Investigation and mitigation of reagent ageing during the continuous flow synthesis of
Andreas Völkl1 and Robin N. Klupp Taylor1*
1Institute of Particle Technology, Friedrich-Alexander University of Erlangen-Nürnberg,
Cauerstr. 4, 91058 Erlangen, Germany
*Corresponding author: email@example.com
Patchy particles, which comprise a core particle of one material partially covered by a thin layer
of another, offer great promise for tailored particle-based materials and devices in the future.
This study concerns the investigation and improvement of a potentially gram-scalable
continuous flow process based on a cascade of two T-mixers for the synthesis of thin, conformal
silver patches on colloidal silica particles. The starting point of the work is the observation that
on seemingly random occasions the process results in patches having undesired spiky
protrusions. Through a systematic study of reagent ageing supported by scanning electron
microscopy, optical spectroscopy and chemical equilibrium calculations it can be ascertained
that the desired product morphology is only obtained through carbon dioxide absorption by the
dilute ammonia solution used in the reaction. It is shown that the ageing process can be emulated
for fresh reagents by addition of ammonium nitrate or nitric acid.
Continuous flow synthesis; Static T-mixer; Heterogeneous nucleation and growth; Absorption
of carbon dioxide by aqueous ammonia; Localized surface plasmon resonance; Chemical
Patchy particles are a class of particles that have distinct surface regions with different
composition or functionality (Duguet et al., 2016; Li et al., 2020). A well-known special case
of such a particle is the so-called Janus particle which has two equal-size regions with distinct
surface properties (Walther and Müller, 2013; Sundararajan et al., 2018; Marschelke et al.,
2020). Due to the anisotropic distribution and large number of possible combinations of core
particle and patch properties, patchy particles are attractive candidates for a wide range of
applications. In particular, significant research revolves around biomedical applications
(Rabanel et al., 2019; Wong et al., 2019; Mirza and Saha, 2020), sensing (Bradley and Zhao,
2016; Dao et al., 2016; Luong et al., 2020) and self-assembly (Glotzer and Solomon, 2007;
Pawar and Kretzschmar, 2010; Li et al., 2011; Ravaine and Duguet, 2017). However, most
methods to produce patchy particles involve complex functional molecules or polymers (Wang
et al., 2012; Löbling et al., 2016; Mirza and Saha, 2020) or immobilization at interfaces (Pawar
and Kretzschmar, 2009; Bradley and Zhao, 2016; Jalilvand et al., 2018; Bradley et al., 2020).
Examples of scalable or continuous processes for patchy and Janus particle synthesis remain
rare in the literature (Tian et al., 2017; Zhang et al., 2017; Cui et al., 2021).
In our research group we have developed processes based on the heterogeneous nucleation and
surface conformal growth of metal patches on dielectric core particles (Bao et al., 2011; Bao et
al., 2014). Such particles combine the advantages of the patchy morphology with tunable optical
properties due to the localized surface plasmon resonance (Motl et al., 2014). Our synthetic
approach is attractive due to its simplicity, scalability and the fact that no prior surface
modification or functionalization of the core particles has to be carried out. In the case of silver
patches on colloidal silica nanospheres, which is based on a Tollens-like reduction of silver by
formaldehyde at alkaline pH, we successfully migrated from a batch (Bao et al., 2011) to a
continuous flow process that utilizes a static T-mixer (Meincke et al., 2017). As already
demonstrated for single phase metal nanoparticles (Thiele et al., 2015; Deshpande and
Kulkarni, 2018; Yang et al., 2018), such a process provides better control over mixing, resulting
in improved homogeneity and reproducibility compared to batch synthesis. We have recently
demonstrated that by operating the reported continuous setup for longer times, several grams
of product can be easily generated. Moreover, in order to improve the patch yield i.e., the
fraction of particles carrying a patch, we found it necessary to use high concentrations of most
reaction components. However, this brought the disadvantage that many core particles were
significantly or even completely covered by silver. Thus, in our currently preferred process, on
which the present article is based, we use a cascade of two mixers (Fig. 1a, top), similar to
multi-mixer setups in microfluidic chips (Köhler et al., 2008), and arrange the conditions at the
first and second mixer such that heterogeneous nucleation is promoted and suppressed,
respectively. This is achieved by adjusting the ammonia concentration at each mixer. As we
have shown in our earlier detailed investigation of the role of this reaction component in the
patch formation process (Sadafi et al., 2018), the balance of free and ammonia complexed silver
ions has a profound influence on heterogeneous nucleation and on the morphology of patch
growth (Fig. 1a, bottom). While lower concentrations favor nucleation and diffusion limited
growth (Fig. 1b), higher concentrations lead to less nucleation and promote reaction limited
growth (Fig. 1c). Thus, besides enabling high yields of silver patches to be achieved, our
application of the T-mixer cascade shown in Fig. 1a also results in the formation of cup-like
patches associated with the latter growth mode. This is especially important for applications
based on the plasmonic properties of the patches, because these structures have considerably
more pronounced resonances than patches with a dendritic morphology. Nevertheless, Fig. 1d
and 1e illustrate a problem with our continuous flow process and this represents the main
motivation for the present article. These images were obtained from two samples produced
under apparently identical conditions in the two T-mixer cascade setup. Both images show the
silica core particles coated with cup-like patches with varying degrees of coverage. However,
in the case of Fig. 1d, some of the patches have additionally grown spiky protrusions, and some
homogeneously nucleated silver nanoparticles, also with a spiky morphology can be seen. Such
structures are completely absent from the sample in Fig. 1e. Interestingly, these different
morphologies appeared randomly. Since we could eliminate reagent purity and concentration,
or post-synthetic processes (washing, sample preparation) as being responsible we considered
one variable which we previously did not closely control - the time between reagent preparation
and the start of the reaction run. While we did not expect our silver source, silica particles or
reducing agent to age significantly, we considered it plausible that our fairly dilute ammonia
solutions may change due to exposure to ambient air.
