Introduction

Just over a decade ago, the experimental demonstration of low-loss propagating hyperbolic phonon-polaritons (HPhPs) in hexagonal boron nitride (h-BN)1 marked the beginning of a new era in infrared (IR) nanophononics and sub-diffractional waveguiding2. Phonon-polaritons—well known as light propagation modes strongly coupled to atomic vibrations—occur in polar dielectrics and signal intense interaction between light and lattice dynamics. In specific frequency ranges where this coupling is strongest, light is almost entirely reflected by the dielectric, forming the so-called Reststrahlen bands (RBs). When extended to two-dimensional (2D) materials, these RBs became highly valuable for nanophotonics due to their pronounced anisotropy, originating from strong in-plane covalent bonds and weak out-of-plane van der Waals (vdW) interlayer interactions. This anisotropy can be so pronounced that the dielectric permittivity tensors in 2D vdW crystals exhibit opposite signs along directions parallel and perpendicular to the stacked layers. As a result, a unique dispersion relation emerges, where the isofrequency surface becomes a hyperboloid3,4, giving rise to the HPhP designation. Within this framework, HPhP dispersion can be passively tuned by adjusting the number of monolayers1 or by modifying the interlayer spacing through intercalation5. This tunability is enabled by the fact that light propagates within the volume of the 2D crystals, making HPhPs a crucial element for light transport in future nanophononic circuits.

These discoveries in h-BN nanophotonics sparked a surge of interest in identifying new nanomaterials capable of supporting HPhPs. Notable examples include h-BN1, α-V2O55, α-MoO36,7, Bi2Se38,9, SnO210, TeO211, GeS12, HfS2, and HfSe213—all members of the vdW family with active HPhP modes spanning from IR to far-IR. These materials now lead a revolutionary front in nanotechnology, where HPhP energy flow can be precisely tailored. Canalized light in α-MoO314,15,16,17 and α-V2O518 twisted bilayers, topological modulation of polaritons dispersion in h-BN metasurfaces19, lower-symmetry shear HPhPs20, and naturally canalized light in LiV2O521 are all examples of engineered 2D systems expanding the functional landscape for HPhP-based mid-to-far-IR applications. However, this revolution has not yet reached deep THz frequencies (e.g., 0.3–3 THz) and as a result, none of the IR nanophotonics advances described above are currently being explored for long-wavelength HPhPs. This gap can be attributed to both the limited knowledge of 2D materials capable of supporting HPhPs with high quality factors and the experimental challenges associated with operating nanoscale analytical techniques in the THz range.

Here, we introduce lead iodide (PbI2) as a model 2D material that supports extremely confined HPhPs in the deep THz range, with propagation efficiency comparable to benchmark 2Ds in the IR range. We used scattering-type Scanning Near-field Optical Microscopy (s-SNOM)22,23,24,25, an ultra-microscopy technique capable of probing the nano-optical properties of materials26,27, to obtain real-space THz imaging and spectral analysis of 2D PbI2 flakes with excellent amplitude contrast and precise agreement with theory.

