ArticlePDF Available

Determination of the yield, mass and structure of silver patches on colloidal silica using multiwavelength analytical ultracentrifugation

August 2021
Journal of Colloid and Interface Science 607(8)

DOI:10.1016/j.jcis.2021.08.161

Authors:

Johannes Walter

Friedrich-Alexander-University Erlangen-Nürnberg

Lukas Pflug

Friedrich-Alexander-University Erlangen-Nürnberg

Thaseem Thajudeen

Indian Institute of Technology Goa, Goa, India

Show all 11 authorsHide

Anisotropic nanoparticles offer considerable promise for applications but also present significant challenges in terms of their characterization. Recent developments in the electroless deposition of silver patches directly onto colloidal silica particles have opened up a simple and scalable synthesis method for patchy particles with tunable optical properties. Due to the reliance on patch nucleation and growth, however, the resulting coatings are distributed in coverage and thickness and some core particles remain uncoated. To support process optimization, new methods are required to rapidly determine patch yield, thickness and coverage. Here we present a novel approach based on multiwavelength analytical ultracentrifugation (MWL-AUC) which permits simultaneous hydrodynamic and spectroscopic characterization. The patchy particle colloids are produced in a continuous flow mixing process that makes use of a KM-type micromixer. By varying the process flow rate or metal precursor concentration we show how the silver to silica mass ratio distribution derived from the AUC-measured sedimentation coefficient distribution can be influenced. Moreover, through reasoned assumptions we arrive at an estimation of the patch yield that is close to that determined by arduous analysis of scanning electron microscopy (SEM) images. Finally, combining MWL-AUC, electrodynamic simulations and SEM image analysis we establish a procedure to estimate the patch thickness and coverage.

Content uploaded by Robin N Klupp Taylor

Content may be subject to copyright.

Determination of the yield, mass and structure of silver patches on colloidal silica using

multiwavelength analytical ultracentrifugation

Thomas Meincke, Johannes Walter, Lukas Pflug, Thaseem Thajudeen, Andreas Völkl, Paola

Cardenas Lopez, Maximilian Uttinger, Michael Stingl, Satoshi Watanabe, Wolfgang Peukert

and Robin Klupp Taylor*

T. Meincke, J. Walter, A. Völkl, P.I. Cardenas Lopez, M.J. Uttinger, Prof, W. Peukert, Prof.

Robin N. Klupp Taylor

Institute of Particle Technology, Friedrich-Alexander University of Erlangen-Nürnberg,

Cauerstrasse 4, 91058 Erlangen, Germany

E-mail: robin.klupp.taylor@fau.de

Dr. L. Pflug, Prof. M. Stingl

Applied Mathematics 2, Friedrich-Alexander University of Erlangen-Nürnberg, Cauerstrasse

11, 91058 Erlangen, Germany

Prof. T. Thajudeen

Mechanical Engineering Department, IIT Goa, Ponda, Goa, 403401, India

Prof. S. Watanabe

Chemical Engineering Department, Kyoto University, Katsura, Nishikyo, Kyoto, 615-8510

Japan

Keywords: ultracentrifugation, localized surface plasmon resonance, patchy particles,

continuous flow synthesis, multidimensional characterization

Abstract

Anisotropic nanoparticles offer considerable promise for applications but also present

significant challenges in terms of their characterization. Recent developments in the

electroless deposition of silver patches directly onto colloidal silica particles have opened up a

simple and scalable synthesis method for patchy particles with tunable optical properties. Due

to the reliance on patch nucleation and growth, however, the resulting coatings are distributed

in coverage and thickness and some core particles remain uncoated. To support process

optimization, new methods are required to rapidly determine patch yield, thickness and

coverage. Here we present a novel approach based on multiwavelength analytical

ultracentrifugation (MWL-AUC) which permits simultaneous hydrodynamic and

spectroscopic characterization. The patchy particle colloids are produced in a continuous flow

mixing process that makes use of a KM-type micromixer. By varying the process flow rate or

metal precursor concentration we show how the silver to silica mass ratio distribution derived

from the AUC-measured sedimentation coefficient distribution can be influenced. Moreover,

through reasoned assumptions we arrive at an estimation of the patch yield that is close to that

determined by arduous analysis of SEM images. Finally, combining MWL-AUC,

electrodynamic simulations and SEM image analysis we establish a procedure to estimate the

patch thickness and coverage.

1. Introduction

Patchy particles represent a special class of anisotropic colloids.[1] In the present context, this

anisotropy is manifested through the presence of regions of different material at the surface of

a colloidal spherical core particle. Due to this arrangement, patchy particles can exhibit the

bulk and interfacial physicochemical properties of two different materials. The type of patchy

particle of concern in the present work comprises a colloidal silica core particle, produced by

the Stöber method,[2] having a certain fraction of its surface coated by a spherical cap of silver

up to a few tens of nanometers in thickness (Figure 1, upper row), produced by electroless

plating. Such types of patchy or Janus particles, the latter defined as having a single patch

with 50% coverage are typically synthesized by template based processes.[3–9] These permit

the fabrication of well-defined, monodisperse patchy or Janus particles but lack considerably

in their scalability. Since anisotropic and especially patchy particles are highly promising for

several fields of applications like sensors,[10] functional films,[9,11] environment protection,[12]

in medicine,[13,14] e.g. for drug delivery[15] and cancer treatment[16,17] or the design of self-

assembled hierarchical particle structures,[1,18–24] scalable fabrication techniques are urgently

sought.

Figure 1. Schematic showing 14 nm thick silver patches with various coverages (cpatch) on

255 nm diameter silica spheres (upper row) and the equivalent patch volume core-shell

particles with shell thickness tshell and having the same silver to silica mass ratio (ϕ) as the

patch particles (lower row).

The recently introduced colloidal method for patchy particle synthesis used in the present

work follows the principle of heterogeneous nucleation and surface conformal growth of

electroless reduced silver on sub-micrometer silica spheres.[25] Due to the effect of mixing and

the stochastic nature of nucleation, some core particles fail to nucleate a patch, and the

patches that do form are distributed in terms of their coverage and thickness. Recently our

original batch process was successfully transferred into a continuous T-mixer based flow

process.[26] The main reason for doing this was to reduce the influence of mixing on the patch

formation reaction and thus obtain more narrowly dispersed patches. By variation of reaction

parameters it was shown that both patchy and almost complete silver coatings, i.e. where the

patch grows so far that it leads to a closed concentric shell, could be reproducibly synthesized.

In addition, it was shown how the combination of optical UV-Vis-NIR spectroscopy, SEM

(scanning electron microscopy) image analysis and FEM (finite element method) based

simulations of the silver patchy particle extinction spectra allow the characterization of a

given sample according to its patch yield (defined as the fraction of all core particles

possessing a patch).[26] Unfortunately, due to very strong dependency of the patch extinction

spectrum on morphology, such optical patch yields can only be derived by UV-Vis-NIR

spectroscopy within a set of samples showing very similar optical properties. Moreover, an

additional reference sample of known statistical patch yield, as derived by SEM image

analysis independently, is required. In turn, the result of such image analysis is highly

dependent on sample quality, orientation of individual patchy particles and to some extent on

the operator of the microscope. Moreover, it is highly challenging to obtain the patch

coverage distribution, a critical parameter to be tuned for applications. Hence, additional

characterization techniques are needed in order to generate a rapid and full understanding of

the patch coverage and thickness distributions and patch yield.

