Process Window for Seeded Growth of Arrays of Quasi-Spherical Substrate-Supported Au Nanoparticles

The controlled growth of surface-supported metal nanoparticles (NPs) is essential to a broad range of applications. To this end, we explore the seeded growth of highly ordered arrays of substrate-supported Au NPs through a fully orthogonal design of experiment (DoE) scheme applied to a reaction system consisting of HAuCl4, citrate, and hydrogen peroxide. Scanning electron microscopy in combination with digital image analysis (DIA) is used to quantitatively characterize the resultant NP populations in terms of both particle and array features. The effective optical properties of the NP arrays are additionally analyzed using spectroscopic ellipsometry (SE), allowing characteristics of the localized surface plasmon resonances (LSPRs) of the arrays to be quantified. We study the dependence of the DIA- and SE-extracted features on the different reagent concentrations through modeling using multiple linear regression with backward elimination of independent variables. A process window is identified for which uniform arrays of quasi-spherical Au NPs are grown over large surface areas. Aside from reagent concentrations the system is highly sensitive to the hydrodynamic conditions during the deposition. This issue is likely caused by an Au precursor mass-transport limitation of the reduction reaction and it is found that agitation of the growth medium is best avoided to ensure a macroscopically even deposition. Parasitic homogeneous nucleation can also be a challenge and was separately studied in a full DoE scheme with equivalent growth media but without substrates, using optical tracking of the solutions over time. Conditions yielding quasi-spherical surface-supported NPs are found to also be affiliated with strong tendencies for parasitic homogeneous nucleation and thereby loss of Au precursor, but addition of polyvinyl alcohol can possibly help alleviate this issue.

: SEM-inspection at 5k (a-d) and 200k (e-f) magnification of sample parent wafers after the HAuCl4-loading of the BCP template. Panels i-p constitute a verification of the seed particle arrays after ashing, dicing and stripping the S1813 protective coating. Two seed reference samples per parent wafer were inspected.
= 1 kV was used in panels a-p. The image in panel q is from a spare edge piece from wafer 3 inspected at = 5 -note the difference in apparent particle size and noise level. Comment: N and ϕ are the number of counted NPs and the fractional area coverage of NPs respectively. Area , Max FD , FF , Ellips. and d are the mean NP cross-sectional area, max Ferret diameter, fill factor and center-to-center interparticle distance respectively while s Area , s FD max , s FF , s Ellips. and s d are the corresponding standard deviations. s θ is the standard deviation of the discrepancy between the angle spanned by the centroids of the two nearest neighbours and that of a perfect hexagonal geometry. t eff , E LSPR , ε 2,LSPR and FWHM are the modelled effective medium thickness, LSPR peak position, magnitude and width respectively. MSE is the mean square error of the SE model. a Peak position, height and width were extracted from one sample from each parent wafer. b The DIA was perfomed on SEM-images aquired at V acc = 1 kV and 200k magnification for all samples except the last entry when the images analyzed were taken at 5 kV and 400k magnification.  Figure S3: Compilation of SEM images, acquired at IP1, of the samples included in the DoE scheme. The seed reference sample was also inspected at = 5 and originates from parent wafer 3. The scalebar equals 100 nm.
5 Figure S4: Compilation of SEM images, acquired at IP2, of the samples included in the DoE scheme. The seed reference sample was also inspected at = 5 and originates from parent wafer 3. This figure is identical to Figure 3 in the paper and is included here for completeness and ease of comparison. The scalebar equals 100 nm. Figure S5: Compilation of SEM images, acquired at IP3, of the samples included in the DoE scheme. The seed reference sample was also inspected at = 5 and originates from parent wafer 3. The scalebar equals 100 nm. Area , Max FD , FF , Ellips. and d are the mean NP cross-sectional area, max Ferret diameter, fill factor and center-to-center interparticle distance respectively while s Area , s FD max , s FF , s Ellips. and s d are the corresponding standard deviations. s θ is the standard deviation of the discrepancy between the angle spanned by the centroids of the two nearest neighbours and that of a perfect hexagonal geometry. t eff , E LSPR , ε 2,LSPR and FWHM are the modelled effective medium thickness, LSPR peak position, magnitude and width respectively. MSE is the mean square error of the SE model. t 0 , τ , A and K are the parameters of the function fitted to ΔI/I 0 . Adj. R 2 is a goodness of fit measure.  Comment: The DIA was perfomed on SEM-images aquired at Vacc = 5 kV and 200k magnification. ΔN/N 0 and ϕ are the relative change in counted NPs and the fractional area coverage of NPs respectively. Area , Max FD , FF , Ellips. and d are the mean NP cross-sectional area, max Ferret diameter, fill factor and center-to-center interparticle distance respectively while s Area , s FD max , s FF , s Ellips. and s d are the corresponding standard deviations. s θ is the standard deviation of the discrepancy between the angle spanned by the centroids of the two nearest neighbours and that of a perfect hexagonal geometry.       ) and lowest HAuCl4 ( 4 = 50 ) DoE levels, maintaining the total volume of 40 mL while not including any H2O2. Ran for 60 min without change in medium appearance, upon which 5 mL 31% H2O2 was added and the appearance changed rapidly (<2 min). Figure S22: Compilation of data acquired from the DoE centerpoint replicates for both the samples and the unseeded media for the purpose of gauging the reproducibility of the SMNPG protocol. Panels a-e are plots of DIA-extracted features form SEM images acquired at IP2 on the DoE samples. Panel f is ( )/ 0 ) of the corresponding unseeded media. Panel g) is a compilation of SEM images from the DoE centerpoint replicate samples at all three inspection points (IP1-3). NB: The sample run here labeled "31" is actually sample DoE run 1. The scale bar equals 100 nm.

