1. Introduction

Dielectric loaded surface plasmon waveguides (DLSPPWs) are comprised of a dielectric ridge deposited onto a metallic film [1–5] or strip [6]. At telecom frequencies, single-mode DLSP-PWs feature a typical cross-section of 0.5 × 0.5 μm² compatible with their interfacing with a silicon on insulator dielectric circuitry [7–9]. Beyond compact passive devices, DLSPPWs are also of interest for dynamically controlled devices. For example, stand-alone DLSSPW based thermo-optical switches have been demonstrated [10–12] and integrated into SOI motherboards for high bit rate traffic applications [13, 14]. So far, thermal control of DLSPPWs is one of the most practical way to activate dynamic plasmonic based components. Except in very few previous works [15], this activation has been achieved by flowing a current along the metallic parts of the plasmonic waveguides. However, a plasmonic waveguide can also be a good light-heat converter when illuminated at frequency highly absorbed by the metal of interest. Photo-thermal effect is then an alternative approach for the activation of thermo-optical DL-SPPW devices which can be of practical interest for low-power consumption devices. Indeed, this specific type of ”all-optical” heating can be implemented to minimize power consumption possibly down to a level that will be difficult to reach with electrical activation. For example, as far as micro-devices are considered, photo-thermal activation can be operated using highly focused beams. Moreover, the devices can be specifically engineered to become highly absorbing at the pump frequency resulting in an optimum light-heat conversion.

In this work, we investigate the use of polymers doped with gold nanoparticles (NPs) to minimize the photo-activation power of DLSPPW thermo-optical components. Photo-thermal properties of NPs [16–18] have attracted much interest in the recent years for applications in fields ranging from nanomedicine (for a review see for example [19]), biological applications [20], material processing [21,22], micro-fluidic [23], nanphotonics [24,25] but NP-doped materials have been investigated only little in the context of dynamically controlled integrated optics devices. The DLSPPW devices we consider here are stand-alone 1 × 2 switches relying on a Multi-Mode Interferometers (MMI) design previously investigated theoretically and experimentally in Refs [26] and [27] respectively. Unlike these previous studies, the activation of the MMI switches is performed by using a pump beam with a frequency within the absorption band of the embedded NPs. On the basis of leakage radiation microscopy images, the photo-activated thermo-optical switches are characterized over the telecom frequency range. We show that the power necessary to photo-activate the NP-doped switches with a gold/polymer volume ratio in the range 0.5% is reduced by a factor 2.5 compared to undoped switches. This factor is attributed to the enhanced absorption of the pump beam within the NP-doped polymer and not to a change of the thermo-optical properties of the polymer in the presence of the gold nanoparticles. Our experimental results indicate that significant further minimization of the activation power is still possible.

2. Samples fabrication

The samples were fabricated by electron-beam lithography applied to NP-doped or undoped polymethyl methacrylate (PMMA) layer spin-coated onto a 80 nm-thick gold layer thermally evaporated onto a glass substrate. The NP-doped PMMA was obtained by dissolving solid PMMA (molecular weight 120K) into toluene and by subsequently mixing the resulting solution with commercially available (Aldrich, Ref. 660434) gold nanoparticles (diameter ranging from 3 to 5nm) dispersed into toluene as well. When spin-coated onto a gold layer, the NP-doped PMMA layers feature pronounced colors ranging from light green to dark pink depending on layer thickness. Prior to the exposure, the PMMA layers were baked on a hot plate. For the NP-doped PMMA solution used for the fabrication of the samples investigated in the following, the volume occupied by the gold NP into the baked doped PMMA layers was estimated to be 0.52% of the solid polymer volume. This volume ratio has been computed from the density (0.9308) and gold concentration (2 % w/v) of the NP solution and from the known density of the 120K PMMA (1.118) and the experimentally measured concentration (0.22g/mL) of solid PMMA into the prepared PMMA-toluene solution. The volume ratio of 0.52% corresponds to a doping level of 2 × 10¹⁷ NP per cm³ of solid polymer and a side-to-side average distance between the NPs in the range of 10 to 15nm. PMMA being a positive electron-beam resin, the switches were fabricated by exposing and eventually dissolving areas surrounding the plasmonic waveguides. In spite of the presence of the NPs, the optimum exposure doses (165μC/cm²) were found to be similar for the doped and undoped PMMA layers.

