Plasmon-driven synthesis of individual metal@semiconductor core@shell nanoparticles

Most syntheses of advanced materials require accurate control of the operating temperature. Plasmon resonances in metal nanoparticles generate nanoscale temperature gradients at their surface that can be exploited to control the growth of functional nanomaterials, including bimetallic and core@shell particles. However, in typical ensemble plasmonic experiments these local gradients vanish due to collective heating effects. Here, we demonstrate how localized plasmonic photothermal effects can generate spatially confined nanoreactors by activating, controlling, and spectroscopically following the growth of individual metal@semiconductor core@shell nanoparticles. By tailoring the illumination geometry and the surrounding chemical environment, we demonstrate the conformal growth of semiconducting shells of CeO2, ZnO, and ZnS, around plasmonic nanoparticles of different morphologies. The shell growth rate scales with the nanoparticle temperature and the process is followed in situ via the inelastic light scattering of the growing nanoparticle. Plasmonic control of chemical reactions can lead to the synthesis of functional nanomaterials otherwise inaccessible with classical colloidal methods, with potential applications in nanolithography, catalysis, energy conversion, and photonic devices.


Supplementary Note 1: Ensemble Au@CeO 2 core@shell nanoparticle synthesis.
The Au@CeO 2 core@shell nanoparticle synthesis is slow at room temperature, while leading to a CeO 2 shell thickness of 10 nm upon heating to 90 °C for 1 h. We confirm the growth of ceria shell around the Au nanoparticles at high temperatures by observing a red-shift of the plasmon resonance in extinction spectroscopy. Furthermore, we utilize electron microscopy and energy dispersive X-ray characterization to visualize the formed ceria shell.
Supplementary Fig. 1: a, Schematic representation of the temperature-activated synthesis of Au@CeO 2 core@shell nanoparticles 1 . b, Ensemble extinction spectra of Au nanoparticles mixed with the Ce 3+ -EDTA solution, kept at room temperature at time = 0 h (brown), time =1 h (green), and when heated to 90 °C using a conventional hot plate for 1 h (blue). c, Bright field scanning transmission electron micrograph (BF-STEM) and Au Lα1 (brown) and Ce Lα1 (blue) energy dispersive X-ray (EDX) maps of Au@CeO 2 core@shell nanoparticles synthesized at 90 °C using a conventional hot plate for 1 h. The temperature increase on a nanoparticle surface under laser illumination is dictated by the absorption cross-section and the radius of the nanoparticle, for a fixed laser intensity.
Under a 532 nm continuous wave laser irradiation, Au nanospheres in water generate the highest temperature when they possess a diameter of 79 nm. The Au nanospheres we synthesize have a diameter of 66 nm, which can generate 92% of the temperature generated for 79 nm particles, at any given intensity (See also Eq 1 in main text). From equation 1 in the main text, the temperature increase on a nanoparticle is dependent on the illumination intensity. We vary the laser intensity to tune the nanoparticle surface temperature, which in turn dictates the size of the photothermally grown core@shell nanoparticles.
Supplementary Fig. 4: The photothermal temperature increase on the nanoparticle surface with respect to the ambient temperature is plotted as a function of the applied laser intensity (See also Eq 1 from main text). Here, we use an absorption cross-section of 1.37 × 10 -14 m 2 corresponding to 66 nm Au nanoparticles and an average thermal conductivity of 1 W/m/K (κ water = 0.6 and κ quartz = 1.4 W/m/K). Note, the temperature dependence of the thermal conductivity is not considered in these calculations. Au nanoparticles deposited on a quart substrate appear as bright green scatterers, when observed under a dark-field microscope. After photothermal growth of a dielectric shell on these nanoparticles, their plasmon resonance red-shifts, thereby changing the color of these green scatterer to yellowish red.
Supplementary Fig. 5: (Top) Dark-field images of Au nanospheres deposited on a quartz substrate before the photothermal growth of CeO 2 shell. These plasmonic particles are green in color. The particles that are numbered are used to grow CeO 2 shell photothermally. Here, one particle is irradiated at a time. (Bottom) Dark-field images of photothermally grown Au@CeO 2 core@shell nanospheres. Scalebars correspond to 10 µm.
These hierarchical structures after irradiation display scattering in yellow and red, depending on the shift in LSPR. The magnitude of LSPR shift is dependent on the thickness of the grown ceria shell. We vary the thickness of the shell by varying the laser intensity. The numbered particles are irradiated using different laser intensities. For example, particle 1,2 and 3 are irradiated at a power density of ~1.3 mW/µm 2 , particle 4 is irradiated at 5 mW/µm 2 and particles 5 and 6 are irradiated at ~1.2 mW/µm 2 . As a dielectric shell is grown around Au nanospheres, their LSPR red-shifts with a simultaneous increase in its scattering cross-section. Thus, in our single-particle experiments, by estimating the size from the measured LSPR of the initial Au nanoparticle before irradiation and by measuring the LSPR red-shift after the shell growth, one can estimate the thickness of the grown semiconducting shell. We perform several control experiments to attribute the growth of ceria shell around Au nanoparticles to photothermal heating.
• Supplementary Note 7.1: Absence of shell growth in dark corroborated by electron-

