Nanoscopy of surface polarization with oblique dipole orientations Journal: opticajournal Publication type: Research Article V. G. M. Duarte, 1,2,* D. A. Miranda, 3 D. F. P. Cunha, 2,5 M. I. Vasilevskiy, 2,5 N. Asger Mortensen, 3,4 A. J. Chaves, 1,5 and N. M. R. Peres 2,3,5 Abstract We present a general electromagnetic description for dipoles confined to surfaces with oblique dipole moment orientations, extending the conventional in-plane (IP) and out-of-plane (OOP) treatments. This description is useful for describing localized polarization in, e.g. , van der Waals heterostructures, thin films of molecular aggregates, and metal-dielectric interfaces. The theory is suitable for any material with vanishingly thin thickness relative to the light wavelength, independent of the geometry of the material and the media interfacing it. We apply the formalism to a uniaxial excitonic sheet, covering a large number of two-dimensional (2D) materials and organic thin films. Our theory reveals pairs of polaritonic resonances originating from the IP and OOP components of the excitonic dipole moment. The formalism suggests experimentally accessible signatures of dipole moment orientation, enhanced by near-field probes. This work proposes a unified language for the description of 2D materials, thin films and interfaces with anisotropic dipolar responses. 1 Introduction The advances of optical microscopy allowed matter to be probed with light at reduced spatial and temporal scales. In particular, scattering-type scanning near-field optical microscopy (s-SNOM) [ 58 ] was successfully applied to various systems, including semiconductors [ 8 ] , organic [ 57 ] , and biological samples [ 61 ] , revealing intrinsic properties of matter [ 3 ] . For example, the evanescent component of the electric field originated in the s-SNOM tip can be used to excite surface plasmon-polaritons [ 77 ] , phonon-polaritons [ 35 ] , and exciton-polaritons [ 34 ] . Additionally, synchrotron facilities can be used as a source for synchrotron infrared nanospectroscopy [ 5 ] . A natural starting point for studying light-matter interactions is the electric dipole [ 36 ] , which represents the first-order response of a neutral charge distribution in the presence of an external field. Frenkel excitons, Wannier–Mott excitons, and quantum surface states, hosted respectively by molecules, semiconductors, and metals, interact with light through their electric dipoles. Frenkel excitons are found in organic materials, and have dipole moments oriented following the molecule orientation [ 7 ] . Wannier–Mott excitons play the leading role in two-dimensional (2D) materials, and can emit light whose polarization lies within the 2D material plane or perpendicular to it [ 67 ] . At last, metallic quantum surface states behave as a series of multipolar corrections to induced charge. These corrections are described through Feibelman’s surface response formalism (SRF) [ 17 ] , where the parameter d_{\perp} represents the out-of-plane component of the induced dipole near the metal surface, originating from the charge distribution of the quantum surface states. The description of light interaction with excitons is presented in the seminal work of Hopfield [ 33 ] . However, as pointed elsewhere [ 68 ] , the dielectric screening reduces both the exciton binding energy and oscillator strengths, making the presence of excitons negligible in most systems [ 30 ] . This can be bypassed in nanostructures such as quantum dots [ 62 ] , or 2D materials [ 68 ] . The latter have garnered great attention in recent years due to their optical properties and tunability via environment or layer stacking in van der Waals (vdW) heterostructures. For monolayers of MoS 2 and other transition metal dichalcogenides, the optical properties in the visible range are dominated by the in-plane exciton peak, which is split into two by virtue of spin-orbit coupling. 