Photoemission electron microscopy of exciton-polaritons in thin WSe 2 waveguides Tobias Eul, Miwan Sabir, Victor DeManuel-Gonzalez, Florian Diekmann, Kai Rossnagel, Michael Bauer Abstract y Exciton-polaritons emerging from the interaction of photons and excitons in the strong coupling regime are intriguing quasiparticles for the potential exchange of energy during light-matter interaction processes such as light harvesting. The coupling causes an energy anti-crossing in the photon dispersion centered around the exciton resonance, i.e., a Rabi splitting between a lower and upper energetic branch. The size of this splitting correlates with the coupling strength between the exciton and the photonic modes. In this work, we investigate this coupling between excitons and photonic waveguide modes excited simultaneously in thin-film flakes of the transition-metal dichalcogenide WSe 2 . Using a Photoemission electron microscope, we are able to extract the dispersion of the transverse electric and magnetic modes propagating through these flakes as well as extract the energy splitting. Ultimately, our findings provide a basis for the investigation of the propagation of exciton-polaritons in the time-domain via time-resolved photoemission. keywords Light-Matter Interaction, Strong Coupling, Exciton-Polaritons, Photoemission Electron Microscopy, WSe 2 † † email: eul@physik.uni-kiel.de † † affiliation: Institute of Experimental and Applied Physics, Kiel University, 24098 Kiel, Germany † † affiliation: Institute of Experimental and Applied Physics, Kiel University, 24098 Kiel, Germany † † affiliation: Kiel Nano, Surface and Interface Science KiNSIS, Kiel University, 24118 Kiel, Germany † † affiliation: Institute of Experimental and Applied Physics, Kiel University, 24098 Kiel, Germany † † affiliation: Kiel Nano, Surface and Interface Science KiNSIS, Kiel University, 24118 Kiel, Germany 1 Introduction Excitons, bound electron-hole pairs, are the focus of numerous experimental and theoretical research efforts based on their potential for light-driven energy harvesting. From the many semiconducting materials exhibiting excitonic resonances, transition-metal dichalcogenides (TMDC) have garnered special interest in recent years since they exhibit excitons with large binding energies ( 1 ; 2 ; 3 ; 4 ) . Embedding TMDC films or even monolayers within micrometer-sized cavities creates a suitable environment for a strong interaction between excitons and photons ( 5 ; 6 ; 7 ; 8 ; 9 ; 10 ) . This interaction facilitates the formation of the light-matter quasiparticle termed exciton-polaritons ( 11 ; 12 ) . Its hybrid nature promises new avenues for the energy exchange between light and matter and thus new technological developments. The transfer of energy via the propagation of such a quasiparticle is especially interesting. Therefore, the interaction of excitons with propagating light modes rather than spatially confined cavity photons may provide another suitable environment for exciton-polaritons. Essentially, thin films of TMDC materials themselves host the necessary waveguiding modes, which are able to interact with the intrinsic exciton resonances ( 13 ) . Recent near-field studies involving scanning-near field microscopy (SNOM) ( 14 ; 15 ) as well as cathode luminescence (CL) ( 16 ) focused on this particular approach for exciton-polaritons and discussed their spectroscopic signatures in thin films of WSe 2 and MoSe 2 . The dispersion of these polaritons exhibits a distinct energy anti-crossing centered around the exciton resonance, as illustrated in Fig. 1 (a). Figure 1: Exciton-polariton dispersion and experimental scheme. (a) Schematic illustration of the exciton-polariton dispersion. (b) Experimental setup with thin WSe 2 flakes situated on a transparent glass substrate. A laser illuminates the flake from the bottom, where it excites the exciton-polaritons as well as generates the photoelectrons, which enter the PEEM lens system. Photoemission electron microscopy (PEEM) is another experimental technique that demonstrates significant sensitivity to the electric near-field and surface waves in both metallic ( 17 ; 18 ; 19 ; 20 ; 21 ; 22 ; 23 ; 24 ; 25 ; 26 ) and photonic materials ( 27 ; 28 ; 