Unveiling α‐MoO₃ as a High‐Index, Low‐Loss Anisotropic Material: Bridging the Gap From Mid‐Infrared to Ultraviolet Nanophotonics

ABSTRACT

Modern nanophotonics demands light manipulation at a subwavelength scale, a capability historically constrained by the isotropic dielectric response of conventional optical semiconductors. Here, we establish α‐MoO₃ as a high‐index, low‐loss anisotropic material that bridges the gap between mid‐infrared polaritonics and short‐wavelength nanophotonics. Integrating spectroscopic ellipsometry, first‐principles calculations, and near‐field optical microscopy, we determine its broadband 250–1700 nm anisotropic optical constants and combine them with a phonon model in mid‐infrared to 30 µm. While the mid‐infrared phonon polariton properties of α‐MoO₃ are well established in the literature, the present work provides the first experimental characterization of its complete anisotropic dielectric tensor in the ultraviolet‐to‐near‐infrared regime, conceptually bridging these two spectral domains in which this material finds application. The pronounced out‐of‐plane birefringence 0.35 suppresses evanescent crosstalk beyond competing van der Waals (vdW) waveguide materials, and the high in‐plane anisotropy enables zero‐order waveplates at wavelengths as short as 325 nm. As a practical demonstration, we prepared nanowaveguide which exhibits wavelength‐selective light guidance at visible range. Together, these results reposition α‐MoO₃ from a mid‐infrared specific material to a versatile wide‐bandgap platform bridging mid‐infrared polaritonics and ultraviolet‐visible nanophotonics.

α‐MoO₃ is established as a high‐index, low‐loss anisotropic platform bridging mid‐infrared polaritonics and ultraviolet‐visible nanophotonics. Spectroscopic ellipsometry and calculations reveal refractive indices exceeding 3.9, with the largest in‐plane birefringence among transparent layered crystals at 425–500 nm, enabling zero‐order UV waveplates, visible nanowaveguides, and crosstalk‐suppressed photonic integration.

1. Introduction

The emergence of van der Waals (vdW) materials has significantly influenced the development of modern photonics. While cubic semiconductors like silicon [1], gallium phosphide [2], and titanium dioxide [3] serve as foundational platforms for modern optics, their isotropic dielectric response restricts the degree of freedom in controlling light polarization and electromagnetic field localization. Early vdW photonics focused on high‐symmetry hexagonal lattices [4, 5, 6, 7, 8, 9, 10, 11, 12], which, despite excellent integration capabilities [13, 14], remain optically isotropic within their basal planes. The scope of van der Waals photonics has expanded to encompass low‐symmetry crystal systems (orthorhombic, monoclinic, triclinic), where a dielectric tensor (ε_xx ≠ ε_yy ≠ ε_zz ) facilitates the realization of hyperbolic phonon polaritons [15, 16, 17] and direction‐dependent refractive indices [18, 19, 20]. This intrinsic anisotropy enables various functionalities, such as biosensing [21, 22], waveplates [23, 24, 25, 26], chirality‐preserving mirrors [7, 27, 28], and sub‐diffractional light guiding [7, 29, 30, 31, 32] without the requirement for complex lithographic patterning. However, many highly anisotropic materials are characterized by a fundamental trade‐off between dissipative losses in the visible (VIS) spectral range and relatively low refractive indices, which can restrict effective optical confinement [33].

While alpha‐phase of molybdenum trioxide (α‐MoO₃) is well‐documented as a platform for mid‐infrared (MIR) hyperbolic phonon polaritons [34, 35], its optical behavior in the short‐wavelength limit has received less systematic attention. Despite its wide bandgap (E _g > 2.9 eV) and recent advancements in the large‐scale synthesis of thin films with out‐of‐plane anisotropy [36, 37, 38], comprehensive experimental data on its anisotropic complex refractive index remain limited. Consequently, currently available optical constants often rely on isotropic thin‐film approximations [36, 37] or theoretical predictions [39], which constrain the precise development of α‐MoO₃‐based polarization‐sensitive devices in the visible regime [40, 41, 42].