2D material monolayers also support out-of-plane excitations, but the optical response of those is weaker than that of their in-plane counterparts by a couple orders of magnitude [ 27 ] . Nonetheless, the optical response of out-of-plane excitons can be probed by polarization-resolved photoluminescence [ 79 , 67 ] or near-field spectroscopy [ 41 ] . Such out-of-plane excitons can also have their emission enhanced by the Purcell effect through coupling with plasmonic cavities [ 4 ] . Furthermore, vdW heterostructures can support interlayer excitons [ 20 ] , which can possess both static permanent dipoles [ 44 ] and an out-of-plane transition dipole component [ 76 ] . The static permanent dipoles of interlayer excitons in vdW heterostructures are responsible for an out-of-plane Stark shift [ 37 ] and dipole-dipole interactions [ 44 ] . When the orbital angular momentum component of the band edges is zero in the direction perpendicular to the 2D plane, the polarization of the photons is perpendicular to the plane [ 63 ] . This can be used to design vdW heterostructures with out-of-plane photon emission [ 65 ] . Strong out-of-plane photon absorption and emission were observed in monoelemental, binary 2D materials and vdW heterostructures [ 70 ] . Besides 2D materials, organic molecules can also exhibit exciton formation. In particular, some molecule species can stack in various geometries and lead to delocalized exciton formation across several molecules [ 7 , 71 ] . Contrary to 2D materials, organic molecular aggregates are highly disordered, but their excitons have dipole moments with fixed and well defined orientation with respect to the aggregate axis. Compared to their monomer counterparts, these molecular aggregates lead to narrower, redshifted resonances and stronger absorption, associated to the formation of large dipole moments that surpass those found in 2D materials [ 49 , 43 ] and can lead to exciton-photon coupling potentially reaching the ultrastrong coupling regime [ 23 , 15 ] . In fact, J-aggregates were one of the first systems to exhibit superradiance under realistic experimental conditions [ 60 ] . The description of excitons in such organic materials, as well as their interaction with light, dates back to Kasha’s dimer model [ 39 , 38 ] , which describes the fundamental mechanism behind the exciton-exciton coupling and the redshifted resonances. Low-dimensional J-aggregate layers have been employed in films [ 40 ] and core-shell nanoparticles [ 22 ] . In charged metals, the classical assumption is that the charge is confined to the surface, typically forming an interface with vacuum or a dielectric. However, Feibelman and coworkers proposed a multipole series expansion of the induced charge at an interface [ 17 ] . Effectively, the first-order term of this correction corresponds to a dipole-polarized sheet and is responsible for the observation of nonlocal electrodynamic effects in diverse nanophotonics systems [ 51 ] . Feibelman’s SRF introduces the so-called d -parameters, related to charge and current density centroid positions defined relative to the metal-dielectric interface. They account for charge spill-out, as well as nonlocal screening, leading to corrections in plasmon dispersion that can be measured. For this reason, Feibelman’s SRF underpins the current understanding of mesoscopic electrodynamics in the field of quantum plasmonics, and provides a perspective into the physics of polarized interfaces between media [ 10 , 73 ] . The description of dipoles at surfaces typically consider purely in-plane (IP) or out-of-plane (OOP) orientations. Nonetheless, a theory for surface confined dipoles pointing at oblique directions has been made several decades ago, in the context of monolayers of strongly polarized molecules [ 31 , 19 , 18 ] . This theory shows that the continuity of the electric field tangential to a surface is broken when an excess polarization perpendicular to the surface is present, an atypical result that is commonly absent in electrodynamics textbooks [ 36 ] . Despite the generality of this discontinuity, until now it has only flourished in the context of higher order nonlinearities [ 46 , 54 , 14 ] . The equation for the jump discontinuity of the tangential electric field is in agreement with Feibelman’s SRF, truncated at first-order. The key unifying idea is that an oblique dipole polarization confined to a surface, henceforth denominated as a polarization sheet, generates a singularity in the electric displacement field D_{\perp} perpendicular to the sheet. In this paper, we build upon the theory of Refs. [ 31 , 19 , 18 ] with a description of polarization sheets that is independent of the geometry of the sheet and the media interfacing it. The paper is organized as follows: in Sec. 2 we present the general electromagnetic description of s. Sec. 3 provides the model for investigation of an anisotropic 2D sheet through s-SNOM. In Sec. 4 we consider an excitonic and uniaxial 2D