29 ; 30 ; 31 ; 32 ; 33 ) . Furthermore, it is possible to measure dispersion relations of these propagating modes by tuning the wavelength of the exciting laser ( 27 ; 34 ; 35 ; 36 ) . The direct imaging of the surface in PEEM in combination with pump-probe excitation schemes additionally facilitates tracing the propagation of waves and quasiparticles along the surface in the time domain ( 37 ; 38 ; 21 ; 39 ) . This makes it a promising method for tracing the propagation of exciton-polaritons in real-time. Ultimately, time-resolved PEEM may even reveal highly local information on the energy transfer, i.e., the Rabi oscillation, between exciton and waveguide mode on an ultrafast timescale. 2 Results and Discussion In this work, we demonstrate the capability of the PEEM technique to measure the dispersion relation of exciton-polaritons in thin WSe 2 films and determine the coupling strength between the A-exciton and photonic waveguide modes from this data. Fig. 1 (b) depicts our experimental geometry. A laser with a tunable center wavelength in the range of 695900\text{\,}\mathrm{nm} corresponding to photon energies E_{\mathrm{Ph}} between 1.38\text{\,}\mathrm{eV} and 1.78\text{\,}\mathrm{eV} illuminates the TMDC sample from the bottom, and allows us to spectrally probe the dispersion around the A exciton in WSe 2 at E_{\mathrm{ex}}=$1.6\text{\,}\mathrm{eV}$ ( 14 ; 16 ) . A high electric field accelerates the photoelectrons carrying the information on the quasiparticle interactions into the PEEM column. The work function of WSe 2 , which must be overcome in the photoemission process, is in the order of 4.3\text{\,}\mathrm{eV} ( 40 ) and is considerably higher than our photon energies. We reduce the work function of our samples using a procedure that is customary for the PEEM-method ( 41 ; 42 ; 35 ; 25 ; 36 ) : We evaporate a fraction of a monolayer of an alkali metal (K) on the surface until it is low enough to enable the emission of photoelectrons via two-photon photoemission (2PPE). In this limit, we are able to generate photoelectrons with the same laser pulses that excite the exciton-polariton. In principle, our study would also be possible without lowering the work function, using three-photon photoemission. We discuss the substantial limitations of this approach in the Supplementary information (SI). We prepared the WSe 2 flakes on indium-tin-oxide (ITO) coated glass substrates via the mechanical exfoliation method. This method yields flakes with clean, flat surfaces and well defined edges albeit with varying thicknesses ( 20150\text{\,}\mathrm{nm} ) and lateral dimensions ( 550\text{\,}\mathrm{\SIUnitSymbolMicro m} ). Prior to the PEEM experiments, we measured the thickness of the flakes with an atomic force microscope (AFM). Fig. 2 (a) shows the AFM image for an exemplary WSe 2 flake, which exhibits a thickness of approximately 55\text{\,}\mathrm{nm} as well as sizes of 30\text{\,}\mathrm{\SIUnitSymbolMicro m} and 25\text{\,}\mathrm{\SIUnitSymbolMicro m} in x- and y-direction, respectively. A more detailed description of our methods follows the main text below. Figure 2: Waveguide mode excitation in WSe 2 investigated using PEEM. (a) AFM image of a WSe 2 flake with a thickness of approximately 55\text{\,}\mathrm{nm} . (b) PEEM image of the same flake recorded with a photon energy of 4.9\text{\,}\mathrm{eV} . The dashed circle (FOV) indicates the field of view of the detailed measurement in (c,d). The dashed square (ROI) indicates the region of interest for the extraction of the wave patterns. (c,d) PEEM images recorded with photon energies of 1.38\text{\,}\mathrm{eV} (left) and 1.72\text{\,}\mathrm{eV} (right) with the polarization parallel and perpendicular to the upper edge of the flake, respectively. (e,f) Extracted wave patterns for parallel and perpendicular polarization as a function of excitation photon energy, respectively. The dashed line indicates the energy E_{\mathrm{Ex}} of the A-exciton of WSe 2 . The corresponding PEEM image in Fig. 2 (b), recorded with a mercury discharge lamp at a photon energy of E_{\mathrm{ph}}=$4.9\text{\,}\mathrm{eV}$ , shows the topography based on work function differences of the substrate and the WSe 2 flake. The dashed circle (FOV) in Fig. 2 (b) highlights the field of view, where we zoomed in on the sample to get close up images of the optical near-field of the waveguide modes. The dashed square (ROI) on the other hand marks the region of interest used for the subsequent quantitative analysis. For the experiment, we focus our laser onto the edge of the flake visible at the border of the ROI. Here, the light can couple into different waveguide modes, depending on the in-plane polarization with respect to this edge. A guided mode propagates through a planar waveguide by internal reflection upon the top and bottom interfaces. The polarization dependent phase change upon this reflection determines the character of a mode in the waveguide. Considering the two typical cases of excitation with s- and p-polarized light with respect to the waveguide geometry, two types of modes can be excited in a planar waveguide. S-polarized light couples to transverse electric (TE) modes, where the electric field is perpendicular to the propagation direction. P-polarized light couples to transverse magnetic (TM) modes, where the magnetic field is perpendicular to the propagation direction. Since the wave propagates perpendicular to the excitation edge, we couple into a TE-mode for a laser polarization parallel to the edge. Similarly, the laser polarization perpendicular to the edge excites a TM-mode. When these waveguide modes propagate through the flake, its electric field interferes with that of the second photon in the 2PPE-process resulting in a wave pattern in the photoemission contrast on the surface ( 43 ; 20 ) . To investigate the dispersion relation of these modes, we excite the flake edge with varying photon energies around the exciton resonance and examine the photoemitted wave patterns. Figs. 2 (c,d) show exemplary images for the photon energies E_{\mathrm{ph}}=$1.38\text{\,}\mathrm{eV}$ and E_{\mathrm{ph}}=$1.72\text{\,}\mathrm{eV}$ , respectively, recorded with the polarization parallel (top) and perpendicular (bottom) to the edge. In the case of parallel polarization, i.e. for excitation of a TE-mode, we see a distinct wave pattern for the low photon energy, whereas just one single period of a wave is visible for the large photon energy. This result was to be expected based on the absorption properties of WSe 2 above the energy E_{\mathrm{Ex}}\approx$1.6\text{\,}\mathrm{eV}$ of the A-exciton in WSe 2 ( 14 ) . This absorption influence should be even more apparent in our experiment due to the nonlinear 2PPE excitation scheme. The wave patterns also clearly exhibit different periods hinting at the dispersion of the photonic mode. For excitation of a TM-mode on the other hand, the wave pattern remains observable for both photon energies. Figs. 2 (e,f) illustrate this more clearly in a comparison of the wave patterns at all measured photon energies. For the data comparison, we sum up the intensity in the ROI of the PEEM image along the direction of the edge. As expected, the waves for the TE-mode in (e) show a decreasing propagation length with increasing photon energy. Surprisingly, in the case of the TM-mode in (f) the propagation length does not differ significantly over the measured spectral region. Since the exciton resonance is largely responsible for the absorption in this spectral region, it appears as if the waveguide mode excited via perpendicular polarization does not interact with the exciton, at least for this particular thickness of the flake. This observation will be discussed further below. Figure 3: Waveguide mode dispersion. Dispersion curves extracted from the Fourier transformation of measured and simulated wave patterns. The wave numbers k are relative to the wave number of light in vacuum k_{0} . (a,c) Experimental data for TE- and TM-mode, respectively. (b,d) FDTD simulations TE- and TM-mode, respectively. Simulated data for a flake thickness of 55\text{\,}\mathrm{nm} . The next step of the analysis is to directly extract the dispersion of the photonic modes from the wave patterns. By applying the Fast Fourier Transform (FFT) method to the wave patterns, we can analyze their periods in terms of the amplitude in Fourier space depending on the wavenumber k relative to the wavenumber k_{0} of