In this work, we present α‐MoO₃ as a high‐index platform that connects mid‐infrared polaritonic functionalities with visible‐range all‐dielectric nanophotonics. By integrating spectroscopic ellipsometry on exfoliated flakes with first‐principles calculations and micro‐reflection spectroscopy, we resolve the broadband (250–1700 nm) anisotropic optical response of α‐MoO₃. The material exhibits a high refractive index (n > 2.4 in the VIS and n > 3.8 in the ultraviolet (UV) regions) alongside a broad transparency window and pronounced in‐ and out‐of‐plane birefringence reaching 0.1 and 0.35, respectively. These results are further corroborated by scattering‐type scanning near‐field optical microscopy (s‐SNOM), which allows for the direct visualization of directional waveguide modes. A comparative analysis with other van der Waals materials confirms that α‐MoO₃ enables ultra‐compact waveguide architectures with suppressed crosstalk. Thus, we experimentally cover the 250–1700 nm range and show that the same material possesses exceptional optical properties across a far broader spectral window, from the ultraviolet to the near‐infrared, thereby connecting its established MIR identity with a class of applications—visible and UV integrated nanophotonics.

2. Results

2.1. Structural Anisotropy of α‐MoO₃

To establish a rigorous framework for the broadband optical characterization of α‐MoO₃, we first analyze its crystal structure and determine the crystallographic orientation of the exfoliated flakes. α‐MoO₃ crystal lattice has the orthorhombic space group Pmcn (No. 62) and lattice parameters a = 3.70 Å, b = 3.96 Å, and c = 13.86 Å [43, 44, 45]. The crystal structure consists of distorted MoO₆ octahedra, where each Mo⁶⁺ ion is coordinated by six oxygen atoms [46]. These octahedra share corners and edges to form sheets stacked along the c‐axis, held together by weak van der Waals interactions that facilitate mechanical exfoliation. In our experimental configuration, the a‐ and b‐axes lie within the plane of the flake, while the c‐axis is perpendicular to the layers (Figure 1a).

FIGURE 1.

(a) Crystal structure of α‐MoO₃. (b) Parallel‐polarized micrographs of α‐MoO₃ flake at orientations from 0° to 90°. (c) Experimental and (d) calculated Raman spectra along a and b axes respectively. Horizontal dashed lines indicate the orientation of crystallographic axes. Vertical dashed lines indicate the silicon mode. (e) Experimental and (f) calculated Raman spectra along the two crystalline axes (a and b). Vertical grey gradient lines indicate A _g, B _1g, and silicon mode.

This atomic‐scale structural anisotropy manifests macroscopically as a birefringence, clearly visible in polarization‐resolved optical microscopy. Figure 1b displays optical micrographs of a mechanically exfoliated α‐MoO₃ flake on a Si/SiO₂ substrate. As the sample is rotated from 0° to 90° relative to the incident linear polarization, the flake exhibits a dramatic color shift, providing visual evidence of the stark contrast in the refractive index along the principal in‐plane axes.

To unambiguously verify the crystallographic orientation and phase purity, we performed angle‐resolved polarized Raman (ARPR) spectroscopy measurements and ab initio calculations (see Methods for details). The Raman spectrum is dominated by the high‐frequency modes associated with terminal Mo─O bonds, while the lower‐frequency region reflects collective vibrations of the octahedral framework and interlayer interactions [14, 47]. These modes correspond to the Raman‐active vibrations allowed by the D2h point group symmetry. The experimental and theoretical ARPR spectroscopy heatmaps are shown in Figure 1c,d, together with corresponding spectra along the a and b axes (Figure 1e,f). The experimental results show excellent agreement with our theoretical predictions, confirming the high crystalline quality of the samples. Consistent with Raman spectroscopic characterization and previous reports [14, 47, 48], A _g vibrational modes at 156 and 818 cm⁻¹ and B _1g mode at 282 cm⁻¹ are used to identify the crystallographic orientation of α‐MoO₃. The intensity of these modes depends on the alignment of the crystallographic axes, enabling lattice indexing for the integration of α‐MoO₃ into anisotropic optoelectronic structures.

2.2. Anisotropic Dielectric Function of α‐MoO₃

Having established the crystallographic orientation of the flakes, we next quantify the anisotropic dielectric function ɛ = ɛ ₁ + iɛ ₂ of α‐MoO₃ across the ultraviolet‐near infrared (NIR) spectral range (250–1700 nm), where ɛ ₁ and ɛ ₂ denote the real and imaginary parts of the dielectric function, respectively. Moreover, the principal optical axes [18] have been confirmed utilizing Mueller matrix measurements (see Methods for details). We performed spectroscopic ellipsometry on mechanically exfoliated flakes transferred onto a silicon substrate. The samples were rotated to align the incident plane of the light parallel to the crystallographic a and b axes. Configuration allows to minimize cross‐polarization, therefore leading to the measurement of standard ellipsometric parameters Ψ and ∆ instead of the Mueller matrix. The resulting parameters of the flake (thickness = 483 nm) at different incident angles are presented in Figure 2a,b. The validity of Ψ, Δ parameterization was further confirmed by performing Mueller matrix measurements for 11 elements (M_12, M₁₃, M₁₄, M₂₁, M₂₂, M₂₃, M₂₄, M₃₁, M₃₂, M₃₃, M₃₄) at visible range and verifying that the off‐diagonal elements M₁₃, M₁₄, M₂₃, M₂₄, M₃₁, M₃₂ were close to zero at the assigned axis orientations (details are presented in Supporting Note S1). The solid lines denote the measured data, while the dashed lines represent the fit curves (see Methods for details). Distinct oscillatory behavior is observed in both Ψ and ∆ spectra, indicating strong optical anisotropy and multiple interband transitions.

