Acs Chemmater 0c04688

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

pubs.acs.

org/cm Comments

A Comparison of the Stochastic and Deterministic Approaches in a


Nucleation−Growth Type Model of Nanoparticle Formation
Cite This: Chem. Mater. 2021, 33, 5430−5436 Read Online

ACCESS Metrics & More Article Recommendations *


sı Supporting Information

F or the last two decades, the research group of Richard that a stochastic simulation of nucleation is necessary and then
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Finke has been developing fairly general kinetic models of a possible expansion of the simulation to the deterministic
nanoparticle formation,1−11 which can be viewed as initial limit (where the experimental evidence is almost exclusively
attempts at a rational control of nanoparticle size distribution. gained) could lead to a realistic model of any mechanism.13
The general approach of the lumped kinetic models used by
Downloaded via 103.99.197.132 on November 14, 2022 at 15:01:17 (UTC).

Such a stochastic−deterministic comparison has well-known


Finke and co-workers was criticized in this journal by James
Martin,12 and a response was published by the Finke group.13 theoretical foundations:37,38 Kurtz’s theorem16,37 states that
The present comment has been written to address a point that the usual deterministic kinetic approach is the infinite volume
was brought up by both authors12,13 in some form. Martin12 limit of the continuous time discrete state (CDS) stochastic
called for a consideration of single nucleation events, for which approach, which is most commonly used and will be employed
stochastic kinetics is the appropriate method. Finke, White- here. However, Kurtz’s theorem has two limitations. First, it is
head, and Watzky13 explicitly called for using the stochastic not valid for open systems.32 This is not relevant for the
approach in really quantitative models. The problem in the present analysis because the model here at least formally
background is that classical (deterministic) chemical kinetics
works with concentrations as continuous functions of time at describes a closed nanoparticle formation system. More
the macroscopic or ensemble level, whereas in reality matter is importantly, it is typically the expectation (or mean) of the
composed of particles and all molecular events actually occur stochastic approach, which is basically the same as an average
atomistically or molecularly between discrete particles. A but defined with mathematically precision,16 that can be
mathematical approach to handle this problem, called compared directly with results from the deterministic
stochastic chemical kinetics, has already been developed calculations. The problem lies in the fact that the stochastic
during the last 100 years and is in common use for solving expectation is occasionally a very unlikely or even impossible
certain kinetic problems,14−16 yet its general application seems
state in a single experiment.16,39 Phenomena similar to this are
to be surprisingly rare.
In detailed, nonlumped models of nanoparticle formation, known in the description of autocatalytic reactions, too.24,25
the number of different species is usually extremely large as The kinetic models of nanoparticle formation are often
each nanoparticle of a particular size should be considered as a described as autocatalytic1,2,6,7,10 which might give some
chemical entity on its own, whose reactivity differs at least reason for concern in this respect.
slightly from nanoparticles of identical chemical compositions The present comment directly addresses the issue of the
but different sizes.17−22 Given the very high number of connection between the stochastic and the deterministic
different species in the system, it is natural that the amounts of approaches to kinetics specifically for the nucleation−growth
substances for most of them are very low. These are exactly the
type mechanisms of nanoparticle formation. Results from the
conditions under which considerations based on deterministic
kinetics, where amounts of substances and concentrations are two approaches will be compared in two different systems: one
considered to be continuous variables, may be misleading or for which the exact analytical solution was reported recently
entirely erroneous.16 It should be noted that the appearance of (eq 1 below)21 and another one for which a method to find the
such problems is possible in such cases but not inevitable. numerical deterministic solution was developed specifically in
There are a number of examples where the stochastic kinetic this work (eq 10 further on). In both cases, the stochastic
approach yielded insightful results including autocatalytic description is based on Monte Carlo simulations,16 which
processes,23 autocatalytic extinction,24 chiral symmetry break- consider the models at the single nucleation event level.
ing,25−27 a mechanism-based interpretation of the Soai
reaction,28 the origins of biomolecular chirality29 and the
effect of parity violation in it,30,31 an analysis of the Frank Received: December 8, 2020
model,32 and details of single-enzyme catalysis.33−36 Revised: May 7, 2021
It is clear that the best way of addressing the concern raised Published: June 25, 2021
about the single-event nucleation by Martin12 is to compare
the results of deterministic and stochastic kinetic calculations
in the same chemical model. Indeed, Finke’s response13 noted
© 2021 The Authors. Published by
American Chemical Society https://doi.org/10.1021/acs.chemmater.0c04688
5430 Chem. Mater. 2021, 33, 5430−5436
Chemistry of Materials