In the present work we describe systematic experiments which aim to unravel the mysterious
appearance of spiky nanostructures in our product. A key contributing aspect is revealed to be
the age of the ammonia solutions used, with aged ammonia leading to smoother patches.
Fig. 1. The starting point of this work. a) Schematic diagram of the T-mixer cascade used to
produce silver patches on silica (top) and plot of silver speciation at the ammonia concentrations
corresponding to the two T
-mixers (bottom). (b, c)
SEM images of silver on silica patchy
produced using low (b) and high (c) ammonia concentration during growth.(d, e
images of typical products of the two T
mixer cascade using apparently identical reaction
Considering the possible speciation in our system, we propose a mechanism to support our
observation and demonstrate how conditions similar to those afforded by aged ammonia can be
artificially created, resulting in spike-free patches irrespective of ageing time.
2. Material and Methods
The SiO2 core particles (Monospher M500, Merck Chemicals GmbH, Germany) used had a
median diameter of 480 nm. Silver nitrate powder (≥ 99%), formaldehyde solution (37%),
ammonia solution (32%), ammonium nitrate powder (≥ 98%) and nitric acid (≥ 65%) were
purchased from Carl Roth GmbH & Co. KG (Germany). For our experiments, stock solutions
of silver nitrate (100 mM), ammonium nitrate (1 M) and nitric acid (0.1 M) were prepared with
a PureLab Flex ultrapure water system (ELGA LabWater, Veolia Water Technologies
Deutschland GmbH, Germany) and used over a longer period. Directly before a set of
experiments a portion of the ammonia stock was diluted to 2% with ultrapure water and a core
particle suspension (2 mg/ml) was prepared from the dry powder. Formaldehyde was used as
2.2 Continuous flow synthesis of silver patches on silica core particles
Fig. 2 shows a schematic diagram of the continuous flow setup used in this work. Three reagent
reservoirs were freshly prepared in glass beakers (typically 50 – 80 mL) before each
experiment. Reservoir 1 (R 1) contained the core silica particles, silver nitrate and
2. Schematic representation of the double T-mixer setup.
formaldehyde while reservoirs 2 and 3 contained aqueous ammonia solutions with low and high
The reagent solutions were pumped from the reservoirs using a peristaltic pump with four
channels (Reglo ICC, Ismatec, Cole-Parmer GmbH, Germany), each providing a flowrate of 30
mL/min. Reservoir 1 was pumped with 2 channels (a T-connector) being used to combine the
flows after the pump, while reservoirs 2 and 3 (R 2 and R 3) were pumped with 1 channel each.
Prior to reaching the T-mixer cascade, the reagent flows passed through stainless steel coils,
which, along with the mixers and the residence time reactor which follow them, were immersed
in a water bath set to 50 °C. This temperature is preferred in order to obtain good patch yields
(Bao et al., 2011), and already implemented in our first generation continuous flow setup
(Meincke et al., 2017). The flows from reservoirs 1 and 2 were passed to the inlets of the first
of two identical polyamide T-mixers (inner diameter 2 mm). The outlet of this mixer was
directly connected to the inlet of the second T-mixer using a short piece of polyamide tubing
(T to T distance approximately 25 mm). The second inlet of this T-mixer received the flow
from reservoir 3. The outlet of the second T-mixer was connected to a 5 meter long residence
time reactor (polyamide tubing, inner diameter 3 mm). Samples of about 26 ml were collected
during steady state operation of the flow reactor directly at the outlet of the residence tube and
immediately mixed with 6.5 g sucrose (20 wt.-%) for stabilization.
Table 1 shows what we will refer to as standard concentrations. It should be noted that these
are expressed as concentrations after the first and second T-mixers and not as the original
reservoir concentrations. In the present work, ageing refers to free contact of the ammonia
containing reservoirs with air in the fume hood following solution preparation for a certain
amount of time ranging from 5 minutes to 3 hours prior to running the continuous flow process.
During ageing experiments, reservoir 1 was still prepared directly before each run. In attempts
to replicate the effect of ammonia solution ageing for non-aged solutions, additional
experiments were conducted in which various amounts of NH4NO3 and HNO3 were added to
Table 1: Standard concentrations used in this work. The concentrations of reservoir 1
components are expressed “conc. 1 / conc. 2”, corresponding to the concentrations after the first
and second T-Mixers respectively. The concentrations of reservoirs 2 and 3 are expressed as
the concentrations after the first and second T-Mixers respectively.
31.3 / 23.5
266.7 / 200.0
101.8 / 76.4
2.3 Characterization methods
Scanning electron microscopy was performed on a Gemini SEM500 instrument (Carl Zeiss
NTS GmbH) using the InLens detector at a typical working distance of around 3 mm and 10
kV acceleration voltage. UV-VIS-NIR spectroscopy was performed using a Lambda 950 a
PerkinElmer Lambda 950 spectrophotometer (PerkinElmer Inc., USA) together with UV-
cuvettes with 10 mm path length (Brand GmbH & Co KG).