sheet on top of a bulk substrate and examine the resultant reflectivity, surface polariton modes, and s-SNOM spectra. The main conclusions are laid out in Sec. 5 . Figure 1: Schematic representation of a surface S , representing a 2D sheet of arbitrary geometry, and the optical configurations considered in this work. (a) Oblique and lateral views of the surface, showing the surface polarization \mathbf{P}^{s} , the outward normal \hat{\mathbf{n}} , and the sheet thickness (much smaller than the light wavelength \lambda , thereby allowing the surface treatment). Inset: zoomed-in view of an arbitrary point in the sheet, showing the normal coordinate r_{\perp} taken along the unit vector \hat{\mathbf{n}} and the tangential and normal components of a field \mathbf{f} . (b) Same as the inset in panel (a), indicating the susceptibility tensors \overleftrightarrow{\chi}^{\pm} as defined in the text. (c) Geometry considered for optical response calculations. Plane wave fields impinge on the sheet with a wavevector \mathbf{k} and incidence angle \theta_{\mathrm{inc}} . (d) Same as panel (c), with the addition of a metallic tip to scatter part of the incident light back to a detector in the far-field region. The tip interacts with the sheet generating a point-dipole p , oriented towards the z axis for an elongated tip geometry. 2 Electrodynamics of polarization sheets It is well established that the continuity of the electric field tangential to a metal-dielectric interface is broken by an OOP surface dipole density \bm{\pi} [ 73 ] , mediated by the parameter d_{\perp} . A similar discontinuity was identified by Heinz [ 31 ] and Felderhof & Marowsky [ 19 , 18 ] in the context of monolayers of strongly polarized molecules, where the tangential electric field discontinuity is related to the OOP component of the surface polarization field \mathbf{P}^{s} associated to the monolayer. We will show that these formalisms provide equivalent descriptions of polarization confined to a surface. 2.1 Boundary conditions We consider a surface s of arbitrary geometry [Fig. 1 (a)], hosting charges and currents that generate electromagnetic fields around s . The charge density \rho is related to the current density \mathbf{j} via the continuity equation, written as \nabla\cdot\mathbf{j}=i\omega\rho in the frequency domain. We associate a surface polarization field to this current density via \mathbf{j}(\mathbf{r}+r_{\perp}\hat{\mathbf{n}},\omega)=-i\omega\mathbf{P}^{s}(\mathbf{r},\omega)\delta(r_{\perp}), (1) for any point \mathbf{r}\in s and a coordinate r_{\perp} taken along the direction of the outward normal \hat{\mathbf{n}} relative to s . The charge density may then be written as \rho=-(\nabla\cdot\mathbf{P}^{s})\delta(r_{\perp})-\mathbf{P}^{s}\cdot\nabla\delta(r_{\perp}). (2) The fields generated by the surface s can polarize its vicinities. To this induced polarization we attribute a field \mathbf{P} , from which it follows that the displacement field in the vicinity of s relates to the electric field as \mathbf{D}=\epsilon_{0}\mathbf{E}+\mathbf{P} . The polarization field \mathbf{P} is indeed the conventional polarization field considered when writing the Maxwell equations in matter, whereas \mathbf{P}^{s} is a non-classical field attributed to electromagnetic sources confined to a surface. The Maxwell’s equations read \displaystyle\nabla\cdot\mathbf{D} \displaystyle= \displaystyle-(\nabla\cdot\mathbf{P}^{s})\delta(r_{\perp})-\mathbf{P}^{s}\cdot\nabla\delta(r_{\perp}), (3a) \displaystyle\nabla\cdot\mathbf{B} \displaystyle= \displaystyle 0, (3b) \displaystyle\nabla\times\mathbf{E} \displaystyle= \displaystyle i\omega\mathbf{B}, (3c) \displaystyle\nabla\times\mathbf{H} \displaystyle= \displaystyle-i\omega\mathbf{P}^{s}\delta(r_{\perp})-i\omega\mathbf{D}. (3d) Before deriving the BCs, we shall focus on the atypical dipolar term, -\mathbf{P}^{s}\cdot\nabla\delta(r_{\perp}) . For smooth surfaces s , effects of curvature become negligible in an infinitesimal interval 0^{-}0 with wavevector \mathbf{k}_{\mathrm{inc}}=q\mathrm{inc}\hat{\mathbf{x}}-k_{z,\mathrm{inc}}\hat{\mathbf{z}} . An additional contribution is generated from direct reflection into the sample before scattering in the tip. Then, E_{0,z}=E_{inc,z}+rE_{inc,z}e^{2ik_{z,\mathrm{inc}}z_{0}}, (22) where r is the Fresnel’s reflection coefficient of the sheet. Substituting Eqs ( 21a , 22 ) in Eq. ( 21b ) results in p=\alpha E_{\mathrm{inc},z}\left(1+re^{2ik_{z,\mathrm{inc}}z_{0}}\right)+\alpha