light in free space. Since the wave patterns are not ideal signals with decay and constant offset, we apply a high-pass filter prior to the FFT, which removes low-frequency contributions from the signal. Fig. 3 shows the results of this method in dispersion plots consisting of the extracted Fourier signal after normalization to its respective maximum for each photon energy and subsequent interpolation. The normalization is necessary, since the experiment conditions such as focal spot size, intensity and, most importantly for the 2PPE process, the pulse durations vary for each individual photon energy. Images (a,c) show the experimental data obtained from the wave patterns in Figs. 2 (e,f), respectively. The dispersion for the TE-mode in (a) clearly shows one of the characteristic signatures of strong coupling between exciton and photonic mode, i.e., the lower polariton branch bending in the vicinity of the exciton resonance. The upper branch is not visible due to the strong absorption in this spectral region. Examining the result for the TM-mode in (c), we first of all notice the same feature as in (a), albeit with a low intensity. This stems from the laser polarization being not perfectly perpendicular to the edge of the WSe 2 flake. More importantly, we see a linearly dispersing feature that does not show any deviation from its straight line around the exciton resonance. As already indicated above, this mode obviously does not interact with the exciton. To explain our observations, we additionally simulated the experiment by means of the finite-difference time-domain method ( 44 ) and using the optical constants for WSe 2 from Ref. ( 45 ) . We use a broadband gaussian source, which covers the spectral range of the experimental data, and focus it on an edge of a thin film of WSe 2 . Equivalently, we extract the electric field wave patterns from a frequency monitor situated in the WSe 2 film and subsequently apply the FFT method. Here, we forego the normalization of the Fourier amplitude for each individual photon energy to retain the information about the absorption. Images (b,d) show the simulated dispersions corresponding to the experimental data in (a,c), i.e., a thickness of 55\text{\,}\mathrm{nm} and E_{\parallel} , E_{\perp} , respectively. They reproduce the experiment reasonably well, albeit with a difference in the absolute values of k/k_{0} , which is probably due to differences in the optical constants for the simulations and the actual sample. For an excitation with E_{\parallel} in (b), the lower branch disperses similarly to the one obtained from the experiment. However, here we can additionally see the upper branch with a significantly reduced intensity and larger linewidth than the lower branch. The analysis of the individual electric field components (see Fig. S3 in the SI) confirms our initial assignment between laser polarization and the two mode types within the waveguide. The largest contribution to the total electric field, and therefore the dispersion signal, comes from the component parallel to the excitation edge. Since the waveguide mode propagates in the direction perpendicular to the edge, we can attribute the dispersive feature to a transverse electric (TE) mode. Similarly, we determine from the simulation data in Fig. 3 (d) and Fig. S3 that the light with E_{\perp} indeed couples into a transverse magnetic (TM) mode. As in the experiment, the dispersion in this case is a straight line, again without any sign of interaction with the exciton. The experimental data and simulation in Fig. 3 (c) and (d) also exhibits another feature at k/k_{0}=1 , i.e., a mode that follows the dispersion of light in free space. This additional feature stems from the excitation source. Additional simulations in the SI and similar results in Ref. 14 show that for increasing flake thickness the exciton also starts to couple to the TM-mode. Interestingly, for both the TE- and the TM-mode, there seems to be a threshold for the waveguide thickness that determines the coupling to the exciton. For the TE-mode this thickness threshold lies between 20\text{\,}\mathrm{nm} and 30\text{\,}\mathrm{nm} , whereas the waveguide needs to be thicker than 60\text{\,}\mathrm{nm} for the