In the present study, the combination of particle sedimentation analysis, through analytical

ultracentrifugation (AUC) with simultaneous spectroscopic characterization, is shown to be a

very promising non-microscopical technique for determining the disperse morphology of

patchy particles.

An AUC equipped with a multiwavelength detector (MWL-AUC) allows the separation of

different particle fractions according to their sedimentation properties and the extinction

spectra corresponding to those fractions extracted. Using this technique, we have recently

demonstrated the derivation of species-specific extinction spectra for gold nanorods enabling

the retrieval of the 2D (length and width) size distributions with excellent agreement to TEM

data.[27] For the case of a population of silver patches grown on monodisperse core particles as

used in this work, the relative effective density of each particle is determined solely by the

mass of patch material attached to the core. This means that, in principle, the patch yield can

be readily identified since the slowest sedimenting fraction will be that of uncoated core

particles. In practice, however, the strong difference in optical extinction between the bare

and coated core particles prohibits this. However, provided that a median patch to core mass

ratio for the coated particles can be measured, and by making reasonable assumptions about

the reaction stoichiometry, it is possible to arrive at an estimate of the patch yield which

avoids the need to detect the uncoated particles. Moreover, if patches are assumed to have a

constant thickness and density of bulk silver their mass will depend only on the thickness and

coverage of the patch. Due to the highly morphology-dependent plasmon resonance of silver

nanostructures,[28] dispersity in thickness and coverage at constant effective density would be

associated with broadened plasmonic features. The benefit of MWL-AUC is that by

separating the particles according to density and measuring optical spectra of resulting

fractions the latter effect can be rationalized, leading to an understanding of the width of the

patch distribution. Hence, MWL-AUC is expected to be a valuable technique able to

determine patch yield, thickness and coverage, essential information for the optimization of

the synthetic process in order to target certain morphologies and properties in the future.

Moreover, the measurements also allow the determination of process relevant aspects like the

weight ratio of patches per particle and the reaction yield of the silver reduction reaction.

In this article it is demonstrated for the first time how MWL-AUC can be used to obtain

realistic values of yield and disperse geometrical parameters of a distribution of

asymmetrically coated nanoparticles (NPs). The theoretical background on the

characterization methodology is provided and it is shown for two exemplary process

parameter variation studies how MWL-AUC can assist with bringing an in-depth

understanding on this promising class of new material.

2. Theoretical background for analytical ultra-centrifugation and data evaluation

2.1. Theory of sedimentation and gravitational sweep analysis

When exposed to a centrifugal field, particles will sediment or float according to their size,

density and shape. Moreover, particles will diffuse depending on temperature, solvent

viscosity and their hydrodynamic diameter. AUC allows to measure the sedimentation and

diffusion properties of NPs by means of sedimentation velocity (SV) experiments with high

resolution and excellent statistical confidence. For large, rapidly sedimenting colloids, such as

the patchy particles of the present work, gravitational sweep (GS) experiments are an

effective alternative.[27,29–31] In such case, the measurement position is kept constant while the

rotor speed is continuously increased. Even though diffusion information is not accessible by

GS experiments, much larger NPs and much broader particle size distributions (PSDs) can be

studied. Besides, for the large particle sizes (> 20 nm) typically studied in GS runs, diffusion

plays a negligible role.

The most important quantity derived by an AUC experiment is the sedimentation coefficient.

It is defined as the sedimentation velocity of a particle divided by the centrifugal acceleration.

The sedimentation coefficient s is a function of the time t, rotor angular velocity , radial

measurement position r and the meniscus position rm:

=



 (1)

During a MWL-AUC GS experiment, the optical transmission is monitored at multiple

wavelengths as a function of the sedimentation coefficient. GS data can be evaluated using the

software HDR-MULTIFIT, which provides the extinction weighted density or cumulative

distributions of the sedimentation coefficient.[31] As the density distributions are extinction

weighted, integration immediately provides the associated extinction spectra. This allows the

derivation of the spectra of individual species in a mixture as has been shown before for

plasmonic NPs.[27,31] Notably, this data is equivalent to band sedimentation, for which details

are discussed in Section VI of the Supplementary Material. However, GS experiments require

no special centrepieces, are less prone to uncertainties, at least for the given boundary

conditions, and are thus the method of choice for patchy particle characterization.

The analysis in HDR-MULTIFIT is based on a model independent approach and basically

applies Equation 1 while correcting for radial dilution effects.[30,31] As data acquired by the

MWL-AUC is noisy, smoothing has to be applied. Direct boundary modelling (DBM)

constitutes an interesting alternative to the model independent approach as excessive

smoothing is not necessary. Here, the ls-g*(s) approach is based on the method developed by

Schuck and Rossmanith,[32] and was previously also transferred to analytical centrifugation.[33]

In this work we also applied this technique to GS data. In short, the amplitudes of the

simplified solution of the Lamm equation, neglecting all diffusional effects, are fitted to the

experimental data by means of a least-squares approach. Regularization is used and results in

a continuous sedimentation coefficient distribution. One drawback of the DBM is that it is

computationally more expensive. For further details the reader is referred to the original

publication and our previous work on this topic.

The sedimentation equivalent particle diameter d can be calculated using Stokes’ equation:

=

 (2)

where  is the particle density,  is the solvent density and η is the solvent viscosity. In the

present work the cumulative distributions of the sedimentation coefficient only represent the

fraction of silver-coated particles. This is due to the similar refractive indices of bare silica

and the 33 wt% aqueous sucrose solution used which lead to the uncoated silica particles not

being detected at the low concentration used.

2.2. Data evaluation strategy for patchy particles

2.2.1. General strategy

As stated in the previous section, AUC is sensitive to the size and density of the analyte, since

it measures its sedimentation rate. In the present work the overall size and mass of

monodisperse silica NPs are increased by the formation of a silver patch on their surface.

Since silver is roughly five times denser than colloidal silica, the presence of a patch results in

more pronounced sedimentation rates, which can be detected by AUC. However, for a patchy

particle, the change in size is typically small compared to the bare particle as the patch

thickness is about one order of magnitude smaller than the core particle diameter (Figure 1).

This is the main reason why light scattering or particle tracking techniques being sensitive to

the size or diffusion properties are of limited applicability when studying patchy particles of

the type discussed here. Nevertheless, the patch does slightly increase the hydrodynamic

diameter and, in addition, due to its asymmetrical shape will alter the friction the particle

experiences as it sediments. However, regarding the latter effect, we demonstrated through

simulations that the asymmetric distribution of silver on the silica cores has a negligible effect

on the sedimentation behavior (Figure S1 in the Supplementary Material). This opens up the

possibility to analyze the AUC data according to a simpler core-shell model.

