Cysteine-encoded chirality evolution in plasmonic rhombic dodecahedral gold nanoparticles

Chiral plasmonic nanostructures have opened up unprecedented opportunities in optical applications. We present chirality evolution in nanoparticles focusing on the crystallographic aspects and elucidate key parameters for chiral structure formation. From a detailed understanding of chirality formation, we achieved a morphology (432 Helicoid IV) of three-dimensionally controlled chiral plasmonic nanoparticles based on the rhombic dodecahedral shape. The role of the synthesis parameters, seed, cysteine, cetyltrimethylammonium bromide and ascorbic acid on chiral formation are studied, and based on this understanding, the systematic control of the chiral structure is presented. The relation between the modulated chiral structure factors and optical response is further elucidated by electromagnetic simulation. Importantly, a new optical response is achieved by assembling chiral nanoparticles into a film. This comprehensive study of chiral nanoparticles will provide valuable insight for the further development of diverse chiral plasmonic nanostructures with fascinating properties.

line, right) spectra of gold chiral nanoparticles with corresponding models. Different structural parameters were studied: the particle size (b, #1-6); core size (c, #7-11); degree of bending of the edge (d, #12-16); protrusion of the edge (e, #17-20); width of the edge (f, #21-24); width of the edge (design 2) (g, #25-28); and width of the edge with increased protrusion (h, #29-32). The trend of the transition spectra from a small number to a large number of models is also indicated by the direction of the arrow. process was used to vary the attached amount of nanoparticles on the substrate (see Methods). An increasing color contrast was observed with increasing nanoparticle density. A reflective gold color was generated for the case of the substrate treated with the dip-coating procedure. (b) SEM images showing the particle density of the substrate prepared in (a). (c) CD spectra of the substrate with different particle densities. With increasing density, the two peaks with different ratios from the solution case gradually increase due to the coupling between the particles. (d, e) Polarization-resolved transmission images of a substrate coated with achiral nanoparticles (d) and 432 helicoid IV (e). Starting from the cross-polarized condition (0°), the angle of the analyzer was changed from -1.5° to 1.5° in the case of the achiral sample and -1.6° to 1.6° in the case of the 432 helicoid IV sample. The helicoid substrate shows an asymmetrical color transition depending on the angle, while the achiral sample displays only a symmetrical change.

Supplementary Discussion 1
Chiroptical response in plasmonic nanostructure can be generated by several mechanisms. Generally, the mechanism of CD signal generation can be categorized by 1) plasmon-induced chirality, 2) intrinsically chiral nanoparticle structure, and 3) nanoparticle assembly with handed configurations 1,2 . In order to identify which mechanism dominates the chiral response of 432 helicoid IV, we performed series of control experiments (Supplementary Figure 3). The first case can be obtained by incorporation of chiral organic molecules to the plasmonic structure. Therefore, here we conducted and analyzed the additional experiment by focusing on the organic ligand of the 432 helicoid IV surface. For experimental demonstration of plasmonic-induced chirality, two methods, attaching chiral ligand 3,4 and embedding chiral molecule in plasmonic structure [5][6][7] , can be applied. Cys and L-GSH, the spectral features of CD were rarely changed and only intensity decrease observed. Also, further addition of L-Cys molecule on the surface, did not bring about enhancement of the g-factor. The reduced g-factor intensity in the L-Cys and D-Cys cases might be originated from the instability of the nanoparticles due to the shortening of the ligand by ligand exchange. When the L-Cys molecule detached by using NaBH 4 (10 mM NaBH 4 with 5 min incubation, washing, repeating 3 times), no significant g-factor change was observed. These marginal effect on the g-factor with the addition and detachment of chiral ligands indicate that the chiral ligand is not the main contributing factor for CD generation.

Embedment of chiral ligand
To prepare nanostructures in which chiral molecules are embedded, a pre-adsorption process is necessary. By adsorbing chiral molecules on the plasmonic nanoparticle and sequentially growing this pre-treated nanoparticle, the chiral molecules can be located or entrapped inside the plasmonic structure. To determine whether this mechanism plays a major role in chiral signal generation, we pre-adsorbed Cys molecule and grow this pre-treated seed in growth solution. First, we added Cys molecule (the same concentration with original synthesis process) into cuboctahedron particle and aged for 1 hour. Second, the pre-treated particle was grown in the growth solution without containing Cys molecule. At the stage of second growth, two type of pre-treated particle were prepared, one with the centrifuged to get rid of remaining Cys and one without centrifuged. Two different concentration of Cys were tested, but in all cases, only small g-factors and random shape of nanoparticle were observed (Supplementary Figure 3b and c). The deep dents observed in the nanoparticles are probably due to the adsorption of Cys. This difference depending on the preparation method suggests that embedding of molecules may not be the cause of CD signal generation.