Results and discussion

PbI2 is a vdW semiconductor with a direct bandgap of 2.34 eV, atomically organized as a layered CdI2-type structure (space group \({{\rm{P}}}\overline{3}{{\rm{m}}}1\), Fig. 1a). It is also known as the rare mineral Bustamentite, and its most common stacking forms are 2H and 4H. The 4H polymorph typically forms in samples grown from the melt, via precipitation, or through sublimation, whereas the 2H polymorph is generally produced via hydrothermal or sol-gel synthesis. In our case, crystalline PbI2 hexagonal micro-platelets (Fig. 1f, g) were synthesized via a hydrothermal process (“Methods,” Supplementary Information, Section 5), a simple and accessible method for producing high-quality 2D PbI2 crystals. Compared with h-BN, PbI2 exhibits negligible isotopic variability, enabling sustainable large-scale crystal growth and ensuring reproducibility in future device applications. While standard h-BN is typically composed of approximately 20% 10B and 80% 11B, resulting in significant mass variation between standard and isotopically purified h-BN, iodide is monoisotopic, and lead consists of only three naturally occurring isotopes, predominantly 207Pb with minimal variation. Consequently, PbI2 exhibits a natural mass fluctuation of less than ±0.5%, a critical factor for achieving high phonon Q-factors and ensuring devices’ future reproducibility. Due to its composition of heavy elements, PbI2 has been proposed as a material for X-ray and gamma-ray detectors28,29 and more recently as a precursor in the fabrication of highly efficient perovskite solar cell materials, including CH3NH3PbI3, CH(NH2)2PbI3, and CsPbI3. Generally an excess of PbI2 is present in these materials as protrusions30,31 and can serve as a passivation layer32,33. The optical properties of PbI2, including Raman scattering and far-infrared behavior, were first investigated in the 1960–1970s and in subsequent studies34,35,36,37,38. Concerning its optical phonon modes, two are Raman-active (Eg and A1g), and two are infrared-active (A2u and Eu). The Eu mode corresponds to an in-plane vibration (Fig. 1e), excited when the electric field lies within the plane (perpendicular to the crystal c-axis, Fig. 1e inset). We experimentally revisited the Raman spectrum of the 2D PbI2 as shown in Fig. 1d. Besides the expected Raman-active modes Eg and A1g, we observed the infrared-active A2u mode, whose presence is detectable due to its coupling with the A1g mode39,40. These modes correspond to vibrations perpendicular to the plane (parallel to the c-axis, see Fig. 1d inset).The A2u(TO) mode was earlier studied by FTIR reflection measurements41. Additionally, we characterized the Eu(TO) mode applying far-field time-domain THz spectroscopy (TDS) on a large PbI2 crystal in transmission geometry (Supplementary Information, Section 7) and confirmed a strong absorption band centered at 1.58 THz (53 cm−1), as shown in Fig. 1e. As shown in Fig. 1b, c, a strong anisotropy is observed between the phonon frequencies of 1.55 THz (\({\nu }_{{E}_{u}}\)(TO)) and 2.85 THz (\({\nu }_{{A}_{2u}}\)(TO)). As a result, PbI2 behaves as a hyperbolic material of type II in the frequency range of 1.55–3.03 THz (52–101 cm−1, Fig. 1b) and of type I between 3.24 and 3.45 THz (108–115 cm−1, Fig. 1c). A surface phonon-polariton (SPhP) band is observed in the overlapping region between RB-1 and RB-2, where the real part of the permittivity tensor components are both negative (Fig. 1c). Notably, the hyperbolic band type II in PbI2 is rather broad pointing to a very large dielectric anisotropy. As a consequence, PbI2 exhibits enhanced ionicity, a property linked to the Born Effective Charge (BEC) and therefore associated with the degree of electron transfer in chemical bonds and polarization response to atomic displacements42. This particular frequency dependence of the PbI2 permittivity tensor arises from the combination of strong anisotropy and a very broad RB associated with its ionic character. Further in this work, we qualitatively assess ionicity using the Lyddane–Sachs–Teller (LST) relation \({\varepsilon }_{0}/{\varepsilon }_{\infty }={\omega }_{LO}^{2}/{\omega }_{TO}^{2}\)43.

Fig. 1: PbI2 crystal structure, vibrational response, and dielectric properties in the THz range.
Fig. 1: PbI2 crystal structure, vibrational response, and dielectric properties in the THz range.
Full size image

a Monolayer (ML) arrangement of the atomic lattices of PbI2 with top and side views of its unit cell. b Overview of its real part of the permittivity and delimitation of hyperbolic, surface, and Reststrahlen bands. Ranges for the Reststrahlen bands type I (RB-1) and type II (RB-2) are indicated. c Detailed view of the permittivity in the range between 95 and 120 cm−1 comprising hyperbolic and surface phonon-polaritons (SPhP). d Raman spectrum of PbI2 flakes with indication of the active modes. e Far-field THz Time-Domain Spectroscopy (TDS) absorption spectrum of an isolated PbI2 crystal fitted by a phenomenological model (Supplementary Information, Section 7). f Optical microscopy image of the PbI2 crystals transferred to a Si substrate. g Zoomed optical image with fine details of the 2D crystals (area within the white dashed square in f). Scale bars in (f, g) represent 50 and 10 μm, respectively.