Scanning Electron Microscope (SEM) images of the switches fabricated following this process with the NP-doped PMMA resin are shown in Fig. 1. The MMI switches we considered are equipped with an input grating coupler (period= 2.45μm) (see Fig. 1(a)) allowing a very efficient excitation of the device by means of a moderately focused beam (spot size = 30μm) impinging at an angle of incidence of 30° for free-space wavelengths ranging from 1500nm to 1600nm [28]. The switch is comprised of the single-mode DLSPPW (labeled as SM in Fig. 1(b)) and a multi-mode waveguide with a typical length of 140μm and finally a Y-junction at the output end of the device. For a switch relying on a dual-mode interference, the optimum length L of the multi-mode waveguide is given by $2L={\lambda }_0/\left|\mathrm{\Delta }{n}_e^{(1)}-\mathrm{\Delta }{n}_e^{(2)}\right|$ where λ₀ is the free space wavelength and where Δn_e⁽ⁱ⁾ (i = 1, 2) denote the change of effective index of each mode involved in the switching process for a given change in temperature.

 

Fig. 1 (a) SEM image of a NP-doped plasmonic MMI switch (scale bar=50μm). The dark regions are coated with polymer, light-gray regions correspond to bare gold film. (b) Zoom onto the region corresponding to the dashed perimeter shown in (a) showing the single-mode (SM) and multi-mode (MM) waveguides (Scale bar=10μm). (c) Detail of the transition between the single-mode and multi-mode regions (Scale bar= 1 μm). The nanoparticles are clearly visible on the DLSPPWs.

Download Full Size | PDF

The excitation of the MMI waveguide is achieved by a straight single-mode DLSPPW laterally shifted with respect to the multi-mode waveguide as shown on the SEM image displayed in Fig. 1(c). This lateral shift is used to control the relative amplitude of the modes excited in the MMI region. The presence of the NP is clearly visible in Fig. 1(b) and 1(c). In particular, we note in Fig. 1(c) that the NPs have coagulated during the fabrication process. The NP distribution observed at the top surface of the polymer was also visible on the SEM images of the vertical side walls of the waveguides indicating that the NPs are distributed in the volume of the doped PMMA layer. We note however, that, from the SEM images of the waveguide side-walls, we could only hardly estimate the NP density in the polymer as a function of the distance to the gold film. Finally, an atomic force microscope imaging of the waveguides reveals a roughness at the top surface of the doped waveguides with an average peak-to-peak height of 12nm significantly degraded compared to undoped structures (average peak-to-peak roughness around 4nm).

3. Radiation leakage characterizations of the MMI swicthes

3.1. Experimental set-up and MMI imaging

The impact of the presence of the NP onto the waveguiding properties of the doped DLSP-PWs is evaluated by using the leakage radiation microscope (LRM) set-up shown in Fig. 2(a). The near-infrared excitation of the switches was achieved using a fiber-focuser illuminating the grating coupler described in the previous section. The optimum coupling was obtained for a position of the polarization loops corresponding to a transverse magnetic (TM) orientation of the incident field. For the photo-thermal activation of the switches, a pump beam (CW free-space wavelength 532nm) was focused by two perpendicularly oriented cylindrical lenses onto the plasmonic devices. By adjusting carefully the distance between the two cylindrical lenses, an elliptically shaped pump beam with a typical size of 100×30μm² was obtained. This elliptical beam was implemented in this work in order to match the elongated shape of MMI switches.

 

Fig. 2 (a) Leakage radiation microscope set-up used for the observation and the photo-thermal activation of the switches. (b) (resp. (c)) Typical LRM images at λ₀ =1580nm of a doped (resp. undoped) MMI waveguide (scale bar=100μm). (c) Single-exponential fits (solid lines) of the experimental LRM intensity (dots) along the doped and undoped MMI waveguide axis.