Reduced g-factor at 0.3 mM Cys case
In the case of nanoparticle grown under the 0.3 mM Cys concentration (discussed in later part, Fig. 4), which is expected to include more Cys molecules on the surface than 0.2 mM and 0.1 mM case, the resultant particle shows significantly decreased g-factor (Fig. 4). In spite of the increased organic molecules in the plasmonic nanostructure, the result of a reduced g-factor suggests that a well-defined chiral structure plays an important role in determining the g-factor.
4. Effect of the CTAB and AA concentration on g-factor We studied the effect of CTAB and ascorbic acid (AA) concentration on g-factor while the Cys concentration remains the same (Fig. 5a). The CTAB and AA are achiral ligands and these factors affect the dimension of nanoparticle, e.g. CTAB changes the distance between chiral edges and AA modifies the height of chiral edge (Supplementary Figure 14a). In this experimental set, the g-factors were changed more than 10-fold by changing the amount of this achiral ligand without changing any amount of Cys. The SEM images of the particles obtained by varying the CTAB and AA show that the g-factor varies greatly as the distance or height between the chiral edges changes, even if the degree of bending does not change. This experimental evidence directly shows that formation of defined structure is critical for chiroptical response.

Supplementary Discussion 2
In the case of 432 helicoid I, the cubic shape of seed nanoparticle is transformed into {321} polyhedral shape The 432 helicoid II also showed a crystal structure which is close to the {321} index at the beginning of growth (10 min). However, since GSH was present and attached from the beginning, less growth was observed along the <100> direction and distorted edges were observable after 20 minutes of growth.
For the 432 helicoid III, the growth direction is different from the 432 helicoid I and II as it starts from octahedron seed composed of {111} planes. As the {111} plane is gradually added, the outline of the RD appears in 6 min of growth, and with further addition of {111} plane, distorted edges based on the (221) polyhedral shape appears.
In the case of 432 helicoid IV, since the growth was slowed by the doubled seed amount, the particle size was relatively small after 6 minutes of growth. The characteristic outline of the RD shows that the (100) face of the seed has been changed to the (110) facet. Cys, which is present at the beginning of growth, interacts with the crystallographic surface that appears in the transition between the (100) and (110) planes. Therefore, these differences create a distorted edge based on the RD outline.

Model construction
To build a structural model of 432 helicoid IV nanoparticles, we carefully analyzed the particles in all different directions based on SEM images. In the case of chiral nanoparticles synthesized under 0.1 mM cysteine, we observed characteristic outlines corresponding to a rhombic dodecahedron (RD) with slight distortion. Depending on the viewing direction of the RD, three different outlines, an elongated hexagon for the (110) axis, a hexagon for the (111) axis, and a square for the (100) axis, are observed (Fig. 2). The axis of the chiral nanoparticle was confirmed by TEM analysis. Based on the RD structure, we carefully designed chiral models by changing the edges of the nanoparticles. Three simplified models that describe the chiral twist of the edge in different ways were simulated and showed spectral features that corresponded well with experimental results ( Supplementary   Fig. 10).

Analysis of structural factors for optical activity
In this work, the diverse structural evolution of chiral nanoparticles was demonstrated. By varying the geometrical parameters in the simulation model, we studied the correlation between structural factors in nanoparticles with a chiroptical response. Models #2 and #3 in Supplementary Fig. 10 were selected in this structural-optical relationship study because of the ease of deformation in the chiral edge dimensions.
(a) Particle size ( Supplementary Fig. 15b) With increasing nanoparticle size, the extinction, CD, and g-factor linearly increased. As an increased particle size provides stronger dipolar modes and even high-order multipolar modes, this increase resulted in stronger extinction and chiro-optical responses.
(b) Core size ( Supplementary Fig. 15c) In this modification, only the size of the achiral core part was increased, while the width of the chiral structure remained the same. Although CD increased due to size increments, the g-factor decreased with increasing achiral core size. This result indicates that the proportion of the chiral edge in the entire nanostructure is important in achieving a high g-factor.
(c) Degree of bending of the edge (Supplementary Fig. 15d) Increasing the degree of bending of the edge resulted in a significant change in the chiral response. While the extinction spectrum remained almost unchanged, the CD and g-factor were enhanced with the increase of the twist in the case of the #12 -#15 models. For example, a 16° increment of the bending angle led to a 3-fold enhancement of the g-factor. This trend explains the enhancement of the g-factor with increasing Cys concentration in the experimental results. However, in the case of the #16 model, a substantially decreased g-factor and CD spectra were observed. This result is probably due to strongly perturbed plasmonic modes and charge distribution near the two close (but opposite) chiral edges. Additionally, this might be the reason for the reduced CD intensity in the 0.3 mM Cys case.
(d) Protrusion of the chiral edge ( Supplementary Fig. 15e) As the chiral edge protruded, steep increases in CD and g-factor were observed. Increasing the height of the protrusion not only increases the proportion of chiral structures in the particle but also increases the depths of the chiral gaps in the nanoparticles. The importance of the deep chiral gap in the CD signal was also pointed out in our previous paper. 8 This large enhancement of the g-factor with protrusion of the edge may originate from the strong dimeric coupling between two edges separated by the plasmonic gap. Increasing the height of the protrusion by a factor of 4 (changes from #17 to #20) resulted in a 10-fold increase in the g-factor. In a single parameter variation, an increased protrusion resulted in the greatest enhancement in the g-factor (Supplementary Fig. 15i).
(e) Width of the edge (Supplementary Fig. 15f and g) To simulate the effect of the width, we tested two structural models with different modulations of the outer boundary. In both cases, widening the width of the edge by changing the curvature of the outer boundary resulted in enhancement of the g-factor. However, excessive expansion of the outer boundary led to a decrease in the gfactor, which might have originated from coupling of close edges, similar to the #16 model case.