pubs.acs.org/cm Comments

THE NUCLEATION−GROWTH TYPE KINETIC


i yz
kg − 2k n

lnjjjj −k g[M]0 t z
z
⟨r ⟩ = r0
z
MODEL AND THE DETERMINISTIC APPROACH 2k gk n + k g(k g − 2k n)e−k g[M]0 t

k 2k n + (k g − 2k n)e {
3 kg
−kn +
In the nucleation−growth type model1,5,17 considered here, M 2(k g − 2k n)(1 − e−k g[M]0 t ) (4)
will denote a single monomeric unit of the nanoparticle,
whereas Ci will stand for a nanoparticle containing exactly i Parameter r0 here is a reference size: the hypothetical sphere-
monomeric units. The two reactions in the model are second equivalent size of a single monomeric unit as a nanoparticle.
order nucleation (kn) and also overall second order particle So the deterministic kinetic approach to the chemical model
growth (kg,i), whose rate constant is directly proportional to in eq 1 leads to a full solution in the form of analytical formulas
the number of monomeric units in the particle (i). for the concentration and particle sizes. Equation 2 gives the
kn
system of simultaneous ordinary differential equations that
2M → C2 rn = k n[M]2 follows from the model, whereas eqs 3 and 4 show the
kg, i
analytical solutions. It should be noted that it is highly
Ci + M → Ci + 1 i≥2 rg, i = ikg[M][Ci] exceptional that this analytical solution could be found for eq
2.21 Normally, such analytical solutions are only available for
(1)
the simplest systems.43 Unlike in the typical cases of
In this equation, ri denotes the rate of the corresponding deterministic kinetics, the numerical solution of eq 2 is
reaction step. This model was primarily developed to interpret impossible by the usual integration methods43 because it has
the time dependence of the average particle size in a simplified an infinitely high number of coupled dependent variables (i.e.,
but fully quantitative manner in solution systems. Unfortu- concentrations).


nately, the kinetics of nanoparticle formation is not studied
with a sufficient time resolution in the majority of the relevant THE STOCHASTIC APPROACH: METHODS
articles, but the usefulness of the model was still demonstrated
in various systems including the formation of zirconiumox- In this section, the stochastic kinetic approach to the model
oalkoxy nanoparticles,40 the formation of amino-PEG covered described in eq 1 will be developed. As usual in the continuous
gold nanoparticles41 (the interpretation itself is given in a later time discrete state approach,16 the system will be described
article for these two sets of data21), and the formation of TiO2 using discrete molecule numbers instead of continuous
nanoparticles from the hydrolysis and condensation of concentrations, and the connection between the two models
titanium(IV)-bis(ammonium-lactato)dihydroxide under basic will be made using the full volume (V) of the system.16
conditions.42 These three examples imply that the model itself The number of monomeric unit molecules in the initial state
could be a general one for solutions. will be denoted n. Symbol c1(t) will mean the number of
The system of simultaneous ordinary differential equations monomeric units at time t, whereas the number of nano-
that describes the investigated chemical model takes the particles with exactly i monomeric units in it will be designated
following form: ci(t). It is seen that the maximum value of i is n (i.e., no
∞ ∞
individual nanoparticle can contain more monomeric units
d[M] than the overall number of monomeric units in the system).
= − 2rn − ∑ rg,j = −2kn[M]2 − kg[M] ∑ j[Cj]
dt j=2 j=2 Because of mass conservation, the following equation holds at
d [C 2 ] any time instant:
= rn − rg,2 = k n[M]2 − 2kg[M][C2]
dt n
d[Ci]
= rg, i − 1 − rg, i = (i − 1)kg[M][Ci − 1] − ikg[M][Ci] i>2 ∑ ici(t ) = n
dt i=1 (5)
(2)
Brackets mean concentrations here (amount of substance Counting the number of possible solutions for eq 5 gives the
divided by volume). The typical initial conditions of number of states (M) in the stochastic model of the process. In
nanoparticle formation are such when, at time t = 0, the a classical subfield within combinatorics, this is called the
initial concentration of the monomeric unit is [M]0, whereas partition problem.44 The number of solutions of eq 5 is called
all of the nanoparticles are absent, i.e., [Ci]0 = 0. In our the partition function of number theory and is abbreviated
previous work,21 detailed mathematical derivations showed p(n).44 No closed form expression is known for calculating
that the concentrations of individual species in the model can p(n). For example, the known value of the partition function
be given exactly as a function of time as follows: for n = 10 000 (ten thousand) is approximately 3.6 × 10106
(exact value is given in the Supporting Information), which
[M]0 kg e−k g[M]0 t renders it hopeless to handle all possible states by simply
[M] =
2k n + (kg − 2k n)e−k g[M]0 t increasing the computational power.
The full form of the CDS approach sets up a master
[M]0 2kgi − 2k n2(i − 1)! (− 1)i − 1(e−k g[M]0 t − 1) equation that is equivalent to the rate equation,16 which has
[Ci] = i