Gp. (23) Before proceeding, we emphasize the assumptions being made when writing the tip’s dipole moment as in Eq. ( 23 ): The dipole moment is restricted to \hat{\mathbf{z}} direction as a phenomenological correction of the model to accommodate for the elongated tip geometry. The last term, \alpha Gp , includes all tip scattering events, whereas the remaining terms are due to field contributions that exist with or without the presence of the tip. In the following, we proceed to rewrite an expression for the dipole moment solely in terms of the incident field E_{\mathrm{inc},z} and an effective polarizability \alpha_{\mathrm{eff}} , which renormalizes the tip to account for the presence of the sheet. The polarizability may be written in the Clausius–Mossotti form \alpha(\omega)=4\pi\epsilon_{0}R_{\mathrm{tip}}^{3}\frac{\epsilon_{\mathrm{tip}}(\omega)-1}{\epsilon_{\mathrm{tip}}(\omega)+2} , for a radius of curvature R_{\mathrm{tip}} and a tip permittivity \epsilon_{\mathrm{tip}}(\omega) [ 2 ] . The factor (1+re^{2ik_{\mathrm{z,inc}}z_{0}}) includes the reflection of the external field on the sheet. The Green’s function G=\omega^{2}\mu_{0}G_{zz}(\mathbf{r}^{\prime},\mathbf{r}^{\prime},\omega) results from self-reflection of the field radiated by the dipole back into itself. Equation ( 23 ) can be rewritten as p=\alpha_{\mathrm{eff}}\left(1+re^{2ik_{\mathrm{z,inc}}z_{0}}\right)E_{\mathrm{inc,z}} , where \alpha_{\mathrm{eff}}=\frac{\alpha}{1-\alpha G}=\frac{1}{\alpha^{-1}-G} (24) is the renormalized polarizability. An expression for G may be derived from the solution of Maxwell’s equations for a point-dipole with BCs at the air-sample interface (see Supplement 1): G=\frac{i}{4\pi\epsilon_{0}}\int_{0}^{\infty}dq\frac{q^{3}}{k_{z}}e^{2ik_{z}z_{0}}r(q,\omega). (25) The back-scattered field is given by the sum of the point-dipole radiation with its reflection on the sheet. Up to leading order, one finds (see Supplement 1) E(\mathbf{R}\rightarrow\infty,\omega)=\frac{ip_{\mathrm{tot}}ke^{ikR}}{2\epsilon_{0}R}, (26) where p_{\mathrm{tot}}(z_{0})=p\left(1+r(q_{\mathrm{inc}},\omega)e^{2ik_{\mathrm{z,inc}}z_{0}}\right) to make sure we include both the reflection of the external field and the reflection of the point-dipole’s field on the sample, as pointed out by Ref. [ 12 ] . We can write p_{\mathrm{tot}}(z_{0})=\alpha_{\mathrm{eff}}(z_{0})E_{\mathrm{inc,z}}\left(1+r(q_{\mathrm{inc}},\omega)\right)^{2}+\mathcal{O}(k_{\mathrm{z,inc}}z_{0}) (27) and the Fourier coefficients ( 20 ) are then given by \sigma_{n}=\frac{ike^{ikR}E_{\mathrm{inc,z}}\left(1+r(q_{\mathrm{inc}},\omega)\right)^{2}}{2\epsilon_{0}RT}\int_{0}^{T}dt~e^{in\Omega t}\alpha_{\mathrm{eff}}[z_{0}(t)]. (28) The signal measured in an s-SNOM detector, by means of proper demodulation techniques [ 13 ] , is proportional to the harmonics \sigma_{n} of the back-scattered field, given by Eq. ( 28 ) in our PDM. To filter out global phases and unimportant scaling factors, the signal can be normalized as \bar{\sigma}_{n}=\frac{\sigma_{n}}{\sigma_{n,\mathrm{subs}}}=\frac{\left(1+r(q_{\mathrm{inc}},\omega)\right)^{2}}{\left(1+r_{\mathrm{subs}}(q_{\mathrm{inc}})\right)^{2}}\frac{\int_{0}^{T}dt~e^{in\Omega t}\alpha_{\mathrm{eff}}[z_{0}(t)]}{\int_{0}^{T}dt~e^{in\Omega t}\alpha_{\mathrm{eff,subs}}[z_{0}(t)]}, (29) where r_{\mathrm{subs}}(q_{\mathrm{inc}})=\frac{\sqrt{\epsilon_{\mathrm{subs}}-\sin^{2}(\theta_{\mathrm{inc}})}}{\epsilon_{\mathrm{subs}}\cos(\theta_{\mathrm{inc}})}, (30) is the reflection coefficient of the bare substrate, obtained setting \chi_{\parallel}^{\pm}=\chi_{\perp}^{\pm}=0 , and \alpha_{\mathrm{eff},\mathrm{subs}} is the polarizability ( 24 ) for r=r_{\mathrm{subs}} . 4 Results In this Section, we simplify the notation by adopting the Cartesian coordinates indicated in Figs. 1 (c,d), as in Sec. 3 . To showcase the formalism, we investigate the optical response of excitons confined to a planar 2D sheet. We assume a symmetric uniaxial sheet through \overleftrightarrow{\chi}^{+}=\overleftrightarrow{\chi}^{-}=\text{diag}(\chi_{\parallel},\chi_{\parallel},\chi_{\perp}) , encompassing a large number of 2D systems [ 51 ] , and use a Lorentzian model for the excitonic contribution to the susceptibility functions, in agreement with the Elliott formula for excitons [ 16 ] : \chi_{\parallel,\perp}(\omega)=\frac{f_{\parallel,\perp}\omega_{0}^{2}}{\omega_{0}^{2}-\omega^{2}-i\omega\gamma}, (31) yielding the usual response of 