From the pioneering studies by Grisel et al.36 and Lucovsky et al.37 in the 1970s, it was evident that the extreme anisotropy in the permittivity of PbI2 would give rise to an ultra-wide RB in the THz range. Figure 1b, c shows the real part of the in-plane (red solid lines) and out-of-plane (blue solid lines) permittivity, calculated using the optical constants reported in refs. 37,38 (see Supplementary Information, Section 3.1). These studies also established that its CdI2-type structure is organized as a lamellar layered material held together by vdW forces. Despite its properties being known for decades, this material is only now being introduced as a promising 2D medium for sub-diffractional THz nanophotonics unveiled by s-SNOM. Figure 2a presents a representative illustration of our THz s-SNOM setup (“Methods” and Supplementary Information, Section 6) used for real-space imaging of HPhPs in PbI2 2D crystals. A ~20 mW THz beam (2.52 THz, 84 cm−1), generated by a CO2-pumped methanol laser, illuminates a long-shaft (70 μm) Pt-coated Si AFM tip (Supplementary Information, Section 8), enabling field confinement at length scales comparable to the tip radius (~40 nm)44,45. The tip-scattered THz signal is detected by a Schottky diode in a homodyne scheme, with background suppression achieved via established25 higher-harmonics demodulation. The top-right inset of Fig. 2a shows an experimental near-field image of a 655 nm-thick PbI2 crystal, revealing an amplitude ripple pattern that was further modeled (bottom-right inset of Fig. 2a) using a multiple-sources model based on Hertzian dipole antennas (HDA, Supplementary Information, Section 3.3)46, supporting the polaritonic nature of the waves.

Fig. 2: Real-space nano-imaging of THz hyperbolic phonon-polaritons (HPhPs) in PbI2.
Fig. 2: Real-space nano-imaging of THz hyperbolic phonon-polaritons (HPhPs) in PbI2.
Full size image

a Graphical illustration of the THz scattering scanning near-field optical microscopy (s-SNOM) scheme accessing HPhPs activity with a long metallic atomic force microscopy (AFM) tip, upper right: near-field amplitude image of a 655 nm-thick flake at 2.52 THz, lower right: corresponding calculated HPhP waves pattern (see Supplementary Information, Section 3.3). Scale bars in the insets represent 2 μm. b Fourier transform (FT) profiles of the near-field image of the 655 nm-thick flake, highlighting sharp peaks centered at q values inversely proportional to the HPhPs wavelength. Inset: Fourier transform image with indication (white dashed lines) of the directions from where the profiles were extracted (see Supplementary Information, Section 2.2). c Near-field images of flakes with various thicknesses (t) measured at 2.52 THz. Scale bars in (c) represent 5 μm. d Near-field amplitude profiles extracted from the nanoscopy images (dashed white line in c) displaying the HPhP wavelength variation with the flake thickness. The confinement factor (CF = λ0/λp) is presented for each thickness. λ0 (119 µm) and λp represent the excitation and polariton wavelengths, respectively. e Frequency–momentum (ω–q) dispersion relation of HPhPs between the transverse optical phonon frequency νEu(TO) = 1.55 THz and the longitudinal phonon frequency νEu(LO) = 3.18 THz calculated for various thicknesses (Supplementary Information, Section 3.1). f HPhP wavelength (λp) estimated from the FT analysis versus crystal thickness (open circles: experimental data, red line: calculation from the dispersion diagram).