2.4 Chemical equilibrium calculations
Calculation of chemical equilibria was done with CHEAQS Next (Version P2019.1, by Wilko
Verweij (Wilko Verweij)) and a tool partially based on CHEAQS’ database developed at our
institute (Haderlein et al., 2017). In CHEAQS, gasses can be introduced to an aqueous system
via a partial pressure, which was 411x10-6 atm for CO2 in this study, based on the global
monthly mean concentration of CO2 in the atmosphere at the time of simulation (Dlugokencky
and Tans). NH4NO3 and HNO3 concentrations were taken as 0.78 mM and 0.67 mM for
calculations corresponding to the first T-mixer and 0.59 mM and 0.50 mM for the second T-
3. Results and Discussion
Based on the hypothesis that different ageing times of our ammonia solution reservoirs is the
cause of the randomly seen differences in patchy particle morphology, we carried out a
systematic variation of this parameter. Fig. 3 shows selected SEM images of the products
resulting from this series along with UV-Vis-NIR extinction spectra for all runs. From the
micrographs it is evident that while freshly prepared ammonia solutions resulted in spiky
morphologies (Fig. 3a), by the time the ammonia solutions had aged for 60 minutes a smooth
morphology was obtained (Fig. 3d).
Even more conclusive evidence for this systematic influence of ammonia ageing is provided by
the extinction spectra (Fig. 3e). The features in these spectra result from the localized surface
plasmon resonances of the silver patches (Motl et al., 2014). Such resonances are extremely
sensitive to nanostructure morphology. Since spectroscopy is an ensemble measurement, even
subtle changes in the average morphology of a collective of plasmonic nanostructures can be
detected. All the spectra in Fig. 3e can be seen to possess three characteristic features: a peak
in the UV region at around 350 - 360 nm, a rising extinction “edge” across the visible region
and a broad high extinction feature in the NIR region. Similar to thin silver nanoparticles such
as nanoprisms or nanodiscs (Jin et al., 2001; Germain et al., 2005; Zhang et al., 2007), the first
of these features can be attributed to the out-of-plane quadrupole resonance of the silver patches
and is mainly dependent on their thickness (Meincke et al., 2017).
In the NIR region patches will possess an in-plane dipole resonance and this is the likely
dominant contribution to the high extinction at long wavelengths, albeit appearing as a rather
flat feature due to the range of patch coverages in the sample. In between the UV and NIR
features, higher order (e.g. quadrupole) in-plane resonances also occur, similar to those seen in
complete plasmonic nanoshells (Halas, 2005). In the case of spiky protrusions on the patches
or homogeneously nucleated silver particles with a spiky morphology, additional contributions
to the extinction spectra are expected. These will be at short wavelengths due to the quadrupolar
transverse resonance of the spikes and at longer wavelengths due to the longitudinal dipolar
resonance (Garcia-Leis et al., 2013; Reyes Gómez et al., 2018). The disappearance of spiky
structures for longer ammonia ageing times shown in Fig. 3a to 3d is clearly reflected by
systematic changes in the extinction spectra in Fig. 3e. Firstly, the UV peak slightly blue-shifts
and narrows due to the removal of the spike transverse resonance at around 400 nm. Then, even
more strikingly, the rise in extinction in the visible region blue-shifts by around 150 nm, a
3. a-d) SEM images
of patchy particle samples produced in discrete runs of the continuous
flow process using ammonia solutions which had been aged for a) 0, b) 5, c) 10 and d) 60
e) UV-Vis-NIR extinction spectra of samples produced using ammonia
for up to 180 minutes. In order to facilitate comparison, the spectra are normalized to the
extinction value at the minimum in the range 400
– 500 nm.
change which can be observed by the naked eye (see Fig. S1 in the Supplementary Material for
a photograph). This shift is apparently due to the loss of spiky structures with their longitudinal
resonances in the NIR and the commensurate increase in the available silver for lateral patch
growing. Indeed, electrodynamic simulations have shown that as the coverage of constant
thickness patches on a certain core size increases, the extinction rises in the center of the
spectrum relative to the dipole features at long wavelengths (Meincke et al., 2017). Moreover,
some higher order resonance peaks fairly insensitive to patch coverage inhomogeneity were
observed. These could explain the faint shoulder at 600 nm and peak at around 750 nm which
appear for longer ammonia ageing times in Fig. 3e. In summary, the changes in the extinction
spectra for patches synthesized following ammonia solution ageing clearly demonstrate the
suppression of spiky nanostructure formation.
We considered how the ageing of ammonia could cause the observed changes in patch
nucleation and growth leading to the variation of morphology and properties seen in Fig. 3. A
possible explanation is the volatility of ammonia which will result in its decreasing
concentration in aqueous solution over time. As already mentioned in the introduction, the
degree of complexation of silver ions is very sensitive to the ammonia concentration (See Fig.
1a, bottom). If ammonia were to be lost from the solution then an increase in the fraction of
non-complexed silver would be expected. However, as shown in Fig. 1b and 1c, such a change
is associated with a transition from cup-like to dendritic patches, something not observed during
our ageing study. Moreover, we have also observed spike formation for the case of an
intermediate ammonia concentration being used (See Fig. S2 in the Supplementary Material).
An alternative, more plausible explanation for the ageing phenomenon relates to the tendency
of solutions of ammonia (and amines in general) to absorb CO2. Indeed, this capability is
significant enough to be exploited on an industrial level, for example as the Chilled Ammonia
Process to reduce CO2 emissions of power plants (Gal; Darde et al., 2009).
In the present case, we suggest that the ammonia reservoirs prepared before the patch synthesis
reaction will naturally take up a certain amount of CO2 over time when in contact with air. CO2
that has been absorbed into an aqueous system will produce carbonic acid, lowering the pH of
the solution. Considering the following reaction for the reduction of silver ions by formaldehyde
and bearing in mind that the concentration of silver nitrate and formaldehyde are not expected
to change during ageing, it is clear that a loss of hydroxide ions will result in a lower reaction
2 ++2+ 2 −→2
0↓+ +2 (1)
We therefore considered whether acidification of non-aged ammonia solutions could replicate
the effect of ageing due to CO2 absorption. For this purpose, different amounts of ammonium
nitrate (NH4NO3) or nitric acid (HNO3) were added to reservoir 2 and synthesis runs performed
with no prior ageing of the ammonia reservoirs. These additives were chosen because their
corresponding cations and anions are already part of the system, thus keeping potential chemical
changes to a minimum.