TM-mode to couple. When analyzing the electric field localization within the waveguide (see Fig. S5 and Fig. S7), it becomes clear that the exciton coupling depends on how strongly the light is confined within the waveguide. In the case of the TE-mode, the waveguides simply become too thin. The electric field is still very localized, but too much of the light resides outside of the waveguide below 30\text{\,}\mathrm{nm} . For the TM-mode however, the electric field amplitudes starts to delocalize for thicknesses below 60\text{\,}\mathrm{nm} , which is why the threshold is different from the TE-mode. To summarize, the data show that for this particular thickness of the WSe 2 flake, we are able to control the exciton-photon coupling with the laser polarization being the experimental control parameter. We can either couple into the TE-mode for strong coupling or into the TM-mode for a freely propagating waveguide mode. Ultimately, we are interested in the coupling strength of the exciton-polariton, which can be determined from the energy splitting in the dispersion curve. For this, we examine another WSe 2 flake with a significantly smaller thickness so that the absorption is overall reduced. As mentioned above, in this thickness range only the TE-mode couples to the exciton. Fig. 4 (a) shows the PEEM image for this flake recorded at E_{\mathrm{ph}}=$1.55\text{\,}\mathrm{eV}$ with the polarization of the laser oriented parallel to the excitation edge at the bottom. The dashed square (ROI) marks the region of interest for the analysis of the wave patterns. Since the absorption is now reduced, we can resolve both the lower and the upper branch of the exciton-polariton in the dispersion curve in Fig. 4 (b). Due to the individual normalization of the signal for each photon energy, we do not see a gap around the exciton resonance. Nevertheless, we can assign an energy splitting between the two branches based on the turning points within the two polariton branches. This becomes more evident when analyzing the corresponding FDTD simulations for a thickness of 30\text{\,}\mathrm{nm} in Figs. 4 (c,d). Figure 4: Strong-coupling energy splitting. (a) PEEM image of a WSe 2 flake recorded with a photon energy of 1.55\text{\,}\mathrm{eV} and laser polarization oriented parallel to the bottom edge. The dashed square (ROI) indicates the region of interest for the extraction of the wave patterns. (b) Dispersion of the TE-mode evaluated from the experimental data. (c,d) Simulated dispersion for a thickness of 30\text{\,}\mathrm{nm} without and with normalizing the signal of each individual frequency. (e) Extracted maximum FFT signals from (c) for determining the energy splitting based on exponential reduction of the branch intensities. (f) Extracted maxima positions from (b) and (d) for determining the energy gaps based on turning points in the dispersion. The two simulated dispersion plots show the same dispersion data without normalization in (c) and with normalization of the signal for each photon energy equivalent to the experimental data in (d). From the data without normalization, it is possible to extract the energy splitting in a straightforward manner by analyzing the intensity distribution of the two branches. Fig. 4 (e) shows the relative maximum intensity of the branches extracted for each photon energy. In both branches we determine the point where the intensity has decreased to the \frac{1}{e} -fraction. We take the energy difference between these two points as a measure for the energy splitting of the dispersion curve. For the examined flake thickness, the analysis yields an energy splitting of \Delta E=$81.5\text{\,}\mathrm{meV}$ . The dashed lines in Fig. 4 (c) mark the extracted limits and the splitting. For the normalized data, we evaluate the energy splitting based on the two turning points in the dispersion curve located between the two branches that are clearly visible in Fig. 4 (d). For the quantitative analysis, we extract the k/k_{0} -values of the maximum in the dispersion for each photon energy and approximate this k(E_{\mathrm{ph}} ) relation by applying a Savitzky-Golay filter on the data points. Fig. 4 (f) shows this relation in orange for the simulated data. The curve