advances.sciencemag.org/cgi/content/full/2/4/e1501227/dc1 Supplementary Materials for Self-assembly of highly efficient, broadband plasmonic absorbers for solar steam generation Lin Zhou, Yingling Tan, Dengxin Ji, Bin Zhu, Pei Zhang, Jun Xu, Qiaoqiang Gan, Zongfu Yu, Jia Zhu Published 8 April 2016, Sci. Adv. 2, e1501227 (2016) DOI: 10.1126/sciadv.1501227 The PDF file includes: Materials Note S1. Template-assisted PVD procedure for self-assembly of Au/NPT. Note S2. Systematic demonstrations of optical absorbance on geometry parameters. Note S3. Antireflection and impedance matching by the nanoporous templates. Note S4. Physical understandings for differences between the simulated and experimental absorbance (Fig. 3). Note S5. Understanding the nonlinear behavior of evaporation rate on light intensity. Note S6. Radiation loss of the steam generation system. Note S7. Advantages of Au/D-NPT absorber for steam generaton: Comparisons with carbon paint or traditional plasmonic absorbers. Note S8. Optical modeling for random gold particles. Note S9. Angular dependence of the Au/D-NPT absorber. Note S10. Electric measurements and potential applications. Fig. S1. Schematic diagrams of the Au/NPT absorber. Fig. S2. Measured absorption spectra of Au/NPT absorbers with different pore diameter D. Fig. S3. Absorbance of the plasmonic absorbers on pore length H and gold film thickness hf. Fig. S4. Schematic diagram of Au/NPT and the propagation direction of light. Fig. S5. Effective index and impedance of nanoporous template calculated by Bruggeman effective medium formula.
Fig. S6. Scheme for the difference between the actual and simulated structures. Fig. S7. Simulated absorbance of the Au/NPT absorber with different particle length hp. Fig. S8. Comparison of the experimental and simulated absorption spectra of the Au/NPT absorber. Fig. S9. Measured temperature of steam as a function of illumination intensity. Fig. S10. The radiation loss of the plasmonic absorber surface. Fig. S11. Advantages of Au/NPT for solar steam generation. Fig. S12. Evaporation comparisons between the Au/D-NPT and carbon black based absorber (carbon nanotube). Fig. S13. Simulated absorbance as a function of particle number N. Fig. S14. Angular dependence of the plasmonic absorber. References (50 56)
Materials Note S1. Template-assisted PVD procedure for self-assembly Au/NPT The limitations of top-down fabrications for plasmonic absorbers (typically Focused Ion Beam and E-beam lithography), namely, fabrication throughput, spatial resolution and scalability, are well documented in the literatures and widely acknowledged in the community of plasmonics. Typical spatial resolution of plasmonic nanostructures that can be achieved by EBL or FIB fabrications is ~ 10 20 nm (19). The system requires extremely high vacuum condition (~10-6 Pa), and have very low throughput (several hours are needed for the fabrication hundred micron size perforated plasmonic structure) (50). Therefore, these top down processes are widely used for academic research, but seldom employed for large scale fabrications. That s exactly the reason that in the past few years there is significant effort to achieve plasmonic structures through bottom up approaches (28, 51). The advantages of bottom up approach on scalability, high throughput, and high spatial resolution (1 2 nm inter-particle distance) is also well documented in the literature, most recently reviewed by several authors recently in literatures (30 33). Particularly in our case, our bottom up approach, combining nanoporous templates with physical vapor deposition, has several unique advantages. 1. This approach can assemble gold particles with unique features, such as graded distribution, closed packed (with a few nanometer gap). These features, not possible by top down approach, are critical for pronounced absorption. 2. The process is highly scalable, especially compared to top down approach. There are essentially only two steps in the process: a) fabrications of nanoporous templates, b) gold deposition by physical vapor deposition (PVD). a) Fabrications of nanoporous templates Actually the alumina-based nanoporous template (or called anodic aluminium oxide) we used in this study has been extensively investigated and widely used for several decades in material sciences and several major industry (more than 8600 literatures with key words anodic aluminium oxide ). For example, currently alumina porous templates have been widely applied as a scalable process in fabrications, optics as well as solar harvesting systems, etc (36, 47, 52). These nanoporous templates are being massively produced and commercially available (with $50 for a 4-inch piece, even cheaper than some of silicon wafers, http://www.nano-star.com ; www.whatman.com). b) Gold deposition by PVD The PVD process has been widely used for large scale manufacture in several major industries, such as semiconductors, displays (http://www.appliedmaterials.com/en-sg/semiconductor/products/pvd/info). Indeed, most of PVD systems require high vacuum environment. However, the requirement of vacuum degree in our preparation (~10-3 Pa) is much weaker compared to the other top-down processes (for example, vacuum requirement for FIB is ~10-6 Pa).