Comment:
The DIA was perfomed on SEM-images aquired at Vacc = 5 kV and 400k magnification. m is the mass of the sample. ΔN/N 0 and ϕ are the relative change in counted NPs and the fractional area coverage of NPs respectively. Area , Max FD , FF , Ellips. and d are the mean NP cross-sectional area, max Ferret diameter, fill factor and center-to-center interparticle distance respectively while s Area , s FD max , s FF , s Ellips. and s d are the corresponding standard deviations. s θ is the standard deviation of the discrepancy between the angle spanned by the centroids of the two nearest neighbours and that of a perfect hexagonal geometry. t SMNPG = 10 min. a Seed reference.

Use of Poly(vinyl alcohol) Instead of, or in Conjunction with, Citrate During SMNPG
As higher was previously observed to be affiliated with more extensive homogenous nucleation (HN) in solution, alternative complexing and capping agents are of interest. In a second follow-up experiment we thus investigated the effect of using poly(vinyl alcohol) (PVA) instead of, or in conjunction with, citrate in our reaction system. PVA is an inexpensive, nontoxic and widely available polymeric capping agent. The experiment consisted of two parts; in the first, unseeded growth media were prepared analogously with those in the corresponding DoE scheme except that PVA was added either instead of or after the citrate from a 1% (w/w) aq. stock solution. Total medium volume (40 mL) was maintained. In the second part, SMNPG was performed, using no growth medium agitation and an extended sample immersion time ( = 10 min), on seed-decorated samples. In running the unseeded media, we found that HN was not an issue (as inferred from Δ ( )/ 0 ) when replacing 400 µL 1% (w/w) citrate(aq) ( = 340 in medium) with an equal volume of 1% (w/w) PVA(aq). Even more interesting is the seeming ability to tune the onset of the Δ ( )/ 0 increase with the concentration of PVA, when added to a SMNPG medium containing with the highest DoE concentration levels of citrate and H2O2 (Figure S25   The results from the unseeded media suggests that higher Au precursor utilization and improved reproducibility might be possible through the addition of PVA when conducting SMNPG on surface supported NPs. To investigate this possibility, we conducted SMNPG on a series of samples using no growth medium agitation, = 340 μM, 4 = 100 μM, 2 2 = 5.06 M and a progressively increasing amount of added PVA. Two controls, for which no citrate and either no or 400 μL 1% PVA was used, were also included for reference ( Figure  S26 a). Digital image analysis (DIA) of SEM images acquired at the sample centers, suggests that the size and ellipticity abruptly increase upon our lowest level of PVA addition and then remains fairly constant with increasing amount of PVA added. We also evaluated using only PVA or neither citrate nor PVA and obtained substantially higher ellipticity values in both cases than when using either citrate alone or the citrate-PVA combination (Table S6 and Figure S27). Moreover, more severe deterioration of the array pattern quality was seen when only using PVA (Figure S26 a). An intriguing phenomenon was observed when inspecting the sample for which citrate and the highest PVA concentration were used. The NPs close to the immediate sample edge exhibit a 'popcorn-like' morphology, although this appearance rapidly transitions to a more spherical one towards the sample center (Figure S26 b). This could potentially stem from a steric hindrance of adsorbed PVA chains, partially blocking reagent access to the NP surface.  All of the samples included in this experiment appear uniform upon visual inspection ( Figure  S25 i-o) and, as with the DoE samples, optical characterization using spectroscopic ellipsometry (SE) was performed ( Figure S28).
To summarize, the use of the combination of PVA and citrate seems like a viable option for increasing the gold precursor utilization. That might, however, come at the cost of a more jagged NP shape in cases of more extensive growth. Looking forward, further optimization of the PVA concentration and/or molecular weight might prove useful in addressing this issue but alternative capping/complexing agents are likely of more interest. One candidate in the latter scenario might be tris-base as used by Li et al. 28

Comment:
The DIA was perfomed on SEM-images aquired at Vacc = 5 kV and 400k magnification. m is the mass of the sample. ΔN/N 0 and ϕ are the relative change in counted NPs and the fractional area coverage of NPs respectively. Area , Max FD , FF , Ellips. and d are the mean NP crosssectional area, max Ferret diameter, fill factor and center-to-center interparticle distance respectively while s Area , s FD max , s FF , s Ellips. and s d are the corresponding standard deviations. s θ is the standard deviation of the discrepancy between the angle spanned by the centroids of the two nearest neighbours and that of a perfect hexagonal geometry. a Seed reference.