The patch morphologies and optical properties of samples produced with various amounts of
NH4NO3 included are shown in Fig. 4. The NH4NO3 concentration was increased from
0.08 mM (0.06 mM) to 0.78 mM (0.59 mM) corresponding to the concentration after the first
(and second) mixer. Similar to the trends observed for ammonia ageing, the addition of
increasing amounts of ammonium nitrate to fresh ammonia solution causes a suppression of
spikes growing from the silver patches and also of homogeneously formed spiky particles (Fig.
4. a-d) SEM images
of patchy particle samples produced in the continuous flow process
using freshly prepared ammonia solutions with
added to Reservoir 2 in order to provide
the following concentrations after the first T
mixer: a) 0 mM, b) 0.39 mM, c) 0.55 mM and d)
e) UV-Vis-NIR extinction spectra of
the full range of samples with different amounts
NH4NO3 added. The concentration shown corresponds to that after the first T-
mixer. Due to
dilution, the concentration after the second T
mixer is 25% lower. In order to facilitate
comparison, the spectra are normalized to the extinction value at the minimum in the range 400
– 500 nm.
4 a-d). The corresponding extinction spectra (Fig. 4e) also show strong similarities to the case
of ammonia ageing (Fig. 3e). Most striking is the large blue-shift of the extinction edge in the
visible spectrum and appearance of a small peak, due to higher order patch resonances, at
around 700 nm. Interestingly, a rather less pronounced change is seen in the UV feature
compared to the ammonia ageing case. Since spiky structures were not produced for larger
amounts of NH4NO3, the lack of significant narrowing of this feature must result from some
other morphological feature, such as a tapered patch thickness or surface roughness, which
cannot be ascertained from SEM images.
Experiments using HNO3 as additive with concentrations up to a slightly lower value of 0.5
mM after the second T-mixer revealed similar trends to NH4NO3 (see Fig. S3 in the
Supplementary Material). One interesting difference was that a more similar trend in the UV
peak to the case of ammonia ageing was observed. We will give a possible explanation for this
Our experiments with acidification appear to confirm that the reduction of pH resulting from
CO2 absorption by ammonia solutions is a key contributing factor in the suppression of spiky
nanostructure formation during patchy particle synthesis in our continuous flow process. In
order to investigate other potentially important influences of the changing reaction conditions,
calculations were performed using the CHEAQS software (Wilko Verweij). These considered
the speciation of aqueous ammonia solutions in the presence of the additives CO2, NH4NO3 or
HNO3 (see Section 2.4 for the partial pressure and concentrations used). Firstly we estimated
the concentration of hydroxide ions after the first and second T-mixer as a function of ammonia
concentration at those positions in the reactor.
The results are shown in Fig. 5. For all three additives, the hydroxide ion concentration
decreases compared to the additive-free reference in both sections of the plot. Considering
reaction (1) this indicates a slower reaction rate during the nucleation phase after the first T-
mixer and growth phase after the second T-mixer. For the case of CO2-absorption, the change
is greater by an order of magnitude as compared to the other two cases, which is an indication
of how much carbon dioxide an ammonia solution can absorb until it reaches equilibrium.
Ageing times of up to three hours were investigated during this work. Although it is not clear
if this corresponds to the equilibrium state in the calculations, for the growth of silver patches
an equilibrium has apparently been achieved because the spectra cease to change significantly
The reduction in pH is not the only factor that may influence the patch growth reaction. As
already noted in our earlier work patch synthesis can only occur when the core particle and
metal precursor ion or complex are oppositely charged (Bao et al., 2014). This will cause the
metal ions to accumulate in the electrical double layer around the core particle, leading to an
increased concentration of precursor ions in the proximity of the surface. For the case of silver
patches on silica particles, there are only two types of cations in the system: silver ions
Fig. 5. Calculated concentration of hydroxide ions after the first (left) and second (right) T-
as a function of ammonia concentrations
at the same positions and for different additives.
partial pressure of CO2 used is 411x10-6 atm; NH4NO3 and HNO3 concentrations at
second mixer are 0.78/0.59 mM and 0.67/0.50 mM respectively.
(complexed and non-complexed) and ammonium ions (NH4+). Considering only the coulombic
attraction between silica and the cations, we assume that Ag+ (and its complexed forms) and
NH4+ within the electrical double layer will have the same relative share as in the bulk solution.
This way, ammonium ions can be seen as a competitor for silver ions within the double layer
as they are not involved in the chemical reaction leading to silver patch formation. Therefore,
variations in concentration of either of these species will lead to a shifting ratio between them,
effectively changing the concentration of silver ions available for the patch reaction close to the
silica surface. Indeed, our speciation calculations revealed that the composition of cations
drastically changes for all of the additives investigated. Fig. 6 shows this result for the ammonia
concentrations used in our experiments (1 mM and 17 mM after the first and second T-mixer
The reference (additive free) case for the first, low ammonia concentration, T-mixer shows that
about 25% of cations are ammonium ions and the remainder are silver-containing ions with
roughly two thirds of these being [Ag(NH3)2]+ and the rest equally divided between [Ag(NH3)]+
and Ag+. Due to a large increase in ammonium resulting from ageing (CO2 absorption) or
Fig. 6. Calculated concentrations of cations at the first (left) and at the second T-mixer (right)
case of no additive, CO2 addition (corresponding to typical ammonia
4NO3 or HNO3 addition at an ammonia concentration of 1 mM and 17
NH4NO3 and HNO3 concentrations at the first/second mixer are 0.78/0.59
addition of NH4NO3 or HNO3 this ratio reverses. Hence, significantly less silver will be
available for patch nucleation. For the second, higher ammonia concentration T-mixer silver-
containing cations (almost exclusively [Ag(NH3)2]+) are already in the minority for the
reference case, with further reduction relative to ammonium occurring for the three additives,
especially CO2. Regarding the latter, it should be noted that absorption of carbon dioxide causes
the formation of several negatively charged carbon-based species, e.g. deprotonated carbonates.