The origin of the near-field amplitude patterns is further investigated through Fourier transform (FT) analysis of the s-SNOM images. Figure 2b shows intensity profiles extracted from the FT of the near-field amplitude image of the 655 nm-thick flake (inset of Fig. 2b). The FT profiles, taken along the three directions indicated by the dashed lines in the q-space map in the inset of Fig. 2b, exhibit strong peaks at ±2.48 × 104 cm−1, indicating a dominant wavelength of 2.53 μm in the analyzed wave pattern. Both real-space and Fourier analyses of these waves (Fig. 2b, d) reveal a periodicity that precisely matches the predicted λp of phonon-polaritons in the RB-2 region of PbI2, thereby confirming PbI2 as a 2D platform for HPhPs. Figure 2c displays near-field amplitude images of PbI2 flakes with various thicknesses, showing a clear variation in the spacing between intensity fringes as a function of thickness. These s-SNOM images reveal pronounced amplitude contrast with a signal-to-noise ratio (SNR) sufficient to resolve at least three oscillation maxima for all investigated thicknesses. All s-SNOM amplitude images presented in Fig. 2c were median leveled and then normalized by an average value of the substrate response. The reproducibility of the waves across all examined crystals indicates a consistent and well-defined crystallinity in the 2D PbI2 produced in this work. These amplitude oscillations are further analyzed in Fig. 2d, where near-field amplitude profiles extracted across the edge of the PbI2 flakes highlight a clear relationship between the crystal thickness and fringe wavelength. To validate the hypothesis of polaritonic activity, we calculated a dispersion diagram for various PbI2 thicknesses within the frequency range of its RB-2, as shown in Fig. 2e. A horizontal dashed line indicates the illumination frequency of 2.52 THz, enabling estimation of the q value for the measured samples. From the dispersion diagram in Fig. 2e, a continuous estimate for λp = 2π/q for thicknesses ranging from 0 to 700 nm is presented in Fig. 2f (red curve), where experimental λp (open circles), extracted from the FFT analysis (Supplementary Information, Section 2.2), show excellent agreement. This confirms that the observed wavelengths correspond to the HPhP wavelength λp; therefore, we determine that the HPhP launching mechanism is predominantly driven by the PbI2 crystal edges in that case.

As previously established in the mid-IR range and extensively studied in h-BN1,47,48,49,50, the underlying assumption is that two mechanisms govern the launching of HPhPs: (i) polariton waves are launched by a metallic tip and reflected back from the crystal edges (round-trip between tip and edge), where the oscillation period corresponds to half of the polariton wavelength, λp/2 (inset in Fig. 2b, bottom left); (ii) HPhPs are launched by the edge propagating as plane waves reaching the tip, where the interference of near-fields from the HPhPs and illuminated tip are re-scaterred into the far-field, forming a standing-wave pattern parallel to the crystal edge with a periodicity of λp (inset in Fig. 2b, top left). In h-BN and other hyperbolic materials in the mid-IR range, the tip typically serves as the primary launcher of polaritons. However, its superior efficiency as a launcher is not always evident, as both mechanisms are often observed within the same measurement. Additionally, tip-launched HPhPs undergo geometric decay and require a round-trip from the edge to be re-coupled to the tip47, increasing the challenge for their observation. A comprehensive FT analysis of the entire set of flakes is presented in the Supplementary Information Section 2.2, where a small contribution of λp/2 is observed when the crystal is analyzed on a dielectric substrate. One important aspect to consider is that our experiments operated with wavelengths nearly an order of magnitude longer than those typically used in the mid-IR, thereby increasing the illumination spot size and, therefore, increasing the interaction cross-section with the sample. This becomes evident in the real-space THz images of the PbI2 flakes, which display a homogeneous scattering response along all edges of the flakes (see Fig. 2c). Moreover, our hydrothermal synthesis method reliably produces crystals with geometries that precisely follow the PbI2 unit cell arrangement, where the edges of the 2D crystals taper to form sharp facets. This is confirmed by scanning electron microscopy, as detailed in the Supplementary Information Section 1. In this configuration, the edges act as efficient HPhP launchers under far-field illumination but, on the other hand, serve as low-efficiency reflectors for internal HPhP waves, making the observation of tip-launched round-trip polaritons more challenging.

To further explore the nanophotonics of PbI2, we employed THz near-field time-domain spectroscopy (nano-TDS) to access the full extent of its RB-2. Our experimental setup was based on integrating a fiber-based TDS system into a s-SNOM nanoscope, as illustrated in Fig. 3a. As a s-SNOM-based technique, nano-TDS enables access to the optical properties of materials in a typical range of 0.5–3 THz in point-spectroscopy mode.