shows a clear local maximum and minimum, which we identify as the k/k_{0} -values of the turning points. The corresponding photon energies at these points yield a value for the energy splitting of \Delta E=$84.4\text{\,}\mathrm{meV}$ . This value is very close to the one extracted from the unnormalized data, which validates this approach. We therefore apply the equivalent procedure to the experimentally determined dispersion curve. The data from this analysis, shown as a blue line in Fig. 4 (f), yields an energy splitting of \Delta E=$94.8\text{\,}\mathrm{meV}$ for the exciton-polariton. This value is in good agreement with the results of a SNOM study on exciton-photon coupling in MoSe 2 flakes ( 15 ) . Conversely, the above mentioned CL study on WSe 2 flakes ( 16 ) discusses larger values of up to 300\text{\,}\mathrm{meV} . This discrepancy stems from two effects present in the CL experiment. Firstly, the excitation with the electron beam in the CL experiment results in much higher exciton densities in the interaction volume with the waveguide mode compared to the light induced excitation used in this study, which enhances the exciton-polariton coupling ( 46 ) . Secondly, electron beams with sufficiently high beam energy generate Cherenkov radiation, which further increases the interaction strength with the excitons ( 47 ) . 3 Conclusion In summary, we studied and analyzed the coupling strength between photonic waveguide modes and the A-exciton in thin-film WSe 2 flakes. By varying the photon energy as well as the polarization of the exciting laser, we were able to extract the dispersion relation of different modes from the near-field wave patterns recorded with PEEM. Additional FDTD simulations reproduced the experimental dispersion curves and provided us another avenue to analyze the strong coupling and the associated energy splitting between the lower and upper polariton branch. The simulations affirmed the nature of the modes, i.e., TE or TM for a polarization parallel or perpendicular to the excitation edge, respectively. With our experiment, we ascertained the challenge in accessing the upper polariton branch using nonlinear photoemission due to a strong absorption in WSe 2 for this spectral region. Nevertheless, we were able to extract the energy splitting from the turning points in the normalized dispersion curve, with \Delta E in the order of 100\text{\,}\mathrm{meV} for the exciton-polariton formed from the TE-mode. In comparison, PEEM provides similar information as the scanning near-field microscopy technique employed in previous work on exciton-polaritons ( 14 ; 15 ) . For example, the extracted energy splitting compares well to the splitting of 100\text{\,}\mathrm{meV} observed with SNOM in thin films of MoSe 2 . While SNOM allows to perform experiment under ambient conditions, PEEM, in combination with varying the laser polarization, is more flexible in selectively exciting the different TE or TM modes. Additionally, the parallel imaging of the surface, in contrast to scanning over it, allows for a faster data acquisition, and therefore facilitates an easier combination with pump-probe techniques for ultrafast time-domain studies. We expect that time-resolved PEEM will provide a new avenue to investigate the propagation of exciton-polaritons as well as the periodic energy exchange between the involved quasiparticles in real time. Experimental Details Photoemission electron microscope The photoemission electron microscope (Focus GmbH) consists of an electrostatic lens system, which images the lateral distribution of photoelectrons emitted from the sample surface. An extractor lens situated directly in front of the sample collects the photoemitted electrons with a high voltage electric field, e.g. V_{\mathrm{ext}}=$10\text{\,}\mathrm{kV}$ for this study. The detector is a combination of a multichannel-plate (MCP), which multiplies the photoelectrons, and a phosphor screen, whose illumination is captured by a CCD-Camera (LaVision). The photoelectrons are being detected in an energy integrated manner with a high spatial resolution of \Delta x\approx$50\text{\,}\mathrm{nm}$ . The overall system is contained within an ultrahigh vacuum chamber with pressures typically