k n(j 2 − 1) ijji − 1 yzz


[2k n + (kg − 2k n)e−k g[M]0 t ] ∏ j = 2 (2k n − kgj − kg) the form given in eq 2 in the present case. In the master
jj z(− 1) j
j − 1z{
i
equation, the probability of a given state (and not the
j = 2 (2k n − kgj − kg)j k
+ [M]0 ∑
ÄÅ ÉÑ
ÅÅ ÑÑ
probability of individual concentrations or molecule numbers)
ÅÅij yz g ÑÑ
ÅÅjj z ÑÑ
is given in a linear, first order set of differential equations. In
ÅÅjj zz ÑÑ
ÅÅjj 2k + (k − 2k )e−k g[M]0 t zzz
k j /(2k n − k g)

ÑÑ
kg
ÅÅk n ÑÑ
the model of eq 1, even stating this master equation in a
ÅÅÇ { ÑÑÖ
− 1
g n concise form runs into substantial difficulties because of the
(3)
complicated and exceptionally large state space described in
The cube-root number-average size of the nanoparticles could the previous paragraph. Despite our attempts, no way to use
also be obtained analytically: the master equation in any meaningful way was found.
5431 https://doi.org/10.1021/acs.chemmater.0c04688
Chem. Mater. 2021, 33, 5430−5436
Chemistry of Materials


pubs.acs.org/cm Comments

In such complicated cases, it is still possible to run Monte THE STOCHASTIC APPROACH: RESULTS AND
Carlo simulations based on CDS kinetics,45−49 an approach COMPARISON
widely known as the Gillespie algorithm.16 In this method, the For maximizing the information content of the graphs
time-dependent propensity pi (or infinitesimal transition presented here and reducing the number of parameters, the
probability) of each step is defined analogously to the results will be shown with scaled variables.43 This technique
deterministic rates: also facilitates comparisons of stochastic and deterministic
results. Instead of time, scaled time will be used here: in the
kn k n c1(t )[c1(t ) − 1] deterministic approach, scaled time can be calculated as
2M → C2 p1 (t ) =
NAV 2 tkg[M]0, whereas the corresponding formula is tkgn/(NAV) for
the stochastic approach. Similarly, the scaled concentration of
kg, i the monomeric unit is [M]/[M]0 in the deterministic approach
Ci + M → Ci + 1 i≥2 and c1/n in the stochastic approach. Finally, the scaled particle
ikg
pi (t ) = c1(t )ci(t ) size is r/r0 in both types of calculations. It should also be noted
NAV (6) that, with the introduction of these scaled parameters, the final
results will depend only on the dimensionless rate constant
Note that the number of different reactions and propensities ratio of kn/kg and not on the individual values of the rate
cannot exceed n, as the highest possible nonzero molecule constants. Therefore, the displayed results fully characterize
number is cn. In these formulas, NA is Avogadro’s constant and the entire meaningful parameter space, so they cover all
V is the volume of the reactor. The inclusion of these possible values of the two rate constants (kn or kg) and the
parameters is necessary for connecting the deterministic and single initial concentration ([M]0) that can be reasonably
stochastic approaches.16 encountered in reality.
The essence of the simulation is that, in each step, two Figure 1 shows the results for the time dependence of the
scaled concentration of the monomeric unit from five
independent, uniformly distributed random numbers are
generated between 0 and 1; these are rnd1 and rnd2. The
first random number is used to increment the time according
to the following equation:

ln rnd1
t new = t old − n
∑ j = 1 pj (t old) (7)