FIGURE 2.

Measured and calculated ellipsometric parameters Ψ and Δ of α‐MoO₃ flake along (a) a‐axis and (b) b‐axis at three incident angles: 45°, 50°, and 55°. (c) Real and (d) imaginary parts of the obtained dielectric function. The inset in panel (d) represents an optical image of the exfoliated α‐MoO₃ flake, acquired during the ellipsometric measurements with the indicated region of interest ROI, a‐ and b‐axes. The scale bar corresponds to 20 µm.

The dielectric function components along all crystalline plane axes are shown in Figure 2c,d. The real and imaginary parts of the dielectric function are presented to demonstrate dielectric behavior and zero loss starting from 423 nm (k = 0.01). This behavior agrees with the wide bandgap nature of α‐MoO₃ reported earlier with E _g ∼2.9 eV [49, 50]. Moreover, the dielectric function reveals significant anisotropy between the in‐plane a‐ and b‐axes and the out‐of‐plane c‐axis (details of ellipsometry process are presented in Supplementary Note S2).

2.3. Anisotropic Optical Constants

The extracted complex refractive indices ñ = n + ik along the optical axes are summarized in Figure 3a. To enhance accuracy, we analyzed data from flakes with different thicknesses. This approach enhances the robustness of the fit because we get two times more spectra for the same optical constants. Ψ and ∆ for 539 nm thickness flake is presented in Supporting Note S3. The inset shows the in‐plane and out‐of‐plane birefringence values, where the grey color indicates the transparency region. Notably, the refractive index along a‐axis exceeds 3.9 in UV and 2.6 in VIS in the low‐absorption regime (k < 0.01) at wavelengths above 423 nm. The material exhibits in‐plane birefringence reaching 0.1 in the transparency window. We also performed ab initio calculations of the anisotropic permittivity tensor utilizing GW (see Methods), as shown in Figure 3a. Experimentally measured optical constants and birefringence are in good agreement with the theoretical calculations. Out‐of‐plane anisotropy is considered as

$$\frac{n_a+n_b}{2}-n_c$$

FIGURE 3.

(a) Optical constants of α‐MoO₃ crystal, obtained via spectral ellipsometry and GW calculations. The inset represents the resulting in‐plane (red line) and out‐of‐plane (blue line) birefringence. The shaded region indicates the transparency band. (b) Experimental polarized micro‐reflection map. Dashed lines label crystallographic axes. (c) Reflection spectra along a (blue line) and b (red line) axes. The solid lines show measured spectra. The dotted lines represent calculated spectra. (d) The comparison of refractive indices of α‐MoO₃ with other photonic materials in their respective transparency bands. The combined measured (e) refractive index and (f) extinction coefficient with adapted previously reported data [54] under open access. Solid lines correspond to measured data in this work. Dotted lines correspond to adapted literature data.

For additional validation of our ellipsometry measurements, we conducted polarized normal incidence reflectance measurements, shown in Figure 3b,c. Representative reflectance spectra maps acquired by rotating the sample stage in the parallel configuration of polarizers are presented in Figure 3b. For an extended wavelength range, the corresponding reflectance spectra are detailed in Supporting Note S4. To further corroborate our findings, we acquired the theoretical reflectance from the ellipsometry measurements. Figure 3c illustrates the excellent agreement between the experimentally measured and calculated reflectance spectra along the principal optical axes. This agreement also provides unambiguous confirmation for the accuracy of extracted optical constants.