To retain a net zero charge in the solution, more ammonium ions need to be produced to
counterbalance additional anions, which explains the difference compared to the other two
Recalling that the hydroxide concentration decreases when additives are included (see Fig. 5),
the increasing dominance of ammonium over silver species shown in Fig. 6 will, after the first
T-mixer, lead to the reaction rate decreasing during the nucleation phase. This seems consistent
with our experimental observation that ageing (CO2 absorption) or acidic additives suppresses
the homogeneous nucleation (of free-standing silver nanoparticles) seen in the non-aged or
additive free samples presented earlier (Fig. 3a and Fig. 4a). After the second T-mixer, where
most patch growth takes place, a lower reaction rate will promote reaction limited crystal
growth which might explain the appearance of smoother i.e. non-spiky coatings. Moreover, the
increase in ammonium concentration may provide a passivating effect with regard to spike
formation. Inhomogeneous charge distributions in a metal nanostructure are known to promote
the formation of anisotropic features like branches and spikes during further growth
(Sadovnikov and Gusev, 2017; Köhler et al., 2021). As we have observed in earlier work, the
silver patches are polycrystalline with a rough surface (Bao et al., 2011). Thus, asperities at the
surface may act, due to their higher electron density, as nucleation sites for spikes. In the case
of a high concentration of ammonium close to the growing silver patch surface, this effect could
be suppressed. Future studies, for instance using liquid cell transmission electron microscopy
(Jung et al., 2019), may be able to resolve this and provide a clearer picture of the mechanism
of spike formation.
A further observation from the data shown in Fig. 6 (first T-mixer) is that the addition of
NH4NO3 leads to a decreasing concentration of non-complexed silver ions (orange species in
Fig. 6), whereas for ageing (CO2 absorption) and the addition of nitric acid the Ag+
concentration is increased compared to the reference case. When looking at the following
chemical equilibria the reason for these changes can be explained:
3+2 ⇌ −+4
++−⇌ 2 (3)
The addition of NH4NO3 will consume hydroxide ions through reaction (2), lowering the pH as
mentioned earlier and increasing the concentration of ammonia. More ammonia in the system
causes a higher degree of complexation, thus decreasing the concentration of non-complexed
silver. This will result in an overall slower reaction rate during the reduction of silver ions and
suppression of diffusion limit growth which promotes spike formation.
For the other two cases, hydrogen ions from nitric or carbonic acid will consume hydroxide
ions via reaction (3), leading to a reduction of the pH. The decreasing OH- concentration will
shift the equilibrium in reaction (2) to the right side, i.e. producing more NH4+. The resulting
lower concentration of ammonia will then cause a smaller degree of complexation of Ag+-ions,
effectively increasing the concentration of non-complexed silver ions. As already noted in the
introduction, such a situation can result in higher reaction rates and dendritic patch formation.
However, here the situation is different due to the issues mentioned above regarding the lowered
pH and competition of the silver species with ammonium. Hence, the reaction rate (and thus
rate of heterogeneous nucleation) may actually only be marginally higher. Nevertheless, this is
possibly the origin of a difference seen between the case of NH4NO3 being added and those of
ageing (CO2 absorption) or HNO3 addition. For the latter, a higher rate of nucleation (due to
more non-complexed silver being available) will produce a larger number of patches compared
to the case of NH4NO3 being added. This larger number of nuclei will grow, with the same
amount of silver, to thinner or less covered patches. This could explain the blue-shifts in the
UV plasmon resonance which were observed for the cases of ammonia ageing (Fig. 3e) and
nitric acid addition (Fig. S3e) and not for NH4NO3 addition (Fig. 4e). This aspect requires
further study in the future, for instance making use of multidimensional characterization
techniques like multi-wavelength analytical ultracentrifugation which can reveal trends in the
yield, thickness and coverage (Meincke et al., 2021).
The objective of this work was to identify and mitigate for the cause of the sporadic appearance
of spiky nanostructures in our patchy particle product. Having confirmed that insufficient
NIR extinction spectra of patchy particle samples produced in discrete runs of
the continuous flow process at various times of reagent ageing where
HNO3 (top) and NH4NO
(bottom) were added with shown concentration to the freshly prepared ammonia solutions for
Reservoir 2 prior to the first run.
ammonia ageing is the cause and that acidification can generate a similar effect to aging, we
verified the improved stability of our process using additives over 180 minutes of reagent
ageing. For this we carried out two sets of synthesis runs with additives included in reservoir 2
to provide concentrations (1st mixer/2nd mixer) of 0.67/0.5 mM HNO3 and 1.33/1 mM NH4NO3,
respectively. While SEM analysis confirmed the absence of spiky nanostructures in all samples
(Fig. S4), the extinction spectra in Fig. 7 show less variation in the main features compared to
the case of no additives being included (see Fig. 3e). In particular, for the case of HNO3 addition
there is negligible change in the UV peak and only ~ 30 nm blue-shift of the main extinction
edge in the visible region over the whole ageing time. This confirms that the addtion of HNO3
at the outset of a long synthesis campaign e.g. for gram-scale production, is preferable in order
to achieve stable synthesis of patchy particles.