Fig. 3: THz nano-TDS analysis.
Fig. 3: THz nano-TDS analysis.
Full size image

a Schematic of the THz s-SNOM TDS experimental setup that employs a fs near-infrared (NIR) source to pump a transmitter (TX) and a receiver (RX) whose signal is digitized by an analog-to-digital converter (ADC). b Near-field complex spectra (2nd harmonic) measured at the center of a PbI2 flake featuring an absorption peak consistent with the Eu(TO) mode. c s-SNOM TDS spectral-linescan across the PbI2 edge (see inset scheme) displaying spatial-spectral amplitude modulations for a 298 nm-thick flake. Two amplitude maxima are recorded as the tip moves away from the edge (position 0 μm), showing a dispersive trend. White dashed lines guide the eyes for the visualization of the amplitude dispersion and the νEu(TO) frequency. d Amplitude profiles extracted from (c) at four different frequencies and fitted to a parameters-retrieval model (Supplementary Information, Section 3.4). e Dispersion diagram of the HPhPs (color map: imaginary part of the Fresnel reflection coefficient rp, white dots: experimental points). f PbI2 HPhPs figure-of-merit (FOM) versus frequency (solid and dashed lines: theoretical FOM; circles and star: extracted from experimental data). Experimental uncertainty bars were calculated via standard error propagation in the fitting procedure described in the Supplementary Information, Section 3.4. Γ is the damping parameter that represents the dielectric losses in the dielectric function (see Supplementary Information, Eq. S4).

Figure 3b presents the nano-TDS point-spectrum acquired at the center of a 298 nm-thick PbI2 crystal deposited onto Si. The near-field amplitude and phase are consistent with the expected THz reflectivity and absorption of PbI2 within this range, showing excellent agreement with the far-field TDS measurements (Fig. 1e). To assess the HPhP behavior in the material, we performed a spatial-spectral analysis as shown in Fig. 3c. Along a line crossing the edge of the 298 nm-thick PbI2 crystal (Fig. 3c inset), multiple near-field amplitude point-spectra were collected, revealing contrast between the Si substrate and the 2D flake. On the PbI2 crystal—starting at its edge (position 0 μm)—the point-spectra exhibit two intensity maxima that consistently shift to lower frequencies as the probe moves away from the edge, in agreement with the dispersive nature of PbI2 HPhPs within the RB-2 region. Conversely, at fixed frequencies, the spacing between the two fringes increases, i.e., λp becomes longer at lower excitation frequencies. This trend is precisely captured in Fig. 3d, where experimental amplitude profiles at selected frequencies (open circles) are extracted from the spatial-spectral map (white arrows in Fig. 3c indicate the analyzed frequencies). We employed a phenomenological model that considers the tip and edge as distinct polariton launchers with different geometric decays (Supplementary Information, Section 3.4) to fit the experimental data (red curve) and extract key HPhP wave parameters, such as amplitude, wavelength, and damping4,5,6,51. The same model was used to extract the HPhP parameters from the real-space images at 2.52 THz shown in Fig. 2c, by fitting the profiles presented in Fig. 2d (Supplementary Information, Section 3.4). Figure 3e shows an overlay of these experimentally derived data points (white dots) onto the theoretical ω–q dispersion of the imaginary part of the Fresnel reflectivity (rp) for the 298 nm-thick PbI2 crystal. The data closely follow the dispersion of the first-order branch (M0) of PbI2 HPhPs, demonstrating good agreement between experiment and theory. These parameters are further used to estimate an experimental figure-of-merit (FOM) related to the propagation efficiency of the observed waves. In line with common practice in this research field \(\frac{{{\rm{Re}}}[q]}{{{\rm{Im}}}[q]}\) has been chosen as an indicator of the quality of the observed phonon polariton. This allows us to compare our results with previous HPhP observations well established in mid-IR frequencies. However, care should be taken not to confuse this FOM with the true quality factor of the guided HPhPs, Q = ωτ, with τ the photon lifetime. This latter can only be corrected by also taking into account the group velocity of the guided HPhP, as \(\tau=1/({{\rm{Im}}}[q]{v}_{g})\). Figure 3f presents FOM values obtained from the nano-TDS (open circles) and s-SNOM narrowband imaging (open star) experiments. These experimental FOM values were compared with theoretical predictions (calculated as the ratio Re[q]/Im[q] for t = 298 nm), showing reasonable agreement. Upon closer inspection, this theoretical estimate of the FOM exhibits a strong dependence on the dielectric losses of PbI2, represented by the damping parameter (Γ) used in the dielectric function. The dashed curve in Fig. 3f shows the FOM calculated using Γ = 0.09, a value adopted from Lucovsky et al.37,41 experimentally extracted from far-field THz reflectance measurements in the 1970s. In contrast, the theoretical estimate based on our near-field results (solid curve in Fig. 3f) indicates substantially lower dielectric losses, with Γ = 0.02 yielding an average FOM curve that aligns well with our data from both broadband TDS and narrowband s-SNOM experiments.