Download Full Size | PDF

The near-infrared signal and the visible pump beam transmitted through the gold film were collected by a high-numerical aperture objective (NA=1.49) forming the image of the surface onto the detector of an InGaAs camera. A longpass filter with a cut-on wavelength of 900nm was used to remove the contribution of the visible light from the LRM images. LRM images of a NP-doped and undoped MMI switch are shown in Fig. 2(b) and 2(c) respectively for an incident free-space wavelength of λ₀=1580nm. The MMI DLSPPWs we consider are expected to support a fundamental symmetric DLSPPW mode with an effective index $n_e^s$ and an higher-order anti-symmetric mode (effective index $n_e^a$) [26] where the symmetric and anti-symmetric nature of the modes refers to the electric field component perpendicular to the thin metal film on which the polymer ridge is deposited. On the LRM images, we observe a two modes beating pattern with a period of p = 6.5μm and p = 6.3μm for the doped and undoped waveguides respectively. The beating period being given by:

3.2. Finite-element analysis of the MMIs

In order to correlate our experimental results to the properties of the MMI waveguides, we perform a finite-element eigenmode analysis. The modeled MMI we consider is made of undoped PMMA and has a width w and height h of respectively 900nm and 400nm extracted from the SEM and AFM characterizations of the fabricated waveguides. As said before, the experimental waveguides are obtained by exposing and dissolving the (doped or undoped) PMMA resin on each side of the waveguides leaving the waveguide itself unexposed. The PMMA, once exposed and developed, is known to create undercut sidewalls. In our situation, these undercuts result in trapezoidal cross-section waveguides. For a reasonable 15°-tilt of the waveguide side walls with respect to the normal, we found an effective index for the symmetric and anti-symmetric mode of respectively ${\text{n}}_e^s=1.251$ and ${\text{n}}_e^a=1.005$ leading to a beating length (at 1580nm) with a period of p = 6.4μm in fair agreement with the experimental value (p = 6.3μm).

The damping distance of each mode can also be extracted from the finite-element analysis and compared to the experimental values. The distribution of the guided power for symmetric and anti-symmetric modes are shown in Fig. 3(b) and 3(c) respectively. For the symmetric mode, the large field amplitude at the surface of the metal leads to a damping distance of L_spp = 37μm significantly smaller than for the anti-symmetric mode (L_spp = 60μm) as already recognized in ref. [26]. With such different decay lengths, the damping distance of the field intensity along the MMI waveguide depends upon the relative amplitude of each mode, a parameter which is controlled experimentally in our situation by the lateral shift of the single-mode DLSPPW exciting the MMI region. For the undoped waveguide shown in Fig. 2(c), the experimental decay (56μm) is close to the decay of the anti-symmetric mode (60μm) suggesting a dominant excitation of the high-order mode at 1580nm. The same conclusion applies for the doped MMI of Fig. 2(b) since the lateral shift of the exciting single-mode DLSPPW was the same for both NP-doped and undoped waveguides considered in Fig. 2. Thus, it is very unlikely that a dominant excitation of the symmetric mode is at the origin of the extra-damping (compared to undoped PMMA) observed for the NP-doped MMI waveguide. Similarly, the absorption of the gold NP at the telecom frequencies is not expected to be the dominant channel for this extra-damping. Relying on Maxwell-Garnett’s approach (which is expected to be reliable far away from the surface plasmon resonance frequency of the NPs [29], we find that the decay length of the symmetric and anti-symmetric modes is decreased by less than 10% when taking into account the absorption of the composite medium with a gold volume fraction of 0.52%. In this respect, the reduced propagation length for the NP-doped waveguides is more probably related to the roughness of waveguides, leading in particular to the decoupling of the DLSPPW modes into the gold/air SPP supported by the bare thin film surrounding the polymer waveguides. This conclusion is supported by the fact that the excitation of gold/air interface SPP modes can be clearly observed on highly saturated LRM images of NP-doped waveguides.

 

Fig. 3 (a) Opto-geometrical parameters for the finite-element analysis of an undoped MMI waveguides, W =900nm, h =400nm (see Fig. 3(a)), n_pmma = 1.489, n_Au = 0.57+i11.7 for a wavelength of λ₀=1580nm. The trapezoidal shaped cross-cut of the DLSPPW accounts for the undercut of the vertical side-walls produced by PMMA processing. (b) (resp. (c)) Spatial distribution of the guided power for the fundamental symmetric (resp. higher-order anti-symmetric) DLSPPW mode.

Download Full Size | PDF

4. Photo-thermal activation of MMI plasmonic switches

A quantitative comparison of the photo-thermal activation power for doped and undoped MMI switches is only possible provided that devices operating an optimized light-heat conversion are used. The waveguides we consider being obtained by processing a positive resin, the reference system (in absence of waveguide) is a polymer thin film deposited onto a gold thin film. Characterizing this reference configuration for the doped and undoped polymer is then necessary for an optimization of the photo-thermal behavior of the MMI switches.