The optical profile of α‐MoO₃ places it within a class of high‐index, low‐loss layered materials exhibiting a refractive index comparable to widely utilized and potentially scalable benchmarks such as MoS₂ [9, 10], hBN [7], TiO₂ [3, 51], calcite [51], and gallium chalcogenides (GaP [2], GaS [52]), as shown in Figure 3d. While α‐MoO₃ does not claim record‐breaking absolute refractive index values, its true distinction, and the motivation for its use in nanophotonics emerges from in‐plane and out‐of‐plane anisotropies. Particularly, it demonstrates the largest in‐plane anisotropy among transparent layered materials within the 425–500 nm spectral range as shown in Supporting Note S5. The pronounced in‐plane optical anisotropy renders α‐MoO₃ a highly promising platform for waveplate applications. The use of strongly anisotropic materials in waveplates enables unique functionality yet results in polarization‐dependent transmittance (linear dichroism) and wavelength selectivity. Therefore, to convert linearly polarization into circular, the incident beam's polarization plane should form an angle of $arctan\sqrt{\frac{T_y}{T_x}}$ with the optical principal axis y. The optical losses and dichroism of α‐MoO₃ waveplates with various thicknesses were systematically evaluated in Supporting Note S6. Notably, the waveplates considered here are substantially thinner than those reported previously [53]. In addition, α‐MoO₃ exhibits a large out‐of‐plane birefringence, reaching values up to 0.35 in the visible spectral range, which highlights its strong potential for waveguiding and integrated photonic applications.

To provide a comprehensive view of the dielectric response from the UV to the MIR (250–30 000 nm), the measured data were combined with previously reported [54] infrared dielectric function. The two datasets were merged solely for broadband visualization. Care was taken to ensure spectral consistency and physical continuity at the stitching boundary (at 1700 nm), where excellent agreement was observed. No artificial fitting procedures were involved beyond ensuring smooth spectral overlap (details in Supporting Note S7). This broadband representation enables systematic analysis of dispersion, optical anisotropy, and polaritonic behavior across multiple spectral regimes. The consolidation of these datasets establishes α‐MoO₃ as a versatile platform for nanophotonics. Moreover, we compared optical constants from different reports, including ab initio calculations [39] and thin films [36, 37] in Supporting Note S8.

The confluence of a high refractive index, negligible dissipative losses, and dual‐axis birefringence transforms α‐MoO₃ into a unique medium for light manipulation. This combination is advantageous for waveguiding applications, where the material's anisotropy can be leveraged to achieve direction‐dependent cut offs of fundamental TE and TM modes [29]. When coupled with the recent progress in the large‐scale synthesis of α‐MoO₃ films [36, 37, 38], these properties position the material as a promising candidate for short‐wavelength integrated photonic circuits.

2.4. Near‐Field Measurements and Waveguide Characteristics of α‐MoO₃

To investigate the waveguiding characteristics of α‐MoO₃ in the VIS to NIR spectral range, we employed scattering‐type scanning near‐field optical microscopy. The experimental layout and sample architecture are illustrated in Figure 4a,b. α‐MoO₃ flakes of two distinct thicknesses: approximately 260 and 850 nm, were mechanically exfoliated onto silicon dioxide substrates, comprising a 285 nm buried oxide layer. Illumination was provided via focused laser sources at discrete wavelengths (710, 850, and 1000 nm).

FIGURE 4.

(a) Schematics of the s‐SNOM experimental configuration for α‐MoO₃ sample on top of the SOI substrate. (b) Optical micrograph of the studied flake with a thickness of 260 nm. (c) Simple waveguiding demonstration at an incident wavelength of λ = 710 nm. The waveguides are excited from the top of the image. Scattered light can be observed on the other side. (d) Amplitude Amp(E) and phase Arg(E) of the electric field at the incident wavelengths of λ = 850 and 1000 nm. (e) Line profiles taken from (d). (f) Fourier transform amplitude of the complex near‐field signal presented in (e). Transfer‐matrix calculations for propagating modes in α‐MoO₃ for flake thicknesses of (g) 260 nm and (h) 848 nm. Blue and green dots correspond to experimental data for TM₀ and TE₀ modes, respectively.

Waveguiding in α‐MoO₃ was first visualized at λ = 710 nm using far‐field optical imaging (Figure 4c). In this experiment, a focused laser beam was directed onto the edge of elongated α‐MoO₃ flakes, while scattered light was monitored at the opposite end of the crystal. The appearance of a bright scattering spot at the distant flake edge indicates that the incident light couples into modes propagating along the flake before being scattered back into the far field.