The strategy pursued for the patchy particle characterization using MWL-AUC in this work is

depicted in Figure 2. First, MWL-AUC provides the sedimentation coefficient distributions

of the patchy particle samples, which can be converted to distributions of the patch-to-core

mass ratio or the patch mass per particle. Relating the median patch mass per particle with the

mass balance allows calculation of the patch yield. Second, MWL-AUC provides spectral

information about the different species being present in the sample. In principle, any species

having a different sedimentation coefficient could be examined. This would result in hundreds

of species per sample. However, the derived spectra would be very noisy in such case and the

data would not be of any use. Thus, only five averaged fractions will be considered for each

sample. Still, future improvements of the detection system will extend the number of

accessible species. As will be shown later in this work, the UV-peak position obtained from

the UV-Vis spectra by MWL-AUC can be related to the patch thickness. This is achieved by

empirical calibration via SEM image analysis. Combination of the patch mass and patch

thickness provides the patch coverage. Hence, in this work a methodology was established

that enables the determination of patch thicknesses and coverages from MWL-AUC.

Figure 2. Strategy established in this work for the multidimensional characterization of

patchy particles using MWL-AUC.

GravitationalSweep

Multiwavelength Analytical

Ultracentrifugation

Sedimentation

coeffici ent

distribution

Silver-silica

mass ratio ()

distribution

Extinction spectra

deconvolutedfro m

selected range of

 distribution

Median patch thick ness t

patch,50

Median patch coverage c

patch,50

Correlat ion of UV-peak posit ion and

thicknes s:

•Calibration via SEM image analy sis

•Comparison with FEM based

simulation of extinction s pectra

UV-peak

positionλ

UV-peak

Median mass of

patches, m

patch,50

Patch yield Y

AUC

Core-sh ell equivalency model Mass balance for Ag and Si O

2.2.2. Determination of the Ag/SiO2 mass ratio distribution from the patchy particle

sedimentation coefficient distribution

The most important equations which allow calculation of the silver to silica mass ratio from

the AUC-determined sedimentation coefficient distribution will be provided in the following

section. The methodology is based on a core-shell equivalency model. In this the volume of a

silver patch is redistributed over the core particle surface as a uniformly thick shell (compare

upper and lower rows in Figure 1). Such an approximation is possible since, as noted above,

the core diameter is very large compared to the patch thickness. As such the presence of the

patch will not introduce significant shape anisotropic effects and differing hydration of the

silver and silica can also be neglected.

To establish the core-shell equivalency model we first define a spherical core particle with

diameter dcore and a patch on it having constant thickness  and volume . From

this we can derive the patch coverage  as the ratio of the patch volume to that of a

hypothetical complete shell with thickness  (Figure 1, upper row):

=





 (3)

We now define the equivalent volume shell particle diameter  as the diameter of a spherical

core-shell particle having a shell with the same volume,  as the patch:

 =

[

]= (4)

From a simple mass balance the effective density of the particle comprising core and patch is

given as:

= 





 (5)

where  and  are the core and patch material densities, respectively. This expression

allows Equation 2 to be rewritten:







18 = 0 (6)

For each sedimentation coefficient in the AUC-measured distribution, Equation 6 is solved to

determine the diameter of the equivalent patch volume core-shell particle defined above. The

density of the uncoated core particle, which is microporous and hydrated, has to be

determined experimentally prior to the patchy particle studies if not known, e.g. by combining

AUC and DLS or SEM. An alternative is Kratky’s method which relies on densitometer

measurements.[34] However, this technique is of limited applicability to microporous particles,

such as Stöber silica as it leads to the partial specific volume which cannot be readily

associated with a macroscopic density.[35–37] The partial specific volume is defined as the

change in volume when adding a certain mass of particles to the solution. Here, we are

interested in the effective density of the non-coated core particle, including all contributions

due to solvent filled and empty pores. Moreover, the size of the non-coated particle must be

determined prior to the patchy particles studies. In the Supplementary Material we show how

this information can be obtained by AUC measurements of silica dispersions in sucrose

solution (Figure S2) and also demonstrate that the finite width of the core PSD negligibly

influences the determined silver to silica mass ratio distribution (Figure S4). In addition, it is

shown that there is no measurable influence of heterogeneity in density for the uncoated core

particles (Figure S3).

Once the equivalent core-shell particle size and core particle size are known, the equivalent

volume shell thickness can be determined (Figure 1, lower row):

 =()/2 (7)

To determine the Ag/SiO2 mass ratio we determine the masses of the silver patch and silica

core:

=

(

) (8)

 =



 (9)

The Ag/SiO2 mass ratio is then calculated via the following expression for each value of d

obtained by solution of Equation 6 for each measured sedimentation coefficient:

=

 =







 (10)

The cumulative sedimentation coefficient distribution can thus be converted to a cumulative

Ag/SiO2 mass ratio distribution, ϕ. In this study the median of this distribution, ϕ50,AUC in

addition to its span will be of primary interest. The latter parameter is defined as:

 =,,

, (11)

where ϕ10,AUC and ϕ90,AUC correspond to the values of ϕ at the 10th, and 90th percentiles of the

cumulative distribution of ϕ, respectively.

2.2.3. Calculation of theoretical patch yield based on Ag/SiO2 mass ratio derived by AUC

data and mass balance

Assuming mass conservation for the absolute mass of silver and silica particles injected into

the reactor within one experiment and 100 % reaction yield for the reduction reaction of silver

nitrate to metallic silver patches, the median isolated absolute mass of all silica particles that

were coated with silver can be calculated:

,  =

, (12)

Knowing the mass of silica particles having a patch from the AUC distribution data and

knowing the total mass of silica particles used for silver coating reaction, allows the definition

of the AUC-determined patch yield:

 =, 

,  (13)

In this work we also determined the statistical patch yield  derived by manual analysis of

SEM images.  is determined by dividing the number of patchy particles  in images

of a sample by the total number of particles  observed. In this analysis we counted at

least 300 particles per sample.

 =

 (14)

2.2.4. Calculation of median patch coverage ,

Provided the patch thickness and median Ag/SiO2 mass ratio are known, an estimate of the

median patch coverage cpatch,50 can be made. From Equations 3 and 10 we obtain:







,















, 











((











)









)

(15)

As will be explained in Section 4.5,  can be obtained from a combination of

spectroscopy, electrodynamic simulations and SEM image analysis.

3. Results and Discussion

3.1. The influence of mixing conditions on the median and width of cumulative Ag/SiO2

mass ratio distribution

To demonstrate the benefit of AUC for the analysis of patchy particles with distributed patch

volumes we consider firstly the effect of mixing, the characteristic time for which can be

selected in our setup through the educt flow rates. Figure 3a and Figure 3b show SEM images

of samples with silver patchy particles synthesized at the highest (20 mL min-1) and the lowest

(2 mL min-1) flow rates used within this study. To illustrate the typical patch morphology,

these images were acquired at positions on the SEM sample holder with very high local patch

yield. More representative images taken at lower magnification for all samples at five

different flow rates are shown in Figure S9 in the Supplementary Material. In both Figure 3a

and Figure 3b silver patches of varying coverages can be seen alongside apparently uncoated

silica particles. While some silica particles appear fully coated, it cannot be ruled out that

these are high coverage patches oriented such that the exposed silica is directed towards the

substrate. Apart from the appearance of homogeneously nucleated silver particles for the

2 mL min-1 flow rate sample, rather little difference in patch thickness and coverage can be

inferred by the naked eye. The similar appearance of the patches can be explained by the

concentrations chosen for synthesis and especially by the very short characteristic mixing

times of ~16 ms and ~1 ms for highest and lowest flow rates, respectively, compared to our

earlier work (see Section VII of the Supplementary Material for details of the characteristic

mixing time determination).[38]

Figure 3. a) and b) SEM images of patchy particles synthesised at the lowest and highest flow

rates used in the study. c) Cumulative extinction weighted distributions of ϕ

at 600 nm

wavelength for samples synthesised at five different flow rates. d) Flow rate dependencies of

the median ϕ50,AUC and span of the ϕ distributions shown in c). e) Flow rate dependencies of

the SEM-determined statistical patch yield Yst and the AUC-determined yield YAUC derived

from ϕ50,AUC,. The scalebars in a) and b) correspond to 200 nm.