In this work we demonstrated how the apparently random appearance of an undesired product
morphology during the continuous flow synthesis of silver patches on colloidal silica spheres
can be understood and mitigated. Our process uses a cascade of two static T-mixers immersed
in a 50 °C water bath, the first mixing silver nitrate, formaldehyde and silica particles with
aqueous ammonia solution, the outlet leading directly to a second mixer where ammonia
solution of a higher concentration is introduced. We showed, through synthetic runs carried out
at different ammonia solution ageing times, and investigation with scanning electron
microscopy and UV-Vis-NIR spectroscopy, that this parameter is related to the disappearance
of undesirable silver protrusions, both on the silver patches and as homogeneously formed silver
particles. Having deduced that this change in morphology was not related to the volatility of
ammonia, we confirmed, using systematic experimentation and chemical equilibrium
calculations, the hypothesis that CO2 absorption by aqueous ammonia solution results in
sufficient change in the reagent composition on mixing to cause the observed effect. Moreover,
it was shown that when suitable amounts of ammonium nitrate or nitric acid were included in
the ammonia solution reservoir being fed to the first T-mixer, the desired patchy particle
morphology and optical properties could be obtained. Importantly, with such an additive
present, the product remained largely unchanged, even when produced at various times over 3
hours of reagent ageing.
This work represents an important step towards upscaling of our process. In the short term, the
findings regarding ageing and its mitigation will be combined with our current activities, which
aim to completely separate the patch growth from nucleation by using a seeding process. This
will enable specific patch dimensions (and associated optical properties) to be targeted. It will
also provide a highly reproducible process to allow us to perform kinetic studies of patch growth
and thereby learn more about possible specific interactions, for instance between ammonium
ions and the silica surface, which were neglected in the present work.
Finally, we note that although our work is based on a rather unique patchy particle system, the
implications of our results and mitigation strategy are significant for any silver nanostructure
synthetic process involving dilute amine solutions, particularly where longer times or larger
reaction volumes are involved.
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).
Bao, H., Bihr, T., Smith, A.-S., Klupp Taylor, R.N., 2014. Facile colloidal coating of polystyrene
nanospheres with tunable gold dendritic patches. Nanoscale 6 (8), 3954–3966.
Bao, H., Peukert, W., Klupp Taylor, R.N., 2011. One-pot colloidal synthesis of plasmonic patchy
particles. Adv. Mater. 23 (22-23), 2644–2649. 10.1002/adma.201100698.
Bradley, L., Ye, D., Luong, H.M., Zhao, Y., 2020. Transition from discrete patches to plasmonic
nanohole array by glancing angle deposition on nanosphere monolayers. Nanotechnology 31
(20), 205301. 10.1088/1361-6528/ab70d2.
Bradley, L., Zhao, Y., 2016. Uniform Plasmonic Response of Colloidal Ag Patchy Particles Prepared by
Swinging Oblique Angle Deposition. Langmuir 32 (19), 4969–4974.
Cui, J., Li, Y., Yuan, H., Gao, N., Feng, K., Li, W., Zhou, K., Yin, X., Li, G., 2021. Gram-scale fabrication of
patchy nanoparticles with tunable spatial topology and chemical functionality. Nano Res.
Dao, T.D., Nghiem, H.L.T., Chen, K., Nagao, T., 2016. Ensemble of gold-patchy nanoparticles with
multiple hot-spots for plasmon-enhanced vibrational spectroscopy, In: Nanostructured Thin Films
IX. SPIE Nanoscience + Engineering, San Diego, California, United States. 26 September 2016.
SPIE, p. 992904–992904.
Darde, V., Thomsen, K., van Well, W.J.M., Stenby, E.H., 2009. Chilled ammonia process for CO2
capture. Energy Procedia 1 (1), 1035–1042. 10.1016/j.egypro.2009.01.137.
Deshpande, J.B., Kulkarni, A.A., 2018. Reaction Engineering for Continuous Production of Silver
Nanoparticles. Chem. Eng. Technol. 41 (1), 157–167. 10.1002/ceat.201700035.
Dlugokencky, E., Tans, P. Global Monthly Mean CO2. NOAA/GML. URL: <gml.noaa.gov/ccgg/trends/>.
Accessed 16 July 2020.
Duguet, É., Hubert, C., Chomette, C., Perro, A., Ravaine, S., 2016. Patchy colloidal particles for
programmed self-assembly. C. R. Chim. 19 (1-2), 173–182. 10.1016/j.crci.2015.11.013.
Gal, E., inventor; EIG Inc., assignee. Ultra cleaning of combustion gas including the removal of CO2.
International Patent WO/2006/022885. 2006 Mar 02.
Garcia-Leis, A., Garcia-Ramos, J.V., Sanchez-Cortes, S., 2013. Silver Nanostars with High SERS
Performance. J. Phys. Chem. C 117 (15), 7791–7795. 10.1021/jp401737y.
Germain, V., Brioude, A., Ingert, D., Pileni, M.P., 2005. Silver nanodisks: size selection via
centrifugation and optical properties. J. Chem. Phys. 122 (12), 124707. 10.1063/1.1865993.
Glotzer, S.C., Solomon, M.J., 2007. Anisotropy of building blocks and their assembly into complex
structures. Nat. Mater. 6, 557–562. 10.1038/nmat1949.
Haderlein, M., Güldenpfennig, A., Segets, D., Peukert, W., 2017. A widely applicable tool for modeling
precipitation processes. Comput. Chem. Eng. 98, 197–208. 10.1016/j.compchemeng.2016.12.007.
Halas, N.J., 2005. Playing with Plasmons: Tuning the optical Resonant properties of metallic
nanoshells. MRS Bull. 30 (5), 362–367. 10.1557/mrs2005.99.
