Toward Ultrathin Gold Nanoshells: Exploring Seed Density, Polymer Additives and Microwave Radiation Effects

. Nanoshells made of a silica core and a gold shell possess an optical response that is sensitive to nanometre-scale variations in shell thickness. The exponential red shift of the plasmon resonance with decreasing shell thickness makes ultrathin nanoshells (less than 10 nm) particularly interesting for broad and tuneable ranges of optical properties. Nanoshells are generally synthesised by coating gold onto seed-covered silica particles, producing continuous shells with a lower limit of 15 nm, as a result of an inhomogeneous droplet formation on the silica surface during the seed regrowth. In this paper, we investigate the effects of three variations of the synthesis protocol to favour ultrathin nanoshells: seed density, polymer additives and microwave treatment. We first maximised gold seed density around the silica core but surprisingly its effect is limited. However, we found that the addition of polyvinylpyrrolidone during the shell synthesis leads to higher homogeneity and a thinner shell and that a post-synthetic thermal treatment using microwaves can further smooth the particle surface. This study brings new insights into the synthesis of metallic nanoshells, pushing the limits of ultrathin shell synthesis.


INTRODUCTION
Metallic nanoshells (or core@shell particles) are nanoparticles made of a dielectric core surrounded by a metallic coating. They have attracted much attention over the past two decades because of their intense optical properties. Independent control over two parameters, i.e. the core radius and shell thickness, offers good tunability over the absorption and scattering crosssections of the particles in the visible and near-infrared region. Therefore, they have been used for many applications such as water desalination and purification, as well as sterilisation, 1 fluorescence enhancement of dye molecules, 2 improvement of dye-functionalised solar cell, 3 and biosensing. 4 In general, the optical extinction of metallic nanoshells red shifts as shell thickness decreases and is located mostly in the near-infrared region for thicknesses below 10 nm, making particles with ultrathin shells of particular interest for biomedical applications in the nearinfrared window. 5,6 The synthesis of gold nanoshells was first described by Oldenburg et al. in three steps: surface amination of silica cores, deposition of Duff's seeds (2-3 nm gold seed) and regrowth of the seeds using a gold plating solution and a reducing agent. 7 The latter is preferentially formaldehyde because it favours a homogeneous seed growth as compared to sodium borohydride or hydroxylamine hydrochloride, 8 and because it is more practical than gaseous CO. 9,10 In terms of the synthesis mechanism, growing seeds eventually touch and merge to form a shell. Previous experiments have shown a qualitative relationship between the seed density and the shell thickness: a n increase i n seed density produces thinner shells. 10,11 However, a quantitative description of this relationship is not straightforward. The formation of a continuous shell results not only from gold plating, but also from an additional process involving seeds coalescing into a smaller number of particles on the surface in the form of droplets. 12 This leads to a commonly observed intermediate state (between particles decorated with seeds and a continuous nanoshells) for which particles have a raspberry-like shape 7,8,13 and whose optical properties are well described by numerical simulations. 12 After regrowth, the synthesis eventually yields continuous shells having at least 15 nm of thickness. 2 Unfortunately, only by getting a thinner continuous shell can the red shifted extinction spectra be obtained.
Since the first report of nanoshells, 7 many variations to the protocol have been proposed to improve or accelerate the synthesis. The most common modifications include adapting the functionalising agent, the reducing agent and the reaction conditions. 11, [14][15][16][17][18] It appears that many parameters have direct effects on the synthesis products, such as smoothness, homogeneity and continuity of the metallic shell, yet few enable the synthesis of an ultrathin continuous shell. For example, the quality of the metallic shell is dependent on the age of the seed solution 19,20 and the gold plating solution 21 , the pH 22 of the latter, as well as the stirring speed during the formation of the shell. 13 On the other hand, ultrathin shells were obtained by specific parameters combination such as high seed density and CO (g) as reducing agent. 10 Overall, because of the high sensitivity of nanoshell synthesis to numerous parameters, to study the effects of a given parameter, it is primordial to synthesise reference samples under the same reaction conditions. In this paper, we produce nanoshells using a seeded-growth process and vary three independent parameters to study how they improve the synthesis of ultrathin shells: seed density, polymer additives and microwave treatment. We first demonstrate that, surprisingly, maximising gold seed density on silica particles does not help to produce ultrathin shells. The seed density on the silica particles was altered by increasing ionic strength during the seed deposition and performing a second functionalisation/deposition step. Although the change in seed density significantly impacts the optical spectra before shell growth, after growth, the spectra are virtually identical. We then study adding a polymer during shell growth to favour the formation of a thin shell. Finally, we study flash microwave radiation treatment as a post-synthetic tool to smooth the nanoshells. Compared to the reference samples, these two innovative approaches led to thinner and smoother shells and exhibited an increased intensity in the extinction spectra in the near infrared.