In Fig. 4, we present an evaluation of PbI2 functioning as an HPhP standing-wave cavity on both dielectric and metallic substrates. Figure 4a, b shows the topography of a 492 nm-thick triangular crystal grown on a 130 nm-thick PbI2 crystal on Si and the corresponding near-field amplitude image acquired at 2.52 THz. The third-harmonic near-field amplitude image (Fig. 4b) reveals high-contrast HPhP interference patterns within the crystal. Similar strong contrast is observed for HPhP waves in a 370 nm-thick PbI2 crystal on Au, where one corner of a triangular flake was analyzed (Fig. 4c, d). A double periodicity is evident in the near-field image of Fig. 4d, from which an amplitude profile (white dashed line) was extracted and fitted using the model described previously (Supplementary Information, Section 3.4). In this case, the function accounts for both tip-launched (Ftip) and edge-launched (Fedge) polaritons, as shown in Fig. 4e. From this fitting, we estimated the polariton wavelength λp for the analyzed PbI2/Au system, which matches well with the theoretical prediction presented in Fig. 4f. The use of a metallic substrate enhances the tip-induced dipole, thereby increasing the SNR and strengthening the coupling between the tip and the PbI2 flake. Consequently, the observed double-periodicity interference pattern confirms the coexistence of both edge- and tip-launched HPhP waves in the experiment conducted on Au substrate (Fig. 4d). Furthermore, Fig. 4b, d reveals outstanding propagation efficiency of the HPhP waves for both Si and Au substrates, demonstrating the versatility of the material for use as a THz resonator or in the development of microscale THz cavities. Finally, we position PbI2 among other prominent hyperbolic 2D materials by comparing key properties, including the FOM Re[q]/Im[q] and the LST ratio, as shown in Fig. 4g. Notably, only PbI2 (THz, FOM = 17), for a 340 nm-thick crystal (Fig. 2c), and h-BN (mid-IR, FOM = 1552) exhibit values exceeding 15. Other emerging hyperbolic materials in the far-IR range—including GeS12, α-MoO36,7, HfSe2, and HfS213—display lower FOM values, with α-MoO3 reaching a maximum of 12.4 in the mid-IR. Figure 4g also reports the LST ratio for each material, with magnitude values indicated below each data point and marker size scaled proportionally for visual comparison. The very high dielectric anisotropy and large ionicity of PbI2 are key advantages, enabling its hyperbolic response across an ultra-wide spectral range. This property is qualitatively assessed by calculating the LST ratio for the type II hyperbolic band of PbI2, which is then compared to other prominent materials in this class. The LST ratio reaches 4.20 for PbI2, while for h-BN (RB-2) and HfSe2 (RB-2)13, this ratio is 1.41 and 3.29, respectively.