Despite the similar outward appearance of the samples prepared at low and high flow rate,

AUC measurements of the sedimentation coefficient at a single wavelength (600 nm) were

able to identify a clear difference in the silver loading of the patchy particles. Here, the

cumulative sedimentation coefficient distributions (Figure S9f) for the samples synthesized at

different flow rates were converted into distributions of the Ag/SiO2 mass ratio according to

the method described in the theoretical section. In addition, the distribution of the bare silica

particles is depicted for reference. The distributions, and their medians and spans are shown in

Figure 3c and Figure 3d, respectively. With increasing flow rate there is an initial slight

increase in ϕ50,AUC. This can be attributed to suppression of the formation of free silver

particles by increasing the flow rate i.e. the better-mixed reactions never reach the critical

supersaturation for homogeneous nucleation. Since more silver becomes available for patch

growth there is a commensurate increase in the median Ag/SiO2 mass ratio for the patchy

particle population. Interestingly, ϕ50,AUC decreases again for further increases in the flow rate.

Since the ratio of silver to silica in the educts is constant, according to Equations 12 and 13,

this decrease in ϕ50,AUC must correspond to an increase in patch yield  as indicated in

Figure 3e (red circles). This trend, which results from better mixing conditions leading to a

higher heterogeneous nucleation rate, is corroborated by the statistical patch yield 

determined manually from SEM images (black squares). Given that errors exist in both

methods of determining the yield, the differences in the absolute values of the yield are less

important. What is significant is that both show a similar trend, when the process variable

(mixing time) is adjusted. This means that the considerably more rapid determination of patch

yield by AUC can be used to optimize the process for maximum yield and arduous analysis of

micrographs can be avoided.

The effect of mixing on the width of the distribution of ϕ is also shown in Figure 3d (blue

squares). While at the lowest flow rate the span is ~1.6, for higher flow rates it is lower and

rather constant in the range 1.33 - 1.39. As verified in Figure S3, any influence of the size

distribution of the silica core particles on this span trend can be neglected. Rather, our

observations can be explained by two effects. For the sample synthesized at a low flow rate,

free silver particles with diameters larger than 100 nm were present (Figure 3b). Such large

silver particles will contribute to the cumulative extinction at 600 nm and thus influence the

sedimentation coefficient distribution at its lower end, which will in turn broaden the

distribution of ϕ. Moreover, as shown in earlier work, lower flow rates lead to a blurring of

the patch nucleation and patch growth periods, resulting in a broader range of patch

volumes.[26] For samples synthesized with higher flow rates corresponding to mixing times of

1 ms to 8 ms, the ϕ distribution width is narrower and the span is rather constant. This results

from the low amount of homogeneously nucleated silver particles, especially those of a size

having a strong extinction at 600 nm, and from better separation of patch nucleation and patch

growth due to the better mixed reaction conditions.

The results shown in Figure 3 therefore confirm that the median and span of the AUC-

determined ϕ distribution enable a simple evaluation of the effect of mixing on patch yield.

Since samples produced at 20 mL min-1 showed negligible free silver particles and had small

values for ϕ50,AUC and span, all further experiments reported here were carried out at this flow

rate.

3.2. Influence of silver nitrate concentration on the ϕ distribution

Silver patches were grown on silica cores particles using silver nitrate concentrations ranging

from 200 µM to 600 µM. Figure 4a and Figure 4b show the SEM images of the samples

synthesized at the lowest and highest concentrations, respectively. By qualitatively comparing

these two images, a difference in patch thickness can be observed. For the higher

concentration the metal coating is generally thicker. Nevertheless, as in the case of the flow

rate variation study described above, an objective assessment of the patch coverage cannot be

obtained from these images. This is also the case when comparing SEM images of samples

from all silver nitrate concentrations used (see Figure S10).

Figure 4. a) and b) SEM images of patchy particles synthesised with the lowest and highest

silver nitrate concentrations used in the study. c) Cumulative extinction weighted distribution

of ϕ at 600 nm wavelength for samples synthesised with five different silver nitrate

concentrations. d) Silver nitrate concentration dependencies of the median ϕ50,AUC and span of

the ϕ distributions shown in c). e) Silver nitrate concentration dependencies of the SEM-

determined patch yield Yst and the AUC-determined yield YAUC derived from ϕ50,AUC,. The

scalebars in a) and b) correspond to 200 nm.

Once again, however, AUC measurements can identify clear trends in the distributions of ϕ

(Figure 4c). Notably, there is a marked shift of the distribution towards larger values of ϕ with

increasing silver nitrate concentration. The extracted ϕ5 0,AUC values, as shown in Figure 4d,

clearly underline this trend which confirms that the patches produced at higher silver nitrate

concentrations contain more silver. Although this might appear an obvious statement, it

should be reminded that the heterogeneous nucleation rate and thus the yield will also be

affected by the silver concentration. Indeed, the derived values of yield  (red circles in

Figure 4e) show an increase up to 500 µM silver nitrate concentration followed by a slight

decrease. This trend is matched by the statistical yield  determined from SEM images

(black squares in Figure 4e).

The decrease in yield at the highest silver concentration indicates that the heterogeneous

nucleation rate is beginning to decrease again. Since homogeneously nucleated silver was not

observed in this sample we suggest that this decrease in yield is due to the mixing time

starting to again play a more significant role at the high silver ion concentration. In other

words, while mixing is rapid enough that in the bulk solution a supersaturation sufficient for

homogeneous nucleation is not achieved, the conditions for heterogeneous nucleation are not

reached uniformly across all silica particles. This in turn leads to the situation of some patches

nucleating and growing (and thus depleting silver) at the same time that others are still yet to

nucleate. The resulting broad patch volume distribution is corroborated by the increase in the

span of the ϕ distribution for the highest silver nitrate concentration compared to the lower

concentrations (blue squares in Figure 4d).

To conclude, in this section we have demonstrated the effective use of AUC at a single

wavelength for the determination of the Ag/SiO2 mass ratio distribution and estimation of the

patch yield for a range of patchy particles samples produced at different silver concentrations.

The yield trends are fully consistent with those determined by arduous image analysis.

3.3. Acquisition of Ag/SiO2 mass ratio dependent optical extinction spectra via MWL-

AUC

In this section we focus on deriving further information on the patchy particles in a sample

from their extinction spectrum. MWL-AUC offers the unique capability to address the optical

properties of different species in a mixture without any further purification and by means of

only a single experiment.[31,39–41] Due to the highly morphology-dependent nature of the

plasmon resonance of silver nanostructures, we expect MWL-AUC to be especially suitable

for studying systems like the present patchy particles. Hence, a comparison between the set of

samples synthesized with increasing silver nitrate concentration and the optical properties of

the particle fractions having different ϕ within the same sample was carried out.

