Supplemental Material Second harmonic generation enhancement at the percolation threshold Stefano De Zuani,* Tobias Peterseim, Audrey Berrier, Bruno Gompf, and Martin Dressel 1. Physikalisches Institut and Research Center SCoPE, Universität Stuttgart, Pfaffenwaldring 57, 70550 Stuttgart, Germany In situ characterization of the gold thin films In-situ reflectance measurements of the thin gold films were taken during e-beam evaporation, in order to determine at which thickness the percolation regime occurs. A scheme of the set up can be seen in Fig. S1a: during the evaporation, the changes of the reflectance due to the increasing amount of gold on the sapphire substrate are measured by a near infrared spectrometer in specular reflection at 45°. The light from a halogen lamp is guided through fibers into the ultra-high vacuum chamber and a similar fiber is used to collect the specular reflected light and guide it to a near infrared spectrometer. The evolution of the reflectance with increasing film thickness is shown for few thicknesses in Fig. S1b. Fig. S1: (a) Scheme of the in-situ reflectance set-up. (b) Evolution of the reflectance of gold films on a sapphire substrate when the layers thickness increases: at smallest film thicknesses the gold film is an insulator, it goes through the percolation around 5 nm and it becomes metallic at larger thicknesses. The slope of the reflectance curve changes across percolation and gives an indication of the material modifications. At the beginning of the deposition, when the film consists of a few well separated particles, it is insulating and its reflectance in the near infrared regime increases with increasing wavenumber (positive slope). At the percolation threshold, when a conductive network has just been formed, the reflectance becomes almost independent of the wavenumber (flat curve). Finally, above the percolation threshold the films become metallic: the reflectance decreases with increasing wavenumber (negative slope) and reaches almost 100% for the thickest film (not shown). Linear properties of the samples: ex-situ spectroscopic ellipsometry and reflectance measurements All the produced films were characterized by ex-situ spectroscopic ellipsometry and reflectance measurements with a Woollam variable angle spectroscopic ellipsometer (VASE) in the spectral range between 300 nm to 2000 nm with a resolution of 1.5 nm and angles of incidence varied from 45° to 65° in steps of 10°. The ellipsometric data were modeled by a general oscillator layer composed of Lorentz contributions and a Drude component starting at the percolation. We chose here an isotropic model although a random distribution of metallic nanoparticles can show optical anisotropy due to the spatial distribution of the particles; however this seems to be only relevant at smaller wavelengths [S1]. In the spectral frequency region considered here, no particular difference on the resulting dielectric constants can be noticed between an isotropic and a uniaxial model. The mean square error (MSE) of the models, used to obtain the optical constants of the samples was around 3. As additional check to test the quality of the fit, the reflectance of each of the samples was extracted from the model for three different angles of incidence (45°, 55°, 65°) and it was compared to the measured reflectance; the model perfectly reproduces the measured reflectance with an error of around 3%. The obtained real and imaginary parts of the dielectric constant of gold films with different thicknesses are shown in Fig. S2. Fig. S2: Real (a) and imaginary parts (b) of the dielectric function of gold films with increasing thickness obtained from the modeled ellipsometric data. As shown in Fig. S2a, in the lowest frequency range up to 5000 cm-1, ɛ1() rises first with film thickness, reaches a maximum at 5.1 nm and then drops rapidly, becoming negative above 5.5 nm. This abrupt change in ɛ1(ω) is characteristic for the percolation threshold where the film becomes metallic and an additional Drude component appears. The permittivity curves are extrapolated to lower frequencies from the modeled of data in the measured frequency range, in order to explicit the influence of the Drude term starting at the percolation threshold. The extrapolation has been made by assuming that no further resonances are present in the non-measured region below 5000 cm-1. The Drude contribution can be also seen in Fig. S2b where the value of the imaginary part of the permittivity ɛ2() shoots up at the lowest frequency region of the spectra (film thickness = 5.3 nm). Two main plasmonic peaks can be also seen in Fig. S2b: as the clusters become