We want to point out that the major difference between PVD and EBL & FIB processes is the throughput. Only ~ 15 min gold deposition by PVD is enough for one sample (as large as twelve inch diameter). In the case of FIB for example, it typically takes several hours to fabricate a several hundred micron sized perforated plasmonic structures (50). As these nanoporous templates (NPT) are being massively produced at low cost, PVD has been widely used in several major industries, it is safe to conclude that the NPT assisted PVD method is scalable and more convenient process for fabrications of plasmonic structures. Finally, we stress that both the performance and fabrication of our absorber are ideally suitable for the solar steam generation as compared to the traditional plasmonic absorber. The double-lens focusing system and solar simulator are not the bottleneck either, for the low optical intensity can be achieved easily. For outdoor applications, the focusing system (which seems to be complex currently) can be replaced by a simple Fresnel lens as employed by Neumann. Therefore, the technological applications can be suggested. Note S2. Systematic demonstrations of optical absorbance on geometry parameters The geometry parameters are judiciously designed in order to satisfy the three requirements to enable efficient and broadband plasmonic absorbers: high density of plasmonic resonant modes, anti-reflection and strong light coupling. The pore diameter D is one of the most critical parameters for the strong absorption, as it determines the density of plasmonic resonant modes, optical path-length as well as anti-reflection, all of which are critical for perfect broadband absorption. Here below, we illustrate experimentally how pore diameter D affects the absorption (Fig. S1 A). Fig. S1. Schematic diagrams of the Au/NPT absorber. (A) The cross sectional schematic diagram, (B) top view schematic diagram. Experimental measurements of absorbers with various pore diameters D are shown in Fig. S2. It is obvious that, the absorption bandwidth is effectively tuned by the pore diameter D.
Fig. S2. Measured absorption spectra of Au/NPT absorbers with different pore sizes D by integrated-sphere-equipped UV-Vis-NIR spectroscopy for different pore diameter D of NPT (porosity f): D = 365 nm (f = 0.59), D = 300 nm (f = 0.38), D = 200 nm (f = 0.19) with the interpore distance L = 450 nm. There are three reasons behind this. 1) Absorbers with larger pore diameter D is more favorable for formation of more close-packed gold nanoparticles with a wide variety of aspect ratios with larger particle length hp during the physical vapor deposition (See Fig. S2 for simulated absorbance for different hp for details), therefore increase the density of plasmonic resonant modes, which is critical for broadband absorption. 2) Large pore-sized templates will have much reduced refractive index, which can help to couple light in with much reduced reflections. 3) Pores with large diameter can enable efficient light coupling within these nanopores, therefore significantly enhance the absorption of gold nanoparticles. Other parameters: Thickness of nanoporous template H and gold film thickness hf As shown in Fig. S3, the most important role of thickness of template H on absorption performance is the extra contribution to absorption for > 6 m, which is due to the intrinsic infrared absorption of alumina. Gold film thickness hf is the least important one, which can be well observed in the simulated absorbance (Fig. S3 B) and measured spectra (Fig. S3 C).
Fig. S3. Absorbance of the plasmonic absorbers on pore length H and gold film thickness h f. (A) calculated absorbance on pore length of NPT H, (B) calculated absorbance on gold film thickness h f, (C) experimental absorbance of Au/NPT with different h f. Note S3. Anti-reflection and impedance matching by the nanoporous templates Impedance matching is crucial for perfect absorbers, and a powerful interpretation for scattering systems under effective medium approximations (18). Different from most of previously reported plasmonic absorbers, our Au/D-NPT absorber is inherently anti-reflective with very small impedance mismatching mainly due to two reasons. 1) The nanoporous template (NPT) with high porosity (as high as 80%) provides low effective index neff (as low as 1.1 for the visible light), indicating small impedance mismatching at the incident interface (see Fig. S4 for propagation
direction of light). Fig. S4. Schematic diagram of Au/NPT and the propagation direction of light. As shown in Fig. S5 A, the effective indices neff of NPT with different porosities are calculated by the Bruggeman effective medium approximation (53). It can be easily understood that neff of NPT can be greatly reduced (compared with bulk alumina) with increased porosity f. Taking the NPT (f=59%) as an example, the effective impedance Z/Z0 = 1/neff and reflectance (according to Fresnel formula) on the air/npt interface are shown in Fig. S5 B. Therefore, small impedance mismatch and much reduced reflectance can be expected (R~ < 1.6% over the 400 5 m, R< 1% over 5 10 m). 2) The absorptive elements (gold particles) of our absorber are distributed along the sidewall of nanopores (~ m in thickness), with very low filling ratio (< 1%, gold with effective thickness ~ 85 nm is distributed in ~ 10 m thick NPT substrate, see inset in Fig. S5 C), therefore has minimum effect on neff under effective medium approximation, as shown in Fig. S5 C.