in the order of p\leq$5\text{\times}{10}^{-10}\text{\,}\mathrm{m\mbox{\normalfontbar}}$ . The system additionally features a wire shaped alkali metal dispenser (SAES Getters) for the evaporation of small doses of potassium onto sample surfaces. For the photoexcitation, the sample can be illuminated by laser sources through two different vacuum chamber entrance windows, either under a grazing incidence of 65\text{\,}\mathrm{\SIUnitSymbolDegree} or under normal illumination from the back side of the sample. The latter case additionally allows a very tight focusing of the laser beam down to 5\text{\,}\mathrm{\SIUnitSymbolMicro m} spot sizes via a lens in close proximity to the sample. More information on this normal incidence configuration can be found in ( 48 ) . Additionally, the PEEM is equipped with a mercury discharge lamp (Focus GmbH) providing continuous wave, unpolarized light with a photon energy of 4.9\text{\,}\mathrm{eV} , which is primarily used for overview images of the sample surface and adjusting based on work function contrasts. Laser system Our laser system is a pulsed Titanium-Sapphire (Ti:Sa) Oscillator system (Tsunami, Spectra Physics) operated at a repetition rate of 80\text{\,}\mathrm{MHz} and pumped with P_{\mathrm{pump}}=$7.5\text{\,}\mathrm{W}$ of the second harmonic of a diode pumped Neodymium-Yttrium Vanadate laser (Millennia Pro 10s, Spectra Physics). The central wavelength of the Ti:Sa output is tunable in the range of 700900\text{\,}\mathrm{nm} . Depending on the output spectrum, the duration of the outgoing laser pulses varies in the range of 100130\text{\,}\mathrm{fs} . Similarly, the output power lies within the range of P_{\mathrm{out}}= 0.51\text{\,}\mathrm{W} . Between the laser system and the PEEM no additional pulse compression was used because the priority of the experiment was a stable pointing of the beam onto the sample while tuning the central wavelength of the laser. Two lenses used as a telescope, the vacuum chamber entrance window, an in-vacuum lens and a half wave plate are present in the beam path and lengthen our pulse duration upon beam transmission. All reflective optics used were optimized for minimal group delay dispersion. Alignment of sample and laser The half wave plate is situated directly before the vacuum chamber entrance window to control the in-plane polarization of the laser upon illuminating the sample under normal incidence. The alignment of the laser with respect to the WSe 2 flake is very crucial due to the rather small spot sizes of the laser in the order of 520\text{\,}\mathrm{\SIUnitSymbolMicro m} at the sample. Ideally, the center of the beam should be on the excitation edge of the flake, to achieve the best coupling to the waveguide mode. Unfortunately, this would result in a very strong overexposure of the edge in the photoemission signal and very low contrast in the rest of the field of view. Therefore, we position the spot in a way that enough of the gaussian profile clips the excitation edge to launch the waveguide mode, while the intensity maximum is in the center of the field of view. This skews the overall intensity of the wave and makes it appear as if the waves decay away from the center of the field of view. Depending on the photon energy this effect is more or less pronounced as evident from Fig. 2 . This is why we apply a high pass Fourier filter to the PEEM images along the propagation direction and normalize the data for each photon energy individually. Sample fabrication Single crystals of 2Hb-WSe 2 were grown by iodine vapor transport. High-purity tungsten wire and selenium pellets were sealed at one end of evacuated quartz ampoules (inner diameter 12\text{\,}\mathrm{mm} , length 200\text{\,}\mathrm{mm} ) in a near stoichiometric ratio with approximately 15\text{\,}\mathrm{\%} selenium excess ( 3\text{\,}\frac{\mathrm{mg}}{\mathrm{cm}} ) together with iodine granules ( 5\text{\,}\frac{\mathrm{mg}}{\mathrm{cm}} ). The ampoules were subjected to a temperature gradient of 60\text{\,}\mathrm{\SIUnitSymbolCelsius} , with the hotter end containing the starting materials at 900\text{\,}\mathrm{\SIUnitSymbolCelsius} . The growth time for crystals forming near the colder end was 900\text{\,}\mathrm{h} . The