2D-confined excitonic resonances at frequency \omega_{0} . The oscillator strengths f_{\parallel} and f_{\perp} modulate the material’s response to incident light in the in- and out-of-plane directions, respectively, and the damping parameter \gamma encompasses all loss sources, thereby modulating the Lorentzian peak’s height and width. We re-parameterize the oscillator strengths as \displaystyle f_{\parallel}=\frac{f_{0}}{2}\cos^{2}(\theta_{\mathrm{dip}}), (32a) \displaystyle f_{\perp}=f_{0}\sin^{2}(\theta_{\mathrm{dip}}), (32b) so that in- and out-of-plane dipoles may be represented by a single parameter: \theta_{\mathrm{dip}}=0^{\circ} (in-plane) and \theta_{\mathrm{dip}}=90^{\circ} (out-of-plane). This choice preserves the trace 2f_{\parallel}+f_{\perp}=f_{0} . One may view this reparameterization as arising from a set of dipoles confined to a plane, each tilted by a polar angle \theta_{\mathrm{dip}} away from the plane and uniformly distributed in azimuth [ 11 ] . In that picture, f_{0} corresponds to the total oscillator strength of the excitonic transition. More generally, however, Eqs. ( 32 ) can be regarded simply as a convenient, trace-preserving reparameterization, independent of any particular microscopic interpretation. Unless stated otherwise, we use \hbar\omega_{0}=2.1 eV and \hbar\gamma=0.02 eV, close to the typical values of TMDs and molecular aggregates with excitonic transitions in the VIS range [ 78 , 75 ] . The oscillator strength f_{0} (with units of length in a 2D configuration) is fixed at f_{0}=2 nm to match the strong dipole moments of molecular aggregates [ 11 ] , and the dipole angle is set to \theta_{\mathrm{dip}}=20^{\circ} to showcase an oblique dipole orientation. 4.1 Reflectance of a uniaxial 2D sheet We assume a planar 2D sheet, as depicted in Fig. 1 (c). The sheet will first be treated as being standing freely in air ( \epsilon_{\mathrm{subs}}\approx 1 ) before considering the case of a real substrate with \epsilon_{\mathrm{subs}}\neq 1 . Since we are interested in capturing the effect of the OOP polarization component, p -polarization will be considered. For a free-standing uniaxial 2D sheet with symmetric response ( \overleftrightarrow{\chi}^{+}=\overleftrightarrow{\chi}^{-} ), the reflection coefficient ( 19b ) simplifies to r_{p}^{\prime}(q,\omega)=\frac{1}{1-\beta_{\parallel}}-\frac{1}{1-\beta_{\perp}}, (33) where \beta_{\perp}=iq^{2}\chi_{\perp}/k_{z} and \beta_{\parallel}=ik_{z}\chi_{\parallel} . This suggests that the IP and OOP contributions can cancel each other out. Figures 2 (a,c) show the evolution of the reflectance spectrum as a function of incidence angle, for IP and OOP dipole orientations ( \theta_{\mathrm{dip}}=0^{\circ} and 90^{\circ} , respectively). For the in-plane reflectance term, maximum dipole-field alignment is achieved when \theta_{\mathrm{inc}}=0^{\circ} , for which k=k_{z} and \beta_{\parallel} stays finite. So the reflectance has a peak at the resonance with broadening mediated by damping. The peak is suppressed as the angle is varied from 0^{\circ} to 90^{\circ} , for which k_{z} goes monotonically from k to 0 and the in-plane dipole component is decoupled. Figure 2: Reflectance spectrum of excitons confined to a plane, for (a,b) in-plane, (c,d) out-of-plane, and (e) oblique dipole moment orientation. The different curves illustrate the influence of the incidence angle \theta_{\mathrm{inc}} . The insets in panels (b,d,e) show the reflectance of the substrate, which is a sole function of \theta_{\mathrm{inc}} . The oscillator strength is f_{0}=2 nm in panels (a,b,c,d) and f_{0}=10 nm in panel (e) to enhance visualization of the peaks compared to the dip at \omega=\omega_{0} . (f) Visibility, defined as (|r|^{2}_{\mathrm{peak}}-|r|^{2}_{\mathrm{dip}})/(|r|^{2}_{\mathrm{peak}}+|r|^{2}_{\mathrm{dip}}) , as a function of f_{0} for different incidence angles. For the out-of-plane reflectance term, maximum alignment is achieved when \theta_{\mathrm{inc}}=90^{\circ} . We have k=q , k_{z}=0 , and \beta_{\perp}\rightarrow\infty . The susceptibility at resonance is given by f_{z}\omega_{0}/(i\gamma) , therefore \beta_{\perp} goes to infinity. Far from the resonance, the susceptibility goes asymptotically to zero, but is finite everywhere, so \beta_{\perp} goes to infinity for any frequency. Hence, both the in-plane and out-of-plane reflectances are resonance-centered peaks, strengthened by field-dipole alignment. The in-plane peak is below 