Seed-covered silica particles
In order to study the effect of seed density on the synthesis of ultrathin gold nanoshells, we synthesised nanoshells using silica particles covered with different seed densities. We used and combined two different approaches to vary and maximise the seed density on the silica particles.
Because the nanoshell synthesis is highly sensitive to different experimental conditions, the different samples were prepared in parallel, to minimise the effect of environmental parameters.
We first pre-synthesised highly monodisperse (polydispersity index less than 2%), ultra-smooth 300 nm silica spheres using a step-by-step growth. 24 Monodispersity improves the ensemble optical properties. 25 After functionalising the silica particles with amine groups by surface condensation of (3-aminopropyl)trimethoxysilane (APTMS), we mixed them with 14-day old seeds under different conditions ( Figure 1a). First, half were mixed without NaCl and half at a salt concentration of 60 mM of NaCl (samples A 1 and B 1 in Figure 1a). Salt decreases the electrostatic repulsion between the particles enabling a higher seed deposition density. 11 UV-Vis spectroscopy confirms the increased density, showing a stronger extinction resulting from a larger number of gold seeds in the sample, the silica particle concentrations being equal ( Figure   1b). Half of these batches were used to grow a nanoshell and the two other halves were then functionalised a second time, following the protocol described in a paper reporting a seed coverage increase from 30 to 60%. 10 In this step, pre-polymerized APTMS is used as a linker to reactivate the first gold-seeded silica core. The seed deposition was then performed without NaCl (sample A 2 , synthesised from A 1 ) or with a salt concentration of 60 mM (sample B 2 , synthesised from B 1 ). We observe that the UV-Vis spectrum of A 2 and B 1 are nearly overlaid meaning that the seed density is very close in both samples (Figure 1b). It demonstrates that performing a double functionalisation/deposition step is roughly equivalent to a single deposition using 60 mM NaCl in terms of achieved seed density. However, the highest density is obtained for the sample that had both treatments: double functionalisation/deposition and salt used during the deposition, as confirmed by UV-Vis and TEM images (Figures 1b-c and S1). In the electron microscopy images, the seed density difference is especially visible on the edge of particles, where we can observe individual isolated seeds for the sample with the lowest density, and a tight seed packing for the sample with the highest density. However, given the low contrast between the seeds and the silica, and the intensity change between different electron microscopy images, it is extremely difficult to provide a precise seed coverage for all the samples. We estimate coverage to be around 30-40 % for A1, 50-60% for A2 and B1, and 60-70 % for B2, based on zoomed-in dark field TEM images (Figures 1c and S1) as performed in previous reports. 10