The second random number is used to decide which of the


steps with nonzero propensity occurs. Step i occurs if and only
if the following inequity is satisfied:
i−1 i
∑ j = 1 pj (t old) ∑ j = 1 pj (t old)
≤ rnd 2 <
n
∑ j = 1 pj (t old)
n
∑ j = 1 pj (t old) Figure 1. Scaled concentration of the monomeric unit in five
(8)
repetitive simulation runs and the deterministic expectation for the
same process for the system of eq 1. Parameters: n = 106, kn/kg = 10−6.
The occurrence of step i causes the following changes in the
molecule numbers:
randomly selected stochastic simulation runs and the
c1(t new ) = c1(t old) − 2; c 2(t new ) = c 2(t old) + 1 corresponding deterministic curve at the fixed values of kn/kg
if i = 1 = 10−6 and n = 106. As expected, the individual Gillespie
simulations are noticeably different, and the deviations are very
c1(t new ) = c1(t old) − 1; ci(t new ) = ci(t old) − 1; much in line with the known time dependence of simple
ci + 1(t new ) = ci + 1(t old) + 1 if i > 1 (9) autocatalytic processes.50 It is also clear that some of the
simulation runs are faster, whereas others are slower than the
These steps are repeated until c1 = 0, which means that all the deterministic expectation. Two more figures showing how the
propensities are zero, i.e., an unchangeable final state is simulations depend on the values of n and kn/kg are found in
reached. the Supporting Information as Figures S1 and S2.
From the same five simulation runs, the time dependence of
A high number of simulations from identical initial
the scaled average size was also determined. These results are
conditions and parameters were done (repetitions) in order
shown in Figure 2 along with the deterministic curve. When
to approximate probability distributions when necessary. The viewing this figure, it should be kept in mind that the size of a
initial number of molecules (n) was usually selected to be particle is proportional to the cube root of the number of
between 106 and 108. This is small compared to the usual monomeric units in it, so a scaled size of 40 actually implies
number of monomeric units in any realistic cases but was 64 000 monomeric units in the particle. The comparison of the
computationally manageable and already gave highly con- runs with each other and the deterministic predictions is very
clusive results for the comparison of the stochastic and similar to the one shown in Figure 1 for the scaled
deterministic approaches. Most simulation were done at kn/kg concentration of the monomeric unit. A close inspection of
≪ 1 (slow nucleation compared to particle growth), because the stochastic results shows that some of the curves are not
otherwise the particles remain very small.21 entirely monotonic. The overwhelming trend is an increase in
5432 https://doi.org/10.1021/acs.chemmater.0c04688
Chem. Mater. 2021, 33, 5430−5436
Chemistry of Materials pubs.acs.org/cm Comments

mentioned, Figure 3 contains 100 points, which was the


number individual runs. In deterministic kinetics, the kinetic
curves are uniquely determined by the initial conditions, so the
half-life is a single value without any distribution. This single
value is shown by a gray vertical line in Figure 3. It is seen that
this line is very close to the time of 50% probability.
Furthermore, a red continuous curve is also included in the
graph, which shows the best fitting normal distribution to the
curve. This line fits the simulation points quite nicely. Small
fluctuations in stochastic kinetics are expected to follow normal
distribution because of the central limit theorem of probability
theory.16 The adherence of the distribution of half-lives to
Figure 2. Scaled average size as a function of scaled time in five normal distribution predicts that the relative fluctuations will
repetitive simulation runs and the deterministic expectation for the be a lot smaller as the initial number of particles is increased.16
same process for the system of eq 1. Parameters: n = 106, kn/kg = 10−6. It is quite noticeable that, even at a low particle number of 108,
no obvious deviations are seen from normal distribution. In
average size as time passes, but occasionally short periods of many cases where the stochastic effects are present even in
decrease can also be seen. This phenomenon has no parallel in macroscopic systems, such deviations are very large at small
the deterministic calculations and is caused by single molecule numbers.16,25,28 In the Supporting Information, two
nucleation events. Whenever a new nucleus is formed, it has more graphs similar to Figure 3 are included as Figures S3 and
a very small size (two monomeric units form a nucleus), so it S4. They show the analogous distributions for different values
may decrease the average somewhat. kn/kg (10−6 and 10−8). The expected half-lives under these
To characterize the magnitude of the deviations expected conditions are different from that shown in Figure 3, but the
because of the stochastic nature of the process, simulation runs general picture and the close adherence to the normal
were started from exactly the same initial state 100 times. distribution is the same.
Drawing a graph similar to Figure 1 or 2 in this case is not As already mentioned in the introduction, one of the most
feasible with 100 runs. Instead, the distribution of a selected important descriptors of a nanoparticle synthesis method is the
property, the monomeric unit half-life (i.e., the time necessary size distribution obtained. This can be calculated for the final
for the concentration of the monomeric unit to fall to one-half state from both stochastic simulations and deterministic
of the initial values), is shown in Figure 3. It is to be noted that formulas. A comparison is shown in Figure 4. As explained

Figure 3. Cumulative distribution of half-lives for the concentration of Figure 4. Final particle size distribution in stochastic simulations and
the monomeric unit for the system of eq 1. Parameters n = 108; kn/kg the deterministic prediction for the system of eq 1. Parameters: n =
= 10−7. 108; kn/kg = 10−6.