1 and has a narrow shape. The out-of-plane peak saturates at 1 and is severely broadened due to the rapid growth of the factor q^{2}/k_{z} in \beta_{\perp}=iq^{2}\chi_{\perp}/k_{z} . This increase in reflectance when the photon and exciton energies match shows that the photon-exciton interaction deflects the photons away. The shape of the peaks, nonetheless, differs due to an asymmetric decomposition of the field into the in-plane and out-of-plane directions. Up to now, the reflectance has been analyzed considering a polarization sheet suspended in air. A more common situation, experimentally speaking, would be to place the layer on top of a substrate. Figures 2 (b,d) show the reflectance of this new configuration, for in-plane and out-of-plane oriented dipoles. The substrate permittivity used was \epsilon_{\mathrm{subs}}(\omega)=2.25 , close to the optical constants of glass and polymers at optical frequency [ 1 ] . The substrate, even though lowly reflective, interferes in the reflectance shifting the peaks up slightly for angles close to 0^{\circ} and largely close to 90^{\circ} , independent of \omega , as the inset in Figs. 2 (b,d,e) shows. For such, one may observe that the reflection coefficient of the pure substrate may be written as Eq. ( 30 ), canceling out the frequency dependence. At last, when a substrate is present and the dipole moments have an oblique orientation, both dipole components become present and we may observe destructive interference in the reflectance channel. Fig. 2 (e) shows the resulting reflectance spectrum. The double peaks show the interference of dipole components. For visualization, f_{0} was increased, representing molecules with stronger dipole moments. To examine the contrast of the peaks compared to the dip at \omega=\omega_{0} , we evaluate the peak visibility, defined as (|r|^{2}_{\mathrm{peak}}-|r|^{2}_{\mathrm{dip}})/(|r|^{2}_{\mathrm{peak}}+|r|^{2}_{\mathrm{dip}}) , as a function of the oscillator strength. For comparison, the typical oscillator strengths of molecular aggregates are close to 2 nm [ 11 ] . The oscillator strengths of conventional 2D materials are usually several times smaller [ 15 ] . Fig. 2 (f) shows that high visibility is achieved only at very large oscillator strengths or angles \theta_{\mathrm{inc}} approaching grazing incidence, where the in- and out-of-plane reflectance components start interfering. 4.2 Surface polariton modes of a uniaxial 2D sheet The loss function, defined as \Im\{r\} , can be used to infer about surface modes excited by optical means, such as electron energy loss spectroscopy (EELS) [ 26 ] . Figs. 3 (a,b) show the loss function of the 2D sheet. At the right of the light line, two evanescent stationary modes are present, above and below the excitonic resonance, respectively. The upper mode resembles that of a p -polarized surface polariton in 2D van der Waals materials [ 21 ] , which only exhibits a polariton branch below the resonance for s -polarized light. Here, by contrast, p -polarized light excites a lower branch with negative group velocity. The insertion of the substrate slightly shifts the branches up and dampens its peaks, but their qualitative aspect is preserved. Figure 3: (a,b) Loss function \Im\{r\} close to the excitonic resonance ( \omega=\omega_{0} ) of a 2D layer of dipoles (a) suspended in air and (b) on top of a thick substrate layer. Red dashed represents the light lines q=\omega/c and q=\omega\sqrt{\epsilon_{\mathrm{subs}}}/c . (c,d,e) Dispersion relation surface modes on a 2D excitonic sheet suspended in air (c) above and (d,e) below the exciton line, for real \omega and complex q . Panel (e) shows the modes on a narrower horizontal scale, highlighting the weakly confined modes. To further examine these surface modes, we solve directly for the poles of the reflection coefficient, which, for a sheet suspended in air, yields \beta_{\parallel}=1 and \beta_{\perp}=1 . We get \kappa_{z}^{t}=-1/\chi_{t} and \kappa_{z}^{z\pm}=(1/\chi_{z}\pm\sqrt{1/\chi_{z}^{2}-k^{2}})/2 (using k_{z}=i\kappa_{z} ). The corresponding modes are q(\omega)=\sqrt{\omega^{2}/c^{2}+\kappa_{z}^{2}} , and we present the resulting \omega vs. \Re\{q\} curves in Figs. 3 (c,d,e). We considered real \omega and, consequently, complex q . For physical meaningfulness, the results were filtered for \Re\{\kappa_{z}\}>0 , keeping the evanescent behavior away from the sheet. The horizontal axis was normalized by k=\omega/c in order