Nanoshell synthesis
To study the effect of the seed density on the synthesis of ultrathin shells, the four previous samples, which have the same silica particle concentration and different seed densities, were used to grow a metallic shell using a gold plating solution and formaldehyde as a reducing agent.
The nanoshell synthesis was performed using aliquots of the four samples of seed-covered particles, which were mixed with different quantities of gold plating solution (100, 800, 1600 and 2666 µL, corresponding to gold salt concentrations of 0.02, 0.16, 0.32 and 0.53 mM, respectively). The optical properties of the different particles produced were measured by UV-Vis spectroscopy (Figure 2a-d). For the lowest gold plating solution volume, we observe a single peak around 520 nm, corresponding to the plasmon resonance of small gold particles. This peak strongly red shifts and intensifies with increasing gold concentration, resulting from both the particle size increase and the plasmon coupling between neighbouring gold particles. 12 This intermediate state corresponds to the formation of gold droplets on the silica surface particles, resembling raspberry-like particles (as schematised in Figure 3d), 7,8,12,13 instead of a uniform growth of seeds. The origin of this uneven particle growth is not completely clear as several phenomena might occur. Steric hindrance could explain this behaviour: the largest surface-bound seeds probably grow faster, as they have a larger reactive surface protruding farther out of the silica surface. As they grow, it becomes even more difficult for small seeds on the surface to react, therefore favouring the continuous and selective growth of the large particles, which eventually dewet to minimize surface energy. Alternatively, Preston et al. noted that some areas initially containing small seeds were later bare, 12 indicating that droplets are formed from several seeds. This decrease in the number of seeds could also result from particle detachment during the regrowth, observed, for example, when using very fresh gold seeds or from seeds not covalently bound to the silica surface. 27 It could also result from the increase of the attractive Van der Waals forces between neighbouring seeds during the seeds growth, forcing some of them to detach to form aggregates. Overall, this behaviour results from various mechanisms and prevents the formation of ultrathin shells. Particles produced with the highest gold plating solution volumes exhibit an increase in extinction in the near-infrared region, corresponding to droplets merging and forming a percolated shell. High-resolution TEM images of the largest particles show a shell of approximately 12 nm on average, mostly continuous but not smooth and exhibiting faults, as evidence of the large droplet formation mechanism (Figures 2e and S2). The shape of the extinction spectra is quite typical 10,12 and differs from numerical simulations which exhibit welldefined peaks of smaller width. The difference might result from roughness and discontinuities in the metallic shell 11,2813,22 leading to a red shift and broadened resonance. 29 Good correlation between the numerical simulations of the particle extinction and the measured UV-Vis spectrum is obtained for large shell thickness (>20 nm) as the defect contribution to the optical response diminishes. 30 Surprisingly, particles made from different seed densities possess very similar optical properties after shell growth (Figure 2a-d). Because of the strong dependence of the optical properties of nanoshells on the shell thickness, 7 this means that all samples are nearly identical.
In other words, it demonstrates that the relationship between the seed density and the achievable shell thickness is more complex than originally thought. The decrease of the shell thickness reaches a plateau when the seed density is higher than 30-40% (value estimated for the sample A 1 made in a single deposition step and no salt 10,26 ). We assume that this behaviour originates from the formation of the droplets, which must be relatively insensitive to the seed density above a certain threshold. This lack of effect from seed density is different to what was observed in previous papers, 10,11 however, it could be accounted for by other parameters. In Zhang and coworkers' paper, 10 the use of CO as reducing agent may produce a different formation mechanism, thus seed coverage density may impact shell thickness with this reducing agent. In García-Soto and coworker's article, 11 differences with our protocol regarding a longer seed deposition process and the seeded-silica particle ageing, could perhaps explain the behavioural difference. We thus pursued alternative methods to decrease shell thickness.