To probe the optical response in greater detail, we performed mapping of the complex near‐field signal at longer wavelengths (see other wavelength images and flakes measurements are presented in Supporting Note S9). Figure 4d shows representative near‐field amplitude and phase images acquired over 260 nm thick flake. Multiple sets of interference fringes emanate from flake edges, with spatially periodic phase fronts clearly visible across both polarizations and thickness regimes. The mechanism underlying fringe formation was further elucidated via line profile extraction perpendicular to the flake boundary (Figure 4e). Both amplitude and phase traces exhibit oscillatory decay with well‐defined spatial frequencies, consistent with interference between counter‐propagating in‐plane guided waves. Figure 4f shows the result of complex Fourier transforms of these profiles. The zero‐frequency peak in the FFT corresponds to the background signal, and other smaller peaks correspond to the transverse magnetic and transverse electric modes scattered on opposite edges of the sample. In the s‐SNOM experiment, the spatial frequencies of the observed fringes are shifted from modal wavenumbers. Consequently, the TM₀ modes propagating in opposite directions produce fringes of two very different apparent effective indices as seen from their respective FFT peaks in Figure 4f. The effective indices of waveguide modes are restored as follows:

$$n_{\mathrm{eff}}=n_{\mathrm{SNOM}}\cos \left(\beta \right)+\cos \left(\alpha \right)\sin \left(\beta +\theta \right)$$

where α and θ are the incident angles with respect to the horizontal plane and the sample edge, respectively, n _eff is the effective index of the mode propagating inside the planar α‐MoO₃ waveguide, n _SNOM is the effective index of fringes obtained by the FFT of near‐field profiles and β is the angle between the waveguide mode wavevector and the normal to the sample edge, which satisfies the following momentum conservation equation:

$$\sin \left(\beta \right)=\frac{\cos \left(\alpha \right)\cos \left(\theta \right)}{n_{\mathrm{eff}}}$$

We compared the extracted effective indices to optical mode dispersion relations computed via transfer‐matrix simulations incorporating the full dielectric tensor of α‐MoO₃ (Figure 4g,h). Experimental data points, extracted across the two flake thicknesses and three excitation wavelengths, align closely with simulated TE₀ and TM₀ branches. The agreement between experimental and theoretical dispersion characteristics underscores the reliability of our near‐field approach for characterizing waveguiding in α‐MoO₃. Moreover, it also verifies the dielectric tensor retrieved from far‐field experimental methods. The fully tabulated anisotropic complex refractive indices spanning the ultraviolet to near‐infrared spectrum are collected in Supporting Note S10.

2.5. Waveguide Simulations

To evaluate the potential of α‐MoO₃ for dense vdW photonic integration, we performed numerical simulations of waveguides based on the extracted optical constants. While the refractive index n of α‐MoO₃ is lower than that of other high‐index materials such as GaS, MoS₂, and GaP (as shown in Figure 3d), its pronounced out‐of‐plane birefringence offers advantages for light manipulation. We compared α‐MoO₃ with transparent, anisotropic high‐index vdW materials, including MoS₂, GaS, and hBN. All simulations were performed at the shortest wavelength operation in a low‐loss regime. A schematic illustration is presented in Figure 5a. Waveguide confinement was analyzed using square‐cross‐section waveguides fully embedded in SiO₂ (Figure 5b). For each material, the waveguide width (w) was chosen such that the exponential decay constant (χ) of the evanescent field into the SiO₂ cladding was fixed at χ = 2.22 µm⁻¹, providing a consistent basis for comparing modal confinement across materials operating at different wavelengths. The mentioned decay constant χ is defined as $\sqrt{n_{\mathrm{eff}}^2-n_s^2}/\lambda$, where n _eff is the effective index of the mode, n _s is the refractive index of the waveguide cladding, and λ is the operating wavelength.

FIGURE 5.

(a) Schematic illustration of waveguides from different transparent vdW materials on a SiO₂ substrate, highlighting relative dimensions under comparable confinement conditions. (b) Required waveguide width as a function of operating wavelength for different vdW materials. All waveguides have square cross sections and are fully embedded in SiO₂. (c) Schematic of the proof‐of‐concept waveguiding experiment: an α‐MoO₃ nanowaveguide of length L = 20.1 µm on a SiO₂(285 nm)/Si substrate. (d) AFM image of the waveguide with a width of 530 nm and a height of 70 nm. (e) Schematic of the waveguide cross‐section. (f and g) Optical images of the waveguide under excitation of the input facet at λ = 532 nm (f) and λ = 785 nm (g). A bright output spot is observed at 532 nm, while no output appears at 785 nm. (h) Simulated transmission spectra of the TE₀ and TE₁ modes of the waveguide. Insets show the mode field distributions at λ = 532 and 785 nm. (i) Comparison of out‐of‐plane anisotropy for layered van der Waals high‐index materials in transparency regions. (j) Crosstalk length or full power transfer length versus separation distance between adjacent waveguides.