this is not analogous to the half-life of first order processes,


which is independent of the initial concentration,43 but can still for Figure 3, a comparison like this should be done using the
be used for comparison purposes here as the graph only shows cumulative distribution function and not the probability
different repetitions of simulation from exactly the same initial density function. So the ordinate of Figure 4 shows F(r),
conditions. Also, it is an important observation in stochastic which is the probability that the size of a randomly selected
kinetics that comparisons of distributions should be done using particle in the final mixture is smaller than r. Only three
the cumulative distribution function rather than the probability simulations are shown in the graph as all three run very close
density function.16,25,28,29,50 In nonexpert use, the probability to each other (they are not even seen separately for most
density function is often preferred as it is the direct equivalent regions of the figure). A much higher number of actual
of a histogram. However, histograms involve highly arbitrary simulation runs were done, but this unique distribution was
categorization of events, which is understood to fundamentally always reached as the final result. Furthermore, the distribution
influence the final picture drawn. The use of cumulative obtained in the stochastic run matches very closely the one
distribution functions avoids these problems entirely. In the predicted by the deterministic approach. Some deviations can
present case, the ordinate of Figure 3 shows the cumulative be seen above the scaled particle size of 60, but these are
distribution functions F(t1/2), which is the probability of arguably small artifacts connected to number representation
obtaining a half-life shorter than t1/2 in a simulation. As already issues in calculating the alternately signed sum of large
5433 https://doi.org/10.1021/acs.chemmater.0c04688
Chem. Mater. 2021, 33, 5430−5436
Chemistry of Materials pubs.acs.org/cm Comments

binomial coefficients in eq 3, a computational problem noted


in several previous works.17,30,31
It should be noted that the initial number of monomeric
units for calculating the results shown in Figure 4 was 108,
which is still a very small one in chemistry. The match of the
result of even a single stochastic run with the deterministic
prediction is very good here. Hence, the result presented in
this section proves that Kurtz’s theorem37 for the nucleation−
growth mechanism with autocatalytic phenomena already gives
practical agreement between the two approaches even at
consideration of just 108 monomers. This agreement will be
even better when a larger initial number of molecules is
considered.


Figure 5. Cumulative distribution of half-lives for the concentration of
the monomeric unit for the system of eq 10. Parameters n = 107; kn/kg
STOCHASTIC−DETERMINISTIC COMPARISON IN A = 5 × 10−8.
SECOND MODEL
As already pointed out, the full analytical deterministic solution
is not known for any other model, and numerical solutions
cannot be obtained either because the number of concen-
trations is infinitely large. However, we could develop a novel
numerical integration method that could be used for the case
when the growth rate constants are independent of the size of
the nanoparticle. The chemical model is the following:
kn
2M → C2 rn = k n[M]2
kg, i
Ci + M → Ci + 1 i≥2 rg, i = kg[M][Ci]
(10)
Figure 6. Final particle size distribution in stochastic simulations and
In this case, the ordinary differential equations are the deterministic prediction for the system of eq 10. Parameters: n =
d[M]
∞ ∞ 107; kn/kg = 5 × 10−7.
= − 2rn − ∑ rg ,j = −2kn[M]2 − kg[M] ∑ [Cj]
dt j=2 j=2

d [C 2 ]
= rn − rg ,2 = k n[M]2 − kg[M][C2]
dt two different simulation runs agree with each other very well
d[Ci] and also agree with the deterministic prediction. The
= rg , i − 1 − rg , i = kg[M][Ci − 1] − kg[M][Ci] i>2
dt (11) distribution itself is very different from the one shown in
This system of differential equations was solved by a Figure 4, no doubt because the two models are substantially
specifically developed numerical method, which is described different. Yet the fact that stochastic and deterministic
in detail in the Supporting Information. It should be predictions for the size distributions are within any reasonable
emphasized that this method is only suitable for the system experimental limit is clear from both Figures 4 and 6.
shown in eq 11 and cannot be generalized.
The stochastic Gillespie simulations could be done very
similarly to the previous case. The only difference was that the
■ SUMMARY
In response to the suggestion of James Martin,12 which called
propensity function pi(t), originally defined in eq 6, had to be for a consideration of single nucleation events, and then the
calculated by a different formula: suggestion of Finke,13 which called for considering the
kg stochastic nature of those single nucleation events, in
pi (t ) = c1(t )ci(t ) nanoparticle formation models, a comparison of the stochastic
NAV (12) (continuous time discrete state) and the more common
Similarly to the previous model, the deterministic and deterministic approaches was presented here in two
stochastic predictions were compared systematically. Figure 5 nucleation−growth type nanoparticle formation models.
shows the distribution of half-lives at kn/kg = 5 × 10−7. The Detectable fluctuations in reaction time were found, which
half-life values are understandably different from those are very similar to those demonstrated in simple autocatalytic
calculated in the previous model, but otherwise the graph is reactions50 and are by no means specific to nanoparticle
very similar to Figure 3: the distribution obtained from 100 formation. The results showed that even at initial monomer
individual runs is described by a normal distribution quite well. molecule numbers as low as 10 7, the stochastic and
The same is shown for two different values of kn/kg in Figures deterministic approaches predict almost identical particle size
S5 and S6 in the Supporting Information. distributions. This comparison validates the deterministic
The final distribution of nanoparticle sizes is shown for one kinetic approach and calculations for the size distribution in
particular value of kn/kg (= 5 × 10−8) in Figure 6. Here n = 107 such systems.
was selected, so the initial number of monomeric units was Rebeka Szabó
even lower than for the analogous data shown in Figure 4. The Gábor Lente orcid.org/0000-0003-2022-2156
5434 https://doi.org/10.1021/acs.chemmater.0c04688
Chem. Mater. 2021, 33, 5430−5436
Chemistry of Materials