Jalilvand, Z., Pawar, A.B., Kretzschmar, I., 2018. Experimental Study of the Motion of Patchy Particle
Swimmers Near a Wall. Langmuir 34 (50), 15593–15599. 10.1021/acs.langmuir.8b03220.
Jin, R., Cao, Y., Mirkin, C.A., Kelly, K.L., Schatz, G.C., Zheng, J.G., 2001. Photoinduced conversion of
silver nanospheres to nanoprisms. Science 294 (5548), 1901–1903. 10.1126/science.1066541.
Jung, W.-G., Park, J.H., Jo, Y.-R., Kim, B.-J., 2019. Growth Kinetics of Individual Au Spiky Nanoparticles
Using Liquid-Cell Transmission Electron Microscopy. J. Am. Chem. Soc. 141 (32), 12601–12609.
Köhler, J.M., Abahmane, L., Wagner, J., Albert, J., Mayer, G., 2008. Preparation of metal nanoparticles
with varied composition for catalytical applications in microreactors. Chem. Eng. Sci. 63 (20),
Köhler, J.M., Kluitmann, J.J., Günther, P.M., 2021. Metal Nanoparticles as Free-Floating Electrodes.
Encyclopedia 1 (3), 551–565. 10.3390/encyclopedia1030046.
Li, F., Josephson, D.P., Stein, A., 2011. Colloidal assembly: the road from particles to colloidal
molecules and crystals. Angew. Chem. Int. Ed. 50 (2), 360–388. 10.1002/anie.201001451.
Li, W., Palis, H., Mérindol, R., Majimel, J., Ravaine, S., Duguet, E., 2020. Colloidal molecules and
patchy particles: complementary concepts, synthesis and self-assembly. Chem. Soc. Rev. 49 (6),
Löbling, T.I., Borisov, O., Haataja, J.S., Ikkala, O., Gröschel, A.H., Müller, A.H.E., 2016. Rational design
of ABC triblock terpolymer solution nanostructures with controlled patch morphology. Nat.
Comm. 7, 12097. 10.1038/ncomms12097.
Luong, H.M., Pham, M.T., Madhogaria, R.P., Phan, M.-H., Larsen, G.K., Nguyen, T.D., 2020. Bilayer
plasmonic nano-lattices for tunable hydrogen sensing platform. Nano Energy 71, 104558.
Marschelke, C., Fery, A., Synytska, A., 2020. Janus particles: from concepts to environmentally
friendly materials and sustainable applications. Colloid Polym. Sci. 298 (7), 841–865.
Meincke, T., Bao, H., Pflug, L., Stingl, M., Klupp Taylor, R.N., 2017. Heterogeneous nucleation and
surface conformal growth of silver nanocoatings on colloidal silica in a continuous flow static T-
mixer. Chem. Eng. J. 308, 89–100. 10.1016/j.cej.2016.09.048.
Meincke, T., Walter, J., Pflug, L., Thajudeen, T., Völkl, A., Cardenas Lopez, P., Uttinger Maximilian,
Stingl, M., Watanabe, S., Peukert, W., Klupp Taylor, R.N., 2021. Determination of the yield, mass
and structure of silver patches on colloidal silica using multiwavelength analytical
ultracentrifugation. J. Colloid Interf. Sci. In press.
Mirza, I., Saha, S., 2020. Biocompatible Anisotropic Polymeric Particles: Synthesis, Characterization,
and Biomedical Applications. ACS Appl. Bio Mater. 3 (12), 8241–8270. 10.1021/acsabm.0c01075.
Motl, N.E., Smith, A.F., DeSantis, C.J., Skrabalak, S.E., 2014. Engineering plasmonic metal colloids
through composition and structural design. Chem. Soc. Rev. 43 (11), 3823–3834.
Pawar, A.B., Kretzschmar, I., 2009. Multifunctional patchy particles by glancing angle deposition.
Langmuir 25 (16), 9057–9063. 10.1021/la900809b.
Pawar, A.B., Kretzschmar, I., 2010. Fabrication, assembly, and application of patchy particles.
Macromol. Rapid Comm. 31 (2), 150–168. 10.1002/marc.200900614.
Rabanel, J.-M., Adibnia, V., Tehrani, S.F., Sanche, S., Hildgen, P., Banquy, X., Ramassamy, C., 2019.
Nanoparticle heterogeneity: an emerging structural parameter influencing particle fate in
biological media? Nanoscale 11 (2), 383–406. 10.1039/c8nr04916e.
Ravaine, S., Duguet, E., 2017. Synthesis and assembly of patchy particles: Recent progress and future
prospects. Curr. Opin. Colloid Interface Sci. 30, 45–53. 10.1016/j.cocis.2017.05.002.
Reyes Gómez, F., Rubira, R.J.G., Camacho, S.A., Martin, C.S., da Silva, R.R., Constantino, C.J.L., Alessio,
P., Oliveira, O.N., Mejía-Salazar, J.R., 2018. Surface Plasmon Resonances in Silver Nanostars.
Sensors 18 (11). 10.3390/s18113821.
Sadafi, F.-Z., Sauerbeck, C., Braunschweig, B., Klupp Taylor, R.N., 2018. On the complex role of
ammonia in the electroless deposition of curved silver patches on silica nanospheres.
CrystEngComm 20 (40), 6214–6224. 10.1039/C8CE00866C.
Sadovnikov, S.I., Gusev, A.I., 2017. Synthesis and characterization of novel stellate sea-urchin-like
silver particles with extremely low density and superhydrophobicity. J. Mater. Chem. A 5 (38),
Sundararajan, P., Wang, J., Rosen, L.A., Procopio, A., Rosenberg, K., 2018. Engineering polymeric
Janus particles for drug delivery using microfluidic solvent dissolution approach. Chem. Eng. Sci.