Comparison of the shell formation mechanism using polyvinylpyrrolidone
Based on our previous conclusions, our aim was to find an approach that minimises the formation of large gold droplets during the nanoshell synthesis, in order to achieve an ultrathin shell. It has been shown that the use of polymers, polyvinylpyrrolidone (PVP) or polyethylenimine, could affect the quality and shell thickness in the synthesis of silver nanoshells. 23,31 For example, polyethylenimine was used as shell growth facilitator and reactionrate regulators, favouring the formation of a thin and continuous silver shell of about 10 nm. 23 Here, we investigated this strategy and studied the effect of polyvinylpyrrolidone (PVP) on the growth of gold nanoshells. PVP is one the most widely used steric agents and exhibits a good affinity for gold as well as silica due to its amphiphilic character. When mixed to core-shell particles, it adsorbs onto the particles and form a layer whose density and thickness depend on the polymer molecular weight. [31][32][33] In our experiment, PVP (40 kg/mol) was added to our seeded silica particles (sample B 1 ) before the regrowth of the seeds. The thickness of PVP layer of such length is difficult to estimate, likely below 10 nm. Upon addition of PVP to the solution, the gold regrowth reaction rate is strongly affected, with a reaction time increasing from a few minutes to a couple of hours. The effect of PVP on the synthesis of thin nanoshells was studied by simultaneously preparing two samples, one with and one without PVP. UV-Vis spectroscopy of the samples shows that the extinction in the near-infrared region is larger for the sample prepared with PVP (Figure 3a). We also observe that the amplitude and the slope of the peak are large in the region from 500 to 800 nm, and that amplitude is higher in the near-infrared region, suggesting a more complete shell. SEM images of the two samples confirm the effect of PVP on the formation mechanism of the shell described above (Figures 3b-c). The particles prepared using PVP exhibit a higher as previously suggested, a slower shell growth offers the possibility of matching the reaction rate and the mixing to obtain a uniform shell growth, and therefore a thinner shell. 23,31 The origin of the reaction rate being slowed down may come from interactions between Au ions and the polymer that form complexes. Also, the adsorbed PVP shell around the silica-gold core@shell likely diminishes the diffusion rate of gold ions compared to bare particles. It is also possible that the PVP shell surrounding the particles favours the gold regrowth parallel to the silica surface.
Not only PVP increases the colloidal stability of the core@shell particles, it may also sufficiently shield the attracting van der Waals forces between the growing seeds, preventing the formation of clusters observed in raspberry-like nanostructures. Overall, PVP plays an important role during the growth process, improving both the smoothness and homogeneity of the obtained shells. The approach presented here opens a simple way to obtain ultra-thin metal-coated silica colloids.

Post-synthesis microwave treatment
Another approach to obtain a percolating shell consists of smoothing the nanoshell postsynthesis. To do so, we performed a thermal treatment of gold nanoshells (regrown from A 2 without PVP) using microwaves. It simply consisted of a flash treatment, at 140°C, for 1 minute.
Under other parameters (different microwave power and internal pressure), particle dewetting was observed (results not shown). Using these optimised conditions, improved optical properties were quickly and easily obtained. The UV-Vis spectroscopy shows that the treatment leads to a moderate increase in the extinction in the near-infrared region (Figure 5a). This is consistent with small gold islands merging into larger and smoother patches, forming a more continuous shell (Figures 5b, c, d and S6). The nature of the effect of microwaves in this process is still unclear, as the interaction between the chemicals and strong electromagnetic field is complex and because several physical parameters are modified simultaneously: localised pressure and temperature and global temperature. However, under appropriate settings, the temperature increase of the sample may occur so that the gold droplets soften without dewetting, leading to a non-quite melted, smoother and thinner, shell. It is also possible that the pressure in the vessel slows down the dewetting, compared to classical heating, offering the possibility of melting the shell before dewetting.