pubs.acs.org/cm Comments

ASSOCIATED CONTENT Particle Formation en Route to Particle Average Size and Size
*
sı Supporting Information
Distribution Understanding and Control. J. Am. Chem. Soc. 2019, 141,
15827−15839.
The Supporting Information is available free of charge at (10) Ö zkar, S.; Finke, R. G. Nanoparticle Formation Kinetics and
https://pubs.acs.org/doi/10.1021/acs.chemmater.0c04688. Mechanistic Studies Important to Mechanism-Based Particle Size
An example of the value of the combinatorial partition Control: Evidence for Ligand-Based Slowing of the Autocatalytic
function, a sample Matlab code used in implementing Surface Growth Step Plus Postulated Mechanisms. J. Phys. Chem. C
Gillespie’s algorithm, and the four additional graphs 2019, 123, 14047−14057.
referred to in the text (PDF) (11) Handwerk, D. R.; Shipman, P. D.; Whitehead, C. B.; Ö zkar, S.;


Finke, R. G. Particle Size Distributions via Mechanism-Enabled
Population Balance Modeling. J. Phys. Chem. C 2020, 124, 4852−
AUTHOR INFORMATION 4880.
(12) Martin, J. D. Particle Size Is a Primary Determinant for
Complete contact information is available at: Sigmoidal Kinetics of Nanoparticle Formation: A “Disproof” of the
https://pubs.acs.org/10.1021/acs.chemmater.0c04688 Finke−Watzky (F-W) Nanoparticle Nucleation and Growth Mech-
anism. Chem. Mater. 2020, 32, 3651−3656.
Notes (13) Finke, R. G.; Watzky, M. A.; Whitehead, C. B. Response to
The authors declare no competing financial interest. “Particle Size Is a Primary Determinant for Sigmoidal Kinetics of

■ ACKNOWLEDGMENTS
This work was supported by the Higher Education Institu-
Nanoparticle Formation: A “Disproof” of the Finke−Watzky (F-W)
Nanoparticle Nucleation and Growth Mechanism. Chem. Mater.
2020, 32, 3657−3672.
tional Excellence Programme of the Ministry of Human (14) É rdi, P.; Tóth, J. Mathematical Models of Chemical Reactions;
Capacities in Hungary, within the framework of the first Manchester University Press: Manchester, U.K., 1989.
thematic programme of the University of Pécs. The National (15) Santillán, M. Chemical Kinetics, Stochastic Processes, and
Irreversible Thermodynamics; Springer: Cham, Switzerland, 2014.
Research, Development, and Innovation Office of Hungary
(16) É rdi, P.; Lente, G. Stochastic Chemical Kinetics − Theory and
also supported this work under Grant No. SNN 125739.