Fig. S5. Effective index and impedance of NPT calculated by Bruggeman effective medium formula. (A) Effective index of NPT with different porosity f. The refractive index of alumina is subtracted from Palik data for comparison (49). (B) The relative impedance and reflectance on the air/npt interface of NPT with porosity f = 0.59. (C) n eff of Au/NPT and NPT with f = 0.59 (Bruggeman formula). The inset is a typical cross sectional SEM diagram for Au/NPT, indicating the penetration depth of gold particles inside nanopores is ~ 10 m. As shown in the detailed analysis above, because of high porosity of NPT and low filling ratio of metal nanoparticles, our plasmonic absorber exhibits only small impedance mismatch with free space, ensuring > 98.4% of incident light can penetrate into the absorber over 400 nm 10 m. It is noted that the impedance-based interpretation is intuitive and powerful for scattering system under the condition of >> d (d is characteristic dimension of scatters, 54). With the decreased d, the scattering of porous template gradually dominates and the overall neff and impedance cannot be well-defined. The impedance and neff calculated above can only serve as guideline for qualitatively understanding in this regime. Note S4. Physical understanding for differences between the simulated and experimental absorbance (Fig. 3) The discrepancy between the simulated and experimental absorbance is a numerical artifact, which is caused by the limited computational resources. Here
below, we show by using more computing resources and a longer computational time to accommodate a thicker structure, the simulation results gradually approaches those observed in experiments. Such differences are often encountered in simulation of random structures. For = 6 10 µm, the simulated absorbance of Au/NPT is much lower than the experimental one mainly due to undervalue the intrinsic absorbance (red dashed line in Fig. 3D) of our nanoporous template. We now provide more detailed analysis. Our plasmonic absorbers have very complex structures, spanning four orders of magnitude difference in length scale for all the components. As shown in Fig. S6, the plasmonic absorber has three components: nanoporous templates (typically 50 m thick in our experiment), Au thin film (< 100 nm) and close-packed Au nanoparticles (with ~ nm sized gap) along the sidewall of pores. This four orders of magnitude difference in length scale make it very challenging for numerical modeling. Fig. S6. Scheme for the difference between the actual and simulated structures. (A) Actual dimension of gold deposited structure. (B) The simplified structure used in the simulated modeling. (C) The characteristic smallest size of the close-packed gold particles. Therefore, in the simulation of Fig. 3, in order to fully reveal the broadband absorptive effects from close-packed nanoparticles, the core part of our plasmonic absorber, the simulated nanoparticles thickness hp as defined in Fig. S6 B is chosen to be round 400 nm, much less than the actual depositing depth (~ 10 m) inside nanopores as well, which would influence the absorption efficiency as well as bandwidth. As shown in Fig. S7, with the increment of particle length hp, gold nanoparticles are more sparsely distributed and thus the total reflectance of Au/NPT can be greatly reduced in more broadband wavelength range. Therefore, more efficient and broadband Au/D-NPT absorber can be demonstrated.
Fig. S7. Simulated absorbance of Au/NPT absorber with different particle length h p. For simplicity, random positioned gold nanoparticles with uniform size r p = 5 nm. The total particle number for Au/NPT (h p = 1 m) and Au/NPT (h p = 1.5 m) is set as 12000 and 18000 respectively with gold deposition quantity conserved. Also for the same reason, only a small section of nanoporous template (500 nm) is selected (see Fig. S6 B), much less than the actual thickness 50 m in the experiment. To confirm this, we performed FDTD calculation (~ 144 hours calculation time) for the Au/NPT nanostructure with H = 10 m (blue solid line in Fig. S8 B, oscillation feature is due to strong FP resonances that make the calculation hard to converge). It is obvious that the absorbance at the wavelength range of 6 10 m increases distinctly as the thickness of nanoporous template increases from 500 nm to 10 m.