microscopy.
In our studies, we confirm the growth of a ceria shell under photothermal heating by performing electron microscopy imaging of the laser irradiated nanoparticles. These nanoparticles are deposited on a SiN TEM membrane. From TEM measurements, we also find that the particles away from the illumination spot do not exhibit any ceria shell growth, despite having the same chemical environment as that of the irradiated particle in Fig. 1D of the main text.  The potential contribution of hot-holes generated under plasmon excitation in driving the Au@CeO 2 core@shell nanoparticles is analyzed, by correlating the Fermi level of Au and the band diagram of CeO 2 1 . Since, we use a 532 nm irradiation laser, which corresponds to an energy of 2.33 eV, the highest energy that a hole could attain is -7.6 eV with respect to the vacuum level. Since the valence band of CeO 2 lies at -7.8 eV with respect to the vacuum level, the holes would not be able to diffuse to the ceria shell under plasmon excitation using 532 nm laser. Moreover, the Schottky nature of the Au-CeO 2 interface would block any possibility of holes being transported to the semiconductor. Also, such holes are generated in the d-band of Au through interband excitation, which is relatively flat in nature.
As such, these holes are generally considered to possess low velocities and their mean free path length is considered to be less than 5 nm 3 . These considerations safely allow us to rule out any significant contribution of hot holes to the plasmon-driven synthesis of Au@CeO 2 core@shell nanoparticles.  In our experiments, the Au nanoparticles are located at the quartz-water interface. To calculate the temperature increases on the nanoparticle under laser irradiation, a precise knowledge on the thermal conductivity of the surroundings is essential, according to equation 1 in main text. Here, we plot the effective thermal conductivity around the nanoparticle at different temperatures. Supplementary Fig. 9: Effective thermal conductivity of the surrounding media of Au nanoparticles constituting water and quartz is plotted as a function of temperature.
In order to calculate the effective thermal conductivity, we take the mean thermal conductivity of water (0.6 W/m/K at 293 K) and quartz (1.4 W/m/K at 293 K). The data for the thermal conductivity of quartz substrate is obtained from the manufacturer website. We also take into account the varying thermal conductivity with temperature, as plotted in Supplementary Fig. 9. The grey line denotes the quadratic fitting of the data points (R 2 = 0.977), according to: In our photothermal growth experiments, we always use a 532 nm cw laser. As the ceria shell grows in size, the plasmon resonance starts to red-shift. As such the absorption cross-section of the nanoparticle at the illumination wavelength decreases. As such, the nanoparticle temperature also decreases according to equation 1, leading to slower shell growth over the illumination period. In this section, we show how ΔPL can be a good proxy for the LSPR of the growing core@shell nanoparticle under laser irradiation.
In our experiments, we measure the inelastic Stokes scattering under nanoparticle illumination, to extract the core@shell growth kinetics (Supplementary Fig. 11a). To obtain ΔPL, we subtract inelastic emission signals at time , with the signal at = 0 ( Supplementary Fig. 11b). Using this approach, we remove background contributions from the underlying quartz substrate and the (negligible) Raman signal of reactants, and isolate the photoluminescence of the plasmonic nanoparticle. As stated in the main text, the photoluminescence of metallic nanoparticles is proportional to their photonic density of states, corresponding to their scattering spectra. Our ΔPL approach generates a curve that is equivalent to the differential scattering spectrum, where the scattering cross-section at time (Au@CeO 2 ), is subtracted with the scattering spectra at = 0 (Au).
At large shell thicknesses, where the plasmon resonance is red-shifted, the maxima of the ΔPL matches exactly with the plasmon resonance of the core@shell nanoparticle ( Supplementary Fig. 11c). At smaller shell thicknesses, a red-shift the maximum of ΔPL compared to the plasmon resonance is expected ( Supplementary Fig. 11d). This deviation can be accounted by correlating the LSPR maximum against the maximum of the differential spectrum ( Supplementary Fig. 11e). In principle, such approximations should allow us to properly estimate the reaction kinetics under irradiation. In reality, accurately estimating the plasmon resonance under irradiation is more challenging, due to photothermal heating of the surrounding medium and multiple photon absorption, as discussed below (and briefly in the main text).
Supplementary Fig. 11: (a,b), Full colormap of the time evolution of the inelastic emission (IE) spectra (a) and ΔPL spectra (b) measured over a period of 15 min, corresponding to the data in Fig. 3a,b (main text). c, FDTD simulations of scattering cross-sections of 62 nm diameter Au (gray line) and corresponding Au@CeO 2 core@shell (teal) nanoparticles with 15 nm shell thickness on a quartz substrate immersed in water. The blue line corresponds to the spectral difference of the scattering cross-sections of Au@CeO 2 and Au nanoparticle, which is henceforth referred as the differential spectrum. The maximum of differential spectrum correlates well to the maximum of Au@CeO 2 scattering cross-section. d, FDTD simulations of scattering cross-sections of 62 nm diameter Au (gray line) and corresponding Au@CeO 2 core@shell (teal) nanoparticles with 2 nm shell thickness on a quartz substrate immersed in water. The blue line corresponds to the spectral difference of the scattering cross-sections of Au@CeO 2 and Au nanoparticle. The maximum of differential spectrum here displays a red-shift compared to the plasmon resonance of the core@shell nanoparticle. e, Correlation of the maximum of the differential spectrum against the corresponding LSPR of Au@CeO 2 core@shell nanoparticles (gray dots), at various ceria shell thicknesses ranging from 2 In our experiments, we find a blue-shift for the max[ΔPL( )] with respect to their LSPR after irradiation ( Fig. 3c in main text), which can be attributed to two factors: 1) large nanoparticle surface temperatures, which can decrease the refractive index of the surrounding medium, which in turn blue-shifts the plasmon resonance and therefore the particle's PL, and 2) high laser intensities that lead to successive photon absorption inside the metal nanoparticle, that can move electron distributions to higher energy levels.