(Mostly) Systems Biological Applications; Springer: Cham, Switzerland,
2014.
REFERENCES (17) Kang, K.; Redner, S.; Meakin, P.; Leyvraz, F. Long-Time
(1) Watzky, M. A.; Finke, R. G. Transition Metal Nanocluster Crossover Phenomena in Coagulation Kinetics. Phys. Rev. A: At., Mol.,
Formation Kinetic and Mechanistic Studies. A New Mechanism Opt. Phys. 1986, 33, 1171−1182.
When Hydrogen Is the Reductant: Slow, Continuous Nucleation and (18) McCoy, B. J.; Madras, G. Evolution to Similarity Solutions for
Fast Autocatalytic Surface Growth. J. Am. Chem. Soc. 1997, 119, Fragmentation and Aggregation. J. Colloid Interface Sci. 1998, 201,
10382−10400. 200−209.
(2) Watzky, M. A.; Morris, A. M.; Ross, E. D.; Finke, R. G. α- (19) McCoy, B. J. A Population Balance Framework for Nucleation,
Synuclein Aggregation Variable Temperature and Variable pH Kinetic Growth, and Aggregation. Chem. Eng. Sci. 2002, 57, 2279−2285.
Data: A Re-Analysis using the Finke−Watzky 2-Step Model of (20) Rempel, J. Y.; Bawendi, M. G.; Jensen, K. F. Insights into the
Nucleation and Autocatalytic Growth. Biochemistry 2008, 47, 10790− Kinetics of Semiconductor Nanocrystal Nucleation and Growth. J.
10800. Am. Chem. Soc. 2009, 131, 4479−4489.
(3) Finney, E. E.; Finke, R. G. Is There a Minimal Chemical (21) Szabó, R.; Lente, G. Full Analytical Solution of a Nucleation-
Mechanism Underlying Classical Avrami-Erofe’ev Treatments of Growth Type Kinetic Model of Nanoparticle Formation. J. Math.
Phase-Transformation Kinetic Data? Chem. Mater. 2009, 21, 4692− Chem. 2019, 57, 616−631.
4705. (22) Zibareva, I. V.; Ilina, L. Y.; Vedyagin, A. A. Catalysis by
(4) Mondloch, J. E.; Finke, R. G. Kinetic Evidence for Bimolecular Nanoparticles: the Main Features and Trend. React. Kinet., Mech.
Nucleation in Supported-Transition-Metal-Nanoparticle Catalyst
Catal. 2019, 127, 19−24.
Formation in Contact with Solution: The Prototype Ir(1,5-COD)-
(23) Delbrück, M. Statistical Fluctuation in Autocatalytic Reactions.
Cl/γ-Al2O3 to Ir(0)∼900/γ-Al2O3 System. ACS Catal. 2012, 2, 298−
J. Chem. Phys. 1940, 8, 120−124.
305.
(24) Drummond, P. G.; Vaughan, T. G.; Drummond, A. J.
(5) Laxson, W. W.; Finke, R. G. Nucleation is Second Order: An
Extinction Times in Autocatalytic Systems. J. Phys. Chem. A 2010,
Apparent Kinetically Effective Nucleus of Two for Ir(0)n Nano-
particle Formation from [(1,5-COD)IrI·P2W15Nb3O62]8− Plus 114, 10481−10491.
Hydrogen. J. Am. Chem. Soc. 2014, 136, 17601−17615. (25) Lente, G. Stochastic Kinetic Models of Chiral Autocatalysis: a
(6) Bentea, L.; Watzky, M. A.; Finke, R. G. Sigmoidal Nucleation General Tool for the Quantitative Interpretation of Total Asymmetric
and Growth Curves Across Nature Fit by the Finke−Watzky Model of Synthesis. J. Phys. Chem. A 2005, 109, 11058−11063.
Slow Continuous Nucleation and Autocatalytic Growth: Explicit (26) Shao, J.; Liu, L. Stochastic Fluctuations and Chiral Symmetry
Formulas for the Lag and Growth Times Plus Other Key Insights. J. Breaking: Exact Solution of Lente Model. J. Phys. Chem. A 2007, 111,
Phys. Chem. C 2017, 121, 5302−5312. 9570−9572.
(7) Iashchishyn, I. A.; Sulskis, D.; Nguyen Ngoc, M.; Smirnovas, V.; (27) Saito, Y.; Sugimori, T.; Hyuga, H. Stochastic Approach to
Morozova-Roche, L. A. Finke−Watzky Two-Step Nucleation− Enantiomeric Excess Amplification and Chiral Symmetry Breaking. J.
Autocatalysis Model of S100A9 Amyloid Formation: Protein Phys. Soc. Jpn. 2007, 76, 044802.
Misfolding as “Nucleation” Event. ACS Chem. Neurosci. 2017, 8, (28) Dóka, É .; Lente, G. Mechanism-based Chemical Understanding
2152−2158. of Chiral Symmetry Breaking in the Soai Reaction. A Combined
(8) Whitehead, C. B.; Ö zkar, S.; Finke, R. G. LaMer’s 1950 Model Probabilistic and Deterministic Description of Chemical Reactions. J.
for Particle Formation of Instantaneous Nucleation and Diffusion- Am. Chem. Soc. 2011, 133, 17878−17881.
Controlled Growth: A Historical Look at the Model’s Origins, (29) Lente, G. The Role of Stochastic Models in Interpreting the
Assumptions, Equations, and Underlying Sulfur Sol Formation Origins of Biological Chirality. Symmetry 2010, 2, 767−798.
Kinetics Data. Chem. Mater. 2019, 31, 7116−713. (30) Lente, G. Stochastic Analysis of the Parity Violating Energy
(9) Handwerk, D. R.; Shipman, P. D.; Whitehead, C. B.; Ö zkar, S.; Differences between Enantiomers and its Implications for the Origin
Finke, R. G. Mechanism-Enabled Population Balance Modeling of of Biological Chirality. J. Phys. Chem. A 2006, 110, 12711−12713.