Fig. S8. (A) Comparison between experimental and simulated absorption spectra of the Au/NPT absorber (subtracted from Fig. 3 in the paper). (B) Calculated absorption spectra of Au/NPT absorber (solid symbol lines) and bare NPT (hollow symbol dashed lines) with different NPT thickness H. The experiment absorbance is shown in black solid line for clear comparison. Finally, the multi-peaks in the simulated absorption spectrum (for < 1 m), which are absent in the experimental spectrum, can be attributed to the simplification of the structure modeling. The actual nonperiodic surface morphology of nanopores is simplified as the periodic hexagonal pore array (Fig. S1 B) in the simulation. Nevertheless, the above simplification will not affect the physical understanding of the broadband plasmonic absorption effect. Note S5. Understanding the nonlinear behavior of evaporation rate on light intensity According to the definition of the solar steam efficiency mh / P, (1) LV in the efficiency is proportional to the evaporation rate m, phase change entropy h LV and illumination light intensity 1/ P in. The evaporation rate m does not increase linearly with Pin (smaller slope for low Pin, larger slope for higher Pin, see Fig. 4E), as m is positively related to the thermal motion or steam temperature Tsteam, which is not linearly increased with Pin (see Fig. R1). For the extreme case, steam temperature will reach a steady state for the open system (99.6 @ 1 atm). Phase change entropy hlv is nonlinear to Pin as well since higher Tsteam refers to lower h LV, as shown in Fig. S9. Therefore, the nonlinear efficiency of steam generation as a function of light intensity can be expected.
Fig. S9. Measured temperature of steam (black square line) as a function of illumination intensity. The blue circle line is the corresponding latent enthalpy h LV of liquid-vapor phase change of water. Note S6. Radiation loss of the steam generation system As it is known that most absorbers are selective and the IR emissive loss does exist. However, the emissive loss of our system is much less compared with the thermopv system due to the much lower operation temperature. We have estimated the radiation loss efficiency of our absorber by using the definition P ( T) P ( T ) / P, (1) rad rad atm amb in Where Prad(T) stands for the radiation loss power of the absorber (analogy to black body radiation from the sample surface to the air space), Patm(Tamb) stands for atmospheric thermal radiation power (corresponding to the air temperature Tamb), and Pin stands for the incident solar power on the sample surface, which can be calculated by integrating the specular black body emission and specular solar irradiations, respectively (7 9). Figure S10 is the maximal estimated radiation loss of our absorber. Fig. S10. The radiation loss of the plasmonic absorber surface. The maximal is estimated by semi-infinite sphere (red circles) and actual solid angle (black squares) integration of black
body emission at the sample temperature. Figure S10 gives the emissive loss towards air space (red circles stand for half space integration with solid angle = 2, black squares stand for the actual solid angle integration = 0.94) The emissive loss is relatively low compared with ThermoPV devices due to low surface temperature of our system (T < 100 ). The radiation losses will decrease distinctly for higher optical concentration, which is consistent with the literature results (7). Note S7. Advantages of Au/D-NPT absorber for steam generaton: comparisons with carbon paint or traditional plasmonic absorbers For the solar steam generation, there are three main advantages of our design. 1) Efficient light absorption The unique close-packed gold particles embedded in porous structure and the highly scalable fabrication process enable our Au/D-NPT absorber the low cost as well as and the most broadband plasmonic absorber (average absorbance ~ 99% over 400 nm 10 m) reported so far, overwhelming most of the multiplexed plasmonic absorbers (5, 10, 19, 20, 23, 24, 55), as shown in Fig. 3E. 2) Efficient thermal to steam generation Because of the porous structures of templates, the entire structure can flow naturally on the top surface of water (See Fig. S11 A-B). Therefore, the energy of light absorbed is focused only on the very top surface of water (Fig. S11 C-D), enables efficient steam generation. For comparison, in the original nanoparticle/solution suspension (8), as nanoparticles are suspended in the water, the entire water body (instead of the top surface in our case) was heated, therefore the efficiency of solar steam generation will not be high. In addition, different from the "lossy-black-paint-type" absorber that uniformly heats the liquid (7), in our case water (around gold particles) is locally heated by our plasmonic absorbers through the near-field enhanced non-radiative plasmon damping (Fig. S11 D). 3) Efficient steam flow: porous structures The porous structure of templates can provide paths for continuous steam flow and fluid flow to hot region (Fig. S11 C).
