (I–V) characteristics under both dark and illuminated conditions using free-space 750 nm illumination at normal incidence. As shown in Figures e and f, the dark current was measured to be 49 pA under a 1 V bias. When exposed to a weak 3 nW light source focused onto a 20-μm diameter spot, the device generated a photocurrent of 0.5 nA. These observations confirm that the WSe₂ photodetector exhibits robust electrical behavior with a clear photocurrent response.

The prefabricated photodetector was subsequently integrated onto the 2D ultrathin waveguide by placing it on the bottom half of the waveguide. The resulting structure, referred to as SID-on-waveguide, consists of a WSe₂ photodetector positioned above the MoS₂ waveguide and encapsulated within an epoxy cladding (Figure S8). To evaluate the functionality of the SID-on-waveguide system, we performed optoelectrical measurements under guided-mode excitation with controlled light polarization. The MoS₂ waveguide exhibits polarization-dependent propagation characteristics: TE modes are efficiently coupled into the waveguide, while transverse magnetic (TM) modes experience minimal confinement.

Figure a shows the I–V characteristics of the photodetector under three conditions: dark (black), TM-polarized excitation (blue), and TE-polarized excitation (red). In both illuminated cases, a 750 nm laser with 34 μW optical power was edge-coupled into the waveguide. At a 0.5 V bias, the measured current increased from 1.43 nA (dark) to 1.94 nA (TM) and further to 18.5 nA (TE). Time-resolved photocurrent measurements under the same conditions, shown in Figure b, further show stable signal generation exclusively under TE excitation. To quantify the selectivity of the photodetector to guided modes, we evaluated the signal-to-noise ratio (SNR), where the signal corresponds to the TE-induced photocurrent (I _TE) and the noise corresponds to the response under TM illumination (I _TM). The SNR is then expressed as the TE-to-TM contrast ratio: (I _TE – I _TM)/(I _TE + I _TM), where both I _TE and I _TM are normalized by subtracting the dark current. Based on the measured values from Figure a, the contrast ratio is approximately 0.942, indicating high polarization selectivity. Figure c shows the temporal response of the device with a rise time (τ_rise) of 7.1 μs and a decay time (τ_decay) of 5.6 μs. Figure d presents the measured responsivity as a function of the optical power. The data follow a power-law dependence, R ∝ P ^–0.54, with a responsivity of 0.087 A/W at 1.8 nW incident power.

4.

On-chip detection of evanescent-field guided light. a) I–V curve of the integrated photodetector under TE (red), TM (blue), and dark (black) conditions. The inset shows a schematic of the on-chip photodetector. b) Time-dependent current signal of the integrated photodetector. c) Response speed of the integrated photodetector. The horizontal dotted lines indicate the 90% (upper) and 10% (lower) levels of photocurrent intensity. d) Responsivity of the integrated photodetector at different TE-polarized guided light optical powers. The inset shows the time-dependent current signal at a power of 1.8 nW. e) Top-view optical microscopy images of the guided waves at TM (top panel) and TE (center) polarization conditions. Optical microscopy image of the guided waves with a laser power of 1.8 nW (bottom), corresponding to the inset of Figure d (scale bar, 40 μm). The dashed boxes represent the electrode positions, and the dashed line on the left corresponds to the position of the monolayer MoS₂.

To visualize guided-mode propagation across the SID we performed top-view dark-field optical imaging of the SID-on-waveguide structure (Figure S9). Figure e presents images obtained by edge-coupling light into the waveguide under different polarization conditions. In the top panel, TM-polarized light with an incident power of 34 μW produces vertically elongated scattering signals near the electrode region but no clear evidence of guided propagation is observed. The middle panel shows the image under TE-polarized excitation at the same optical power. A continuous horizontal trace extends from the input edge across the device, indicating a confined optical mode propagating through the SID. While some scattering remains visible near the electrode interface, the guided mode is sustained throughout the device. The bottom panel displays the response under TE-polarized excitation at an input power of 1.8 nW. Although only a faint scattering signal is observed near the photodetector (black arrow), the device still generates a clear photocurrent (inset of Figure d). These results confirm the successful integration of SIDs on the 2D ultrathin waveguide platform. The SID-on-waveguide configuration preserves robust photodetection capabilities while enabling highly selective, polarization-sensitive detection of guided optical fields.

We present a 2D ultrathin waveguide platform that supports evanescent-field-dominant light propagation using an atomically thin monolayer MoS₂ core. This architecture enables the integration of active devices on the waveguide surface without requiring subwavelength alignment or modification of the core structure. As a model system, we demonstrated a seamlessly integrated WSe₂ photodetector that achieves nanowatt-level sensitivity, microsecond-scale response times, and strong polarization selectivity. The ability to couple guided modes with SIDs, while maintaining a structurally simple waveguide geometry, opens opportunities for incorporating modulators, emitters, and nonlinear optical elements without the need to redesign the waveguide structure. Additionally, the evanescent-field guiding mechanism allows photonic functionality to be defined by the spatial arrangement of devices on the waveguide surface rather than being fixed through lithographic processing. This modularity enables devices to be placed, repositioned, or swapped without altering the underlying optical path, offering a reconfigurable design strategy that resembles the assembly of optical components on an optical table. Such flexibility supports the rapid prototyping and customization of application-specific photonic systems, all built upon a universal 2D waveguide backbone. As photonic integrated circuits continue to scale in complexity and density, the proposed 2D ultrathin waveguide platform offers a versatile and scalable strategy for unifying passive waveguiding with diverse active functionalities.

The data that support the findings of this study are available within the paper and its Supporting Information, or from the corresponding author on reasonable request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.nanolett.5c03190.

• Materials and methods; 2D ultrathin waveguide and seamlessly integrated device (SID) fabrication (epoxy transfer, vdW electrode transfer, wire bonding); optical measurement setups for mode-size and waveguide characterization; simulation procedures and parameter definitions; evanescent-field and optical-overlap factor calculations; wavelength-dependent propagation loss and cladding-index mismatch analyses; AFM/DIC surface characterization; dark-field top-view propagation imaging; confinement factor according to MoS₂ layer number (PDF)

§.

S.K. and C.H. contributed equally to this work.

The authors declare no competing financial interest.