Fig. 4: PbI2 as an HPhP cavity on different substrates and in comparison to other phonon-polaritonic media.
Fig. 4: PbI2 as an HPhP cavity on different substrates and in comparison to other phonon-polaritonic media.
Full size image

a, b AFM and near-field amplitude images of a 492 nm-thick PbI2 crystal on a 130 nm-thick PbI2 crystal on Si measured at 2.52 THz. The near-field image reveals a high-contrast standing-wave pattern of HPhPs within the triangular crystal. Scale bar: 10 μm. c, d AFM and s-SNOM amplitude images of a 370 nm-thick triangular PbI2 flake on Au, highlighting an interference pattern formed by HPhP waves launched by both the tip and crystal edges. Scale bar: 5 μm. e Experimental near-field amplitude profile (open circles) extracted along the white dashed line in (d), fitted by a phenomenological function (red line) that accounts for HPhPs launched by the edge (orange line) and the tip (blue line). f Expected polariton wavelengths as a function of PbI2 thickness when transferred onto an Au substrate for 2.52 THz illumination. The experimental data point (open circle) is obtained from the analysis in (e) (Supplementary Information, Section 3.4). g PbI2 in the context of other selected 2D materials supporting HPhPs across the mid-IR, far-IR, and THz ranges, organized by their key parameters FOM and Lyddane–Sachs–Teller (LST) relation. νTO represents the frequency of the transverse optical (TO) phonon mode. ε0 and ε∞ represent PbI2 dielectric constants at the low and high frequency limits, respectively.

This work introduces a hyperbolic material with compelling nanophotonic activity across an ultra-wide band in the deep terahertz range. Our s-SNOM experimental data exhibit enhanced SNR, enabling high contrast in real-space visualization of THz phonon-polaritons, with excellent agreement with established theory. We measured HPhPs with FOM values reaching 17, positioning PbI2 as a competitive platform alongside h-BN and α-MoO3, now extended into the THz regime. The field confinement also surpasses that of most materials in this class, with a measured λ0/λp of 264 for a 144 nm-thick PbI2. This confinement factor can exceed 300 for flakes thinner than 100 nm under the same illumination (2.52 THz). From a production perspective, 2D PbI2 is facile to synthesize, composed of relatively abundant elements, and exhibits chemical stability under standard ambient conditions. Compared to mechanical exfoliation, its common growth methods offer scalability and potential for practical integration. As a semiconductor, it also supports carrier injection—whether electrical, optical, or via doping—enabling coupling to phonons and paving the way for active components like modulators and detectors. As a vdW material, PbI2 can also be stacked with other 2D materials53,54,55, further expanding its range of potential applications. In conclusion, PbI2 emerges as a compelling platform for THz nanophotonics, offering both fundamental advantages and a pathway toward practical device integration. Its ability to support long-wavelength HPhPs, with inherently lower momentum requirements, eases many of the coupling challenges that currently limit higher-order polaritonic modes in the mid-IR regime56,57. This property positions PbI2 as a candidate for enabling compact on-chip light-routing elements and advanced wave manipulation concepts. Moreover, the recent introduction of multifunctional 2D materials such as PdSe2 for IR detection58 highlights the technological momentum of polaritonic circuitry, which can be naturally extended into the THz domain with PbI2. Altogether, PbI2 provides a versatile material foundation for enabling polariton-mediated phenomena and driving the development of next-generation THz optoelectronic and nanophotonic devices.

Methods

PbI2 crystals synthesis

Crystals were grown using a simple hydrothermal technique and deposited onto a high-resistivity silicon wafer for the Raman and s-SNOM experiments, except for the experiment in Fig. 4c, d that we used Au(100 nm)/Si as a substrate. Larger flakes were grown in a solution following the same technique and transferred above a hole drilled in a metallic foil for transmission THz TDS (see Supplementary Information Section 5 for details).

Far-field pre-characterization

Raman spectroscopy measurements on PbI2 flakes (Fig. 1d) were performed using a micro-photoluminescence and Raman setup based on a LabRAM HR confocal system (Horiba Jobin-Yvon). For Raman experiments, we employed a 633 nm laser source for excitation. The scattered light was analyzed using an 800 mm focal length monochromator, with detection provided by a CCD camera covering the 0.3–1.06 μm spectral range. We used a notch filter to reject Rayleigh-scattered light below 50 cm−1. Far-field THz transmission spectra (Fig. 1e) were acquired using a commercial time-domain spectroscopy (TDS) system (TeraSmart, Menlo Systems GmbH), which employs a pulsed THz source based on a photoconductive antenna excited by a femtosecond laser, along with a synchronously gated photoconductive detector. An oscillating delay line was used to reconstruct the temporal profile of the electric field. A large PbI2 flake (~4 μm-thick) was mounted over a 500 μm pinhole in a 100 μm-thick metallic plate. The THz beam was focused and collected using TPX lenses with a 38 mm diameter and 50 mm focal length. All measurements were performed under a dry nitrogen atmosphere to minimize water-vapor absorption. The spectra were obtained by averaging 1000 scans (120 ps-long) with a 33 fs time step and subsequently normalized to a reference measurement taken through an identical empty pinhole. Further details on the setup and measurement procedure are provided in the Supplementary Information Section 7.