To corroborate these simulations experimentally, we prepared a proof‐of‐concept device: an α‐MoO₃ nanowaveguide with a width of 530 nm, a height of 70 nm, and a length of L = 20.1 µm on a SiO₂ (285 nm)/Si substrate (Figure 5c–e). A laser beam was focused onto the input facet of the waveguide, and the scattered light was collected at the opposite facet. At λ = 532 nm, a bright output spot appears at the distant facet (Figure 5f), evidencing guided propagation along the entire structure. By contrast, at λ = 785 nm no output is observed (Figure 5g). This wavelength‐selective response is reproduced by transmission simulations of the studied geometry (Figure 5h): at 532 nm, the waveguide supports a well‐confined TE₀ mode, whereas at longer wavelengths the mode approaches cutoff and delocalizes from the waveguide core, suppressing the transmission at 785 nm. Thus, even a sub‐100‐nm‐thick α‐MoO₃ nanowaveguide routes visible light over tens of micrometers, providing a device‐level validation of the extracted optical constants and waveguide simulations.

Although α‐MoO₃ possesses a lower refractive index than GaS and MoS₂ (Figure 3d), its high intrinsic birefringence facilitates enhanced light manipulation. For comparison, the out‐of‐plane anisotropy of these materials in their respective transparency regions is summarized in Figure 5c. GaS exhibits a higher refractive index, but lower anisotropy compared to α‐MoO₃, whereas hBN displays a lower refractive index alongside higher anisotropy. MoS₂ surpasses α‐MoO₃ in both parameters, but its operation is restricted to longer wavelengths. The integration density of photonic circuits is ultimately limited by evanescent coupling between neighboring waveguides. To quantify this effect, the full power transfer (crosstalk) length (L _ct) was calculated as a function of the separation distance (d) between adjacent waveguides (Figure 5j). In these simulations, the waveguides were positioned on a SiO₂ substrate with air cladding, and both the width and height of the waveguides were optimized for each separation distance to achieve the highest crosstalk distance in the single‐mode regime. All calculations were performed at wavelengths where the extinction coefficient satisfies k < 0.01, corresponding to the onset of absorption for each respective material.

The results demonstrate that for a fixed acceptable crosstalk threshold, α‐MoO₃ allows for smaller waveguide separations and higher achievable integration density. These findings highlight the suitability of α‐MoO₃ for compact, low‐loss, and densely integrated planar vdW photonic circuits. This capability allows for a substantial reduction in the physical footprint of waveguide structures while maintaining minimal signal interference, positioning α‐MoO₃ as a building block for ultra‐compact, high‐density planar photonic circuits in the visible regime.

3. Discussion

The central finding of this work is a fundamental repositioning of α‐MoO₃ within the photonic landscape. Prior to this study, the material was considered almost through its mid‐infrared identity [42, 54]. The ultraviolet and visible ranges had remained uncharacterized despite being the spectral window where α‐MoO₃ reaches its highest refractive indices and strongest birefringence. Our results reveal that the same crystal supporting phonon polaritons at 10 µm simultaneously delivers n > 3.9 in the UV, n > 2.6 in the visible, and the largest in‐plane birefringence among transparent layered materials at 425–500 nm. GW first‐principles calculations, applied here for the first time to α‐MoO₃ with spin‐orbit coupling, correctly reproduce the optical constants. The unified broadband dielectric tensor spanning 250 nm to 30 µm, constructed by combining our new UV‐to‐NIR data with established MIR phonon models [54], is the first continuous optical portrait of this material from the ultraviolet to the terahertz.

These findings open application domains inaccessible without the UV optical constants. The monotonically increasing in‐plane birefringence enables zero‐order quarter‐wave plates at 325–375 nm using sub‐200‐nm‐thick flakes, extending visible‐frequency demonstrations into the near‐UV. As a practical demonstration we prepared nanowaveguide (70 nm × 530 nm) which demonstrates wavelength‐selective guidance at visible range with quantitative agreement between experiment and simulation, confirming that the extracted dielectric tensor is accurate at the device level. From an integration standpoint, the large out‐of‐plane birefringence (0.35) suppresses evanescent crosstalk beyond competing van der Waals waveguide materials at equivalent waveguide separations, making α‐MoO₃ architecturally advantageous for densely integrated visible photonic circuits. Together with recent advances in large‐scale α‐MoO₃ synthesis [36, 37, 38], these results demonstrate the feasibility of wide‐bandgap van der Waals photonic circuits spanning the ultraviolet through the near‐infrared.