to highlight degree of confinement of the surface modes. An upper branch with positive group velocity ( \kappa_{z}^{t} ) and a lower branch with negative group velocity ( \kappa_{z}^{z+} ) were obtained. The additional branch ( \kappa_{z}^{z-} ) approaches a free-photon mode, with low confinement since cq/\omega\sim 1 ; hence, it does not form a properly bound surface polariton mode. We further verified the low confinement of the \kappa_{z}^{z-} through the \omega vs. \Im\{q\} curves, showing that the \kappa_{z}^{z-} present lower \Im\{q\} than \kappa_{z}^{z+} across the entire spectrum. In fact, taking into account the group velocity v_{g}=d\omega/dq , we verified that |\Im\{q\}d\omega/dq|<\gamma/2 for \kappa_{z}^{z-} whereas \Im\{q\}d\omega/dq>\gamma/2 for \kappa_{z}^{z+} , thereby confirming the low damping of \kappa_{z}^{z-} modes. 4.3 s-SNOM characterization of oblique dipole orientation As previously discussed, reflectance experiments lack visibility for electric dipoles with low oscillator strength. Near-field spectroscopy techniques can make dipolar coupling more efficient, allowing for highly visible excitations at subwavelength scale. In this Section, we investigate the s-SNOM response of excitons confined to a planar sheet. To allow for direct comparison with Sec. 4.1 , the sheet is placed on top of a substrate with the same optical characterization: \epsilon_{\mathrm{subs}}(\omega)=2.25 . In the following, we examine the modulus s_{n}=|\bar{\sigma}_{n}| and phase \varphi_{n}=\arg(\bar{\sigma}_{n}) of the normalized field harmonics \bar{\sigma}_{n} . We used numerical integration for the calculation of the effective polarizability ( 24 ) and the normalized harmonics ( 29 ). Details on how these integrals were performed are given in Supplement 1. There, we show that the integrals in Eq. ( 29 ) do not depend on the value of the tapping frequency \Omega . The s-SNOM tip is modeled using a gold sphere of radius R_{\mathrm{tip}}=25 nm, with a Lorentz–Drude model dielectric function that includes corrections in the NIR-VIS range [ 59 ] : \epsilon^{\mathrm{tip}}(\omega)=\epsilon^{\infty}-\frac{\gamma\omega^{\mathrm{tip}}_{0}}{\omega^{2}+i\omega\gamma}+i\left(\frac{\omega_{1}^{\mathrm{tip}}}{\omega-\Omega_{1}^{\mathrm{tip}}}+\frac{{\omega_{1}^{\mathrm{tip}}}^{*}}{\omega+{\Omega_{1}^{\mathrm{tip}}}^{*}}\right), (34) with \epsilon^{\infty}=3.9199 , \hbar\gamma=0.0893 eV, \hbar\omega^{\mathrm{tip}}_{0}=875.79 eV, \hbar\omega_{1}^{\mathrm{tip}}=3.0701+2.9306i eV, and \hbar\Omega_{1}^{\mathrm{tip}}=2.7326-0.69021i eV. For the tip oscillation, we use z_{0}(t)=h_{0}+A[1-\cos(\Omega t)] . Since reducing tip-sample separation enhances the s-SNOM signal [ 12 ] , we set the closest separation between tip apex and sample to h_{0}=26 nm, respecting the geometrical restriction h_{0}>R_{\mathrm{tip}} . The amplitude A was set to 50 nm in agreement with [ 1 , 56 ] . We verified that the incidence angle does not play a relevant qualitative or quantitative role in the resulting signals (see Supplement 1), therefore we fix \theta_{\mathrm{inc}}=70^{\circ} as it is common to use oblique incidence angles in experiments [ 8 , 53 ] . Figure 4: (a) Modulus and (b) phase of 2nd harmonic demodulated signals of the scattered field, normalized to substrate, for different dipole angles \theta_{\mathrm{dip}} . (c,d) Visibility of the left (c) and right (d) peaks of panel (a), defined as (s_{2,\mathrm{peak}}-s_{2,\mathrm{dip}})/(s_{2,\mathrm{peak}}+s_{2,\mathrm{dip}}) . Figures 4 (a) and (b) show the n=2 demodulated harmonics of the 2D excitonic sheet on substrate, for different angles \theta_{\mathrm{dip}} . Large contrast is observed, with signal peaks reaching up to 10 times the pure substrate signal. This is a consequence not only of the large oscillator strength here considered ( f_{0}=2 nm) to match those of strongly bounded dipoles in molecular aggregates [ 11 ] , but also of the substrate being considered. As Fig. 4 already suggested, the two peaks are related to the in- and out-of-plane surface modes mentioned in Sec. 4.2 , given by the poles of the reflection coefficient. As \theta_{\mathrm{dip}} is varied from 5^{\circ} to 90^{\circ} , shifting the molecules away from the sheet’s plane, the peak below \omega_{0} rises. The peak above \omega_{0} decreases with increasing \theta_{\mathrm{dip}} and broadens in width, with its center position shifting towards larger frequencies. The phase spectrum also shows a