Figure 5a shows raw spectra obtained by deconvoluting data corresponding to all values of ϕ

for the five samples studied. Due to the high level of noise on these MWL-AUC acquired

spectra, we verified their validity with measurements of the same samples on a standard dual-

beam spectrophotometer (Figure S7 in the Supplementary Material). While minor differences

are observed due to the lower spectral resolution of the MWL-AUC, the agreement of the

spectra is very good. The higher deviations seen at wavelengths > 600 nm are due to the low

signal-to-noise-ratio of the MWL-AUC detector in that range (see marked increase in noise in

Figure 5a). Most importantly, it is clear that the amount of noise in the spectral region of

interest (UV-peak) does not preclude a quantitative analysis.

Figure 5. a) Raw spectra for patchy particle samples produced at different silver nitrate

concentrations and obtained by deconvoluting MWL-AUC data over the full range of

measured sedimentation coefficients. The UV-peak in the spectral region from 320 nm to 380

nm is fit with an asymmetric double sigmoidal function. b) Dependency of the UV-peak

position obtained from the spectra in a) on the median Ag/SiO2

mass ratio of the five samples.

c) Deconvoluted MWL-AUC spectra of five fractions of the patchy particle sample

synthesised at 600 µM silver nitrate concentration. Each spectrum corresponds to a fraction of

patchy particles with increasing median Ag/SiO2

mass ratios which contribute equally with

20% to the cumulative extinction at 600 nm. The legend corresponds to the percentage ranges

of the total extinction. For the sake of clarity, the spectra are plotted with y-offset. d) UV-peak

positions of MWL-AUC extinction spectra for samples synthesised with different silver

nitrate concentrations for fractions of the Ag/SiO2 mass ratio with equal 20% contributions to

the total extinction at 600 nm. The spectra used in the analysis are shown in c) and Figure

S12. The UV-peak positions were obtained by fitting an asymmetric double sigmoidal

function in the wavelength range 320 to 380 nm. The abscissa labels correspond to the

percentage ranges of the total extinction.

Figure 5a indicates that with increasing silver nitrate concentration the extinction increases

and shifts towards shorter wavelengths and some peaks appear in the spectral region around

500 to 600 nm. These observations are consistent with both the already-observed patch yield

increase (Figure 4e) and FEM simulated spectra of patches with increasing coverage or

thickness (Figure S11 in the Supplementary Material). A further, and, as we shall see, highly

useful feature of these spectra is the peak between 320 nm and 360 nm (hereafter referred to

as the “UV-peak”). Figure 5b shows that the position of this peak, determined from the

MWL-AUC spectra in Figure 5a by fitting with an asymmetric double sigmoidal function,

shifts monotonously from about 333 nm to 345 nm with increasing ϕ50,AUC. The position of

this spectral feature therefore appears to be strongly correlated with the amount of silver

attached to the silica sphere. This dependency is also observed when the spectra of individual

fractions of a sample are investigated. Figure 5c shows, exemplarily for the sample

synthesized at 600 µM silver nitrate, the deconvoluted and smoothed extinction spectra of

fractions within five intervals of cumulatively increasing ϕ which contribute almost equally

with ~20% relative extinction each at 600 nm i.e. the extinction of the fractional spectra are

equal at this wavelength. Figure S12 shows the same analysis for the other silver

concentration samples with similar trends apparent throughout. The lowest ϕ fraction of each

sample generally has a spectrum with just the UV-peak and a broad peak at around 650-700

nm. With an increase in concentration the spectra show the appearance of features between

400 nm and 600 nm with increasing ϕ fraction. As with the complete sample spectra of Figure

5a we analyzed the position of the UV-peak for all five fractions of each sample. Figure 5d

shows the UV-peak position plotted against the cumulative fraction for each sample. It is clear

that the individual fractions generally show an increase (red-shift) in UV-peak position for

both increasing silver nitrate concentration and fraction. For the heaviest fraction (80-100%)

there is a levelling off or even a blue-shift of this peak which seems to contradict the trend in

Figure 5b. There may be various reasons for this observation. First, the deconvoluted

spectrum is in general noisier at the upper end of the distribution of ϕ which makes it difficult

to determine the exact peak position. This is because of the higher absolute extinction

measured and thus lower signal-to-noise ratio at the beginning of the measurement, where all

species are still present in the dispersion. Second, the interval with highest sedimentation

coefficients might also be influenced by the presence of agglomerates of patchy particles

having themselves a lower ϕ and thus a smaller UV-peak. Similarly, particles having multiple

thin patches might contribute to the extinction of this fraction.

In conclusion, in this section we have shown that useful spectral information can be obtained

for ranges of the Ag/SiO2 mass ratio distribution by deconvoluting GS data acquired by

MWL-AUC. From these spectra we see that there is a particularly strong correlation between

the MWL-AUC determined UV-peak position and Ag/SiO2 mass ratio for the silver patchy

particles considered in this work.

3.4. Determination of patch thickness and coverage from MWL-AUC data

To generate a deeper understanding of the meaning of the UV-peak, FEM based simulations

of the extinction spectra of idealized cup-like patchy particles having patch coverages

between 10% and 90% and patch thicknesses between 5 nm to 35 nm were performed (see

Section IX of the Supplementary Material for more details). The simulated silver patches

were assumed to have a perfect smooth surface, constant thickness and regular sharp edges.

To obtain each wavelength point, multiple simulations covering different particle orientations

and light polarizations were averaged. The simulated extinction spectra of the UV region for

patchy particles with different patch coverage and thickness indicated that the UV-peak

position is almost completely independent of the patch coverage (Figure S11a). This is

consistent with the fact that this feature has been attributed to the out-of-plane quadrupole

resonance.[42] Hence we can infer that the correlation between the UV-peak position and the

Ag/SiO2 mass ratio (Figure 5b) implies that the patch thickness must increase with increasing

silver nitrate concentration. To obtain an expression for the relation between the UV-peak

position and patch thickness we first plotted the average position of the peak in the

simulations against the patch thickness used as input (see Figure S14). The resulting curve

was fitted with a Boltzmann sigmoid function leading to an expression relating UV-peak

position to “optical” patch thickness ,. We used this expression to estimate the patch

thickness of the samples in the silver nitrate concentration series. Rather than taking the UV-

peak position from the spectra in Figure 5a we analyzed spectra obtained by summating the

three central fractions i.e. 20%-80% in Figure 5c and S12. This was done so as to avoid

potential errors introduced by patchy particles at the extremities of the cumulative ϕ

distribution, particularly the heaviest fraction, as already noted above. Figure S12 shows the

fitting of the UV-peaks in these spectra with an asymmetric double sigmoidal function. The

resulting UV-peak positions were used to estimate , for each silver nitrate

concentration. The results are plotted in Figure 6a (blue squares) and indicate a monotonous

increase in thickness with concentration from 13.2 nm up to 24.1 nm.