Decreasing refractive index of surrounding medium:
The high temperatures generated on the surface of the nanoparticles can change the refractive index of the surrounding water medium. This lowering of refractive index of the surrounding medium, leads to a blue-shift of the plasmon resonance. Since the photoluminescence of a nanoparticle spectrally follows its plasmon resonance, the decreasing refractive index of surrounding medium at high temperatures leads to a blue-shift of the PL. Mie theory calculations for a 66 nm diameter Au nanoparticle shows that the plasmon resonance blue-shifts by ~15 meV when the particle is heated from room temperature (293 K, refractive index of water -1.333) to 220 °C (493 K, refractive index of water -1.293).

Successive photon absorption: Previously, Link et al. have reported a blue-shift in the inelastic
photoluminescence of Au nanorods with respect to their corresponding elastic dark-field scattering spectra 4 . They reported a blue-shift of 16-23 meV, depending on the interband excitation wavelength. They argue that such blue-shift is attributed to the higher energy of hot-charge carrier distributions under higher excitation powers. Under such laser intensities, successive photon absorption takes places before the excited hot-charge carriers can relax in to the lattice. As such, the overall energy of the hot carrier distributions is raised, leading to a blue-shift of the photoluminescence spectra. In their experiments, they calculate the average time intervals between successive photon absorption to be between 100 -500 fs, which matches the hot-electron thermalization time scales. Furthermore, in their experiments they calculate an increase in surface temperatures of less than 30 K, which rules out nanoparticle reshaping induced blue-shift of photoluminescence. In our experiments, the average time interval between two photon absorption is in the order of 5 fs. Such higher laser intensities in our experiments will also lead to higher energy distributions of hot-charge carriers, which can contribute to blue-shift of the plasmon resonance.

PL measurements.
By plotting the max[ΔPL] as a function of time for the growing core@shell nanoparticle, we can estimate the absorption cross-section, diameter, and ΔT of the growing nanoparticle as a function of time, by correlating them with FDTD calculations. Such insights can be useful to study the kinetics of nanoparticle syntheses and also to design novel nanoparticle syntheses. Using FDTD simulations, we can estimate the photothermal ceria shell growth on Au nanorods. By measuring the LSPR of the Au nanorods before and after irradiation, we can calculate the red-shift in the plasmon resonance, which can then be used to estimate the ceria shell thickness by assuming a constant refractive index of ceria. Supplementary Fig. 13: (a) FDTD simulation of the scattering cross-sections of a 25 × 59 nm gold nanorod (gray) and a Au@CeO 2 core@shell nanorod (orange) with 0.75 nm shell thickness. A 16 nm LSPR red-shift is observed for the growth of 0.75 nm shell thickness. In both the calculations, the nanorods are considered to be at the quartz-water interface. (b) Experimentally measured dark-field scattering spectra of Au and Au@CeO 2 core@shell nanoparticles corresponding to the data in Fig. 4b (main text). The orange shaded area denotes the spectral range shown in Fig. 4b   The absorption cross-section of these rods at 532 nm is less than that of Au nanospheres, by roughly an order of magnitude. As such, we expect a much lower photothermal temperature increase on these nanorods, which explains the measured small ceria shell thickness. The photothermal growth of CeO 2 can be followed by using our ΔPL approach, while in the case of ZnO and ZnS, it is challenging to track the growth as the inelastic emission intensity decreases with increasing time. The decrease in emission intensity indicates non-radiative quenching of the PL of the Au nanoparticle by the growing ZnO and ZnS semiconductor.
Since the bandgap of ZnO and ZnS are above the illumination energy, the luminescence exhibited from these materials should originate from the defects present in them. We postulate that defects in ZnO