clear distinction of dipole orientations, however it displays two increasing peaks as the dipole orientation is shifted away from the surface. In Figs. 4 (c) and (d) we present the visibility (s_{2,\mathrm{peak}}-s_{2,\mathrm{dip}})/(s_{2,\mathrm{peak}}+s_{2,\mathrm{dip}}) as a function of f_{0} , showing much larger peak-to-dip contrast at lower oscillator strengths when compared to those of the bare reflectance measurements [Fig. 2 (f)]. This result reveals the potential of s-SNOM to identify and characterize dipole moment orientations with higher resolution and contrast. As a last remark, we would like to bring attention to the shape of the s_{2} curves in Fig. 4 (a). The \theta_{\mathrm{dip}}=90^{\circ} presents a shape that is quite resembling of a Fano resonance, with a dip close to s_{2}=0 at the discrete dipolar resonance \omega=\omega_{0} . As punctuated in related works [ 6 , 1 ] , this indicates an interference of the dipolar surface with the tip in analogy with that of a discrete state with a continuum of states in quantum mechanics. Indeed, we verified that, due to renormalization of the tip’s polarizability by interaction with the dipolar surface, a Fano-like resonance can be verified through the spectral features of the effective polarizability [Eq. ( 24 )], which are present at \theta_{\mathrm{dip}}=90^{\circ} (see Supplement 1). The in-plane dipole component does not couple efficiently with the tip’s point-dipole, however its correspondent s_{2} curve shape also displays a slight asymmetry due to influence of the substrate. The s_{2} shapes for oblique oriented dipoles are a consequence of phase interference between the in-plane- and out-of-plane-projected field components, which, as Sec. 4.1 suggested, generates a mix of destructive and constructive interferences in the spectral range of interest. For this reason, the s_{2} curve shapes for 0<\theta_{\mathrm{dip}}<90^{\circ} are quite more complicated, despite preserving the peak trends as already discussed. 5 Conclusion To account for the out-of-plane dipole orientation, Felderhof’s formalism of linear polarization sheets was reviewed and generalized to an arbitrary sheet geometry and interfacing media. By doing so, we developed a flexible model that provides a unified theoretical language. It applies to 2D sheets arising either from surface-confined excitonic resonances in molecular layers and vdW heterostructures or from plasmonic resonances at metal–dielectric interfaces described by Feibelman’s SRF. Our formalism provides a stepping stone into the development of a complete mesoscopic description of surface polarization fluctuations in 2D-confined regions, bridging the gap between macro- and microscopic electrodynamics. We demonstrated the characterization of average dipole orientation in a layer of excitonic transition dipoles confined to two-dimensions. By examining a planar 2D sheet on a substrate of weak reflectivity, we were able to show distinct reflectance peaks clearly associated with exciton-polaritons originating from to in- and out-of-plane dipole contributions, which interfere destructively when the dipole moments are obliquely oriented. These peaks achieve higher visibility when employing near-field techniques — here exemplified through a s-SNOM setup—, which generates Fano-like resonances that capture the effect of surface modes in the sheet. Funding. São Paulo Research Foundation (2023/14231-6, 2024/05040-5); Independent Research Fund Denmark (2032-00045B). Portuguese Foundation for Science and Technology (UID/04650/2025, PTDC/FIS-MAC/2045/2021). Danish National Research Foundation (Project No. DNRF165). Acknowledgments. V. G. M. D. acknowledges a PhD. Scholarship from the São Paulo Research Foundation (FAPESP), Processes No. 2023/14231-6 and 2024/05040-5. N. M. R. P. and N. A. M. acknowledge the Independent Research Fund Denmark (grant no. 2032-00045B). N. M. R. P. also acknowledges support from the Portuguese Foundation for Science and Technology (FCT) in the framework of the Strategic Funding UID/04650/2025 and by the project PTDC/FIS-MAC/2045/2021. The Center for Polariton-driven Light–Matter Interactions (POLIMA) is funded by the Danish National Research Foundation (Project No. DNRF165). Disclosures. The authors declare no conflicts of interest. Data Availability. All data supporting the findings of this study are available within the article. Supplemental document. See Supplement 1 for supporting content.