4.1. Optical properties of the NP-doped PMMA layers

With the aim of extracting the optical properties of the NP-doped polymer, we first characterize the optical response of the NP in solution into toluene diluted PMMA. A dual beam spectrophotometer has been used to record the normal incidence absorbance spectra displayed in Fig. 4(a). These spectra have been recorded by comparing the transmission through 1 mm-thick calibrated cells filled with NP-doped PMMA solutions with NP concentration corresponding to gold volume fraction of 0.0038% and 0.0074%. For these measurements, a cell filled with undoped PMMA was placed on the path of the reference beam of the spectrophotometer. As expected for gold NP embedded into a liquid matrix with a refractive index close to 1.5, the NP-doped solution exhibits a maximum absorbance at a free-space wavelength around 520nm. Starting from a NP-doped polymer solution diluted into toluene with a gold volume ratio of 0.04%, NP-doped PMMA films have been prepared by spin-coating at different speeds onto bare (without gold film) glass substrates. After spin-coating, and prior to absorbance measurements, the PMMA films were baked according to the process used for electron-beam lithography leading to a gold volume fraction of 0.52% in the solid PMMA layers.

 

Fig. 4 (a) Absorbance spectra of NP-PMMA solutions with gold volume fractions of 38 part per million (ppm) and 74 ppm. (b) Absorbance spectra for NP-doped PMMA layers of different thicknesses deposited on a glass substrate. The gold volume fraction of solid PMMA is 5200ppm (0.52%).

Download Full Size | PDF

The absorbance spectra shown in Fig. 4(b) have been obtained for baked NP-doped PMMA films deposited on the glass substrate with a bare glass slide placed on the reference arm of the spectrophotometer. On these spectra, we note a red-shift of the maximum absorbance of about 40nm compared to the NP-doped PMMA solutions. This red-shift results from the combined effect of the refractive index increase of the polymer host matrix when the solvent is evaporated and from the fact that the NP particles are closely-packed in the solid PMMA layer as previously noticed on the waveguide SEM images (see Fig. 1). Despite the red-shift of the absorbance peak, we note that the 532nm pump beam is still within the absorbance peak of the NP-doped solid PMMA layers. The radius of the NP we use in this study being very small compared to visible wavelengths, their scattering efficiency is more than three order of magnitude smaller than their absorption efficiency over the whole visible range [30]. Thus, the spectra of the NP-doped PMMA layers are characteristic of the absorption of the composite medium. The absorbance Abs measured with the dual-beam of the spectro-photometer is defined as $Abs=-{\text{log}}_{10}\frac{I}{I_{\mathit{ref}}}$ where I and I_ref are the powers transmitted through the coated and the bare substrate respectively. Neglecting the scattering of the NPs and the change of reflectivity due to the presence of the PMMA layer (compare to the bare substrate), the power I can be expressed as I = I_ref exp(−2κ_pk₀h) where k₀ is the incident wavevector, h is the thickness of the doped layer and where κ_c is the extinction coefficient of the doped polymer defined as n̂_p = n_p + iκ_p with n̂_p the complex effective index of the composite medium. Knowing the thickness h of the NP-doped PMMA layers from AFM measurements, the effective extinction coefficient κ_p can be evaluated from:

We have considered so far NP-doped PMMA layers deposited on a glass substrate whereas the MMI switches are implemented onto optically thick gold films. The quantity of interest for the photo-activation of the MMI switches is obviously the amount of incident pump light which is absorbed by the system. This quantity is difficult to extract from absorbance spectra in the presence of metal due to the high gold film reflectivity. However, by recording the transmittance T and the near-normal incidence reflectance R spectra of PMMA layers deposited onto gold coated glass substrates, the absorption can be extracted according to A = 1− (R + T) (assuming once again that the scattering of NPs is negligible). This measurement procedure is illustrated in Fig. 5(a) showing the experimental transmittance and reflectance spectra along with the experimental absorption spectrum in the case of a bare gold film (thickness=80nm) evaporated onto a glass substrate. These experimental spectra are compared to the theoretical ones (dashed-lines) computed with the dielectric function of gold extracted from ref. [31]. For this configuration, we note that only 23% of the incident light is converted into heat at 532nm making light-heat conversion rather inefficient at this wavelength for this configuration. The same measurements have been performed for undoped and NP-doped PMMA layers with different thickness. For each PMMA layer, the absorption at 532nm has been computed and plotted in Fig. 5(b). For the undoped PMMA, the experimental values are in good agreement with the computed absorption when the refractive index of PMMA at 532nm is 1.502 in agreement with PMMA specifications. Note that for the undoped PMMA layer thicknesses leading to an anti-reflecting effect (for example h = 403nm), the maximum absorption level is about 38±2%. For NP-doped PMMA layers, a fairly good agreement between computed and experimental values is obtained when the complex refractive index at 532nm is 1.57 + i0.031. In agreement with Maxwell-Garnett effective medium approach, the real refractive index of NP-doped PMMA layers is larger than for undoped layers, however the refractive index of 1.57 is about 0.02 larger compared to the refractive index returned by the Maxwell-Garnett theory. In spite of the rather low gold volume ratio, Maxwell-Garnett approach fails to provide an accurate value for the real effective index of the composite medium likely due to the coagulation of the NPs [29]. Finally, we note that for a NP-doped PMMA layer with a thickness of h = 396nm, the absorption is about 79±2.5%, two-fold larger than for an undoped layer of same thickness. Hence, light-heat conversion for optimized layers is improved by a factor 2 for doped polymer.