4. Experimental Section

4.1. Sample Preparation

Samples with α‐MoO₃ flakes were prepared by mechanically exfoliating the bulk α‐MoO₃ crystal (acquired from 2D Semiconductors, Inc.) onto the desired substrate (silicon, Si/SiO₂, glass). An optical microscope Nikon ECLIPSE LV150N was employed to visually identify exfoliated flakes of suitable size.

4.2. First‐Principles Calculations

Calculations of optical and dynamical properties were performed using the Vienna ab initio package (VASP) [55, 56]. For these calculations, we used experimental crystal structure parameters of α‐MoO₃ crystal [45]. To calculate optical properties, we first obtained the one‐electron ground‐state wavefunctions from a single‐shot DFT calculation. Crucially, spin‐orbit coupling (SOC) was explicitly included in this step to accurately account for relativistic effects associated with the heavy transition metal (Mo) atoms. These wavefunctions were then used to initialize the GW run. Next, we calculated the imaginary and real parts of the frequency dependent dielectric function within the many‐body GW approximation and derived the refractive indices and extinction coefficients of the material. The cutoff energy for the plane wave basis was set to 550 eV, while the first Brillouin zone was sampled with a 𝛤‐centered 6 × 2 × 6 grid. The exchange correlation effects were described with a generalized gradient approximation (Perdew–Burke–Ernzerhof functional, PBE‐xc), and the behavior of the wavefunctions in the core region was reconstructed with the projector augmented wave pseudopotentials of a GW type. We used 504 energy bands in the DFT and GW calculations.

The second‐order force constants were calculated for a 3 × 3 × 1 (144 atoms) supercell with an atomic displacement value of 0.01 Å using a non‐analytical term correction to the dynamical matrix. K‐meshes of 13 × 12 × 4 and 2 × 2 × 2 divisions, respectively, were used for the primitive cell and the α‐MoO₃ supercell. The convergence parameter of the self‐consistent electron cycle (EDIFF) was 10⁻⁸ eV. The PBE xc‐functional, together with Grimme‐D3 correction, was used for all calculations of Raman spectra. The kinetic energy cutoff for the basis set of plane waves was chosen to be 550 eV. DFPT was used to calculate the static dielectric tensor. Intensities were computed based on the Phonopy‐Spectroscopy code with our modifications [57]. The widths of all computed peaks were chosen to be 5 cm⁻¹.

4.3. Angle‐Resolved Polarized Raman Spectroscopy

ARPR‐spectroscopy measurements of the samples were performed using an alpha300 RA confocal Raman‐AFM microscope (WITec, Ulm, Germany). Raman spectra were acquired with the 100× Zeiss EC Epiplan‐Neofluar objective (0.9, Differential Interference Contrast (DIC)). A 532 nm laser with a power of 1 mW was used for sample excitation, yielding a power density of approximately 14 mW/µm2 (a spot diameter of ∼300 nm). ARPR spectra were recorded at 5° intervals over a complete 360° rotation (72 angles were recorded in total) using the parallel configuration of the ARPR system (i.e., with the polarizer aligned parallel to the analyzer). A grating with 1200 lines/mm was employed, and the backscattered light was detected using a back‐illuminated deep‐depletion Charge‐Coupled Device (CCD) cooled to −60°C, yielding a registered spectral resolution of approximately 0.5 cm⁻¹. Each spectrum was acquired with an integration time of 5 s per angle. Spectral processing was performed using WITec Project 7 software.

4.4. Spectroscopic Ellipsometry

To analyze the anisotropic in‐plane optical constants of α‐MoO₃ flakes, we utilized the Accurion EP4 imaging spectroscopic ellipsometer (Park Systems GmbH, Göttingen) in the rotating compensator mode. The measurements of the flakes were performed with the plane of incidence aligned with the crystallographic a‐axis and b‐axis, respectively, which allowed for quasi‐isotropic optical recording of Ψ and Δ. The orientation of crystallographic axes of the flakes was determined and confirmed using Mueller matrix measurements at 400, 650, and 900 nm, done with a ∼5° step over a complete 360° rotation of the table where the sample with the flake was installed. Once, for the table rotation value, the measured Mueller indices: (M13, M14), (M23, M24), (M31, M32) are equal to 0 than any other table rotation value, that angle was assigned as the first axis (and that axis orientation was additionally confirmed with the axis orientation using ARPR spectroscopy). The second axis was assigned by rotating the table 90° clockwise (or anticlockwise). The final assignment of the crystallographic a‐axis and b‐axis to the determined axes is done during the modeling step, after the spectroscopic ellipsometry measurements are done. Ellipsometry spectra of Ψ and Δ were recorded in the spectral range from 250 nm to 1700 nm for three incident angles: 45°, 50°, and 55°.