CONCLUSIONS
Although the first synthesis of metallic nanoshells appeared 20 years ago, the synthesis of ultrathin shells with interesting extinction in the near-infrared remains challenging. We show that by combining a high ionic strength during the seed deposition with a double functionalisation/deposition step, we obtain a very high seed density on silica particles. However, we demonstrate that an increased seed density does not yield thinner shells, which we explain using the synthesis mechanism. In fact, seed regrowth is associated with the formation of less numerous, but larger droplets and their following regrowth eventually produces inhomogeneous nanoshells whose thickness is larger than 15 nm on average.
Once establishing that the limiting factor in obtaining ultrathin shells is not the seeddeposition, but the regrowth step, we turned our attention to this process. To achieve a more homogeneous seed regrowth, PVP was added prior to the gold plating solution. This widely used steric stabilising agent prevented seeds dewetting into thicker shells, allowing a much more homogeneous coverage of smaller particles. We also employed a post-synthetic microwave treatment of the nanoshells to smooth the surface of the metallic shell. By using a flash treatment, gold particle melting, which would inevitably lead to dewetting, was avoided, successfully producing a percolated shell. The combined modifications maintained precise control over morphology and our findings could also be extended to other metals to help improve the synthesis of ultrathin Ag or Cu nanoshells of high interest for energy saving applications.
Silica particles were synthesised following the step-by-step published protocol. 19 Synthesis of the silica seeds. An L-arginine aqueous solution (6 mM, 100 mL) was placed in a 150 mL double-walled vial, equipped with a reflux condenser. When the temperature stabilised at 60 °C, 10 mL of TEOS was added. The heating/stirring system was stopped when the TEOS upper phase disappeared.
First regrowth of the silica seeds. Ethanol (99%, 455 mL), ammonia (28%, 33 mL) and an aqueous dispersion of seeds (10 mL) were mixed and placed under stirring. TEOS (20 mL) was added to the solution with a syringe pump at the rate of 0.5 mL/h. Once finished, another 20 mL of TEOS was added using the same protocol. The as-obtained silica particles were approximately 125 nm in diameter. They were washed by two cycles of centrifugation/redispersion in ethanol (9,000 g, 10 min, 55 mL).
Second regrowth of the silica seeds. Ethanol (99%, 450 mL), ammonia (28%, 34 mL), water (10 mL) and the previous silica seed dispersion (5 mL) were added and stirred. Successively 20 mL, 15 mL and 8 mL of TEOS were added to the solution with a syringe pump at the rate of 0.5 mL/h. The synthesised particles were approximately 320 nm in diameter. They were purified by three cycles of centrifugation redispersion in ethanol (8,000 g, 10 min, 55 mL). The final volume of the silica nanoparticle dispersion was increased to 250 mL with ethanol and the silica concentration was measured to be 58.45 mg/mL by the dried extracts method. b) Surface amination of the silica particles.
The previous silica particle dispersion (8.55 mL corresponding to 0.5 g) was diluted in 200 mL of ethanol. APTMS (5 mL) was added to the mixture and stirred overnight. The dispersion was then heated to reflux overnight. The aminated silica nanoparticles were purified by three cycles of centrifugation/redispersion in ethanol (8,000 g, 10 min, 25 mL). Upon the final cycle, the particles were dispersed in 40 mL of ethanol with a final measured silica concentration of 19.5 mg/mL. c) Gold seed deposition on the silica particle surface.
Synthesis of gold seeds according to the Duff's protocol. 34 Water (227.5 mL) was mixed with 0.2 M NaOH (7.5 mL) and 5 mL of diluted THPC (120 μL of the stock solution diluted in 10 mL of water) for 15 min. Then, 5 mL of HAuCl 4 (25 mM) was quickly added and the solution changed from colourless to brown immediately, indicating the formation of gold seeds. The asprepared gold-seed solution was kept at 4°C for 14 days before use.
First deposition of the gold seeds onto the silica core (samples A 1 and B 1 ). APTMSfunctionalised silica nanoparticles (1.5 mL) were added to 38.5 mL of gold seed solution without NaCl (A 1 ) or with an additional 3.478 mL of 1 M NaCl (B 1 ). The dispersions were slowly agitated overnight. Excess gold seeds were removed by two cycles of centrifugation/redispersion in water (6000 g, 10 min, 40 mL). After the final cycles, the particles were redispersed in water (10 mL).
Preparation of pre-polymerized APTMS solution. This synthesis was adapted from a published protocol. 10  The excess reactants were removed by two cycles of centrifugation/redispersion in water (from 1000 to 4000 g depending on shell thickness, 5 min, 10 mL). After the final cycles, the particles were redispersed in water (10 mL).

Funding Sources
This work was jointly supported by the Solvay company and the LabEx AMADEus (ANR-10-LABX-42) in the framework of IdEx Bordeaux (ANR-10-IDEX-03-02), i.e. the Investissements d'Avenir program of the French government managed by the Agence Nationale de la Recherche.

Competing interests
The author(s) declare no competing interests.       Silica particles covered with different gold seed densities. a) Scheme of the different parameters used during the seed deposition onto APTMS-functionalised silica particles. The four samples produced are then used to grow core@shell particles. b) UV-Vis spectroscopy of the four different seed-covered silica particle samples. c) TEM images of the samples corresponding to the three different extinction curves. Scale bars represent 50 nm.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.