Figure 6. a) Patch thickness versus silver nitrate concentration obtained by SEM image

analysis (black crosses) and two different methods of processing UV-peak data from MWL-

AUC spectra. The blue squares correspond to patch thickness values obtained using an FEM-

simulation determined relation (see Figure S14) while the red circles are for patch thickness

values obtained using Equation 16, an empirical fit of the UV-peak position versus SEM-

determined patch thickness relation (see Figure S17). For the UV-peak position derived data

the original spectra were deconvoluted from the MWL-AUC data for the central 20%-80%

fraction of the cumulative extinction at 600 nm. Fitting is shown in Figure S13. b) Patch

coverage versus silver nitrate concentration determined using patch thicknesses from UV-

peak positions through Equation 16 (red circles in a). c) Patch thickness and d) patch coverage

determined for individual fractions of the concentration series samples and plotted against the

Ag/SiO2 mass ratio. Horizontal error bars correspond to the mass ratio ranges of the fractions.

For clarity the lightest fraction was taken from a starting mass ratio corresponding to 2% of

the cumulative extinction at 600 nm while the heaviest fraction was truncated at a mass ratio

corresponding to 98% of the cumulative extinction at 600 nm.

To verify the accuracy of the determined patch thicknesses we carried out arduous, systematic

image analysis of SEM micrographs. The procedure followed is described in Section VIII of

the Supplementary Material and illustrated with Figure S15 showing the various stages of

processing for one image. Briefly, we made use of the fact that for all samples across the

silver nitrate concentration range used, a large proportion of the patchy particles had high

coverage patches, and, for many of those, the exposed silica was hidden from view. For our

image analysis we selected only those particles and assumed their diameter corresponded to

the core particle diameter plus twice the patch thickness. Analysis of at least 120 particles in

each sample, along with a sample of only core particles, enabled us to determine mass

weighted PSDs (Figure S16a). These were fitted with Boltzmann sigmoidal functions and

thus the cumulative mass weighted distributions of , could be determined (Figure

S16b). The median values of these distributions are plotted in Figure 6a (black crosses). It is

evident that the patch thicknesses estimated using the UV-peak position and the simulation-

derived relation are somewhat lower than the “real” SEM-determined patch thicknesses.

There are several reasons why this might be the case. On the one hand, it is very likely due to

the fact that the simple geometry used in the FEM simulation does not reflect reality. As seen

in the SEM images of Figure 4a and 5a, the patches are neither smooth nor single crystalline.

On the other hand, we cannot rule out at this stage that they are not exactly concentric i.e.

their thickness tapers towards their edges. Hence it was recognized that rather than using a

simulation-based calibration of the UV-peak to patch thickness relation, an empirical

expression should be developed. We obtained this by plotting the UV-peak positions

(determined in Figure S13) against the SEM-determined median patch thicknesses

(determined in Figure S16b). Figure S17 in the Supplementary Material shows these data

mostly lie on a line slightly offset to the simulated data. We performed a linear fit to the

empirical data, ignoring the outlying data point corresponding to the 200 µM sample. The

latter exclusion is reasonable since the largest error in the SEM image analysis is likely in this

sample due to it having the thinnest patches. Moreover, the UV-peak is the least prominent of

the five samples, making its analysis most inaccurate. The linear fit of Figure S17 provided

the following empirical relation between the UV-peak position  and the patch

thickness ,:

, = 1.067  340.9  (16)

From this relation a new estimate of the patch thickness from the MWL-AUC spectra could

be made. As shown in Figure 6a (red circles) this data closely follows the SEM-measured

patch thickness. This good agreement is of course due to the fact that the latter data set was

used to develop the empirical relation.

Once the median silver to silica mass ratio ϕ and median patch thickness is known, the

median patch coverage can be estimated from Equation 15. This was done for the five

samples of the silver nitrate concentration series using the patch thickness as determined from

Equation 16. The results are shown in Figure 6b. While the median patch coverages are rather

high (between 86% and 93% for concentrations between 300 µM and 600 µM) it is not

possible to identify a clear trend. This is rather unsurprising since SEM images for all samples

(Figure S10) indicate predominantly high coverage patches. In other words, the patch

coverage is not strongly dependent on the silver nitrate concentration.

To demonstrate that information on the dispersity of patchy particle morphological features

can be acquired by MWL-AUC we applied the relations for patch thickness (Equation 16) and

coverage (Equation 15) to fractionized spectroscopic data from the silver nitrate concentration

series of Figure 5c and S10. The full derived data is listed in Table S4 in the Supplementary

Material and is plotted in Figure 6c (patch thickness) and 6d (patch coverage). These plots

show the Ag/SiO2 mass ratio on the abscissa and horizontal error bars show the width of the

fractions. Thus it is possible to identify that for increasing silver nitrate concentration the

dispersity of patch thicknesses over all fractions increases from being rather narrow at 13 to

17 nm for the 200 µM sample to being rather broad at 20 to 34 nm for the 600 µM sample.

Due to the inverse relation between thickness and coverage in Equation 15, the variation of

patch coverage is rather complementary to the thickness. Thus the lowest concentration

sample has a large spread of coverages over a given mass ratio range, while the highest

concentration sample has a comparatively narrower range of coverages. It can also be seen

that the heaviest fraction in each sample has a coverage of above 100%. As already noted

above, this fraction is likely to contain aggregates of patchy particles or multi-patch particles

and the unexpectedly low thickness leads to this non-physical coverage value.

The above analysis demonstrates that realistic trends in patch thickness and coverage and

information on the dispersity of these parameters within individual samples can be obtained

by MWL-AUC measurements alone. However, more work is required in the future to refine

our approach. As shown in the electrodynamic simulations, the NIR region of the spectrum

contains information about the patch coverage (Figure S11a). Hence, upgrading of the

spectrometer in our setup to obtain spectra deeper into the NIR will open up an additional

means to further improve the determination of patch coverage. Overall, due to the fact that it

avoids arduous microscopy studies, we believe our method will greatly assist with the

systematic exploration and optimization of process variables in order to target other patchy

particle morphologies not discussed here e.g. narrowly dispersed, small coverage patches or

even multiple patch particles.

4. Conclusion

In this work, a method based on multiwavelength analytical ultracentrifugation (MWL-AUC)

for the characterization of silver patches on silica spheres, synthesized via continuous a flow

process, has been shown. This technique helps to rationally investigate the patch yield in

addition to the amount and geometrical distribution of silver on the silica particles.

First, the use of AUC measurements to experimentally investigate the effect of the process

flow rate was demonstrated. The combination of a low median Ag/SiO2 mass ratio and a

narrow width of the mass ratio distribution indicated a high sample quality regarding patch

yield and patch uniformity for the sample synthesized at highest flow rate, i.e. shortest

characteristic mixing time.

In the second part, patchy particles were synthesized at constant high flow rate and with

increasing silver nitrate concentration. The results showed that the patch yield in addition to

the median Ag/SiO2 mass ratio can be estimated from AUC measurements. The trends in

these parameters compared favourably with those obtained by arduous analysis of electron

micrographs.