4.5. Polarization‐Resolved Reflectance

The experimental setup for polarization‐resolved reflectance mapping was based on a trinocular microscope (RX50M), which was modified by the polarizer and an analyzer [58]. Data were recorded across the 550–850 nm wavelength range. An 20× magnification objective with a numerical aperture of 0.5 (SOPTOP MPlanFL) was utilized to collect the reflected signal from a spot size of <20 µm. The collected signal was directed from one of the oculars to a splitting system that coupled the light to a camera and a spectrometer (Optosky ATP5020P). An optical fiber with a 100 µm core diameter (Thorlabs M96L02) was used to connect the spectrometer. All measurements were conducted at room temperature. The spectral data were processed, and the in‐plane principal optical axes were identified according to the methods [59].

4.6. Atomic Force Microscopy

The thicknesses of the studied α‐MoO₃ flakes were determined using atomic force microscopy. Measurements were performed in tapping mode with a WITec alpha300 RA system equipped with a NanoWorld ARROW‐FMR probe (resonant frequency 75 kHz; spring constant 2.8 N/m). The scanning speed, scan size, and resolution were tailored to each study area to ensure optimal visualization quality. AFM image processing was conducted using Gwyddion software.

4.7. Scattering‐Type Scanning Near‐Field Optical Microscopy

Near‐field optical measurements were performed using a commercial s‐SNOM system (NeaSNOM, Neaspec GmbH, Germany) operated in reflection mode. The system utilizes an atomic force microscopy (AFM) tip illuminated by a focused illumination source, collecting the scattered near‐field signal for amplitude and phase‐resolved imaging with a spatial resolution of ∼20 nm. A single off‐axis parabolic mirror was employed for both focusing the incident light onto the probe and collecting the backscattered near‐field signal. The probe consisted of platinum‐coated silicon tips (ARROW‐NCPt‐50), oscillating vertically at a frequency of approximately 280 kHz with an amplitude of ∼80 nm. Two distinct illumination sources were used as the excitation sources in two separate experimental series: (1) a tunable monochromatic NIR‐laser (1475–1600 nm) operated at six discrete wavelengths with 25 nm step was utilized for the sample with thickness of ∼850 nm, and (2) a Ti:sapphire (titanium‐doped sapphire) femtosecond laser system (TiC, AVESTA) tunable from 700 nm to 1000 nm, and operated at discrete wavelengths of 710, 850, and 1000 nm, was used for measurements of the sample with thickness of ∼260 nm. Near‐field signals were extracted by demodulating the detector output at higher harmonics (1 Ω–4 Ω) of the tip oscillation frequency; all reported data correspond to the third harmonic (3Ω), effectively suppressing far‐field background contributions. Signal detection was carried out using a pseudo‐heterodyne interferometric scheme, in which the reference beam was phase‐modulated by an oscillating mirror to enable phase‐resolved acquisition of the near‐field amplitude and phase.

4.8. Numerical Calculations for Waveguides

We performed numerical calculations using a previously established method [60], combining COMSOL Multiphysics and MATLAB through a live link. All simulations used a simple 2D geometry in COMSOL as the solver, while MATLAB handled geometry optimization. Specifically, we used a binary search algorithm to determine the waveguide width for a constant χ, and a modified gradient descent algorithm to optimize waveguide widths for maximum crosstalk length at a fixed core spacing.

Supporting Information

Supporting Information contains additional figures and tabulated optical constants.

Author Contributions

Gleb I. Tselikov, Aleksey V. Arsenin, Kostya S. Novoselov, and Valentyn S. Volkov suggested and directed the project. Adilet N. Toksumakov and Nikolay V. Pak prepared samples. Adilet N. Toksumakov, Dmitriy V. Grudinin, Georgy A. Ermolaev, Alexey P. Tsapenko, Anton A. Minnekhanov, Konstantin V. Kravtsov, Nikolay V. Pak., Arslan B. Mazitov, Andrey A. Vyshnevyy, Aleksandr S. Slavich, and Ivan A. Kruglov performed the measurements and calculations and analyzed the data. Adilet N. Toksumakov and Alexey P. Tsapenko wrote the original manuscript. All authors reviewed and edited the paper. All authors contributed to the discussions and commented on the paper.

Funding

No funding was received for conducting this study.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: adma74228‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.