THz s-SNOM imaging

s-SNOM images (Figs. 2c and 4a–d) were acquired using a commercial near-field nanoscope (NeaSNOM, attocube systems GmbH) illuminated by a continuous-wave (CW) terahertz source based on a CO2 laser-pumped molecular laser (FIRL100, Edinburgh Instruments Ltd.). The THz cavity was filled with low-pressure methanol (CH3OH) vapor, with the CO2 laser tuned to the 9P36 line to generate lasing at 2522.78 GHz. Under optimal pressure conditions, the typical THz output power, measured using a thermopile, was approximately 70 mW. The beam was directed toward the s-SNOM system, where a 12 μm-thick Mylar beamsplitter divided it into two paths: a reference arm and the tip-sample interaction stage. Parabolic optics focused the THz radiation onto a metallic AFM probe, which acts as an antenna to confine a portion of the incident field at its apex, enabling nanoscale THz imaging. We estimated ~20 mW to reach the tip-sample stage. For the s-SNOM images shown in Fig. 2c, we used 70 μm-long Pt-coated Si probes (Vmicro SAS, 270 kHz) with an apex radius of approximately 40 nm (Supplementary Information, Section 8), operating with tapping amplitudes around 100 nm and an integration time of 20 ms per pixel. For the images in Fig. 4a–d, we used 80 μm-long full-metal probes (80 kHz, 25PtIr200B-H40, Rocky Mountain Nanotechnology LLC) with tapping amplitudes of approximately 120 nm and a pixel integration time of 10 ms. All s-SNOM measurements were performed using a homodyne scheme (reference mirror fixed in an optimized position to maximize the near-field signal). Background suppression was achieved by demodulating the signal at the third harmonic of the cantilever’s resonance frequency. Signal detection was performed using a room-temperature Schottky diode detector/mixer (WM86, Virginia Diodes Inc.).

THz s-SNOM nanospectroscopy

THz nano-TDS experiments were done using a commercial instrument (THz-neaSCOPE+S, attocube systems GmbH) equipped with an integrated THz time-domain spectroscopy module (TeraSmart, Menlo Systems GmbH). THz pulses covering the spectral range 0.5–2.5 THz are generated and coherently detected by a pair of photoconductive antennas. A conductive AFM tip with a shaft length of 80 μm and a typical average tip radius of ~40 nm (25PtIr200B-H40, Rocky Mountain Nanotechnology LLC), operated in non-contact tapping mode (at a nominal frequency of 80 kHz), acts as a nanofocusing probe, allowing for simultaneous capture of sample topography together with measurements of the scattered near-field signal in a single scan of the sample. Experimental data presented in Fig. 3b, c were acquired in a nitrogen-purged environment with relative humidity below 5% to minimize the presence of water-vapor absorption lines in the detected THz spectra. Background subtraction was done via higher harmonic demodulation at the second harmonic of the tip tapping frequency. The point-spectrum in Fig. 3b was acquired in the middle of a 298 nm-thick PbI2 flake, where only the dielectric response was detected by the tip (no back-reflected polariton waves from the edges at this position). The spectrum was averaged from 10 interferograms. Spectra-linescan in Fig. 3c was measured across the edge of the same flake, where 25 point-spectra were acquired along the 10 μm-long dashed line indicated in the Fig. 3c inset. For each point-spectrum, we averaged 5 interferograms. For all nano-TDS experiments, the probe amplitude was kept at 150 nm, pixel time was set as 50 ms, and spectral resolution was 0.03 THz (30 ps scanning delay). All spectra were normalized by a reference spectrum taken on an Au/Si substrate.