Fig. S11. Advantages of Au/NPT for solar steam generation. (A B) Optical photograph of the experimental setup of the solar steam generation. The red arrow refers to the sample floating on water. (C) Schematic diagram for key points of our plasmonic absorber for solar steam generation. (D) Plasmonic effect that exists in the process of solar steam generation. Figure S12 shows a control solar steam experiment between Au/D-NPT and carbon nanotube forests (CNTs, absorbance ~ 100%, as shown in Fig. 3E). The overall solar steam efficiency of Au/D-NPT is ~ 90.4%, much higher than the performance of CNTs (73.4%) at 4 sun illumination. It can be ascribed to the strong plasmonic near field enhancement across the entire solar spectrum.
Fig. S12. Evaporation comparisons between Au/D-NPT and carbon-black-based absorber (CNT). The pore radius of D-NPT substrate is D = 365 nm, the solar irradiation is 4 kw/m 2. The ambient temperature is 24 and humidity ~ 42%. Because of the unique features listed above, the overall solar steam efficiency achieved by our plasmonic absorbers can reach 90% under 4-sun illumination, as far as we know, the highest reported so far. Note S8. Optical modeling for random gold particles The most crucial point for the optical modeling is the random distributed gold nanoparticles. Indeed, the random sized and positioned gold nanoparticles greatly increase the difficulty of the numerical simulation. To mimic the graded distribution profile observed in the cross sectional SEM image (Fig. 2F), the total length of particles (hp = 400 nm) is divided into eight sections and the particle number of each section (from gold film to NPT side) is set as 25, 50, 100, 175,., with particle increment linearly increased. The total number of particles inside each pore is thus ~ 2300, the gold quantity of which is approximately conserved with the deposition condition. The number density of gold nanoparticles mainly determines the absorption bandwidth. The dependence of the calculated absorbance spectra with the particle density is shown in Fig. S13. It is as expected that, absorbers with sparse gold particles (N=920) is less absorptive in IR range. As density of particles increases towards N=2300, the absorption edge is red shifted. However, as the particle number increases further (N=4600), the absorption edge turns to be blue shifted. The blue-shifted trend becomes more obvious as N approaches infinity (the Au particles become solid gold tube, as shown in triangle green line). Therefore, there exists an optimized particle density for optimal broadband absorption. Detailed calculations reveal that the absorbers with N=2300 is most close to the overall experimental absorbance (especially for the average absorbance and band edge where the efficiency drops distinctly). Fig. S13. Simulated absorbance as a function of particle number N. FDTD calculated absorbance of Au/NPT nanostructures for different number of gold nanoparticles inside a pore: N=920 (sparse particles), N=2300 (close-packed particles), N=4600 (extremely dense
particles) and smooth gold tube (limiting case of N ). Note S9. Angular dependence of the Au/D-NPT absorber As partially suggested, generation of solar steam will be affected by the angular dependence of the light because absorbance of plasmonic nanostructures are more or less angular dependent (1, 17, 56). Figure S14 A B show the simulated and measured absorption spectra of the plasmonic absorber for different incident angles. In addition, we also have solar steam generation by 1 sun irradiation under different focusing conditions (therefore different angles, see Fig. S14 C). All of the data consistently demonstrate that, the performance of light absorption as well as steam generation has very weak angular dependence of incident light, as our disordered nanoporous templates can efficiently couple light from a wider range of incident angles than ordered nanostructures.
Fig. S14. Angular dependence of the plasmonic absorber. (A) Incident angle dependence of calculated absorbance spectra for the Au/NPT sample: black solid circles for 0 0 and blue hollow circles for 45 0, respectively. (B) The angular dependent UV-Vis-NIR absorption spectra measured by an integrated sphere equipped spectroscopy. (C) The mass change as a function of time for two focusing cases with different incident angles. The insets in B and C show the corresponding schematic diagrams of the measured setups respectively. Note S10. Electric measurements and potential applications Apart from the appealing absorption performance, the excellent electric conductivity of the plasmonic absorber can be suggested. We have measured the resistance of our absorber on both sides (across the radius direction). Our absorber shows good conductivity on the Au side (with a resistance < 100 which may be applied in hot electron generation or photo detector accompanied with a semiconductor contact (5). Although at early stage, we suggest that the highly efficient solar steam generation with Au/NPT absorber under low optical concentration, relatively low temperature and surface local heating has many potential applications, such as brine desalination, absorption chillers, sterilization, chemical purification, etc.