178, 199–210. 10.1016/j.ces.2017.12.013.
Thiele, M., Knauer, A., Csáki, A., Mallsch, D., Henkel, T., Köhler, J.M., Fritzsche, W., 2015. High-
Throughput Synthesis of Uniform Silver Seed Particles by a Continuous Microfluidic Synthesis
Platform. Chem. Eng. Technol. 38 (7), 1131–1137. 10.1002/ceat.201400524.
Tian, L., Li, X., Wan, D., Ali, Z., Zhang, Q., 2017. Large-scale fabrication of polymer ellipsoids with
controllable patches via the viscosity-induced deformation of spherical particles. Polym. Chem. 8
(25), 3774–3777. 10.1039/C7PY00475C.
Walther, A., Müller, A.H.E., 2013. Janus particles: synthesis, self-assembly, physical properties, and
applications. Chem. Rev. 113 (7), 5194–5261. 10.1021/cr300089t.
Wang, Y., Wang, Y., Breed, D.R., Manoharan, V.N., Feng, L., Hollingsworth, A.D., Weck, M., Pine, D.J.,
2012. Colloids with valence and specific directional bonding. Nature 491 (7422), 51–55.
Wilko Verweij. CHEAQS Next. URL: <https://www.cheaqs.eu/index.html>. Accessed 15 April 2020.
Wong, C.K., Chen, F., Walther, A., Stenzel, M.H., 2019. Bioactive Patchy Nanoparticles with
Compartmentalized Cargoes for Simultaneous and Trackable Delivery. Angew. Chem. Int. Ed. 58
(22), 7335–7340. 10.1002/anie.201901880.
Yang, T., Segets, D., Thajudeen, T., Han, Y., Peukert, W., 2018. The effect of mixing on silver particle
morphology in flow synthesis. Chem. Eng. Sci. 192, 254–263. 10.1016/j.ces.2018.07.043.
Zhang, J.H., Liu, H.Y., Zhan, P., Wang, Z.L., Ming, N.B., 2007. Controlling the Growth and Assembly of
Silver Nanoprisms. Adv. Funct. Mater. 17 (9), 1558–1566. 10.1002/adfm.200600727.
Zhang, S.-B., Ge, X.-H., Geng, Y.-H., Luo, G.-S., Chen, J., Xu, J.-H., 2017. From core-shell to Janus:
Microfluidic preparation and morphology transition of Gas/Oil/Water emulsions. Chem. Eng. Sci.
172, 100–106. 10.1016/j.ces.2017.06.031.
Investigation and mitigation of reagent ageing during the continuous flow synthesis of
Andreas Völkl1 and Robin N. Klupp Taylor1*
1Institute of Particle Technology, Friedrich-Alexander University of Erlangen-Nürnberg,
Cauerstr. 4, 91058 Erlangen, Germany
Fig. S1. Photograph of patchy particle suspensions showing the color transition caused by
ageing ammonia reservoirs.
The ageing times were 0 min (left), 15 min (middle) and 60 min
Fig. S2: SEM image of silver patches produced at an ammonia concentration of 5 mM after the
(rather than 17 mM used in the rest of this work). The ammonia solutions were
Fig. S3: a-d) SEM images of patchy particle samples produced in the continuous flow process
using freshly prepared ammonia solutions with HNO
added to Reservoir 2 in order to provide the
following concentrations after the fi
rst T-mixer. a) 0 mM, b) 0.53 mM
, c) 0.60 mM and d) 0.67
. e) UV-Vis-
NIR extinction spectra of the full range of samples with different amounts of
3 added. The concentration shown corresponds to that after the first T-
mixer. Due to dilution,
the concentration after the second T
-mixer is 25% lower. In order to
facilitate comparison, the
spectra are normalized to the extinction value at the minimum in the range 400
– 500 nm.
Fig. S4: SEM images of samples produced of patchy particle samples produced in discrete runs
of the continuous flow process where
HNO3 (a,b) and NH4NO3
(c,d) was added to freshly
prepared ammonia solutions for Reservoir 2 prior to the first run. Images (a,c) show samples
produced by a run carried out immediately after reagent preparation. Images (b,d) show samples
produced after 180 minu
tes of reagent ageing. The additive concentrations used
were 0.67 / 0.50 mM for HNO3 and 1.33 / 1.00 mM for NH4NO3.
Table S1: Synthesis parameters for all samples in the work which were not produced with
standard parameters for cup-like patches. All concentrations except that of NH3 correspond to
concentrations after the first T-mixer and are diluted to 75% after the second mixer.
Standard 266.7 100.8 31.3 1.0 17.0 - -
Fig. 1b 266.7 100.8 31.3 1.0 1.0 - -
Fig. 4b 266.7 100.8 31.3 1.0 17.0 0.39 -
Fig. 4c 266.7 100.8 31.3 1.0 17.0 0.55 -
Fig. 4d 266.7 100.8 31.3 1.0 17.0 0.78 -
266.7 100.8 31.3 1.0 17.0 - 0.67
266.7 100.8 31.3 1.0 17.0 1.33 -
Fig. S2 266.7 100.8 31.3 1.0 5.0 - -
Fig. S3b 266.7 100.8 31.3 1.0 17.0 - 0.53
Fig. S3c 266.7 100.8 31.3 1.0 17.0 - 0.60
Fig. S3d 266.7 100.8 31.3 1.0 17.0 - 0.67
Fig. S4a 266.7 100.8 31.3 1.0 17.0 - 0.67
Fig. S4b 266.7 100.8 31.3 1.0 17.0 - 0.67
Fig. S4c 266.7 100.8 31.3 1.0 17.0 1.33 -
Fig. S4d 266.7 100.8 31.3 1.0 17.0 1.33 -