Since the silver patchy particles possess plasmonic properties which depend sensitively on the

patch thickness and patch coverage, MWL-AUC-determined extinction spectra were

investigated in detail. Making use of FEM-based electrodynamic simulations it was found that

the observed UV-peak, being present in the range of 330 nm to 360 nm, is directly related to

the patch thickness. While simulations could provide a relation between the peak position and

thickness, when applied to real data from MWL-AUC spectra the patch thicknesses were

underestimated compared to SEM-determined thicknesses. Thus, an empirical relationship

was obtained by correlating UV-peak position with patch thicknesses determined by SEM

image analysis. Besides patch thickness, this enabled, with use of the already-known Ag/SiO2

mass ratio, the patch coverage to be estimated. Finally, we demonstrated the true value of

MWL-AUC for analysing the dispersity of this new type of material by determining the

median patch thicknesses and coverages of five individual fractions of samples produced at

different silver nitrate concentrations.

In conclusion, our work presents a powerful, fast and versatile method to analyze complex

plasmonic anisotropic particles by means of just a single measurement. Future work is

addressed to improve the optical equipment installed inside the AUC in order to further

improve the quality of the data obtained, especially at longer wavelengths. This is critical in

order to validate the derived information on the patch coverage, since the corresponding

plasmonic features are mainly located at longer wavelengths.

5. Materials and Methods

Synthesis of colloidal silica particles:

Heat-treated colloidal silica particles, produced by the Stöber method,[43] having a mean

diameter of 255 nm, the same as those used in our earlier work,[26] were applied as core

particles for patchy particles synthesis. Their synthesis proceeded following the method of

Stöber et al.[43] First, a mixture of 497 g absolute ethanol (Carl Roth GmbH, Germany), 26.4 g

aqueous ammonium hydroxide (32 %, Carl Roth GmbH, Germany) and 62.71 g ultrapure

water was prepared. By the subsequent addition of 50 mL tetraethylorthosilicate (98 %,

Merck KGaA, Germany) the formation of colloidal silica was initiated. The reaction was

allowed to proceed for 5 h at room temperature before it was finally stopped by four times

washing of the synthesized silica particles with ultrapure water. After drying of the derived

particles, the powder was calcined at 1000 °C for 6 h.

Continuous flow synthesis and stabilisation with sucrose of patchy- and complete nanoshell

particles: In extension to our earlier work,[26] patchy particles were produced here in a

continuous process using a KM-micromixer (Figure 7), similar to that recently published by

S. Watanabe et al. for gold nanoshell and gold patch synthesis.[44,45] This type of mixer

combines reasonable flow rates for lab-scale synthesis with higher mixing performance

(characteristic mixing times of 1-20 ms) than that of a simple T-mixer used earlier

(characteristic mixing times greater than ~90 ms), as verified by the Villermaux-Dushmann

reaction.[38] Two educt streams were injected into the microreactor via a syringe pump. The

first stream contained aqueous ammonia solution (32 %, Carl Roth GmbH, Germany) at a

concentration of 27 mM and for the second stream a dispersion of colloidal silica (25 mg/L),

silver nitrate (99 %, Carl Roth GmbH, Germany) with concentrations in the range of 200 µM

to 600 µM and 152 mM formaldehyde (37 % aqueous solution, methanol stabilized, Carl Roth

GmbH, Germany) was prepared. As the preferred temperature for high yield silver patch

formation has already been shown to be 50 °C,[25] the microreactor was immersed into a

stirred water bath and two stainless steel heating coils (inner diameter = 1.6 mm, length of

tubing = 2 m) were installed before its inlets to ensure the educt streams were at the correct

temperature prior to mixing. After the reactor, a coiled polyamide tube (inner diameter =

2 mm, length of tubing = 6 m) was used to ensure sufficient residence time for the reduction

reaction of ammoniacal silver nitrate to metallic silver by formaldehyde at all flow rates used.

For each syringe, the flow rates were varied within the range of 2 mL min-1 to 20 mL min-1.

An overview of synthetic conditions used in all experiments can be found in Table S1 in the

Supplementary Material.

Figure 7. Schematic of the KM-microreactor used for the growth of silver patches on silica

NPs.

After synthesis, an aliquot of 2 mL product dispersion was added to 1 g D(+) sucrose (>99,5%

p.a.) in powder form (Carl Roth GmbH, Germany). The dissolution of sucrose was achieved

by short sonication and shaking by hand.

Scanning electron microscopy: SEM samples were prepared directly after synthesis before the

addition of sucrose and without any further purification. 3 µL of product particles were

dropped onto a silicon wafer and were allowed to dry at ambient conditions. The patchy silver

particles were analysed using a ULTRA 55 field emission SEM (Carl Zeiss NTS GmbH,

Germany). Statistical patch yields were obtained by manual counting of particles with and

without a patch and using Equation 14.

UV-Vis-NIR spectroscopy: Since UV-Vis-NIR measurements were done of samples after the

addition of sucrose, a reference sample of 1 g sucrose and 2 mL ultrapure water was prepared.

The extinction spectra were recorded in the wavelength range of 300 nm to 1350 nm using a

PerkinElmer Lambda 950 UV-Vis-NIR spectrophotometer.

Analytical ultracentrifugation: GS experiments of patchy particles were performed using a

modified preparative ultracentrifuge, type Optima L-90K (Beckman Coulter, USA), equipped

with an integrated UV-Vis MWL detector.[39,46] For all runs, titanium centrepieces

(Nanolytics, Germany) with path lengths of 12 mm were used. GS data were acquired at

20 °C and rotor speeds ranging from 2000 rpm to 20000 rpm depending on the sample

investigated. MWL data between 250 nm and 750 nm was recorded and saved in binary

format.[31] Figure S5 exemplarily shows GS raw data for a wavelength of 600 nm for all

samples discussed in this work. The bare silica particles in 33 wt% aqueous sucrose solution

were measured in a SV run in the same experimental setup. The rotor speed was 3000 rpm

and the temperature was 20 °C. It has to be emphasised that such low rotor speeds will not

allow a sucrose density gradient to build up during the GS or SV experiments. Thus, it can be

safely assumed that the sucrose is equally distributed in the measurement cell. Moreover, it

should be pointed out that the necessary low scanning speed of the MWL setup at a typical

rotor speed of 2000 rpm prohibit SV or band sedimentation experiments to be performed as an

alternative for spectral characterization for sedimentation coefficient-resolved fractions of the

patchy particle samples considered in this work.

Data were evaluated with HDR-MULTIFIT (version 1.06).[31] Sedimentation coefficient

distributions were obtained using DBM based on the works by Schuck et al.[32] and Walter et

al.[33] Distributions are either represented as cumulative distributions or as their first

derivative, herein referred to as density distributions according to ISO 9276-1. As noted above

the DBM analysis of GS experiments allows for less noisy distributions compared to the

traditional analysis. Details on the evaluation procedure can be found in the original

publications.[32,33] Most importantly, both strategies result in very similar results and could be

used interchangeably for the given investigations (see Figure S6 and Section IV in the

Supplementary Material for further details). For the extraction of extinction spectra, the model

independent procedure implemented in HDR-MULTIFIT was chosen due to computational

limitations of the DBM analysis when applied to all wavelengths.[31] The SV data obtained for

the bare silica particles was evaluated using the ls-g*(s) model in SEDFIT (version 14.81)

(Figure S2). The sedimentation coefficient distribution of the bare silica particles was

corrected for Mie scattering in HDR-MULTIFIT. Correlation of the AUC and SEM size data

resulted in an effective density of 2.1327 g cm-3 for the bare silica, which was used

throughout this work.