 

Fig. 5 (a) Solid lines: Experimental normal incidence transmittance (T) and near-normal incidence reflectance (R) spectra of a 80nm-thick gold film deposited onto a glass substrate using a titanium adhesion layer with a thickness of 3nm. The Absorption spectra (A) is computed from T and R. Dashed lines: comparison of the experimental spectra with computed spectra using the gold dielectric function tabulated in Ref. [31] and neglecting the effect of the adhesion layer. (b) Dots: Experimental absorption values at 532nm for doped and undoped PMMA layers deposited on a gold film. Solid lines: computed absorption at 532nm for PMMA layers deposited on a gold film. For undoped PMMA, the best fit is obtained for n_PMMA=1.502, for doped PMMA, n_NP–PMMA=1.57+i0.031.

Download Full Size | PDF

4.2. Doped and undoped MMI switches

Based on the results of the previous section, doped and undoped MMIs have been fabricated using the optimized polymer layers. By using these optimized switches, we demonstrate in this last paragraph the minimization of the photo-activation power provided by the NP doping. A direct comparison of the switching performances of different MMIs for a given activation power is difficult since the properties of each device depend critically on its opto-geometrical parameters [26]. Instead, we assess the performances of the MMI switches when photo-activated by comparison with their spectral response in the cold state. In other words, the spectral response in the cold state is used as calibration to characterize the photo-activation of the switches.

We consider first a MMI fabricated in an undoped PMMA layer with a thickness of 406nm. Figure 6(a) shows the intensities I_CROSS and I_BAR at the outputs of the Y-junction obtained by integrating the LRM images over the two ports when sweeping the incident wavelength. The BAR (resp. CROSS) port is located on the same (resp. opposite) side of the MMI waveguide than the laterally shifted single-mode input waveguide. The cold spectrum displayed in Fig. 6(b) is obtained by computing the extinction ratio ER given by:

 

Fig. 6 (a) LRM intensity integrated over the CROSS and the BAR output port of an un-doped MMI switch for a signal sweeping the telecom frequency range. (b) CROSS-BAR Extinction ratio spectrum computed using the results of (a). (c) (resp.(d)) LRM image of the CROSS (resp. BAR) state. (e) For a fixed wavelength of 1540nm, change of extinction ratio at the output port as a function of the pump power. (f) (resp.(g)) LRM image taken at λ₀=1540nm in the hot (resp. cold) state. The elliptically shaped pump (532nm) beam with a power of 40mW is visible in (f). (h) Activation power efficiency obtained by correlating the extinction ratio change in (e) to a wavelength shift in the cold state given in (b).