size of the crystals were up to 5x5 mm 2 in area and 0.1\text{\,}\mathrm{mm} thick. The doping of the semiconducting samples is typically p-type ( 49 ) . We prepare our samples via mechanical exfoliation of the bulk WSe 2 crystal. For this, we press a semiconductor wafer dicing tape on top of the bulk and remove the topmost layers. With a second strip of this tape, we repeat the process on the material stuck on the first tape and afterwards the same with a third tape. Finally, we press the third tape on top of our glass substrate to transfer the WSe 2 layers. The glass substrate (Präzisions Glas & Optik GmbH) is a standard soda-lime-glass with a thickness of 1.1\text{\,}\mathrm{mm} covered with a few nano meters of a SiO 2 passivation layer. On top of this is a sputter deposited thin film of indium-tin oxide with a thickness of 21\text{\,}\mathrm{nm} and a resistance of 100\text{\,}\mathrm{\SIUnitSymbolOhm} . Prior to the PEEM experiments, we evaporate a fraction of a monolayer of potassium onto the sample surface to reduce the work function. For this, we operate the K-dispenser at I=$5.7\text{\,}\mathrm{A}$ for 60\text{\,}\mathrm{s} from a distance of approximately 5.5\text{\,}\mathrm{cm} to the sample. Over time and especially after illumination with the laser, the potassium level on the WSe 2 surface decreases and needs to be refreshed. This effect occurs faster for illumination at the exciton resonance. Finite-Difference Time Domain Simulations The simulation environment spans a volume of 25\text{\,}\mathrm{\SIUnitSymbolMicro m} \times 3\text{\,}\mathrm{\SIUnitSymbolMicro m} \times 4\text{\,}\mathrm{\SIUnitSymbolMicro m} with an uniform 20\text{\,}\mathrm{nm} \times 20\text{\,}\mathrm{nm} \times 100\text{\,}\mathrm{nm} mesh grid in the region of the waveguide. The mesh step in z-direction is quite large to reduce simulation size and time. We ensured its validity by reproducing the results of exemplary simulations with \Delta z = 4\text{\,}\mathrm{nm} . The environment is surrounded by perfect-matching layer boundaries and the total run time per simulation is 200\text{\,}\mathrm{fs} . The WSe 2 waveguide sits on top of a 21\text{\,}\mathrm{nm} thin Indium-Tin Oxide film and a glass substrate filling the simulation environment in the negative z-direction. Its excitation edge is located at x = -4\text{\,}\mathrm{\SIUnitSymbolMicro m} with propagation in the positive x-direction. All materials fill the environment in the y-direction. The source is given by a Gaussian beam with a waist radius of 5\text{\,}\mathrm{\SIUnitSymbolMicro m} at the excitation edge of the waveguide. Its spectrum is centered around 800\text{\,}\mathrm{nm} with a full width at half-maximum of 100\text{\,}\mathrm{nm} and a temporal width of about 16\text{\,}\mathrm{fs} . A 2D-field monitor located at the top surface of the WSe 2 waveguide records the electric field components in the wavelength range of 650950\text{\,}\mathrm{nm} . Atomic Force Microscope We use a commercial atomic force microscope (Veeco, Nanoscope V) and tip (Budgetsensors, TAP 150). The AFM is installed on an anti-vibration table. The aluminum coated tip has a resonance frequency of about 150\text{\,}\mathrm{kHz} and we measure the height of the WSe 2 flakes in tapping mode. Data Availability Statement The data that support the findings of this study are openly available in Zenodo at http://doi.org/10.5281/zenodo.14168816 . Author information Author contributions M.B. conceived the ideas and designed the experiments. T.E. carried out the PEEM experiments, the FDTD simulations and is responsible for the data evaluation. F.D. and K.R. synthesized the WSe 2 crystals. M.S. performed the mechanical exfoliation of the WSe 2 flakes onto the ITO substrates. V.D.G. carried out the AFM measurements. T.E. and M.B. wrote the manuscript. Funding The PEEM study and the sample growth was supported by the German Research Foundation (DFG) through Projects 499426961 and 434434223 (CRC 1461), respectively. 4 Supporting Information Supporting Information Available: Additional results measured in the 3PPE regime and additional simulations of the electric field components, thickness dependent dispersion and the field localizations in the waveguide.