Acknowledgements

The authors gratefully acknowledge the Deutsche Forschungsgemeinschaft (DFG, German

Research Foundation) for funding of the Collaborative Research Centre 1411 “Design of

Particulate Products” (Project-ID 416229255) and the Cluster of Excellence “Engineering of

Advanced Materials” (Project-ID 53244630). The authors thank the Deutscher Akademischer

Austauschdienst (DAAD) and Japan Society for the Promotion of Science (JSPS) for financial

and travel support. Additionally, the authors acknowledge Shuji Ohsaki for his support in

experimental work at the Department of Chemical Engineering, Kyoto University. Thomas

Meincke and Johannes Walter contributed equally to this work.

Received: ((will be filled in by the editorial staff))

Revised: ((will be filled in by the editorial staff))

Published online: ((will be filled in by the editorial staff))

References

[1] Zhang, S. C. Glotzer, Nano Lett. 2004, 4, 1407.

[2] P. P. Ghimire, M. Jaroniec, J. Colloid Interface Sci. 2021, 584, 838.

[3] D. Rodríguez-Fernández, L. M. Liz-Marzán, Part. Part. Syst. Charact. 2013, 30, 46.

[4] M. Frederiksen, V. E. Bochenkov, M. B. Cortie, D. S. Sutherland, J. Phys. Chem. C

2013, 117, 15782.

[5] Z. He, I. Kretzschmar, Langmuir 2012, 28, 9915.

[6] Z. He, I. Kretzschmar, Langmuir 2013, 29, 15755.

[7] L. Bradley, Y. Zhao, Langmuir 2016, 32, 4969.

[8] D. Rodriguez-Fernandez, J. Perez-Juste, I. Pastoriza-Santos, L. M. Liz-Marzan,

ChemistryOpen 2012, 1, 90.

[9] G. K. Larsen, N. Stom, Y. Zhao, RSC Adv 2015, 5, 101257.

[10] M. E. Stewart, C. R. Anderton, L. B. Thompson, J. Maria, S. K. Gray, J. A. Rogers, R. G.

Nuzzo, Chem. Rev. 2008, 108, 494.

[11] B. Ai, Y. Yu, H. Mohwald, G. Zhang, B. Yang, Adv. Colloid Interface Sci. 2014, 206, 5.

[12] Y. Hou, Y. Li, L. Wang, D. Chen, M. Bao, Z. Wang, J. Colloid Interface Sci. 2019, 556,

54.

[13] Y. Yi, L. Sanchez, Y. Gao, K. Lee, Y. Yu, Chem. Mater. 2017.

[14] C. Kaewsaneha, P. Tangboriboonrat, D. Polpanich, M. Eissa, A. Elaissari, ACS Appl.

Mater. Interfaces 2013, 5, 1857.

[15] L.-T.-C. Tran, S. Lesieur, V. Faivre, Expert opinion on drug delivery 2014, 11, 1061.

[16] W. He, J. Frueh, Z. Wu, Q. He, ACS Appl. Mater. Interfaces 2016, 8, 4407.

[17] W. He, J. Frueh, Z. Wu, Q. He, Langmuir 2016, 32, 3637.

[18] F. Chang, S. Ouhajji, A. Townsend, K. Sanogo Lacina, B. G. P. van Ravensteijn, W. K.

Kegel, J. Colloid Interface Sci. 2021, 582, 333.

[19] M. A. Boles, M. Engel, D. V. Talapin, Chem. Rev. 2016, 116, 11220.

[20] B. Bharti, D. Rutkowski, K. Han, A. U. Kumar, C. K. Hall, O. D. Velev, J. Am. Chem.

Soc. 2016, 138, 14948.

[21] G. Loget, A. Kuhn, J. Mater. Chem. 2012, 22, 15457.

[22] A. Walther, A. H. E. Muller, Chem. Rev. 2013, 113, 5194.

[23] A. B. Pawar, I. Kretzschmar, Macromol. Rapid Commun. 2010, 31, 150.

[24] Q. Chen, S. C. Bae, S. Granick, Nature 2011, 469, 381.

[25] H. Bao, W. Peukert, R. N. Klupp Taylor, Adv. Mater. 2011, 23, 2644.

[26] T. Meincke, H. Bao, L. Pflug, M. Stingl, R. N. Klupp Taylor, Chem. Eng. J. 2017, 308,

89.

[27] S. E. Wawra, L. Pflug, T. Thajudeen, C. Kryschi, M. Stingl, W. Peukert, Nat. Commun.

2018, 9, 4898.

[28] X. Lu, M. Rycenga, S. E. Skrabalak, B. Wiley, Y. Xia, Annual review of physical

chemistry 2009, 60, 167.

[29] H. G. Müller, Colloid Polym. Sci. 1989, 267, 1113.

[30] W. Mächtle, Biophysical journal 1999, 76, 1080.

[31] J. Walter, W. Peukert, Nanoscale 2016, 8, 7484.

[32] P. Schuck, P. Rossmanith, Biopolymers 2000, 54, 328.

[33] J. Walter, T. Thajudeen, S. Suss, D. Segets, W. Peukert, Nanoscale 2015, 7, 6574.

[34] O. Kratky, H. Leopold, H. Stabinger, in Part D: Enzyme Structure, Vol. 27, Elsevier

1973, p. 98.

[35] S. Li, Q. Wan, Z. Qin, Y. Fu, Y. Gu, Langmuir 2015, 31, 824.

[36] S. R. Parnell, A. L. Washington, A. J. Parnell, A. Walsh, R. M. Dalgliesh, F. Li, W. A.

Hamilton, S. Prevost, J. P. A. Fairclough, R. Pynn, Soft matter 2016, 12, 4709.

[37] S. Li, Q. Wan, Z. Qin, Y. Fu, Y. Gu, Langmuir 2016, 32, 9180.

[38] J.-M. Commenge, L. Falk, Chemical Engineering and Processing: Process

Intensification 2011, 50, 979.

[39] J. Walter, K. Lohr, E. Karabudak, W. Reis, J. Mikhael, W. Peukert, W. Wohlleben, H.

Colfen, ACS Nano 2014, 8, 8871.

[40] J. Walter, P. J. Sherwood, W. Lin, D. Segets, W. F. Stafford, W. Peukert, Analytical

chemistry 2015, 87, 3396.

[41] E. Karabudak, E. Brookes, V. Lesnyak, N. Gaponik, A. Eychmuller, J. Walter, D. Segets,

W. Peukert, W. Wohlleben, B. Demeler, H. Colfen, Angew. Chem. Int. Ed. 2016, 55,

11770.

[42] K. L. Kelly, E. A. Coronado, L. L. Zhao, G. C. Schatz, J. Phys. Chem. B 2003, 107, 668.

[43] W. Stöber, A. Fink, E. Bohn, J. Colloid Interface Sci. 1968, 26, 62.

[44] S. Watanabe, T. Hiratsuka, Y. Asahi, A. Tanaka, K. Mae, M. T. Miyahara, Part. Part.

Syst. Charact. 2015, 32, 234.

[45] S. Watanabe, Y. Asahi, H. Omura, K. Mae, M. T. Miyahara, Adv. Powder Technol. 2016,

27, 2335.

[46] J. Pearson, J. Walter, W. Peukert, H. Cölfen, Anal. Chem. 2018, 90, 1280.

List of Figure Captions