Download Full Size | PDF

The same experiment has been conducted for a NP-doped MMI switch fabricated in a doped PMMA layer with a thickness of 396nm. The extinction ratio observed for this NP-doped MMI is plotted in Fig. 7(a). The MMI switch is in the BAR state at λ_B=1522nm (ER=+20dB) and in the CROSS state (ER=−18 dB) at λ_C=1558nm. The incident wavelength being fixed to 1536nm, the photo-thermal activation of the MMI leads to a decrease of the extinction ratio as shown in Fig. 7(b). This time, the ER drops from +6dB in the cold state to −4.5 dB for a pump power of only 8mW. The image displayed in Fig. 7(c) shows the difference of the LRM image in the hot and cold state for such a pump power. On this image, the build-up process along the MMI region leading to the switching at the Y-junction is clearly visible. Using once again the cold ER spectrum as a calibration curve, the drop of the ER is correlated to the wavelength shift plotted in Fig. 7(d) with a slope of P = 1.37 nm/mW. Similar characterizations have been repeated for several undoped and NP-doped switches fabricated respectively on the same substrate as the devices considered above. The values of the slope P obtained in each case are summarized in table (1). Based on the values of the slope P, we conclude that the pump power needed for the photo-thermal activation of the NP-doped MMIs is about 2.5 times smaller than for undoped ones, a factor that is even larger than the absorption enhancement obtained for the doped and undoped PMMA unstructured layers. This activation power ratio can in principle originate from the enhanced absorption of the NP-doped polymer and/or an increase of the thermo-optic coefficient of the composite medium. In our situation, the contribution of the thermo-optic coefficient of the NP-doped polymer can be safely neglected. Indeed, the infrared signal being far from the plasmon resonance of NPs, the thermo-optic coefficient (TOC) of the composite medium can be approximated as the volume fraction average of the TOCs of each compound of the composite medium [32]. Using the data of bulk gold [32], we find that the TOC of the doped polymer at 1550nm for a gold volume ratio of 0.52% is changed by less than 2% compared to the undoped one. Thus, the decrease of the photo-thermal activation power is due to the enhanced light-heat conversion for the doped switches.

 

Fig. 7 (a) CROSS-BAR exctinction ratio for a doped MMI switch. (b) For a fixed wavelength of 1536nm, change of extinction ratio at the output port as a function of the pump power. (c) Differences of the LRM images recorded in the hot and cold state for a pumping power of 8mW. (d) Activation power efficiency obtained from (b) and (a).

Download Full Size | PDF

Table 1. Activation efficiency obtained for three different NP-doped and undoped MMI switches

View Table

In fact, not only the total amount of absorbed light is changed when doping the polymer but also the spatial distribution of the heat source density. The volume density of heat sources is given by:

 

Fig. 8 (a) Configuration for the heat source density computation. The parameters are W =900nm, h =400nm and g =2.5μm. The thickness of the gold film is 80nm. (b) Heat source density at 532nm computed for along the vertical (z-axis) of the waveguide for the doped and undoped polymer.

Download Full Size | PDF

5. Conclusion

In summary, we have investigated the photo-thermal activation of 1×2 dielectric loaded plasmonic switches. The switches which rely on a MMI design were fabricated by electron beam lithography using a NP-doped positive resin. For a doping level corresponding to a gold volume ratio of 0.52%, the radiation leakage microscope images of the doped MMI waveguides reveal an decrease by about 30% of the plasmon modes propagation distance attributed to waveguide walls scattering losses. By inspecting the spectral response at room temperature of 140μm-long MMI waveguides, we note a switching between the CROSS and the BAR state with typical extinction ratios around 20dB over a wavelength range of 30±5nm for both doped and un-doped devices. For the purpose of comparing the power needed to thermally activate doped and undoped switches, the thickness of polymer leading to an optimum light-heat conversion when illuminated at 532nm has been determined by analyzing absorbance spectra of homogeneous layers. For an optimized thickness of 400nm, the light-heat conversion efficiency at 532nm is enhanced by a factor 2 by the presence of the NPs.

Next we have developed a procedure to compare the photo-activation power of doped and undoped switches. This procedure is based on the conversion of the extinction switching ratios obtained for different heating powers into wavelength shifts by using the room temperature spectrum of the switch under test. With this approach, which accounts for the switching performances of each device, we show that the power necessary to photo-activate a NP-doped switch is about 2.5 times smaller than for the undoped devices, demonstrating the interest of the NPs in this context of power consumption minimization. The decrease of the activation power is not due to a change of the doped polymer TOC but rather to the enhanced absorption of the pump light for the NP-doped devices and the large power generation into the polymer as shown by the computation of the heat source volume density.

In this study, the polymer material has been engineered to enhance light-heat conversion. Following this approach, many improvements can be suggested such as for example the use of silver NPs which are known to generate about ten times more heat when illuminated at their resonance frequency compared to gold NPs [17]. Finally, and beyond polymer doping, the metal film supporting the plasmon mode of interest can also be patterned to become fully absorbing at the pump frequency leading to an optimum light-heat conversion. Actions in this direction are currently in progress and will be reported elsewhere.