larger, the resonance peaks shift to lower frequencies and they broaden due to an increase in the size dispersion of the islands. The shift of the most pronounced resonances with increasing film thickness is shown in Fig. S3 and the evolution of the Drude components is shown in Table 1. Fig. S3: (a) Imaginary part of the dielectric constant of the gold film with thickness of 5.4 nm with the Drude component and the two more pronounced resonances indicated; (b) shift of the most pronounced resonances with increasing film thickness. Drude components Film thickness ωp/2πc γ (nm) (104 cm-1) (cm-1) 5.3 0.28 1750 5.4 1.81 3870 5.5 2.68 4439 5.7 4.30 3052 5.8 11.13 1095 6.4 18.78 717 Table 1: Drude components of the thin gold films obtained from a Drude fit of the optical data: ω p is the plasma frequency and γ denotes the scattering rate. The plasmon resonances in Fig. S3b shift to lower wavenumber with increasing film thickness. When the Drude component sets in these resonances do not disappear immediately but they become weaker and broader so that it is not possible to distinguish them anymore. At the same time the plasma frequency of the Drude components increases with film thickness and the films become more and more conductive. Influence of the laser pulses on the morphology of the thin gold films Reflectance measurements were carried out before and after laser illumination in order to check the influence of the high power laser pulses on the film structure. It has been recently shown that melting of the gold at the position of the hot spots can induce reshaping of the nanostructures [S2] that in turn can influence the SHG signal. It is therefore important to check that we operate in a regime where the reshaping is negligible. Optical characterization is an extremely sensitive method to determine very small structural changes of the film morphology due to the strong dependence of the plasmon resonances on shape and size of the clusters. In Fig. S4a, the reflectance spectra of samples of different thicknesses before and after the SHG experiment at an average power density of 60 mW/cm2 are shown for an angle of incidence of 45°. Fig. S4: (a) Reflectance of gold films before (solid line) and after (dashed line) they were illuminated by laser pulses (1064 nm) with an average power density of 60 mW/cm2, and (b) Ellipsometric angle Ψ measured at the angle of incidence of 45° of the 5 nm thick gold film before (solid line) and after (dashed line) illumination with the laser. As it can be seen from the spectra displayed in Fig. S4a, basically no changes in the reflectance are observed, which clearly indicates that only very small morphological changes, if any, are induced by the exposure to the laser pulses. In order to confirm this, we compared the ellipsometric angle Ψ measured at the angle of incidence of 45° of the 5 nm thick gold film before (solid line) and after (dashed line) illumination with the laser (Fig. S4b). As we can see the figure confirms the results of Fig. S4a. Only tiny changes above 800 nm can be observed. The very good agreement of both reflectivity curves in Fig. S4a is a further proof than the geometrical modifications are negligible. Increasing the laser power to 225 mW/cm2 leads to slightly larger modifications in the spectra (not shown), but it was checked that the resulting SHG signal stays constant over time and that it is not modified by any significant reshaping of the clusters due to material melting that would cause a closing of the gaps in between the clusters. Skin depth of the gold thin films The skin depth δ of gold was calculated using the equation (S1) [S3]: 𝛿= 𝑐 𝜔𝑘 (S1) Where the extinction coefficient k is extracted from the model of the spectroscopic ellipsometry data of the 30 nm thick gold film. Figure S5: Skin depth of gold calculated using the optical constants extracted from the model of the spectroscopic ellipsometry data of the 30 nm thick gold film. The figure shows how the skin depth is well above the thickness at which the gold thin films percolate over the all visible near-infrared frequency range. * Author to whom correspondence should be addressed. Electronic mail: stefano.de-zuani@pi1.physik.unistuttgart.de References S1. T. W. H. Oates, M. Ranjan, S. Facsko, and H. Arwin, Opt. Express 19, 2014-2028 (2011). S2. V. K. Valev, A. V. Silhanek, Y. Jeyaram, D. Denkova, B. De Clercq, V. Petkov, X. Zheng, V. Volskiy, W. Gillijns, G. A. E. Vandenbosch, O. A. Aktsipetrov, M. Ameloot, V. V. Moshchalkov, and T. Verbiest, Phys. Rev. Lett. 106, 226803 (2011). S3. M. Dressel and G. Grüner, Electrodynamics of Solids (Cambridge University Press, Cambridge, England, 2002).