5435 https://doi.org/10.1021/acs.chemmater.0c04688
Chem. Mater. 2021, 33, 5430−5436
Chemistry of Materials pubs.acs.org/cm Comments

(31) Lente, G. The Effect of Parity Violation on Kinetic Models of


Enantioselective Autocatalyis. Phys. Chem. Chem. Phys. 2007, 9,
6134−6141.
(32) Lente, G.; Ditrói, T. Stochastic Kinetic Analysis of the Frank
model. Stochastic Approach to Flow-Through Reactors. J. Phys. Chem.
B 2009, 113, 7237−7242.
(33) Arányi, P.; Tóth, J. A Full Stochastic Description of the
Michaelis-Menten Reaction for Small Systems. Acta Biochim. Biophys.
Acad. Sci. Hung. 1977, 12, 375−388.
(34) Kou, S. C.; Cherayil, B. J.; Min, W.; English, B. P.; Xie, X. S.
Single-Molecule Michaelis−Menten Equations. J. Phys. Chem. B 2005,
109, 19068−19081.
(35) Rissin, D. M.; Gorris, H. H.; Walt, D. R. Distinct and Long-
Lived Activity States of Single Enzyme Molecules. J. Am. Chem. Soc.
2008, 130, 5349−5353.
(36) Dóka, É .; Lente, G. Stochastic Mapping of the Michaelis-
Menten Mechanism. J. Chem. Phys. 2012, 136, 054111.
(37) Kurtz, T. G. The relationship between stochastic and
deterministic models for chemical reactions. J. Chem. Phys. 1972,
57, 2976−2978.
(38) Gillespie, D. T. Deterministic Limit of Stochastic Chemical
Kinetics. J. Phys. Chem. B 2009, 113, 1640−1644.
(39) Blomberg, C. Fluctuations for good and bad: the role of noise
in living systems. Phys. Life Rev. 2006, 3, 133−161.
(40) Zámbó, D.; Pothorszky, S.; Brougham, D. F.; Deák, A.
Aggregation kinetics and cluster structure of amino-PEG covered gold
nanoparticles. RSC Adv. 2016, 6, 27151−27157.
(41) Labidi, S.; Jia, Z.; Amar, M. B.; Chhor, K.; Kanaev, A.
Nucleation and growth kinetics of zirconium-oxo-alkoxy nano-
particles. Phys. Chem. Chem. Phys. 2015, 17, 2651−2659.
(42) Forgács, A.; Moldován, K.; Herman, P.; Baranyai, E.; Fábián, I.;
Lente, G.; Kalmár, J. Kinetic Model for Hydrolytic Nucleation and
Growth of TiO2 Nanoparticles. J. Phys. Chem. C 2018, 122, 19161−
19170.
(43) Lente, G. Deterministic Kinetics in Chemistry and Systems Biology;
Springer: Cham, Switzerland, 2015.
(44) Barvinok, A. Combinatorics and Complexity of Partition
Functions; Springer: Cham, Switzerland, 2016.
(45) Sipos, T.; Tóth, J.; É rdi, P. Stochastic simulation of complex
chemical reactions by digital computer. I. The model. React. Kinet.
Catal. Lett. 1974, 1, 113−117.
(46) Sipos, T.; Tóth, J.; É rdi, P. Stochastic simulation of complex
chemical reactions by digital computer. II. Applications. React. Kinet.
Catal. Lett. 1974, 1, 209−213.
(47) Gillespie, D. T. A general method for numerically simulating
the stochastic time evolution of coupled chemical reactions. J.
Comput. Phys. 1976, 22, 403−434.
(48) Gillespie, D. T. Exact stochastic simulation of coupled chemical
reactions. J. Phys. Chem. 1977, 81, 2340−2361.
(49) Gillespie, D. T.; Hellander, A.; Petzold, L. R. Perspective:
stochastic algorithms for chemical kinetics. J. Chem. Phys. 2013, 138,
170901.
(50) Lente, G. A novel method to compute the time dependence of
state distributions in the stochastic kinetic description of an
autocatalytic system. Comput. Chem. Eng. 2019, 125, 587−593.

5436 https://doi.org/10.1021/acs.chemmater.0c04688
Chem. Mater. 2021, 33, 5430−5436

You might also like